1This is gcc.info, produced by makeinfo version 6.5 from gcc.texi. 2 3Copyright (C) 1988-2020 Free Software Foundation, Inc. 4 5 Permission is granted to copy, distribute and/or modify this document 6under the terms of the GNU Free Documentation License, Version 1.3 or 7any later version published by the Free Software Foundation; with the 8Invariant Sections being "Funding Free Software", the Front-Cover Texts 9being (a) (see below), and with the Back-Cover Texts being (b) (see 10below). A copy of the license is included in the section entitled "GNU 11Free Documentation License". 12 13 (a) The FSF's Front-Cover Text is: 14 15 A GNU Manual 16 17 (b) The FSF's Back-Cover Text is: 18 19 You have freedom to copy and modify this GNU Manual, like GNU software. 20Copies published by the Free Software Foundation raise funds for GNU 21development. 22INFO-DIR-SECTION Software development 23START-INFO-DIR-ENTRY 24* gcc: (gcc). The GNU Compiler Collection. 25* g++: (gcc). The GNU C++ compiler. 26* gcov: (gcc) Gcov. 'gcov'--a test coverage program. 27* gcov-tool: (gcc) Gcov-tool. 'gcov-tool'--an offline gcda profile processing program. 28* gcov-dump: (gcc) Gcov-dump. 'gcov-dump'--an offline gcda and gcno profile dump tool. 29* lto-dump: (gcc) lto-dump. 'lto-dump'--Tool for 30dumping LTO object files. 31END-INFO-DIR-ENTRY 32 33 This file documents the use of the GNU compilers. 34 35 Copyright (C) 1988-2020 Free Software Foundation, Inc. 36 37 Permission is granted to copy, distribute and/or modify this document 38under the terms of the GNU Free Documentation License, Version 1.3 or 39any later version published by the Free Software Foundation; with the 40Invariant Sections being "Funding Free Software", the Front-Cover Texts 41being (a) (see below), and with the Back-Cover Texts being (b) (see 42below). A copy of the license is included in the section entitled "GNU 43Free Documentation License". 44 45 (a) The FSF's Front-Cover Text is: 46 47 A GNU Manual 48 49 (b) The FSF's Back-Cover Text is: 50 51 You have freedom to copy and modify this GNU Manual, like GNU software. 52Copies published by the Free Software Foundation raise funds for GNU 53development. 54 55 56File: gcc.info, Node: Top, Next: G++ and GCC 57 58Introduction 59************ 60 61This manual documents how to use the GNU compilers, as well as their 62features and incompatibilities, and how to report bugs. It corresponds 63to the compilers (GCC) version 10.5.0. The internals of the GNU 64compilers, including how to port them to new targets and some 65information about how to write front ends for new languages, are 66documented in a separate manual. *Note Introduction: (gccint)Top. 67 68* Menu: 69 70* G++ and GCC:: You can compile C or C++ programs. 71* Standards:: Language standards supported by GCC. 72* Invoking GCC:: Command options supported by 'gcc'. 73* C Implementation:: How GCC implements the ISO C specification. 74* C++ Implementation:: How GCC implements the ISO C++ specification. 75* C Extensions:: GNU extensions to the C language family. 76* C++ Extensions:: GNU extensions to the C++ language. 77* Objective-C:: GNU Objective-C runtime features. 78* Compatibility:: Binary Compatibility 79* Gcov:: 'gcov'--a test coverage program. 80* Gcov-tool:: 'gcov-tool'--an offline gcda profile processing program. 81* Gcov-dump:: 'gcov-dump'--an offline gcda and gcno profile dump tool. 82* lto-dump:: 'lto-dump'--Tool for dumping LTO 83object files. 84* Trouble:: If you have trouble using GCC. 85* Bugs:: How, why and where to report bugs. 86* Service:: How To Get Help with GCC 87* Contributing:: How to contribute to testing and developing GCC. 88 89* Funding:: How to help assure funding for free software. 90* GNU Project:: The GNU Project and GNU/Linux. 91 92* Copying:: GNU General Public License says 93 how you can copy and share GCC. 94* GNU Free Documentation License:: How you can copy and share this manual. 95* Contributors:: People who have contributed to GCC. 96 97* Option Index:: Index to command line options. 98* Keyword Index:: Index of concepts and symbol names. 99 100 101File: gcc.info, Node: G++ and GCC, Next: Standards, Up: Top 102 1031 Programming Languages Supported by GCC 104**************************************** 105 106GCC stands for "GNU Compiler Collection". GCC is an integrated 107distribution of compilers for several major programming languages. 108These languages currently include C, C++, Objective-C, Objective-C++, 109Fortran, Ada, D, Go, and BRIG (HSAIL). 110 111 The abbreviation "GCC" has multiple meanings in common use. The 112current official meaning is "GNU Compiler Collection", which refers 113generically to the complete suite of tools. The name historically stood 114for "GNU C Compiler", and this usage is still common when the emphasis 115is on compiling C programs. Finally, the name is also used when 116speaking of the "language-independent" component of GCC: code shared 117among the compilers for all supported languages. 118 119 The language-independent component of GCC includes the majority of the 120optimizers, as well as the "back ends" that generate machine code for 121various processors. 122 123 The part of a compiler that is specific to a particular language is 124called the "front end". In addition to the front ends that are 125integrated components of GCC, there are several other front ends that 126are maintained separately. These support languages such as Mercury, and 127COBOL. To use these, they must be built together with GCC proper. 128 129 Most of the compilers for languages other than C have their own names. 130The C++ compiler is G++, the Ada compiler is GNAT, and so on. When we 131talk about compiling one of those languages, we might refer to that 132compiler by its own name, or as GCC. Either is correct. 133 134 Historically, compilers for many languages, including C++ and Fortran, 135have been implemented as "preprocessors" which emit another high level 136language such as C. None of the compilers included in GCC are 137implemented this way; they all generate machine code directly. This 138sort of preprocessor should not be confused with the "C preprocessor", 139which is an integral feature of the C, C++, Objective-C and 140Objective-C++ languages. 141 142 143File: gcc.info, Node: Standards, Next: Invoking GCC, Prev: G++ and GCC, Up: Top 144 1452 Language Standards Supported by GCC 146************************************* 147 148For each language compiled by GCC for which there is a standard, GCC 149attempts to follow one or more versions of that standard, possibly with 150some exceptions, and possibly with some extensions. 151 1522.1 C Language 153============== 154 155The original ANSI C standard (X3.159-1989) was ratified in 1989 and 156published in 1990. This standard was ratified as an ISO standard 157(ISO/IEC 9899:1990) later in 1990. There were no technical differences 158between these publications, although the sections of the ANSI standard 159were renumbered and became clauses in the ISO standard. The ANSI 160standard, but not the ISO standard, also came with a Rationale document. 161This standard, in both its forms, is commonly known as "C89", or 162occasionally as "C90", from the dates of ratification. To select this 163standard in GCC, use one of the options '-ansi', '-std=c90' or 164'-std=iso9899:1990'; to obtain all the diagnostics required by the 165standard, you should also specify '-pedantic' (or '-pedantic-errors' if 166you want them to be errors rather than warnings). *Note Options 167Controlling C Dialect: C Dialect Options. 168 169 Errors in the 1990 ISO C standard were corrected in two Technical 170Corrigenda published in 1994 and 1996. GCC does not support the 171uncorrected version. 172 173 An amendment to the 1990 standard was published in 1995. This 174amendment added digraphs and '__STDC_VERSION__' to the language, but 175otherwise concerned the library. This amendment is commonly known as 176"AMD1"; the amended standard is sometimes known as "C94" or "C95". To 177select this standard in GCC, use the option '-std=iso9899:199409' (with, 178as for other standard versions, '-pedantic' to receive all required 179diagnostics). 180 181 A new edition of the ISO C standard was published in 1999 as ISO/IEC 1829899:1999, and is commonly known as "C99". (While in development, 183drafts of this standard version were referred to as "C9X".) GCC has 184substantially complete support for this standard version; see 185<http://gcc.gnu.org/c99status.html> for details. To select this 186standard, use '-std=c99' or '-std=iso9899:1999'. 187 188 Errors in the 1999 ISO C standard were corrected in three Technical 189Corrigenda published in 2001, 2004 and 2007. GCC does not support the 190uncorrected version. 191 192 A fourth version of the C standard, known as "C11", was published in 1932011 as ISO/IEC 9899:2011. (While in development, drafts of this 194standard version were referred to as "C1X".) GCC has substantially 195complete support for this standard, enabled with '-std=c11' or 196'-std=iso9899:2011'. A version with corrections integrated was prepared 197in 2017 and published in 2018 as ISO/IEC 9899:2018; it is known as "C17" 198and is supported with '-std=c17' or '-std=iso9899:2017'; the corrections 199are also applied with '-std=c11', and the only difference between the 200options is the value of '__STDC_VERSION__'. 201 202 A further version of the C standard, known as "C2X", is under 203development; experimental and incomplete support for this is enabled 204with '-std=c2x'. 205 206 By default, GCC provides some extensions to the C language that, on 207rare occasions conflict with the C standard. *Note Extensions to the C 208Language Family: C Extensions. Some features that are part of the C99 209standard are accepted as extensions in C90 mode, and some features that 210are part of the C11 standard are accepted as extensions in C90 and C99 211modes. Use of the '-std' options listed above disables these extensions 212where they conflict with the C standard version selected. You may also 213select an extended version of the C language explicitly with 214'-std=gnu90' (for C90 with GNU extensions), '-std=gnu99' (for C99 with 215GNU extensions) or '-std=gnu11' (for C11 with GNU extensions). 216 217 The default, if no C language dialect options are given, is 218'-std=gnu11'. 219 220 The ISO C standard defines (in clause 4) two classes of conforming 221implementation. A "conforming hosted implementation" supports the whole 222standard including all the library facilities; a "conforming 223freestanding implementation" is only required to provide certain library 224facilities: those in '<float.h>', '<limits.h>', '<stdarg.h>', and 225'<stddef.h>'; since AMD1, also those in '<iso646.h>'; since C99, also 226those in '<stdbool.h>' and '<stdint.h>'; and since C11, also those in 227'<stdalign.h>' and '<stdnoreturn.h>'. In addition, complex types, added 228in C99, are not required for freestanding implementations. 229 230 The standard also defines two environments for programs, a 231"freestanding environment", required of all implementations and which 232may not have library facilities beyond those required of freestanding 233implementations, where the handling of program startup and termination 234are implementation-defined; and a "hosted environment", which is not 235required, in which all the library facilities are provided and startup 236is through a function 'int main (void)' or 'int main (int, char *[])'. 237An OS kernel is an example of a program running in a freestanding 238environment; a program using the facilities of an operating system is an 239example of a program running in a hosted environment. 240 241 GCC aims towards being usable as a conforming freestanding 242implementation, or as the compiler for a conforming hosted 243implementation. By default, it acts as the compiler for a hosted 244implementation, defining '__STDC_HOSTED__' as '1' and presuming that 245when the names of ISO C functions are used, they have the semantics 246defined in the standard. To make it act as a conforming freestanding 247implementation for a freestanding environment, use the option 248'-ffreestanding'; it then defines '__STDC_HOSTED__' to '0' and does not 249make assumptions about the meanings of function names from the standard 250library, with exceptions noted below. To build an OS kernel, you may 251well still need to make your own arrangements for linking and startup. 252*Note Options Controlling C Dialect: C Dialect Options. 253 254 GCC does not provide the library facilities required only of hosted 255implementations, nor yet all the facilities required by C99 of 256freestanding implementations on all platforms. To use the facilities of 257a hosted environment, you need to find them elsewhere (for example, in 258the GNU C library). *Note Standard Libraries: Standard Libraries. 259 260 Most of the compiler support routines used by GCC are present in 261'libgcc', but there are a few exceptions. GCC requires the freestanding 262environment provide 'memcpy', 'memmove', 'memset' and 'memcmp'. 263Finally, if '__builtin_trap' is used, and the target does not implement 264the 'trap' pattern, then GCC emits a call to 'abort'. 265 266 For references to Technical Corrigenda, Rationale documents and 267information concerning the history of C that is available online, see 268<http://gcc.gnu.org/readings.html> 269 2702.2 C++ Language 271================ 272 273GCC supports the original ISO C++ standard published in 1998, and the 2742011 and 2014 revisions. 275 276 The original ISO C++ standard was published as the ISO standard 277(ISO/IEC 14882:1998) and amended by a Technical Corrigenda published in 2782003 (ISO/IEC 14882:2003). These standards are referred to as C++98 and 279C++03, respectively. GCC implements the majority of C++98 ('export' is 280a notable exception) and most of the changes in C++03. To select this 281standard in GCC, use one of the options '-ansi', '-std=c++98', or 282'-std=c++03'; to obtain all the diagnostics required by the standard, 283you should also specify '-pedantic' (or '-pedantic-errors' if you want 284them to be errors rather than warnings). 285 286 A revised ISO C++ standard was published in 2011 as ISO/IEC 14882:2011, 287and is referred to as C++11; before its publication it was commonly 288referred to as C++0x. C++11 contains several changes to the C++ 289language, all of which have been implemented in GCC. For details see 290<https://gcc.gnu.org/projects/cxx-status.html#cxx11>. To select this 291standard in GCC, use the option '-std=c++11'. 292 293 Another revised ISO C++ standard was published in 2014 as ISO/IEC 29414882:2014, and is referred to as C++14; before its publication it was 295sometimes referred to as C++1y. C++14 contains several further changes 296to the C++ language, all of which have been implemented in GCC. For 297details see <https://gcc.gnu.org/projects/cxx-status.html#cxx14>. To 298select this standard in GCC, use the option '-std=c++14'. 299 300 The C++ language was further revised in 2017 and ISO/IEC 14882:2017 was 301published. This is referred to as C++17, and before publication was 302often referred to as C++1z. GCC supports all the changes in the new 303specification. For further details see 304<https://gcc.gnu.org/projects/cxx-status.html#cxx1z>. Use the option 305'-std=c++17' to select this variant of C++. 306 307 More information about the C++ standards is available on the ISO C++ 308committee's web site at <http://www.open-std.org/jtc1/sc22/wg21/>. 309 310 To obtain all the diagnostics required by any of the standard versions 311described above you should specify '-pedantic' or '-pedantic-errors', 312otherwise GCC will allow some non-ISO C++ features as extensions. *Note 313Warning Options::. 314 315 By default, GCC also provides some additional extensions to the C++ 316language that on rare occasions conflict with the C++ standard. *Note 317Options Controlling C++ Dialect: C++ Dialect Options. Use of the '-std' 318options listed above disables these extensions where they they conflict 319with the C++ standard version selected. You may also select an extended 320version of the C++ language explicitly with '-std=gnu++98' (for C++98 321with GNU extensions), or '-std=gnu++11' (for C++11 with GNU extensions), 322or '-std=gnu++14' (for C++14 with GNU extensions), or '-std=gnu++17' 323(for C++17 with GNU extensions). 324 325 The default, if no C++ language dialect options are given, is 326'-std=gnu++14'. 327 3282.3 Objective-C and Objective-C++ Languages 329=========================================== 330 331GCC supports "traditional" Objective-C (also known as "Objective-C 1.0") 332and contains support for the Objective-C exception and synchronization 333syntax. It has also support for a number of "Objective-C 2.0" language 334extensions, including properties, fast enumeration (only for 335Objective-C), method attributes and the @optional and @required keywords 336in protocols. GCC supports Objective-C++ and features available in 337Objective-C are also available in Objective-C++. 338 339 GCC by default uses the GNU Objective-C runtime library, which is part 340of GCC and is not the same as the Apple/NeXT Objective-C runtime library 341used on Apple systems. There are a number of differences documented in 342this manual. The options '-fgnu-runtime' and '-fnext-runtime' allow you 343to switch between producing output that works with the GNU Objective-C 344runtime library and output that works with the Apple/NeXT Objective-C 345runtime library. 346 347 There is no formal written standard for Objective-C or Objective-C++. 348The authoritative manual on traditional Objective-C (1.0) is 349"Object-Oriented Programming and the Objective-C Language": 350<http://www.gnustep.org/resources/documentation/ObjectivCBook.pdf> is 351the original NeXTstep document. 352 353 The Objective-C exception and synchronization syntax (that is, the 354keywords '@try', '@throw', '@catch', '@finally' and '@synchronized') is 355supported by GCC and is enabled with the option '-fobjc-exceptions'. 356The syntax is briefly documented in this manual and in the Objective-C 3572.0 manuals from Apple. 358 359 The Objective-C 2.0 language extensions and features are automatically 360enabled; they include properties (via the '@property', '@synthesize' and 361'@dynamic keywords'), fast enumeration (not available in Objective-C++), 362attributes for methods (such as 'deprecated', 'noreturn', 'sentinel', 363'format'), the 'unused' attribute for method arguments, the '@package' 364keyword for instance variables and the '@optional' and '@required' 365keywords in protocols. You can disable all these Objective-C 2.0 366language extensions with the option '-fobjc-std=objc1', which causes the 367compiler to recognize the same Objective-C language syntax recognized by 368GCC 4.0, and to produce an error if one of the new features is used. 369 370 GCC has currently no support for non-fragile instance variables. 371 372 The authoritative manual on Objective-C 2.0 is available from Apple: 373 * 374 <https://developer.apple.com/library/archive/documentation/Cocoa/Conceptual/ProgrammingWithObjectiveC/Introduction/Introduction.html> 375 376 For more information concerning the history of Objective-C that is 377available online, see <http://gcc.gnu.org/readings.html> 378 3792.4 Go Language 380=============== 381 382As of the GCC 4.7.1 release, GCC supports the Go 1 language standard, 383described at <https://golang.org/doc/go1>. 384 3852.5 HSA Intermediate Language (HSAIL) 386===================================== 387 388GCC can compile the binary representation (BRIG) of the HSAIL text 389format as described in HSA Programmer's Reference Manual version 1.0.1. 390This capability is typically utilized to implement the HSA runtime API's 391HSAIL finalization extension for a gcc supported processor. HSA 392standards are freely available at 393<http://www.hsafoundation.com/standards/>. 394 3952.6 D language 396============== 397 398GCC supports the D 2.0 programming language. The D language itself is 399currently defined by its reference implementation and supporting 400language specification, described at <https://dlang.org/spec/spec.html>. 401 4022.7 References for Other Languages 403================================== 404 405*Note GNAT Reference Manual: (gnat_rm)Top, for information on standard 406conformance and compatibility of the Ada compiler. 407 408 *Note Standards: (gfortran)Standards, for details of standards 409supported by GNU Fortran. 410 411 412File: gcc.info, Node: Invoking GCC, Next: C Implementation, Prev: Standards, Up: Top 413 4143 GCC Command Options 415********************* 416 417When you invoke GCC, it normally does preprocessing, compilation, 418assembly and linking. The "overall options" allow you to stop this 419process at an intermediate stage. For example, the '-c' option says not 420to run the linker. Then the output consists of object files output by 421the assembler. *Note Options Controlling the Kind of Output: Overall 422Options. 423 424 Other options are passed on to one or more stages of processing. Some 425options control the preprocessor and others the compiler itself. Yet 426other options control the assembler and linker; most of these are not 427documented here, since you rarely need to use any of them. 428 429 Most of the command-line options that you can use with GCC are useful 430for C programs; when an option is only useful with another language 431(usually C++), the explanation says so explicitly. If the description 432for a particular option does not mention a source language, you can use 433that option with all supported languages. 434 435 The usual way to run GCC is to run the executable called 'gcc', or 436'MACHINE-gcc' when cross-compiling, or 'MACHINE-gcc-VERSION' to run a 437specific version of GCC. When you compile C++ programs, you should 438invoke GCC as 'g++' instead. *Note Compiling C++ Programs: Invoking 439G++, for information about the differences in behavior between 'gcc' and 440'g++' when compiling C++ programs. 441 442 The 'gcc' program accepts options and file names as operands. Many 443options have multi-letter names; therefore multiple single-letter 444options may _not_ be grouped: '-dv' is very different from '-d -v'. 445 446 You can mix options and other arguments. For the most part, the order 447you use doesn't matter. Order does matter when you use several options 448of the same kind; for example, if you specify '-L' more than once, the 449directories are searched in the order specified. Also, the placement of 450the '-l' option is significant. 451 452 Many options have long names starting with '-f' or with '-W'--for 453example, '-fmove-loop-invariants', '-Wformat' and so on. Most of these 454have both positive and negative forms; the negative form of '-ffoo' is 455'-fno-foo'. This manual documents only one of these two forms, 456whichever one is not the default. 457 458 Some options take one or more arguments typically separated either by a 459space or by the equals sign ('=') from the option name. Unless 460documented otherwise, an argument can be either numeric or a string. 461Numeric arguments must typically be small unsigned decimal or 462hexadecimal integers. Hexadecimal arguments must begin with the '0x' 463prefix. Arguments to options that specify a size threshold of some sort 464may be arbitrarily large decimal or hexadecimal integers followed by a 465byte size suffix designating a multiple of bytes such as 'kB' and 'KiB' 466for kilobyte and kibibyte, respectively, 'MB' and 'MiB' for megabyte and 467mebibyte, 'GB' and 'GiB' for gigabyte and gigibyte, and so on. Such 468arguments are designated by BYTE-SIZE in the following text. Refer to 469the NIST, IEC, and other relevant national and international standards 470for the full listing and explanation of the binary and decimal byte size 471prefixes. 472 473 *Note Option Index::, for an index to GCC's options. 474 475* Menu: 476 477* Option Summary:: Brief list of all options, without explanations. 478* Overall Options:: Controlling the kind of output: 479 an executable, object files, assembler files, 480 or preprocessed source. 481* Invoking G++:: Compiling C++ programs. 482* C Dialect Options:: Controlling the variant of C language compiled. 483* C++ Dialect Options:: Variations on C++. 484* Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C 485 and Objective-C++. 486* Diagnostic Message Formatting Options:: Controlling how diagnostics should 487 be formatted. 488* Warning Options:: How picky should the compiler be? 489* Static Analyzer Options:: More expensive warnings. 490* Debugging Options:: Producing debuggable code. 491* Optimize Options:: How much optimization? 492* Instrumentation Options:: Enabling profiling and extra run-time error checking. 493* Preprocessor Options:: Controlling header files and macro definitions. 494 Also, getting dependency information for Make. 495* Assembler Options:: Passing options to the assembler. 496* Link Options:: Specifying libraries and so on. 497* Directory Options:: Where to find header files and libraries. 498 Where to find the compiler executable files. 499* Code Gen Options:: Specifying conventions for function calls, data layout 500 and register usage. 501* Developer Options:: Printing GCC configuration info, statistics, and 502 debugging dumps. 503* Submodel Options:: Target-specific options, such as compiling for a 504 specific processor variant. 505* Spec Files:: How to pass switches to sub-processes. 506* Environment Variables:: Env vars that affect GCC. 507* Precompiled Headers:: Compiling a header once, and using it many times. 508 509 510File: gcc.info, Node: Option Summary, Next: Overall Options, Up: Invoking GCC 511 5123.1 Option Summary 513================== 514 515Here is a summary of all the options, grouped by type. Explanations are 516in the following sections. 517 518_Overall Options_ 519 *Note Options Controlling the Kind of Output: Overall Options. 520 -c -S -E -o FILE -x LANGUAGE 521 -v -### --help[=CLASS[,...]] --target-help --version 522 -pass-exit-codes -pipe -specs=FILE -wrapper 523 @FILE -ffile-prefix-map=OLD=NEW 524 -fplugin=FILE -fplugin-arg-NAME=ARG 525 -fdump-ada-spec[-slim] -fada-spec-parent=UNIT -fdump-go-spec=FILE 526 527_C Language Options_ 528 *Note Options Controlling C Dialect: C Dialect Options. 529 -ansi -std=STANDARD -fgnu89-inline 530 -fpermitted-flt-eval-methods=STANDARD 531 -aux-info FILENAME -fallow-parameterless-variadic-functions 532 -fno-asm -fno-builtin -fno-builtin-FUNCTION -fgimple 533 -fhosted -ffreestanding 534 -fopenacc -fopenacc-dim=GEOM 535 -fopenmp -fopenmp-simd 536 -fms-extensions -fplan9-extensions -fsso-struct=ENDIANNESS 537 -fallow-single-precision -fcond-mismatch -flax-vector-conversions 538 -fsigned-bitfields -fsigned-char 539 -funsigned-bitfields -funsigned-char 540 541_C++ Language Options_ 542 *Note Options Controlling C++ Dialect: C++ Dialect Options. 543 -fabi-version=N -fno-access-control 544 -faligned-new=N -fargs-in-order=N -fchar8_t -fcheck-new 545 -fconstexpr-depth=N -fconstexpr-cache-depth=N 546 -fconstexpr-loop-limit=N -fconstexpr-ops-limit=N 547 -fno-elide-constructors 548 -fno-enforce-eh-specs 549 -fno-gnu-keywords 550 -fno-implicit-templates 551 -fno-implicit-inline-templates 552 -fno-implement-inlines -fms-extensions 553 -fnew-inheriting-ctors 554 -fnew-ttp-matching 555 -fno-nonansi-builtins -fnothrow-opt -fno-operator-names 556 -fno-optional-diags -fpermissive 557 -fno-pretty-templates 558 -fno-rtti -fsized-deallocation 559 -ftemplate-backtrace-limit=N 560 -ftemplate-depth=N 561 -fno-threadsafe-statics -fuse-cxa-atexit 562 -fno-weak -nostdinc++ 563 -fvisibility-inlines-hidden 564 -fvisibility-ms-compat 565 -fext-numeric-literals 566 -Wabi-tag -Wcatch-value -Wcatch-value=N 567 -Wno-class-conversion -Wclass-memaccess 568 -Wcomma-subscript -Wconditionally-supported 569 -Wno-conversion-null -Wctor-dtor-privacy -Wno-delete-incomplete 570 -Wdelete-non-virtual-dtor -Wdeprecated-copy -Wdeprecated-copy-dtor 571 -Weffc++ -Wextra-semi -Wno-inaccessible-base 572 -Wno-inherited-variadic-ctor -Wno-init-list-lifetime 573 -Wno-invalid-offsetof -Wno-literal-suffix -Wmismatched-tags 574 -Wmultiple-inheritance -Wnamespaces -Wnarrowing 575 -Wnoexcept -Wnoexcept-type -Wnon-virtual-dtor 576 -Wpessimizing-move -Wno-placement-new -Wplacement-new=N 577 -Wredundant-move -Wredundant-tags 578 -Wreorder -Wregister 579 -Wstrict-null-sentinel -Wno-subobject-linkage -Wtemplates 580 -Wno-non-template-friend -Wold-style-cast 581 -Woverloaded-virtual -Wno-pmf-conversions -Wsign-promo 582 -Wsized-deallocation -Wsuggest-final-methods 583 -Wsuggest-final-types -Wsuggest-override 584 -Wno-terminate -Wuseless-cast -Wvirtual-inheritance 585 -Wno-virtual-move-assign -Wvolatile -Wzero-as-null-pointer-constant 586 587_Objective-C and Objective-C++ Language Options_ 588 *Note Options Controlling Objective-C and Objective-C++ Dialects: 589 Objective-C and Objective-C++ Dialect Options. 590 -fconstant-string-class=CLASS-NAME 591 -fgnu-runtime -fnext-runtime 592 -fno-nil-receivers 593 -fobjc-abi-version=N 594 -fobjc-call-cxx-cdtors 595 -fobjc-direct-dispatch 596 -fobjc-exceptions 597 -fobjc-gc 598 -fobjc-nilcheck 599 -fobjc-std=objc1 600 -fno-local-ivars 601 -fivar-visibility=[public|protected|private|package] 602 -freplace-objc-classes 603 -fzero-link 604 -gen-decls 605 -Wassign-intercept -Wno-property-assign-default 606 -Wno-protocol -Wselector 607 -Wstrict-selector-match 608 -Wundeclared-selector 609 610_Diagnostic Message Formatting Options_ 611 *Note Options to Control Diagnostic Messages Formatting: Diagnostic 612 Message Formatting Options. 613 -fmessage-length=N 614 -fdiagnostics-show-location=[once|every-line] 615 -fdiagnostics-color=[auto|never|always] 616 -fdiagnostics-urls=[auto|never|always] 617 -fdiagnostics-format=[text|json] 618 -fno-diagnostics-show-option -fno-diagnostics-show-caret 619 -fno-diagnostics-show-labels -fno-diagnostics-show-line-numbers 620 -fno-diagnostics-show-cwe 621 -fdiagnostics-minimum-margin-width=WIDTH 622 -fdiagnostics-parseable-fixits -fdiagnostics-generate-patch 623 -fdiagnostics-show-template-tree -fno-elide-type 624 -fdiagnostics-path-format=[none|separate-events|inline-events] 625 -fdiagnostics-show-path-depths 626 -fno-show-column 627 628_Warning Options_ 629 *Note Options to Request or Suppress Warnings: Warning Options. 630 -fsyntax-only -fmax-errors=N -Wpedantic 631 -pedantic-errors 632 -w -Wextra -Wall -Wabi=N 633 -Waddress -Wno-address-of-packed-member -Waggregate-return 634 -Walloc-size-larger-than=BYTE-SIZE -Walloc-zero 635 -Walloca -Walloca-larger-than=BYTE-SIZE 636 -Wno-aggressive-loop-optimizations 637 -Warith-conversion 638 -Warray-bounds -Warray-bounds=N 639 -Wno-attributes -Wattribute-alias=N -Wno-attribute-alias 640 -Wno-attribute-warning -Wbool-compare -Wbool-operation 641 -Wno-builtin-declaration-mismatch 642 -Wno-builtin-macro-redefined -Wc90-c99-compat -Wc99-c11-compat 643 -Wc11-c2x-compat 644 -Wc++-compat -Wc++11-compat -Wc++14-compat -Wc++17-compat 645 -Wc++20-compat 646 -Wcast-align -Wcast-align=strict -Wcast-function-type -Wcast-qual 647 -Wchar-subscripts 648 -Wclobbered -Wcomment 649 -Wconversion -Wno-coverage-mismatch -Wno-cpp 650 -Wdangling-else -Wdate-time 651 -Wno-deprecated -Wno-deprecated-declarations -Wno-designated-init 652 -Wdisabled-optimization 653 -Wno-discarded-array-qualifiers -Wno-discarded-qualifiers 654 -Wno-div-by-zero -Wdouble-promotion 655 -Wduplicated-branches -Wduplicated-cond 656 -Wempty-body -Wno-endif-labels -Wenum-compare -Wenum-conversion 657 -Werror -Werror=* -Wexpansion-to-defined -Wfatal-errors 658 -Wfloat-conversion -Wfloat-equal -Wformat -Wformat=2 659 -Wno-format-contains-nul -Wno-format-extra-args 660 -Wformat-nonliteral -Wformat-overflow=N 661 -Wformat-security -Wformat-signedness -Wformat-truncation=N 662 -Wformat-y2k -Wframe-address 663 -Wframe-larger-than=BYTE-SIZE -Wno-free-nonheap-object 664 -Wno-hsa -Wno-if-not-aligned -Wno-ignored-attributes 665 -Wignored-qualifiers -Wno-incompatible-pointer-types 666 -Wimplicit -Wimplicit-fallthrough -Wimplicit-fallthrough=N 667 -Wno-implicit-function-declaration -Wno-implicit-int 668 -Winit-self -Winline -Wno-int-conversion -Wint-in-bool-context 669 -Wno-int-to-pointer-cast -Wno-invalid-memory-model 670 -Winvalid-pch -Wjump-misses-init -Wlarger-than=BYTE-SIZE 671 -Wlogical-not-parentheses -Wlogical-op -Wlong-long 672 -Wno-lto-type-mismatch -Wmain -Wmaybe-uninitialized 673 -Wmemset-elt-size -Wmemset-transposed-args 674 -Wmisleading-indentation -Wmissing-attributes -Wmissing-braces 675 -Wmissing-field-initializers -Wmissing-format-attribute 676 -Wmissing-include-dirs -Wmissing-noreturn -Wno-missing-profile 677 -Wno-multichar -Wmultistatement-macros -Wnonnull -Wnonnull-compare 678 -Wnormalized=[none|id|nfc|nfkc] 679 -Wnull-dereference -Wno-odr -Wopenmp-simd 680 -Wno-overflow -Woverlength-strings -Wno-override-init-side-effects 681 -Wpacked -Wno-packed-bitfield-compat -Wpacked-not-aligned -Wpadded 682 -Wparentheses -Wno-pedantic-ms-format 683 -Wpointer-arith -Wno-pointer-compare -Wno-pointer-to-int-cast 684 -Wno-pragmas -Wno-prio-ctor-dtor -Wredundant-decls 685 -Wrestrict -Wno-return-local-addr -Wreturn-type 686 -Wno-scalar-storage-order -Wsequence-point 687 -Wshadow -Wshadow=global -Wshadow=local -Wshadow=compatible-local 688 -Wno-shadow-ivar 689 -Wno-shift-count-negative -Wno-shift-count-overflow -Wshift-negative-value 690 -Wno-shift-overflow -Wshift-overflow=N 691 -Wsign-compare -Wsign-conversion 692 -Wno-sizeof-array-argument 693 -Wsizeof-pointer-div -Wsizeof-pointer-memaccess 694 -Wstack-protector -Wstack-usage=BYTE-SIZE -Wstrict-aliasing 695 -Wstrict-aliasing=n -Wstrict-overflow -Wstrict-overflow=N 696 -Wstring-compare 697 -Wstringop-overflow=N -Wno-stringop-truncation 698 -Wsuggest-attribute=[pure|const|noreturn|format|malloc] 699 -Wswitch -Wno-switch-bool -Wswitch-default -Wswitch-enum 700 -Wno-switch-outside-range -Wno-switch-unreachable -Wsync-nand 701 -Wsystem-headers -Wtautological-compare -Wtrampolines -Wtrigraphs 702 -Wtype-limits -Wundef 703 -Wuninitialized -Wunknown-pragmas 704 -Wunsuffixed-float-constants -Wunused 705 -Wunused-but-set-parameter -Wunused-but-set-variable 706 -Wunused-const-variable -Wunused-const-variable=N 707 -Wunused-function -Wunused-label -Wunused-local-typedefs 708 -Wunused-macros 709 -Wunused-parameter -Wno-unused-result 710 -Wunused-value -Wunused-variable 711 -Wno-varargs -Wvariadic-macros 712 -Wvector-operation-performance 713 -Wvla -Wvla-larger-than=BYTE-SIZE -Wno-vla-larger-than 714 -Wvolatile-register-var -Wwrite-strings 715 -Wzero-length-bounds 716 717_Static Analyzer Options_ 718 -fanalyzer 719 -fanalyzer-call-summaries 720 -fanalyzer-checker=NAME 721 -fanalyzer-fine-grained 722 -fanalyzer-state-merge 723 -fanalyzer-state-purge 724 -fanalyzer-transitivity 725 -fanalyzer-verbose-edges 726 -fanalyzer-verbose-state-changes 727 -fanalyzer-verbosity=LEVEL 728 -fdump-analyzer 729 -fdump-analyzer-stderr 730 -fdump-analyzer-callgraph 731 -fdump-analyzer-exploded-graph 732 -fdump-analyzer-exploded-nodes 733 -fdump-analyzer-exploded-nodes-2 734 -fdump-analyzer-exploded-nodes-3 735 -fdump-analyzer-state-purge 736 -fdump-analyzer-supergraph 737 -Wno-analyzer-double-fclose 738 -Wno-analyzer-double-free 739 -Wno-analyzer-exposure-through-output-file 740 -Wno-analyzer-file-leak 741 -Wno-analyzer-free-of-non-heap 742 -Wno-analyzer-malloc-leak 743 -Wno-analyzer-null-argument 744 -Wno-analyzer-null-dereference 745 -Wno-analyzer-possible-null-argument 746 -Wno-analyzer-possible-null-dereference 747 -Wno-analyzer-stale-setjmp-buffer 748 -Wno-analyzer-tainted-array-index 749 -Wanalyzer-too-complex 750 -Wno-analyzer-unsafe-call-within-signal-handler 751 -Wno-analyzer-use-after-free 752 -Wno-analyzer-use-of-pointer-in-stale-stack-frame 753 -Wno-analyzer-use-of-uninitialized-value 754 755 756_C and Objective-C-only Warning Options_ 757 -Wbad-function-cast -Wmissing-declarations 758 -Wmissing-parameter-type -Wmissing-prototypes -Wnested-externs 759 -Wold-style-declaration -Wold-style-definition 760 -Wstrict-prototypes -Wtraditional -Wtraditional-conversion 761 -Wdeclaration-after-statement -Wpointer-sign 762 763_Debugging Options_ 764 *Note Options for Debugging Your Program: Debugging Options. 765 -g -gLEVEL -gdwarf -gdwarf-VERSION 766 -ggdb -grecord-gcc-switches -gno-record-gcc-switches 767 -gstabs -gstabs+ -gstrict-dwarf -gno-strict-dwarf 768 -gas-loc-support -gno-as-loc-support 769 -gas-locview-support -gno-as-locview-support 770 -gcolumn-info -gno-column-info 771 -gstatement-frontiers -gno-statement-frontiers 772 -gvariable-location-views -gno-variable-location-views 773 -ginternal-reset-location-views -gno-internal-reset-location-views 774 -ginline-points -gno-inline-points 775 -gvms -gxcoff -gxcoff+ -gz[=TYPE] 776 -gsplit-dwarf -gdescribe-dies -gno-describe-dies 777 -fdebug-prefix-map=OLD=NEW -fdebug-types-section 778 -fno-eliminate-unused-debug-types 779 -femit-struct-debug-baseonly -femit-struct-debug-reduced 780 -femit-struct-debug-detailed[=SPEC-LIST] 781 -fno-eliminate-unused-debug-symbols -femit-class-debug-always 782 -fno-merge-debug-strings -fno-dwarf2-cfi-asm 783 -fvar-tracking -fvar-tracking-assignments 784 785_Optimization Options_ 786 *Note Options that Control Optimization: Optimize Options. 787 -faggressive-loop-optimizations 788 -falign-functions[=N[:M:[N2[:M2]]]] 789 -falign-jumps[=N[:M:[N2[:M2]]]] 790 -falign-labels[=N[:M:[N2[:M2]]]] 791 -falign-loops[=N[:M:[N2[:M2]]]] 792 -fno-allocation-dce -fallow-store-data-races 793 -fassociative-math -fauto-profile -fauto-profile[=PATH] 794 -fauto-inc-dec -fbranch-probabilities 795 -fcaller-saves 796 -fcombine-stack-adjustments -fconserve-stack 797 -fcompare-elim -fcprop-registers -fcrossjumping 798 -fcse-follow-jumps -fcse-skip-blocks -fcx-fortran-rules 799 -fcx-limited-range 800 -fdata-sections -fdce -fdelayed-branch 801 -fdelete-null-pointer-checks -fdevirtualize -fdevirtualize-speculatively 802 -fdevirtualize-at-ltrans -fdse 803 -fearly-inlining -fipa-sra -fexpensive-optimizations -ffat-lto-objects 804 -ffast-math -ffinite-math-only -ffloat-store -fexcess-precision=STYLE 805 -ffinite-loops 806 -fforward-propagate -ffp-contract=STYLE -ffunction-sections 807 -fgcse -fgcse-after-reload -fgcse-las -fgcse-lm -fgraphite-identity 808 -fgcse-sm -fhoist-adjacent-loads -fif-conversion 809 -fif-conversion2 -findirect-inlining 810 -finline-functions -finline-functions-called-once -finline-limit=N 811 -finline-small-functions -fipa-cp -fipa-cp-clone 812 -fipa-bit-cp -fipa-vrp -fipa-pta -fipa-profile -fipa-pure-const 813 -fipa-reference -fipa-reference-addressable 814 -fipa-stack-alignment -fipa-icf -fira-algorithm=ALGORITHM 815 -flive-patching=LEVEL 816 -fira-region=REGION -fira-hoist-pressure 817 -fira-loop-pressure -fno-ira-share-save-slots 818 -fno-ira-share-spill-slots 819 -fisolate-erroneous-paths-dereference -fisolate-erroneous-paths-attribute 820 -fivopts -fkeep-inline-functions -fkeep-static-functions 821 -fkeep-static-consts -flimit-function-alignment -flive-range-shrinkage 822 -floop-block -floop-interchange -floop-strip-mine 823 -floop-unroll-and-jam -floop-nest-optimize 824 -floop-parallelize-all -flra-remat -flto -flto-compression-level 825 -flto-partition=ALG -fmerge-all-constants 826 -fmerge-constants -fmodulo-sched -fmodulo-sched-allow-regmoves 827 -fmove-loop-invariants -fno-branch-count-reg 828 -fno-defer-pop -fno-fp-int-builtin-inexact -fno-function-cse 829 -fno-guess-branch-probability -fno-inline -fno-math-errno -fno-peephole 830 -fno-peephole2 -fno-printf-return-value -fno-sched-interblock 831 -fno-sched-spec -fno-signed-zeros 832 -fno-toplevel-reorder -fno-trapping-math -fno-zero-initialized-in-bss 833 -fomit-frame-pointer -foptimize-sibling-calls 834 -fpartial-inlining -fpeel-loops -fpredictive-commoning 835 -fprefetch-loop-arrays 836 -fprofile-correction 837 -fprofile-use -fprofile-use=PATH -fprofile-partial-training 838 -fprofile-values -fprofile-reorder-functions 839 -freciprocal-math -free -frename-registers -freorder-blocks 840 -freorder-blocks-algorithm=ALGORITHM 841 -freorder-blocks-and-partition -freorder-functions 842 -frerun-cse-after-loop -freschedule-modulo-scheduled-loops 843 -frounding-math -fsave-optimization-record 844 -fsched2-use-superblocks -fsched-pressure 845 -fsched-spec-load -fsched-spec-load-dangerous 846 -fsched-stalled-insns-dep[=N] -fsched-stalled-insns[=N] 847 -fsched-group-heuristic -fsched-critical-path-heuristic 848 -fsched-spec-insn-heuristic -fsched-rank-heuristic 849 -fsched-last-insn-heuristic -fsched-dep-count-heuristic 850 -fschedule-fusion 851 -fschedule-insns -fschedule-insns2 -fsection-anchors 852 -fselective-scheduling -fselective-scheduling2 853 -fsel-sched-pipelining -fsel-sched-pipelining-outer-loops 854 -fsemantic-interposition -fshrink-wrap -fshrink-wrap-separate 855 -fsignaling-nans 856 -fsingle-precision-constant -fsplit-ivs-in-unroller -fsplit-loops 857 -fsplit-paths 858 -fsplit-wide-types -fsplit-wide-types-early -fssa-backprop -fssa-phiopt 859 -fstdarg-opt -fstore-merging -fstrict-aliasing 860 -fthread-jumps -ftracer -ftree-bit-ccp 861 -ftree-builtin-call-dce -ftree-ccp -ftree-ch 862 -ftree-coalesce-vars -ftree-copy-prop -ftree-dce -ftree-dominator-opts 863 -ftree-dse -ftree-forwprop -ftree-fre -fcode-hoisting 864 -ftree-loop-if-convert -ftree-loop-im 865 -ftree-phiprop -ftree-loop-distribution -ftree-loop-distribute-patterns 866 -ftree-loop-ivcanon -ftree-loop-linear -ftree-loop-optimize 867 -ftree-loop-vectorize 868 -ftree-parallelize-loops=N -ftree-pre -ftree-partial-pre -ftree-pta 869 -ftree-reassoc -ftree-scev-cprop -ftree-sink -ftree-slsr -ftree-sra 870 -ftree-switch-conversion -ftree-tail-merge 871 -ftree-ter -ftree-vectorize -ftree-vrp -funconstrained-commons 872 -funit-at-a-time -funroll-all-loops -funroll-loops 873 -funsafe-math-optimizations -funswitch-loops 874 -fipa-ra -fvariable-expansion-in-unroller -fvect-cost-model -fvpt 875 -fweb -fwhole-program -fwpa -fuse-linker-plugin 876 --param NAME=VALUE 877 -O -O0 -O1 -O2 -O3 -Os -Ofast -Og 878 879_Program Instrumentation Options_ 880 *Note Program Instrumentation Options: Instrumentation Options. 881 -p -pg -fprofile-arcs --coverage -ftest-coverage 882 -fprofile-abs-path 883 -fprofile-dir=PATH -fprofile-generate -fprofile-generate=PATH 884 -fprofile-note=PATH -fprofile-prefix-path=PATH 885 -fprofile-update=METHOD -fprofile-filter-files=REGEX 886 -fprofile-exclude-files=REGEX 887 -fprofile-reproducible=[multithreaded|parallel-runs|serial] 888 -fsanitize=STYLE -fsanitize-recover -fsanitize-recover=STYLE 889 -fasan-shadow-offset=NUMBER -fsanitize-sections=S1,S2,... 890 -fsanitize-undefined-trap-on-error -fbounds-check 891 -fcf-protection=[full|branch|return|none|check] 892 -fstack-protector -fstack-protector-all -fstack-protector-strong 893 -fstack-protector-explicit -fstack-check 894 -fstack-limit-register=REG -fstack-limit-symbol=SYM 895 -fno-stack-limit -fsplit-stack 896 -fvtable-verify=[std|preinit|none] 897 -fvtv-counts -fvtv-debug 898 -finstrument-functions 899 -finstrument-functions-exclude-function-list=SYM,SYM,... 900 -finstrument-functions-exclude-file-list=FILE,FILE,... 901 902_Preprocessor Options_ 903 *Note Options Controlling the Preprocessor: Preprocessor Options. 904 -AQUESTION=ANSWER 905 -A-QUESTION[=ANSWER] 906 -C -CC -DMACRO[=DEFN] 907 -dD -dI -dM -dN -dU 908 -fdebug-cpp -fdirectives-only -fdollars-in-identifiers 909 -fexec-charset=CHARSET -fextended-identifiers 910 -finput-charset=CHARSET -fmacro-prefix-map=OLD=NEW 911 -fmax-include-depth=DEPTH 912 -fno-canonical-system-headers -fpch-deps -fpch-preprocess 913 -fpreprocessed -ftabstop=WIDTH -ftrack-macro-expansion 914 -fwide-exec-charset=CHARSET -fworking-directory 915 -H -imacros FILE -include FILE 916 -M -MD -MF -MG -MM -MMD -MP -MQ -MT 917 -no-integrated-cpp -P -pthread -remap 918 -traditional -traditional-cpp -trigraphs 919 -UMACRO -undef 920 -Wp,OPTION -Xpreprocessor OPTION 921 922_Assembler Options_ 923 *Note Passing Options to the Assembler: Assembler Options. 924 -Wa,OPTION -Xassembler OPTION 925 926_Linker Options_ 927 *Note Options for Linking: Link Options. 928 OBJECT-FILE-NAME -fuse-ld=LINKER -lLIBRARY 929 -nostartfiles -nodefaultlibs -nolibc -nostdlib 930 -e ENTRY --entry=ENTRY 931 -pie -pthread -r -rdynamic 932 -s -static -static-pie -static-libgcc -static-libstdc++ 933 -static-libasan -static-libtsan -static-liblsan -static-libubsan 934 -shared -shared-libgcc -symbolic 935 -T SCRIPT -Wl,OPTION -Xlinker OPTION 936 -u SYMBOL -z KEYWORD 937 938_Directory Options_ 939 *Note Options for Directory Search: Directory Options. 940 -BPREFIX -IDIR -I- 941 -idirafter DIR 942 -imacros FILE -imultilib DIR 943 -iplugindir=DIR -iprefix FILE 944 -iquote DIR -isysroot DIR -isystem DIR 945 -iwithprefix DIR -iwithprefixbefore DIR 946 -LDIR -no-canonical-prefixes --no-sysroot-suffix 947 -nostdinc -nostdinc++ --sysroot=DIR 948 949_Code Generation Options_ 950 *Note Options for Code Generation Conventions: Code Gen Options. 951 -fcall-saved-REG -fcall-used-REG 952 -ffixed-REG -fexceptions 953 -fnon-call-exceptions -fdelete-dead-exceptions -funwind-tables 954 -fasynchronous-unwind-tables 955 -fno-gnu-unique 956 -finhibit-size-directive -fcommon -fno-ident 957 -fpcc-struct-return -fpic -fPIC -fpie -fPIE -fno-plt 958 -fno-jump-tables 959 -frecord-gcc-switches 960 -freg-struct-return -fshort-enums -fshort-wchar 961 -fverbose-asm -fpack-struct[=N] 962 -fleading-underscore -ftls-model=MODEL 963 -fstack-reuse=REUSE_LEVEL 964 -ftrampolines -ftrapv -fwrapv 965 -fvisibility=[default|internal|hidden|protected] 966 -fstrict-volatile-bitfields -fsync-libcalls 967 968_Developer Options_ 969 *Note GCC Developer Options: Developer Options. 970 -dLETTERS -dumpspecs -dumpmachine -dumpversion 971 -dumpfullversion -fcallgraph-info[=su,da] 972 -fchecking -fchecking=N 973 -fdbg-cnt-list -fdbg-cnt=COUNTER-VALUE-LIST 974 -fdisable-ipa-PASS_NAME 975 -fdisable-rtl-PASS_NAME 976 -fdisable-rtl-PASS-NAME=RANGE-LIST 977 -fdisable-tree-PASS_NAME 978 -fdisable-tree-PASS-NAME=RANGE-LIST 979 -fdump-debug -fdump-earlydebug 980 -fdump-noaddr -fdump-unnumbered -fdump-unnumbered-links 981 -fdump-final-insns[=FILE] 982 -fdump-ipa-all -fdump-ipa-cgraph -fdump-ipa-inline 983 -fdump-lang-all 984 -fdump-lang-SWITCH 985 -fdump-lang-SWITCH-OPTIONS 986 -fdump-lang-SWITCH-OPTIONS=FILENAME 987 -fdump-passes 988 -fdump-rtl-PASS -fdump-rtl-PASS=FILENAME 989 -fdump-statistics 990 -fdump-tree-all 991 -fdump-tree-SWITCH 992 -fdump-tree-SWITCH-OPTIONS 993 -fdump-tree-SWITCH-OPTIONS=FILENAME 994 -fcompare-debug[=OPTS] -fcompare-debug-second 995 -fenable-KIND-PASS 996 -fenable-KIND-PASS=RANGE-LIST 997 -fira-verbose=N 998 -flto-report -flto-report-wpa -fmem-report-wpa 999 -fmem-report -fpre-ipa-mem-report -fpost-ipa-mem-report 1000 -fopt-info -fopt-info-OPTIONS[=FILE] 1001 -fprofile-report 1002 -frandom-seed=STRING -fsched-verbose=N 1003 -fsel-sched-verbose -fsel-sched-dump-cfg -fsel-sched-pipelining-verbose 1004 -fstats -fstack-usage -ftime-report -ftime-report-details 1005 -fvar-tracking-assignments-toggle -gtoggle 1006 -print-file-name=LIBRARY -print-libgcc-file-name 1007 -print-multi-directory -print-multi-lib -print-multi-os-directory 1008 -print-prog-name=PROGRAM -print-search-dirs -Q 1009 -print-sysroot -print-sysroot-headers-suffix 1010 -save-temps -save-temps=cwd -save-temps=obj -time[=FILE] 1011 1012_Machine-Dependent Options_ 1013 *Note Machine-Dependent Options: Submodel Options. 1014 1015 _AArch64 Options_ 1016 -mabi=NAME -mbig-endian -mlittle-endian 1017 -mgeneral-regs-only 1018 -mcmodel=tiny -mcmodel=small -mcmodel=large 1019 -mstrict-align -mno-strict-align 1020 -momit-leaf-frame-pointer 1021 -mtls-dialect=desc -mtls-dialect=traditional 1022 -mtls-size=SIZE 1023 -mfix-cortex-a53-835769 -mfix-cortex-a53-843419 1024 -mlow-precision-recip-sqrt -mlow-precision-sqrt -mlow-precision-div 1025 -mpc-relative-literal-loads 1026 -msign-return-address=SCOPE 1027 -mbranch-protection=NONE|STANDARD|PAC-RET[+LEAF 1028 +B-KEY]|BTI 1029 -mharden-sls=OPTS 1030 -march=NAME -mcpu=NAME -mtune=NAME 1031 -moverride=STRING -mverbose-cost-dump 1032 -mstack-protector-guard=GUARD -mstack-protector-guard-reg=SYSREG 1033 -mstack-protector-guard-offset=OFFSET -mtrack-speculation 1034 -moutline-atomics 1035 1036 _Adapteva Epiphany Options_ 1037 -mhalf-reg-file -mprefer-short-insn-regs 1038 -mbranch-cost=NUM -mcmove -mnops=NUM -msoft-cmpsf 1039 -msplit-lohi -mpost-inc -mpost-modify -mstack-offset=NUM 1040 -mround-nearest -mlong-calls -mshort-calls -msmall16 1041 -mfp-mode=MODE -mvect-double -max-vect-align=NUM 1042 -msplit-vecmove-early -m1reg-REG 1043 1044 _AMD GCN Options_ 1045 -march=GPU -mtune=GPU -mstack-size=BYTES 1046 1047 _ARC Options_ 1048 -mbarrel-shifter -mjli-always 1049 -mcpu=CPU -mA6 -mARC600 -mA7 -mARC700 1050 -mdpfp -mdpfp-compact -mdpfp-fast -mno-dpfp-lrsr 1051 -mea -mno-mpy -mmul32x16 -mmul64 -matomic 1052 -mnorm -mspfp -mspfp-compact -mspfp-fast -msimd -msoft-float -mswap 1053 -mcrc -mdsp-packa -mdvbf -mlock -mmac-d16 -mmac-24 -mrtsc -mswape 1054 -mtelephony -mxy -misize -mannotate-align -marclinux -marclinux_prof 1055 -mlong-calls -mmedium-calls -msdata -mirq-ctrl-saved 1056 -mrgf-banked-regs -mlpc-width=WIDTH -G NUM 1057 -mvolatile-cache -mtp-regno=REGNO 1058 -malign-call -mauto-modify-reg -mbbit-peephole -mno-brcc 1059 -mcase-vector-pcrel -mcompact-casesi -mno-cond-exec -mearly-cbranchsi 1060 -mexpand-adddi -mindexed-loads -mlra -mlra-priority-none 1061 -mlra-priority-compact mlra-priority-noncompact -mmillicode 1062 -mmixed-code -mq-class -mRcq -mRcw -msize-level=LEVEL 1063 -mtune=CPU -mmultcost=NUM -mcode-density-frame 1064 -munalign-prob-threshold=PROBABILITY -mmpy-option=MULTO 1065 -mdiv-rem -mcode-density -mll64 -mfpu=FPU -mrf16 -mbranch-index 1066 1067 _ARM Options_ 1068 -mapcs-frame -mno-apcs-frame 1069 -mabi=NAME 1070 -mapcs-stack-check -mno-apcs-stack-check 1071 -mapcs-reentrant -mno-apcs-reentrant 1072 -mgeneral-regs-only 1073 -msched-prolog -mno-sched-prolog 1074 -mlittle-endian -mbig-endian 1075 -mbe8 -mbe32 1076 -mfloat-abi=NAME 1077 -mfp16-format=NAME 1078 -mthumb-interwork -mno-thumb-interwork 1079 -mcpu=NAME -march=NAME -mfpu=NAME 1080 -mtune=NAME -mprint-tune-info 1081 -mstructure-size-boundary=N 1082 -mabort-on-noreturn 1083 -mlong-calls -mno-long-calls 1084 -msingle-pic-base -mno-single-pic-base 1085 -mpic-register=REG 1086 -mnop-fun-dllimport 1087 -mpoke-function-name 1088 -mthumb -marm -mflip-thumb 1089 -mtpcs-frame -mtpcs-leaf-frame 1090 -mcaller-super-interworking -mcallee-super-interworking 1091 -mtp=NAME -mtls-dialect=DIALECT 1092 -mword-relocations 1093 -mfix-cortex-m3-ldrd 1094 -munaligned-access 1095 -mneon-for-64bits 1096 -mslow-flash-data 1097 -masm-syntax-unified 1098 -mrestrict-it 1099 -mverbose-cost-dump 1100 -mpure-code 1101 -mcmse 1102 -mfix-cmse-cve-2021-35465 1103 -mfdpic 1104 1105 _AVR Options_ 1106 -mmcu=MCU -mabsdata -maccumulate-args 1107 -mbranch-cost=COST 1108 -mcall-prologues -mgas-isr-prologues -mint8 1109 -mdouble=BITS -mlong-double=BITS 1110 -mn_flash=SIZE -mno-interrupts 1111 -mmain-is-OS_task -mrelax -mrmw -mstrict-X -mtiny-stack 1112 -mfract-convert-truncate 1113 -mshort-calls -nodevicelib -nodevicespecs 1114 -Waddr-space-convert -Wmisspelled-isr 1115 1116 _Blackfin Options_ 1117 -mcpu=CPU[-SIREVISION] 1118 -msim -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer 1119 -mspecld-anomaly -mno-specld-anomaly -mcsync-anomaly -mno-csync-anomaly 1120 -mlow-64k -mno-low64k -mstack-check-l1 -mid-shared-library 1121 -mno-id-shared-library -mshared-library-id=N 1122 -mleaf-id-shared-library -mno-leaf-id-shared-library 1123 -msep-data -mno-sep-data -mlong-calls -mno-long-calls 1124 -mfast-fp -minline-plt -mmulticore -mcorea -mcoreb -msdram 1125 -micplb 1126 1127 _C6X Options_ 1128 -mbig-endian -mlittle-endian -march=CPU 1129 -msim -msdata=SDATA-TYPE 1130 1131 _CRIS Options_ 1132 -mcpu=CPU -march=CPU -mtune=CPU 1133 -mmax-stack-frame=N -melinux-stacksize=N 1134 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects 1135 -mstack-align -mdata-align -mconst-align 1136 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt 1137 -melf -maout -melinux -mlinux -sim -sim2 1138 -mmul-bug-workaround -mno-mul-bug-workaround 1139 1140 _CR16 Options_ 1141 -mmac 1142 -mcr16cplus -mcr16c 1143 -msim -mint32 -mbit-ops 1144 -mdata-model=MODEL 1145 1146 _C-SKY Options_ 1147 -march=ARCH -mcpu=CPU 1148 -mbig-endian -EB -mlittle-endian -EL 1149 -mhard-float -msoft-float -mfpu=FPU -mdouble-float -mfdivdu 1150 -melrw -mistack -mmp -mcp -mcache -msecurity -mtrust 1151 -mdsp -medsp -mvdsp 1152 -mdiv -msmart -mhigh-registers -manchor 1153 -mpushpop -mmultiple-stld -mconstpool -mstack-size -mccrt 1154 -mbranch-cost=N -mcse-cc -msched-prolog 1155 1156 _Darwin Options_ 1157 -all_load -allowable_client -arch -arch_errors_fatal 1158 -arch_only -bind_at_load -bundle -bundle_loader 1159 -client_name -compatibility_version -current_version 1160 -dead_strip 1161 -dependency-file -dylib_file -dylinker_install_name 1162 -dynamic -dynamiclib -exported_symbols_list 1163 -filelist -flat_namespace -force_cpusubtype_ALL 1164 -force_flat_namespace -headerpad_max_install_names 1165 -iframework 1166 -image_base -init -install_name -keep_private_externs 1167 -multi_module -multiply_defined -multiply_defined_unused 1168 -noall_load -no_dead_strip_inits_and_terms 1169 -nofixprebinding -nomultidefs -noprebind -noseglinkedit 1170 -pagezero_size -prebind -prebind_all_twolevel_modules 1171 -private_bundle -read_only_relocs -sectalign 1172 -sectobjectsymbols -whyload -seg1addr 1173 -sectcreate -sectobjectsymbols -sectorder 1174 -segaddr -segs_read_only_addr -segs_read_write_addr 1175 -seg_addr_table -seg_addr_table_filename -seglinkedit 1176 -segprot -segs_read_only_addr -segs_read_write_addr 1177 -single_module -static -sub_library -sub_umbrella 1178 -twolevel_namespace -umbrella -undefined 1179 -unexported_symbols_list -weak_reference_mismatches 1180 -whatsloaded -F -gused -gfull -mmacosx-version-min=VERSION 1181 -mkernel -mone-byte-bool 1182 1183 _DEC Alpha Options_ 1184 -mno-fp-regs -msoft-float 1185 -mieee -mieee-with-inexact -mieee-conformant 1186 -mfp-trap-mode=MODE -mfp-rounding-mode=MODE 1187 -mtrap-precision=MODE -mbuild-constants 1188 -mcpu=CPU-TYPE -mtune=CPU-TYPE 1189 -mbwx -mmax -mfix -mcix 1190 -mfloat-vax -mfloat-ieee 1191 -mexplicit-relocs -msmall-data -mlarge-data 1192 -msmall-text -mlarge-text 1193 -mmemory-latency=TIME 1194 1195 _eBPF Options_ 1196 -mbig-endian -mlittle-endian -mkernel=VERSION 1197 -mframe-limit=BYTES -mxbpf 1198 1199 _FR30 Options_ 1200 -msmall-model -mno-lsim 1201 1202 _FT32 Options_ 1203 -msim -mlra -mnodiv -mft32b -mcompress -mnopm 1204 1205 _FRV Options_ 1206 -mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64 1207 -mhard-float -msoft-float 1208 -malloc-cc -mfixed-cc -mdword -mno-dword 1209 -mdouble -mno-double 1210 -mmedia -mno-media -mmuladd -mno-muladd 1211 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic 1212 -mlinked-fp -mlong-calls -malign-labels 1213 -mlibrary-pic -macc-4 -macc-8 1214 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move 1215 -moptimize-membar -mno-optimize-membar 1216 -mscc -mno-scc -mcond-exec -mno-cond-exec 1217 -mvliw-branch -mno-vliw-branch 1218 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec 1219 -mno-nested-cond-exec -mtomcat-stats 1220 -mTLS -mtls 1221 -mcpu=CPU 1222 1223 _GNU/Linux Options_ 1224 -mglibc -muclibc -mmusl -mbionic -mandroid 1225 -tno-android-cc -tno-android-ld 1226 1227 _H8/300 Options_ 1228 -mrelax -mh -ms -mn -mexr -mno-exr -mint32 -malign-300 1229 1230 _HPPA Options_ 1231 -march=ARCHITECTURE-TYPE 1232 -mcaller-copies -mdisable-fpregs -mdisable-indexing 1233 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld 1234 -mfixed-range=REGISTER-RANGE 1235 -mjump-in-delay -mlinker-opt -mlong-calls 1236 -mlong-load-store -mno-disable-fpregs 1237 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas 1238 -mno-jump-in-delay -mno-long-load-store 1239 -mno-portable-runtime -mno-soft-float 1240 -mno-space-regs -msoft-float -mpa-risc-1-0 1241 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime 1242 -mschedule=CPU-TYPE -mspace-regs -msio -mwsio 1243 -munix=UNIX-STD -nolibdld -static -threads 1244 1245 _IA-64 Options_ 1246 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic 1247 -mvolatile-asm-stop -mregister-names -msdata -mno-sdata 1248 -mconstant-gp -mauto-pic -mfused-madd 1249 -minline-float-divide-min-latency 1250 -minline-float-divide-max-throughput 1251 -mno-inline-float-divide 1252 -minline-int-divide-min-latency 1253 -minline-int-divide-max-throughput 1254 -mno-inline-int-divide 1255 -minline-sqrt-min-latency -minline-sqrt-max-throughput 1256 -mno-inline-sqrt 1257 -mdwarf2-asm -mearly-stop-bits 1258 -mfixed-range=REGISTER-RANGE -mtls-size=TLS-SIZE 1259 -mtune=CPU-TYPE -milp32 -mlp64 1260 -msched-br-data-spec -msched-ar-data-spec -msched-control-spec 1261 -msched-br-in-data-spec -msched-ar-in-data-spec -msched-in-control-spec 1262 -msched-spec-ldc -msched-spec-control-ldc 1263 -msched-prefer-non-data-spec-insns -msched-prefer-non-control-spec-insns 1264 -msched-stop-bits-after-every-cycle -msched-count-spec-in-critical-path 1265 -msel-sched-dont-check-control-spec -msched-fp-mem-deps-zero-cost 1266 -msched-max-memory-insns-hard-limit -msched-max-memory-insns=MAX-INSNS 1267 1268 _LM32 Options_ 1269 -mbarrel-shift-enabled -mdivide-enabled -mmultiply-enabled 1270 -msign-extend-enabled -muser-enabled 1271 1272 _M32R/D Options_ 1273 -m32r2 -m32rx -m32r 1274 -mdebug 1275 -malign-loops -mno-align-loops 1276 -missue-rate=NUMBER 1277 -mbranch-cost=NUMBER 1278 -mmodel=CODE-SIZE-MODEL-TYPE 1279 -msdata=SDATA-TYPE 1280 -mno-flush-func -mflush-func=NAME 1281 -mno-flush-trap -mflush-trap=NUMBER 1282 -G NUM 1283 1284 _M32C Options_ 1285 -mcpu=CPU -msim -memregs=NUMBER 1286 1287 _M680x0 Options_ 1288 -march=ARCH -mcpu=CPU -mtune=TUNE 1289 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040 1290 -m68060 -mcpu32 -m5200 -m5206e -m528x -m5307 -m5407 1291 -mcfv4e -mbitfield -mno-bitfield -mc68000 -mc68020 1292 -mnobitfield -mrtd -mno-rtd -mdiv -mno-div -mshort 1293 -mno-short -mhard-float -m68881 -msoft-float -mpcrel 1294 -malign-int -mstrict-align -msep-data -mno-sep-data 1295 -mshared-library-id=n -mid-shared-library -mno-id-shared-library 1296 -mxgot -mno-xgot -mlong-jump-table-offsets 1297 1298 _MCore Options_ 1299 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates 1300 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields 1301 -m4byte-functions -mno-4byte-functions -mcallgraph-data 1302 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim 1303 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment 1304 1305 _MeP Options_ 1306 -mabsdiff -mall-opts -maverage -mbased=N -mbitops 1307 -mc=N -mclip -mconfig=NAME -mcop -mcop32 -mcop64 -mivc2 1308 -mdc -mdiv -meb -mel -mio-volatile -ml -mleadz -mm -mminmax 1309 -mmult -mno-opts -mrepeat -ms -msatur -msdram -msim -msimnovec -mtf 1310 -mtiny=N 1311 1312 _MicroBlaze Options_ 1313 -msoft-float -mhard-float -msmall-divides -mcpu=CPU 1314 -mmemcpy -mxl-soft-mul -mxl-soft-div -mxl-barrel-shift 1315 -mxl-pattern-compare -mxl-stack-check -mxl-gp-opt -mno-clearbss 1316 -mxl-multiply-high -mxl-float-convert -mxl-float-sqrt 1317 -mbig-endian -mlittle-endian -mxl-reorder -mxl-mode-APP-MODEL 1318 -mpic-data-is-text-relative 1319 1320 _MIPS Options_ 1321 -EL -EB -march=ARCH -mtune=ARCH 1322 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips32r3 -mips32r5 1323 -mips32r6 -mips64 -mips64r2 -mips64r3 -mips64r5 -mips64r6 1324 -mips16 -mno-mips16 -mflip-mips16 1325 -minterlink-compressed -mno-interlink-compressed 1326 -minterlink-mips16 -mno-interlink-mips16 1327 -mabi=ABI -mabicalls -mno-abicalls 1328 -mshared -mno-shared -mplt -mno-plt -mxgot -mno-xgot 1329 -mgp32 -mgp64 -mfp32 -mfpxx -mfp64 -mhard-float -msoft-float 1330 -mno-float -msingle-float -mdouble-float 1331 -modd-spreg -mno-odd-spreg 1332 -mabs=MODE -mnan=ENCODING 1333 -mdsp -mno-dsp -mdspr2 -mno-dspr2 1334 -mmcu -mmno-mcu 1335 -meva -mno-eva 1336 -mvirt -mno-virt 1337 -mxpa -mno-xpa 1338 -mcrc -mno-crc 1339 -mginv -mno-ginv 1340 -mmicromips -mno-micromips 1341 -mmsa -mno-msa 1342 -mloongson-mmi -mno-loongson-mmi 1343 -mloongson-ext -mno-loongson-ext 1344 -mloongson-ext2 -mno-loongson-ext2 1345 -mfpu=FPU-TYPE 1346 -msmartmips -mno-smartmips 1347 -mpaired-single -mno-paired-single -mdmx -mno-mdmx 1348 -mips3d -mno-mips3d -mmt -mno-mt -mllsc -mno-llsc 1349 -mlong64 -mlong32 -msym32 -mno-sym32 1350 -GNUM -mlocal-sdata -mno-local-sdata 1351 -mextern-sdata -mno-extern-sdata -mgpopt -mno-gopt 1352 -membedded-data -mno-embedded-data 1353 -muninit-const-in-rodata -mno-uninit-const-in-rodata 1354 -mcode-readable=SETTING 1355 -msplit-addresses -mno-split-addresses 1356 -mexplicit-relocs -mno-explicit-relocs 1357 -mcheck-zero-division -mno-check-zero-division 1358 -mdivide-traps -mdivide-breaks 1359 -mload-store-pairs -mno-load-store-pairs 1360 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls 1361 -mmad -mno-mad -mimadd -mno-imadd -mfused-madd -mno-fused-madd -nocpp 1362 -mfix-24k -mno-fix-24k 1363 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400 1364 -mfix-r5900 -mno-fix-r5900 1365 -mfix-r10000 -mno-fix-r10000 -mfix-rm7000 -mno-fix-rm7000 1366 -mfix-vr4120 -mno-fix-vr4120 1367 -mfix-vr4130 -mno-fix-vr4130 -mfix-sb1 -mno-fix-sb1 1368 -mflush-func=FUNC -mno-flush-func 1369 -mbranch-cost=NUM -mbranch-likely -mno-branch-likely 1370 -mcompact-branches=POLICY 1371 -mfp-exceptions -mno-fp-exceptions 1372 -mvr4130-align -mno-vr4130-align -msynci -mno-synci 1373 -mlxc1-sxc1 -mno-lxc1-sxc1 -mmadd4 -mno-madd4 1374 -mrelax-pic-calls -mno-relax-pic-calls -mmcount-ra-address 1375 -mframe-header-opt -mno-frame-header-opt 1376 1377 _MMIX Options_ 1378 -mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu 1379 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols 1380 -melf -mbranch-predict -mno-branch-predict -mbase-addresses 1381 -mno-base-addresses -msingle-exit -mno-single-exit 1382 1383 _MN10300 Options_ 1384 -mmult-bug -mno-mult-bug 1385 -mno-am33 -mam33 -mam33-2 -mam34 1386 -mtune=CPU-TYPE 1387 -mreturn-pointer-on-d0 1388 -mno-crt0 -mrelax -mliw -msetlb 1389 1390 _Moxie Options_ 1391 -meb -mel -mmul.x -mno-crt0 1392 1393 _MSP430 Options_ 1394 -msim -masm-hex -mmcu= -mcpu= -mlarge -msmall -mrelax 1395 -mwarn-mcu 1396 -mcode-region= -mdata-region= 1397 -msilicon-errata= -msilicon-errata-warn= 1398 -mhwmult= -minrt -mtiny-printf 1399 1400 _NDS32 Options_ 1401 -mbig-endian -mlittle-endian 1402 -mreduced-regs -mfull-regs 1403 -mcmov -mno-cmov 1404 -mext-perf -mno-ext-perf 1405 -mext-perf2 -mno-ext-perf2 1406 -mext-string -mno-ext-string 1407 -mv3push -mno-v3push 1408 -m16bit -mno-16bit 1409 -misr-vector-size=NUM 1410 -mcache-block-size=NUM 1411 -march=ARCH 1412 -mcmodel=CODE-MODEL 1413 -mctor-dtor -mrelax 1414 1415 _Nios II Options_ 1416 -G NUM -mgpopt=OPTION -mgpopt -mno-gpopt 1417 -mgprel-sec=REGEXP -mr0rel-sec=REGEXP 1418 -mel -meb 1419 -mno-bypass-cache -mbypass-cache 1420 -mno-cache-volatile -mcache-volatile 1421 -mno-fast-sw-div -mfast-sw-div 1422 -mhw-mul -mno-hw-mul -mhw-mulx -mno-hw-mulx -mno-hw-div -mhw-div 1423 -mcustom-INSN=N -mno-custom-INSN 1424 -mcustom-fpu-cfg=NAME 1425 -mhal -msmallc -msys-crt0=NAME -msys-lib=NAME 1426 -march=ARCH -mbmx -mno-bmx -mcdx -mno-cdx 1427 1428 _Nvidia PTX Options_ 1429 -m32 -m64 -mmainkernel -moptimize 1430 1431 _OpenRISC Options_ 1432 -mboard=NAME -mnewlib -mhard-mul -mhard-div 1433 -msoft-mul -msoft-div 1434 -msoft-float -mhard-float -mdouble-float -munordered-float 1435 -mcmov -mror -mrori -msext -msfimm -mshftimm 1436 1437 _PDP-11 Options_ 1438 -mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10 1439 -mint32 -mno-int16 -mint16 -mno-int32 1440 -msplit -munix-asm -mdec-asm -mgnu-asm -mlra 1441 1442 _picoChip Options_ 1443 -mae=AE_TYPE -mvliw-lookahead=N 1444 -msymbol-as-address -mno-inefficient-warnings 1445 1446 _PowerPC Options_ See RS/6000 and PowerPC Options. 1447 1448 _PRU Options_ 1449 -mmcu=MCU -minrt -mno-relax -mloop 1450 -mabi=VARIANT 1451 1452 _RISC-V Options_ 1453 -mbranch-cost=N-INSTRUCTION 1454 -mplt -mno-plt 1455 -mabi=ABI-STRING 1456 -mfdiv -mno-fdiv 1457 -mdiv -mno-div 1458 -march=ISA-STRING 1459 -mtune=PROCESSOR-STRING 1460 -mpreferred-stack-boundary=NUM 1461 -msmall-data-limit=N-BYTES 1462 -msave-restore -mno-save-restore 1463 -mstrict-align -mno-strict-align 1464 -mcmodel=medlow -mcmodel=medany 1465 -mexplicit-relocs -mno-explicit-relocs 1466 -mrelax -mno-relax 1467 -mriscv-attribute -mmo-riscv-attribute 1468 -malign-data=TYPE 1469 1470 _RL78 Options_ 1471 -msim -mmul=none -mmul=g13 -mmul=g14 -mallregs 1472 -mcpu=g10 -mcpu=g13 -mcpu=g14 -mg10 -mg13 -mg14 1473 -m64bit-doubles -m32bit-doubles -msave-mduc-in-interrupts 1474 1475 _RS/6000 and PowerPC Options_ 1476 -mcpu=CPU-TYPE 1477 -mtune=CPU-TYPE 1478 -mcmodel=CODE-MODEL 1479 -mpowerpc64 1480 -maltivec -mno-altivec 1481 -mpowerpc-gpopt -mno-powerpc-gpopt 1482 -mpowerpc-gfxopt -mno-powerpc-gfxopt 1483 -mmfcrf -mno-mfcrf -mpopcntb -mno-popcntb -mpopcntd -mno-popcntd 1484 -mfprnd -mno-fprnd 1485 -mcmpb -mno-cmpb -mhard-dfp -mno-hard-dfp 1486 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc 1487 -m64 -m32 -mxl-compat -mno-xl-compat -mpe 1488 -malign-power -malign-natural 1489 -msoft-float -mhard-float -mmultiple -mno-multiple 1490 -mupdate -mno-update 1491 -mavoid-indexed-addresses -mno-avoid-indexed-addresses 1492 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align 1493 -mstrict-align -mno-strict-align -mrelocatable 1494 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib 1495 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian 1496 -mdynamic-no-pic -mswdiv -msingle-pic-base 1497 -mprioritize-restricted-insns=PRIORITY 1498 -msched-costly-dep=DEPENDENCE_TYPE 1499 -minsert-sched-nops=SCHEME 1500 -mcall-aixdesc -mcall-eabi -mcall-freebsd 1501 -mcall-linux -mcall-netbsd -mcall-openbsd 1502 -mcall-sysv -mcall-sysv-eabi -mcall-sysv-noeabi 1503 -mtraceback=TRACEBACK_TYPE 1504 -maix-struct-return -msvr4-struct-return 1505 -mabi=ABI-TYPE -msecure-plt -mbss-plt 1506 -mlongcall -mno-longcall -mpltseq -mno-pltseq 1507 -mblock-move-inline-limit=NUM 1508 -mblock-compare-inline-limit=NUM 1509 -mblock-compare-inline-loop-limit=NUM 1510 -mstring-compare-inline-limit=NUM 1511 -misel -mno-isel 1512 -mvrsave -mno-vrsave 1513 -mmulhw -mno-mulhw 1514 -mdlmzb -mno-dlmzb 1515 -mprototype -mno-prototype 1516 -msim -mmvme -mads -myellowknife -memb -msdata 1517 -msdata=OPT -mreadonly-in-sdata -mvxworks -G NUM 1518 -mrecip -mrecip=OPT -mno-recip -mrecip-precision 1519 -mno-recip-precision 1520 -mveclibabi=TYPE -mfriz -mno-friz 1521 -mpointers-to-nested-functions -mno-pointers-to-nested-functions 1522 -msave-toc-indirect -mno-save-toc-indirect 1523 -mpower8-fusion -mno-mpower8-fusion -mpower8-vector -mno-power8-vector 1524 -mcrypto -mno-crypto -mhtm -mno-htm 1525 -mquad-memory -mno-quad-memory 1526 -mquad-memory-atomic -mno-quad-memory-atomic 1527 -mcompat-align-parm -mno-compat-align-parm 1528 -mfloat128 -mno-float128 -mfloat128-hardware -mno-float128-hardware 1529 -mgnu-attribute -mno-gnu-attribute 1530 -mstack-protector-guard=GUARD -mstack-protector-guard-reg=REG 1531 -mstack-protector-guard-offset=OFFSET -mprefixed -mno-prefixed 1532 -mpcrel -mno-pcrel -mmma -mno-mmma 1533 1534 _RX Options_ 1535 -m64bit-doubles -m32bit-doubles -fpu -nofpu 1536 -mcpu= 1537 -mbig-endian-data -mlittle-endian-data 1538 -msmall-data 1539 -msim -mno-sim 1540 -mas100-syntax -mno-as100-syntax 1541 -mrelax 1542 -mmax-constant-size= 1543 -mint-register= 1544 -mpid 1545 -mallow-string-insns -mno-allow-string-insns 1546 -mjsr 1547 -mno-warn-multiple-fast-interrupts 1548 -msave-acc-in-interrupts 1549 1550 _S/390 and zSeries Options_ 1551 -mtune=CPU-TYPE -march=CPU-TYPE 1552 -mhard-float -msoft-float -mhard-dfp -mno-hard-dfp 1553 -mlong-double-64 -mlong-double-128 1554 -mbackchain -mno-backchain -mpacked-stack -mno-packed-stack 1555 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle 1556 -m64 -m31 -mdebug -mno-debug -mesa -mzarch 1557 -mhtm -mvx -mzvector 1558 -mtpf-trace -mno-tpf-trace -mtpf-trace-skip -mno-tpf-trace-skip 1559 -mfused-madd -mno-fused-madd 1560 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard 1561 -mhotpatch=HALFWORDS,HALFWORDS 1562 1563 _Score Options_ 1564 -meb -mel 1565 -mnhwloop 1566 -muls 1567 -mmac 1568 -mscore5 -mscore5u -mscore7 -mscore7d 1569 1570 _SH Options_ 1571 -m1 -m2 -m2e 1572 -m2a-nofpu -m2a-single-only -m2a-single -m2a 1573 -m3 -m3e 1574 -m4-nofpu -m4-single-only -m4-single -m4 1575 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al 1576 -mb -ml -mdalign -mrelax 1577 -mbigtable -mfmovd -mrenesas -mno-renesas -mnomacsave 1578 -mieee -mno-ieee -mbitops -misize -minline-ic_invalidate -mpadstruct 1579 -mprefergot -musermode -multcost=NUMBER -mdiv=STRATEGY 1580 -mdivsi3_libfunc=NAME -mfixed-range=REGISTER-RANGE 1581 -maccumulate-outgoing-args 1582 -matomic-model=ATOMIC-MODEL 1583 -mbranch-cost=NUM -mzdcbranch -mno-zdcbranch 1584 -mcbranch-force-delay-slot 1585 -mfused-madd -mno-fused-madd -mfsca -mno-fsca -mfsrra -mno-fsrra 1586 -mpretend-cmove -mtas 1587 1588 _Solaris 2 Options_ 1589 -mclear-hwcap -mno-clear-hwcap -mimpure-text -mno-impure-text 1590 -pthreads 1591 1592 _SPARC Options_ 1593 -mcpu=CPU-TYPE 1594 -mtune=CPU-TYPE 1595 -mcmodel=CODE-MODEL 1596 -mmemory-model=MEM-MODEL 1597 -m32 -m64 -mapp-regs -mno-app-regs 1598 -mfaster-structs -mno-faster-structs -mflat -mno-flat 1599 -mfpu -mno-fpu -mhard-float -msoft-float 1600 -mhard-quad-float -msoft-quad-float 1601 -mstack-bias -mno-stack-bias 1602 -mstd-struct-return -mno-std-struct-return 1603 -munaligned-doubles -mno-unaligned-doubles 1604 -muser-mode -mno-user-mode 1605 -mv8plus -mno-v8plus -mvis -mno-vis 1606 -mvis2 -mno-vis2 -mvis3 -mno-vis3 1607 -mvis4 -mno-vis4 -mvis4b -mno-vis4b 1608 -mcbcond -mno-cbcond -mfmaf -mno-fmaf -mfsmuld -mno-fsmuld 1609 -mpopc -mno-popc -msubxc -mno-subxc 1610 -mfix-at697f -mfix-ut699 -mfix-ut700 -mfix-gr712rc 1611 -mlra -mno-lra 1612 1613 _System V Options_ 1614 -Qy -Qn -YP,PATHS -Ym,DIR 1615 1616 _TILE-Gx Options_ 1617 -mcpu=CPU -m32 -m64 -mbig-endian -mlittle-endian 1618 -mcmodel=CODE-MODEL 1619 1620 _TILEPro Options_ 1621 -mcpu=CPU -m32 1622 1623 _V850 Options_ 1624 -mlong-calls -mno-long-calls -mep -mno-ep 1625 -mprolog-function -mno-prolog-function -mspace 1626 -mtda=N -msda=N -mzda=N 1627 -mapp-regs -mno-app-regs 1628 -mdisable-callt -mno-disable-callt 1629 -mv850e2v3 -mv850e2 -mv850e1 -mv850es 1630 -mv850e -mv850 -mv850e3v5 1631 -mloop 1632 -mrelax 1633 -mlong-jumps 1634 -msoft-float 1635 -mhard-float 1636 -mgcc-abi 1637 -mrh850-abi 1638 -mbig-switch 1639 1640 _VAX Options_ 1641 -mg -mgnu -munix 1642 1643 _Visium Options_ 1644 -mdebug -msim -mfpu -mno-fpu -mhard-float -msoft-float 1645 -mcpu=CPU-TYPE -mtune=CPU-TYPE -msv-mode -muser-mode 1646 1647 _VMS Options_ 1648 -mvms-return-codes -mdebug-main=PREFIX -mmalloc64 1649 -mpointer-size=SIZE 1650 1651 _VxWorks Options_ 1652 -mrtp -non-static -Bstatic -Bdynamic 1653 -Xbind-lazy -Xbind-now 1654 1655 _x86 Options_ 1656 -mtune=CPU-TYPE -march=CPU-TYPE 1657 -mtune-ctrl=FEATURE-LIST -mdump-tune-features -mno-default 1658 -mfpmath=UNIT 1659 -masm=DIALECT -mno-fancy-math-387 1660 -mno-fp-ret-in-387 -m80387 -mhard-float -msoft-float 1661 -mno-wide-multiply -mrtd -malign-double 1662 -mpreferred-stack-boundary=NUM 1663 -mincoming-stack-boundary=NUM 1664 -mcld -mcx16 -msahf -mmovbe -mcrc32 1665 -mrecip -mrecip=OPT 1666 -mvzeroupper -mprefer-avx128 -mprefer-vector-width=OPT 1667 -mmmx -msse -msse2 -msse3 -mssse3 -msse4.1 -msse4.2 -msse4 -mavx 1668 -mavx2 -mavx512f -mavx512pf -mavx512er -mavx512cd -mavx512vl 1669 -mavx512bw -mavx512dq -mavx512ifma -mavx512vbmi -msha -maes 1670 -mpclmul -mfsgsbase -mrdrnd -mf16c -mfma -mpconfig -mwbnoinvd 1671 -mptwrite -mprefetchwt1 -mclflushopt -mclwb -mxsavec -mxsaves 1672 -msse4a -m3dnow -m3dnowa -mpopcnt -mabm -mbmi -mtbm -mfma4 -mxop 1673 -madx -mlzcnt -mbmi2 -mfxsr -mxsave -mxsaveopt -mrtm -mhle -mlwp 1674 -mmwaitx -mclzero -mpku -mthreads -mgfni -mvaes -mwaitpkg 1675 -mshstk -mmanual-endbr -mforce-indirect-call -mavx512vbmi2 -mavx512bf16 -menqcmd 1676 -mvpclmulqdq -mavx512bitalg -mmovdiri -mmovdir64b -mavx512vpopcntdq 1677 -mavx5124fmaps -mavx512vnni -mavx5124vnniw -mprfchw -mrdpid 1678 -mrdseed -msgx -mavx512vp2intersect 1679 -mcldemote -mms-bitfields -mno-align-stringops -minline-all-stringops 1680 -minline-stringops-dynamically -mstringop-strategy=ALG 1681 -mmemcpy-strategy=STRATEGY -mmemset-strategy=STRATEGY 1682 -mpush-args -maccumulate-outgoing-args -m128bit-long-double 1683 -m96bit-long-double -mlong-double-64 -mlong-double-80 -mlong-double-128 1684 -mregparm=NUM -msseregparm 1685 -mveclibabi=TYPE -mvect8-ret-in-mem 1686 -mpc32 -mpc64 -mpc80 -mstackrealign 1687 -momit-leaf-frame-pointer -mno-red-zone -mno-tls-direct-seg-refs 1688 -mcmodel=CODE-MODEL -mabi=NAME -maddress-mode=MODE 1689 -m32 -m64 -mx32 -m16 -miamcu -mlarge-data-threshold=NUM 1690 -msse2avx -mfentry -mrecord-mcount -mnop-mcount -m8bit-idiv 1691 -minstrument-return=TYPE -mfentry-name=NAME -mfentry-section=NAME 1692 -mavx256-split-unaligned-load -mavx256-split-unaligned-store 1693 -malign-data=TYPE -mstack-protector-guard=GUARD 1694 -mstack-protector-guard-reg=REG 1695 -mstack-protector-guard-offset=OFFSET 1696 -mstack-protector-guard-symbol=SYMBOL 1697 -mgeneral-regs-only -mcall-ms2sysv-xlogues 1698 -mindirect-branch=CHOICE -mfunction-return=CHOICE 1699 -mindirect-branch-register 1700 1701 _x86 Windows Options_ 1702 -mconsole -mcygwin -mno-cygwin -mdll 1703 -mnop-fun-dllimport -mthread 1704 -municode -mwin32 -mwindows -fno-set-stack-executable 1705 1706 _Xstormy16 Options_ 1707 -msim 1708 1709 _Xtensa Options_ 1710 -mconst16 -mno-const16 1711 -mfused-madd -mno-fused-madd 1712 -mforce-no-pic 1713 -mserialize-volatile -mno-serialize-volatile 1714 -mtext-section-literals -mno-text-section-literals 1715 -mauto-litpools -mno-auto-litpools 1716 -mtarget-align -mno-target-align 1717 -mlongcalls -mno-longcalls 1718 1719 _zSeries Options_ See S/390 and zSeries Options. 1720 1721 1722File: gcc.info, Node: Overall Options, Next: Invoking G++, Prev: Option Summary, Up: Invoking GCC 1723 17243.2 Options Controlling the Kind of Output 1725========================================== 1726 1727Compilation can involve up to four stages: preprocessing, compilation 1728proper, assembly and linking, always in that order. GCC is capable of 1729preprocessing and compiling several files either into several assembler 1730input files, or into one assembler input file; then each assembler input 1731file produces an object file, and linking combines all the object files 1732(those newly compiled, and those specified as input) into an executable 1733file. 1734 1735 For any given input file, the file name suffix determines what kind of 1736compilation is done: 1737 1738'FILE.c' 1739 C source code that must be preprocessed. 1740 1741'FILE.i' 1742 C source code that should not be preprocessed. 1743 1744'FILE.ii' 1745 C++ source code that should not be preprocessed. 1746 1747'FILE.m' 1748 Objective-C source code. Note that you must link with the 1749 'libobjc' library to make an Objective-C program work. 1750 1751'FILE.mi' 1752 Objective-C source code that should not be preprocessed. 1753 1754'FILE.mm' 1755'FILE.M' 1756 Objective-C++ source code. Note that you must link with the 1757 'libobjc' library to make an Objective-C++ program work. Note that 1758 '.M' refers to a literal capital M. 1759 1760'FILE.mii' 1761 Objective-C++ source code that should not be preprocessed. 1762 1763'FILE.h' 1764 C, C++, Objective-C or Objective-C++ header file to be turned into 1765 a precompiled header (default), or C, C++ header file to be turned 1766 into an Ada spec (via the '-fdump-ada-spec' switch). 1767 1768'FILE.cc' 1769'FILE.cp' 1770'FILE.cxx' 1771'FILE.cpp' 1772'FILE.CPP' 1773'FILE.c++' 1774'FILE.C' 1775 C++ source code that must be preprocessed. Note that in '.cxx', 1776 the last two letters must both be literally 'x'. Likewise, '.C' 1777 refers to a literal capital C. 1778 1779'FILE.mm' 1780'FILE.M' 1781 Objective-C++ source code that must be preprocessed. 1782 1783'FILE.mii' 1784 Objective-C++ source code that should not be preprocessed. 1785 1786'FILE.hh' 1787'FILE.H' 1788'FILE.hp' 1789'FILE.hxx' 1790'FILE.hpp' 1791'FILE.HPP' 1792'FILE.h++' 1793'FILE.tcc' 1794 C++ header file to be turned into a precompiled header or Ada spec. 1795 1796'FILE.f' 1797'FILE.for' 1798'FILE.ftn' 1799 Fixed form Fortran source code that should not be preprocessed. 1800 1801'FILE.F' 1802'FILE.FOR' 1803'FILE.fpp' 1804'FILE.FPP' 1805'FILE.FTN' 1806 Fixed form Fortran source code that must be preprocessed (with the 1807 traditional preprocessor). 1808 1809'FILE.f90' 1810'FILE.f95' 1811'FILE.f03' 1812'FILE.f08' 1813 Free form Fortran source code that should not be preprocessed. 1814 1815'FILE.F90' 1816'FILE.F95' 1817'FILE.F03' 1818'FILE.F08' 1819 Free form Fortran source code that must be preprocessed (with the 1820 traditional preprocessor). 1821 1822'FILE.go' 1823 Go source code. 1824 1825'FILE.brig' 1826 BRIG files (binary representation of HSAIL). 1827 1828'FILE.d' 1829 D source code. 1830 1831'FILE.di' 1832 D interface file. 1833 1834'FILE.dd' 1835 D documentation code (Ddoc). 1836 1837'FILE.ads' 1838 Ada source code file that contains a library unit declaration (a 1839 declaration of a package, subprogram, or generic, or a generic 1840 instantiation), or a library unit renaming declaration (a package, 1841 generic, or subprogram renaming declaration). Such files are also 1842 called "specs". 1843 1844'FILE.adb' 1845 Ada source code file containing a library unit body (a subprogram 1846 or package body). Such files are also called "bodies". 1847 1848'FILE.s' 1849 Assembler code. 1850 1851'FILE.S' 1852'FILE.sx' 1853 Assembler code that must be preprocessed. 1854 1855'OTHER' 1856 An object file to be fed straight into linking. Any file name with 1857 no recognized suffix is treated this way. 1858 1859 You can specify the input language explicitly with the '-x' option: 1860 1861'-x LANGUAGE' 1862 Specify explicitly the LANGUAGE for the following input files 1863 (rather than letting the compiler choose a default based on the 1864 file name suffix). This option applies to all following input 1865 files until the next '-x' option. Possible values for LANGUAGE 1866 are: 1867 c c-header cpp-output 1868 c++ c++-header c++-cpp-output 1869 objective-c objective-c-header objective-c-cpp-output 1870 objective-c++ objective-c++-header objective-c++-cpp-output 1871 assembler assembler-with-cpp 1872 ada 1873 d 1874 f77 f77-cpp-input f95 f95-cpp-input 1875 go 1876 brig 1877 1878'-x none' 1879 Turn off any specification of a language, so that subsequent files 1880 are handled according to their file name suffixes (as they are if 1881 '-x' has not been used at all). 1882 1883 If you only want some of the stages of compilation, you can use '-x' 1884(or filename suffixes) to tell 'gcc' where to start, and one of the 1885options '-c', '-S', or '-E' to say where 'gcc' is to stop. Note that 1886some combinations (for example, '-x cpp-output -E') instruct 'gcc' to do 1887nothing at all. 1888 1889'-c' 1890 Compile or assemble the source files, but do not link. The linking 1891 stage simply is not done. The ultimate output is in the form of an 1892 object file for each source file. 1893 1894 By default, the object file name for a source file is made by 1895 replacing the suffix '.c', '.i', '.s', etc., with '.o'. 1896 1897 Unrecognized input files, not requiring compilation or assembly, 1898 are ignored. 1899 1900'-S' 1901 Stop after the stage of compilation proper; do not assemble. The 1902 output is in the form of an assembler code file for each 1903 non-assembler input file specified. 1904 1905 By default, the assembler file name for a source file is made by 1906 replacing the suffix '.c', '.i', etc., with '.s'. 1907 1908 Input files that don't require compilation are ignored. 1909 1910'-E' 1911 Stop after the preprocessing stage; do not run the compiler proper. 1912 The output is in the form of preprocessed source code, which is 1913 sent to the standard output. 1914 1915 Input files that don't require preprocessing are ignored. 1916 1917'-o FILE' 1918 Place output in file FILE. This applies to whatever sort of output 1919 is being produced, whether it be an executable file, an object 1920 file, an assembler file or preprocessed C code. 1921 1922 If '-o' is not specified, the default is to put an executable file 1923 in 'a.out', the object file for 'SOURCE.SUFFIX' in 'SOURCE.o', its 1924 assembler file in 'SOURCE.s', a precompiled header file in 1925 'SOURCE.SUFFIX.gch', and all preprocessed C source on standard 1926 output. 1927 1928'-v' 1929 Print (on standard error output) the commands executed to run the 1930 stages of compilation. Also print the version number of the 1931 compiler driver program and of the preprocessor and the compiler 1932 proper. 1933 1934'-###' 1935 Like '-v' except the commands are not executed and arguments are 1936 quoted unless they contain only alphanumeric characters or './-_'. 1937 This is useful for shell scripts to capture the driver-generated 1938 command lines. 1939 1940'--help' 1941 Print (on the standard output) a description of the command-line 1942 options understood by 'gcc'. If the '-v' option is also specified 1943 then '--help' is also passed on to the various processes invoked by 1944 'gcc', so that they can display the command-line options they 1945 accept. If the '-Wextra' option has also been specified (prior to 1946 the '--help' option), then command-line options that have no 1947 documentation associated with them are also displayed. 1948 1949'--target-help' 1950 Print (on the standard output) a description of target-specific 1951 command-line options for each tool. For some targets extra 1952 target-specific information may also be printed. 1953 1954'--help={CLASS|[^]QUALIFIER}[,...]' 1955 Print (on the standard output) a description of the command-line 1956 options understood by the compiler that fit into all specified 1957 classes and qualifiers. These are the supported classes: 1958 1959 'optimizers' 1960 Display all of the optimization options supported by the 1961 compiler. 1962 1963 'warnings' 1964 Display all of the options controlling warning messages 1965 produced by the compiler. 1966 1967 'target' 1968 Display target-specific options. Unlike the '--target-help' 1969 option however, target-specific options of the linker and 1970 assembler are not displayed. This is because those tools do 1971 not currently support the extended '--help=' syntax. 1972 1973 'params' 1974 Display the values recognized by the '--param' option. 1975 1976 LANGUAGE 1977 Display the options supported for LANGUAGE, where LANGUAGE is 1978 the name of one of the languages supported in this version of 1979 GCC. If an option is supported by all languages, one needs to 1980 select 'common' class. 1981 1982 'common' 1983 Display the options that are common to all languages. 1984 1985 These are the supported qualifiers: 1986 1987 'undocumented' 1988 Display only those options that are undocumented. 1989 1990 'joined' 1991 Display options taking an argument that appears after an equal 1992 sign in the same continuous piece of text, such as: 1993 '--help=target'. 1994 1995 'separate' 1996 Display options taking an argument that appears as a separate 1997 word following the original option, such as: '-o output-file'. 1998 1999 Thus for example to display all the undocumented target-specific 2000 switches supported by the compiler, use: 2001 2002 --help=target,undocumented 2003 2004 The sense of a qualifier can be inverted by prefixing it with the 2005 '^' character, so for example to display all binary warning options 2006 (i.e., ones that are either on or off and that do not take an 2007 argument) that have a description, use: 2008 2009 --help=warnings,^joined,^undocumented 2010 2011 The argument to '--help=' should not consist solely of inverted 2012 qualifiers. 2013 2014 Combining several classes is possible, although this usually 2015 restricts the output so much that there is nothing to display. One 2016 case where it does work, however, is when one of the classes is 2017 TARGET. For example, to display all the target-specific 2018 optimization options, use: 2019 2020 --help=target,optimizers 2021 2022 The '--help=' option can be repeated on the command line. Each 2023 successive use displays its requested class of options, skipping 2024 those that have already been displayed. If '--help' is also 2025 specified anywhere on the command line then this takes precedence 2026 over any '--help=' option. 2027 2028 If the '-Q' option appears on the command line before the '--help=' 2029 option, then the descriptive text displayed by '--help=' is 2030 changed. Instead of describing the displayed options, an 2031 indication is given as to whether the option is enabled, disabled 2032 or set to a specific value (assuming that the compiler knows this 2033 at the point where the '--help=' option is used). 2034 2035 Here is a truncated example from the ARM port of 'gcc': 2036 2037 % gcc -Q -mabi=2 --help=target -c 2038 The following options are target specific: 2039 -mabi= 2 2040 -mabort-on-noreturn [disabled] 2041 -mapcs [disabled] 2042 2043 The output is sensitive to the effects of previous command-line 2044 options, so for example it is possible to find out which 2045 optimizations are enabled at '-O2' by using: 2046 2047 -Q -O2 --help=optimizers 2048 2049 Alternatively you can discover which binary optimizations are 2050 enabled by '-O3' by using: 2051 2052 gcc -c -Q -O3 --help=optimizers > /tmp/O3-opts 2053 gcc -c -Q -O2 --help=optimizers > /tmp/O2-opts 2054 diff /tmp/O2-opts /tmp/O3-opts | grep enabled 2055 2056'--version' 2057 Display the version number and copyrights of the invoked GCC. 2058 2059'-pass-exit-codes' 2060 Normally the 'gcc' program exits with the code of 1 if any phase of 2061 the compiler returns a non-success return code. If you specify 2062 '-pass-exit-codes', the 'gcc' program instead returns with the 2063 numerically highest error produced by any phase returning an error 2064 indication. The C, C++, and Fortran front ends return 4 if an 2065 internal compiler error is encountered. 2066 2067'-pipe' 2068 Use pipes rather than temporary files for communication between the 2069 various stages of compilation. This fails to work on some systems 2070 where the assembler is unable to read from a pipe; but the GNU 2071 assembler has no trouble. 2072 2073'-specs=FILE' 2074 Process FILE after the compiler reads in the standard 'specs' file, 2075 in order to override the defaults which the 'gcc' driver program 2076 uses when determining what switches to pass to 'cc1', 'cc1plus', 2077 'as', 'ld', etc. More than one '-specs=FILE' can be specified on 2078 the command line, and they are processed in order, from left to 2079 right. *Note Spec Files::, for information about the format of the 2080 FILE. 2081 2082'-wrapper' 2083 Invoke all subcommands under a wrapper program. The name of the 2084 wrapper program and its parameters are passed as a comma separated 2085 list. 2086 2087 gcc -c t.c -wrapper gdb,--args 2088 2089 This invokes all subprograms of 'gcc' under 'gdb --args', thus the 2090 invocation of 'cc1' is 'gdb --args cc1 ...'. 2091 2092'-ffile-prefix-map=OLD=NEW' 2093 When compiling files residing in directory 'OLD', record any 2094 references to them in the result of the compilation as if the files 2095 resided in directory 'NEW' instead. Specifying this option is 2096 equivalent to specifying all the individual '-f*-prefix-map' 2097 options. This can be used to make reproducible builds that are 2098 location independent. See also '-fmacro-prefix-map' and 2099 '-fdebug-prefix-map'. 2100 2101'-fplugin=NAME.so' 2102 Load the plugin code in file NAME.so, assumed to be a shared object 2103 to be dlopen'd by the compiler. The base name of the shared object 2104 file is used to identify the plugin for the purposes of argument 2105 parsing (See '-fplugin-arg-NAME-KEY=VALUE' below). Each plugin 2106 should define the callback functions specified in the Plugins API. 2107 2108'-fplugin-arg-NAME-KEY=VALUE' 2109 Define an argument called KEY with a value of VALUE for the plugin 2110 called NAME. 2111 2112'-fdump-ada-spec[-slim]' 2113 For C and C++ source and include files, generate corresponding Ada 2114 specs. *Note (gnat_ugn)Generating Ada Bindings for C and C++ 2115 headers::, which provides detailed documentation on this feature. 2116 2117'-fada-spec-parent=UNIT' 2118 In conjunction with '-fdump-ada-spec[-slim]' above, generate Ada 2119 specs as child units of parent UNIT. 2120 2121'-fdump-go-spec=FILE' 2122 For input files in any language, generate corresponding Go 2123 declarations in FILE. This generates Go 'const', 'type', 'var', 2124 and 'func' declarations which may be a useful way to start writing 2125 a Go interface to code written in some other language. 2126 2127'@FILE' 2128 Read command-line options from FILE. The options read are inserted 2129 in place of the original @FILE option. If FILE does not exist, or 2130 cannot be read, then the option will be treated literally, and not 2131 removed. 2132 2133 Options in FILE are separated by whitespace. A whitespace 2134 character may be included in an option by surrounding the entire 2135 option in either single or double quotes. Any character (including 2136 a backslash) may be included by prefixing the character to be 2137 included with a backslash. The FILE may itself contain additional 2138 @FILE options; any such options will be processed recursively. 2139 2140 2141File: gcc.info, Node: Invoking G++, Next: C Dialect Options, Prev: Overall Options, Up: Invoking GCC 2142 21433.3 Compiling C++ Programs 2144========================== 2145 2146C++ source files conventionally use one of the suffixes '.C', '.cc', 2147'.cpp', '.CPP', '.c++', '.cp', or '.cxx'; C++ header files often use 2148'.hh', '.hpp', '.H', or (for shared template code) '.tcc'; and 2149preprocessed C++ files use the suffix '.ii'. GCC recognizes files with 2150these names and compiles them as C++ programs even if you call the 2151compiler the same way as for compiling C programs (usually with the name 2152'gcc'). 2153 2154 However, the use of 'gcc' does not add the C++ library. 'g++' is a 2155program that calls GCC and automatically specifies linking against the 2156C++ library. It treats '.c', '.h' and '.i' files as C++ source files 2157instead of C source files unless '-x' is used. This program is also 2158useful when precompiling a C header file with a '.h' extension for use 2159in C++ compilations. On many systems, 'g++' is also installed with the 2160name 'c++'. 2161 2162 When you compile C++ programs, you may specify many of the same 2163command-line options that you use for compiling programs in any 2164language; or command-line options meaningful for C and related 2165languages; or options that are meaningful only for C++ programs. *Note 2166Options Controlling C Dialect: C Dialect Options, for explanations of 2167options for languages related to C. *Note Options Controlling C++ 2168Dialect: C++ Dialect Options, for explanations of options that are 2169meaningful only for C++ programs. 2170 2171 2172File: gcc.info, Node: C Dialect Options, Next: C++ Dialect Options, Prev: Invoking G++, Up: Invoking GCC 2173 21743.4 Options Controlling C Dialect 2175================================= 2176 2177The following options control the dialect of C (or languages derived 2178from C, such as C++, Objective-C and Objective-C++) that the compiler 2179accepts: 2180 2181'-ansi' 2182 In C mode, this is equivalent to '-std=c90'. In C++ mode, it is 2183 equivalent to '-std=c++98'. 2184 2185 This turns off certain features of GCC that are incompatible with 2186 ISO C90 (when compiling C code), or of standard C++ (when compiling 2187 C++ code), such as the 'asm' and 'typeof' keywords, and predefined 2188 macros such as 'unix' and 'vax' that identify the type of system 2189 you are using. It also enables the undesirable and rarely used ISO 2190 trigraph feature. For the C compiler, it disables recognition of 2191 C++ style '//' comments as well as the 'inline' keyword. 2192 2193 The alternate keywords '__asm__', '__extension__', '__inline__' and 2194 '__typeof__' continue to work despite '-ansi'. You would not want 2195 to use them in an ISO C program, of course, but it is useful to put 2196 them in header files that might be included in compilations done 2197 with '-ansi'. Alternate predefined macros such as '__unix__' and 2198 '__vax__' are also available, with or without '-ansi'. 2199 2200 The '-ansi' option does not cause non-ISO programs to be rejected 2201 gratuitously. For that, '-Wpedantic' is required in addition to 2202 '-ansi'. *Note Warning Options::. 2203 2204 The macro '__STRICT_ANSI__' is predefined when the '-ansi' option 2205 is used. Some header files may notice this macro and refrain from 2206 declaring certain functions or defining certain macros that the ISO 2207 standard doesn't call for; this is to avoid interfering with any 2208 programs that might use these names for other things. 2209 2210 Functions that are normally built in but do not have semantics 2211 defined by ISO C (such as 'alloca' and 'ffs') are not built-in 2212 functions when '-ansi' is used. *Note Other built-in functions 2213 provided by GCC: Other Builtins, for details of the functions 2214 affected. 2215 2216'-std=' 2217 Determine the language standard. *Note Language Standards 2218 Supported by GCC: Standards, for details of these standard 2219 versions. This option is currently only supported when compiling C 2220 or C++. 2221 2222 The compiler can accept several base standards, such as 'c90' or 2223 'c++98', and GNU dialects of those standards, such as 'gnu90' or 2224 'gnu++98'. When a base standard is specified, the compiler accepts 2225 all programs following that standard plus those using GNU 2226 extensions that do not contradict it. For example, '-std=c90' 2227 turns off certain features of GCC that are incompatible with ISO 2228 C90, such as the 'asm' and 'typeof' keywords, but not other GNU 2229 extensions that do not have a meaning in ISO C90, such as omitting 2230 the middle term of a '?:' expression. On the other hand, when a 2231 GNU dialect of a standard is specified, all features supported by 2232 the compiler are enabled, even when those features change the 2233 meaning of the base standard. As a result, some strict-conforming 2234 programs may be rejected. The particular standard is used by 2235 '-Wpedantic' to identify which features are GNU extensions given 2236 that version of the standard. For example '-std=gnu90 -Wpedantic' 2237 warns about C++ style '//' comments, while '-std=gnu99 -Wpedantic' 2238 does not. 2239 2240 A value for this option must be provided; possible values are 2241 2242 'c90' 2243 'c89' 2244 'iso9899:1990' 2245 Support all ISO C90 programs (certain GNU extensions that 2246 conflict with ISO C90 are disabled). Same as '-ansi' for C 2247 code. 2248 2249 'iso9899:199409' 2250 ISO C90 as modified in amendment 1. 2251 2252 'c99' 2253 'c9x' 2254 'iso9899:1999' 2255 'iso9899:199x' 2256 ISO C99. This standard is substantially completely supported, 2257 modulo bugs and floating-point issues (mainly but not entirely 2258 relating to optional C99 features from Annexes F and G). See 2259 <http://gcc.gnu.org/c99status.html> for more information. The 2260 names 'c9x' and 'iso9899:199x' are deprecated. 2261 2262 'c11' 2263 'c1x' 2264 'iso9899:2011' 2265 ISO C11, the 2011 revision of the ISO C standard. This 2266 standard is substantially completely supported, modulo bugs, 2267 floating-point issues (mainly but not entirely relating to 2268 optional C11 features from Annexes F and G) and the optional 2269 Annexes K (Bounds-checking interfaces) and L (Analyzability). 2270 The name 'c1x' is deprecated. 2271 2272 'c17' 2273 'c18' 2274 'iso9899:2017' 2275 'iso9899:2018' 2276 ISO C17, the 2017 revision of the ISO C standard (published in 2277 2018). This standard is same as C11 except for corrections of 2278 defects (all of which are also applied with '-std=c11') and a 2279 new value of '__STDC_VERSION__', and so is supported to the 2280 same extent as C11. 2281 2282 'c2x' 2283 The next version of the ISO C standard, still under 2284 development. The support for this version is experimental and 2285 incomplete. 2286 2287 'gnu90' 2288 'gnu89' 2289 GNU dialect of ISO C90 (including some C99 features). 2290 2291 'gnu99' 2292 'gnu9x' 2293 GNU dialect of ISO C99. The name 'gnu9x' is deprecated. 2294 2295 'gnu11' 2296 'gnu1x' 2297 GNU dialect of ISO C11. The name 'gnu1x' is deprecated. 2298 2299 'gnu17' 2300 'gnu18' 2301 GNU dialect of ISO C17. This is the default for C code. 2302 2303 'gnu2x' 2304 The next version of the ISO C standard, still under 2305 development, plus GNU extensions. The support for this 2306 version is experimental and incomplete. 2307 2308 'c++98' 2309 'c++03' 2310 The 1998 ISO C++ standard plus the 2003 technical corrigendum 2311 and some additional defect reports. Same as '-ansi' for C++ 2312 code. 2313 2314 'gnu++98' 2315 'gnu++03' 2316 GNU dialect of '-std=c++98'. 2317 2318 'c++11' 2319 'c++0x' 2320 The 2011 ISO C++ standard plus amendments. The name 'c++0x' 2321 is deprecated. 2322 2323 'gnu++11' 2324 'gnu++0x' 2325 GNU dialect of '-std=c++11'. The name 'gnu++0x' is 2326 deprecated. 2327 2328 'c++14' 2329 'c++1y' 2330 The 2014 ISO C++ standard plus amendments. The name 'c++1y' 2331 is deprecated. 2332 2333 'gnu++14' 2334 'gnu++1y' 2335 GNU dialect of '-std=c++14'. This is the default for C++ 2336 code. The name 'gnu++1y' is deprecated. 2337 2338 'c++17' 2339 'c++1z' 2340 The 2017 ISO C++ standard plus amendments. The name 'c++1z' 2341 is deprecated. 2342 2343 'gnu++17' 2344 'gnu++1z' 2345 GNU dialect of '-std=c++17'. The name 'gnu++1z' is 2346 deprecated. 2347 2348 'c++20' 2349 'c++2a' 2350 The next revision of the ISO C++ standard, planned for 2020. 2351 Support is highly experimental, and will almost certainly 2352 change in incompatible ways in future releases. 2353 2354 'gnu++20' 2355 'gnu++2a' 2356 GNU dialect of '-std=c++20'. Support is highly experimental, 2357 and will almost certainly change in incompatible ways in 2358 future releases. 2359 2360'-fgnu89-inline' 2361 The option '-fgnu89-inline' tells GCC to use the traditional GNU 2362 semantics for 'inline' functions when in C99 mode. *Note An Inline 2363 Function is As Fast As a Macro: Inline. Using this option is 2364 roughly equivalent to adding the 'gnu_inline' function attribute to 2365 all inline functions (*note Function Attributes::). 2366 2367 The option '-fno-gnu89-inline' explicitly tells GCC to use the C99 2368 semantics for 'inline' when in C99 or gnu99 mode (i.e., it 2369 specifies the default behavior). This option is not supported in 2370 '-std=c90' or '-std=gnu90' mode. 2371 2372 The preprocessor macros '__GNUC_GNU_INLINE__' and 2373 '__GNUC_STDC_INLINE__' may be used to check which semantics are in 2374 effect for 'inline' functions. *Note (cpp)Common Predefined 2375 Macros::. 2376 2377'-fpermitted-flt-eval-methods=STYLE' 2378 ISO/IEC TS 18661-3 defines new permissible values for 2379 'FLT_EVAL_METHOD' that indicate that operations and constants with 2380 a semantic type that is an interchange or extended format should be 2381 evaluated to the precision and range of that type. These new 2382 values are a superset of those permitted under C99/C11, which does 2383 not specify the meaning of other positive values of 2384 'FLT_EVAL_METHOD'. As such, code conforming to C11 may not have 2385 been written expecting the possibility of the new values. 2386 2387 '-fpermitted-flt-eval-methods' specifies whether the compiler 2388 should allow only the values of 'FLT_EVAL_METHOD' specified in 2389 C99/C11, or the extended set of values specified in ISO/IEC TS 2390 18661-3. 2391 2392 STYLE is either 'c11' or 'ts-18661-3' as appropriate. 2393 2394 The default when in a standards compliant mode ('-std=c11' or 2395 similar) is '-fpermitted-flt-eval-methods=c11'. The default when 2396 in a GNU dialect ('-std=gnu11' or similar) is 2397 '-fpermitted-flt-eval-methods=ts-18661-3'. 2398 2399'-aux-info FILENAME' 2400 Output to the given filename prototyped declarations for all 2401 functions declared and/or defined in a translation unit, including 2402 those in header files. This option is silently ignored in any 2403 language other than C. 2404 2405 Besides declarations, the file indicates, in comments, the origin 2406 of each declaration (source file and line), whether the declaration 2407 was implicit, prototyped or unprototyped ('I', 'N' for new or 'O' 2408 for old, respectively, in the first character after the line number 2409 and the colon), and whether it came from a declaration or a 2410 definition ('C' or 'F', respectively, in the following character). 2411 In the case of function definitions, a K&R-style list of arguments 2412 followed by their declarations is also provided, inside comments, 2413 after the declaration. 2414 2415'-fallow-parameterless-variadic-functions' 2416 Accept variadic functions without named parameters. 2417 2418 Although it is possible to define such a function, this is not very 2419 useful as it is not possible to read the arguments. This is only 2420 supported for C as this construct is allowed by C++. 2421 2422'-fno-asm' 2423 Do not recognize 'asm', 'inline' or 'typeof' as a keyword, so that 2424 code can use these words as identifiers. You can use the keywords 2425 '__asm__', '__inline__' and '__typeof__' instead. '-ansi' implies 2426 '-fno-asm'. 2427 2428 In C++, this switch only affects the 'typeof' keyword, since 'asm' 2429 and 'inline' are standard keywords. You may want to use the 2430 '-fno-gnu-keywords' flag instead, which has the same effect. In 2431 C99 mode ('-std=c99' or '-std=gnu99'), this switch only affects the 2432 'asm' and 'typeof' keywords, since 'inline' is a standard keyword 2433 in ISO C99. 2434 2435'-fno-builtin' 2436'-fno-builtin-FUNCTION' 2437 Don't recognize built-in functions that do not begin with 2438 '__builtin_' as prefix. *Note Other built-in functions provided by 2439 GCC: Other Builtins, for details of the functions affected, 2440 including those which are not built-in functions when '-ansi' or 2441 '-std' options for strict ISO C conformance are used because they 2442 do not have an ISO standard meaning. 2443 2444 GCC normally generates special code to handle certain built-in 2445 functions more efficiently; for instance, calls to 'alloca' may 2446 become single instructions which adjust the stack directly, and 2447 calls to 'memcpy' may become inline copy loops. The resulting code 2448 is often both smaller and faster, but since the function calls no 2449 longer appear as such, you cannot set a breakpoint on those calls, 2450 nor can you change the behavior of the functions by linking with a 2451 different library. In addition, when a function is recognized as a 2452 built-in function, GCC may use information about that function to 2453 warn about problems with calls to that function, or to generate 2454 more efficient code, even if the resulting code still contains 2455 calls to that function. For example, warnings are given with 2456 '-Wformat' for bad calls to 'printf' when 'printf' is built in and 2457 'strlen' is known not to modify global memory. 2458 2459 With the '-fno-builtin-FUNCTION' option only the built-in function 2460 FUNCTION is disabled. FUNCTION must not begin with '__builtin_'. 2461 If a function is named that is not built-in in this version of GCC, 2462 this option is ignored. There is no corresponding 2463 '-fbuiltin-FUNCTION' option; if you wish to enable built-in 2464 functions selectively when using '-fno-builtin' or 2465 '-ffreestanding', you may define macros such as: 2466 2467 #define abs(n) __builtin_abs ((n)) 2468 #define strcpy(d, s) __builtin_strcpy ((d), (s)) 2469 2470'-fgimple' 2471 2472 Enable parsing of function definitions marked with '__GIMPLE'. 2473 This is an experimental feature that allows unit testing of GIMPLE 2474 passes. 2475 2476'-fhosted' 2477 2478 Assert that compilation targets a hosted environment. This implies 2479 '-fbuiltin'. A hosted environment is one in which the entire 2480 standard library is available, and in which 'main' has a return 2481 type of 'int'. Examples are nearly everything except a kernel. 2482 This is equivalent to '-fno-freestanding'. 2483 2484'-ffreestanding' 2485 2486 Assert that compilation targets a freestanding environment. This 2487 implies '-fno-builtin'. A freestanding environment is one in which 2488 the standard library may not exist, and program startup may not 2489 necessarily be at 'main'. The most obvious example is an OS 2490 kernel. This is equivalent to '-fno-hosted'. 2491 2492 *Note Language Standards Supported by GCC: Standards, for details 2493 of freestanding and hosted environments. 2494 2495'-fopenacc' 2496 Enable handling of OpenACC directives '#pragma acc' in C/C++ and 2497 '!$acc' in Fortran. When '-fopenacc' is specified, the compiler 2498 generates accelerated code according to the OpenACC Application 2499 Programming Interface v2.6 <https://www.openacc.org>. This option 2500 implies '-pthread', and thus is only supported on targets that have 2501 support for '-pthread'. 2502 2503'-fopenacc-dim=GEOM' 2504 Specify default compute dimensions for parallel offload regions 2505 that do not explicitly specify. The GEOM value is a triple of 2506 ':'-separated sizes, in order 'gang', 'worker' and, 'vector'. A 2507 size can be omitted, to use a target-specific default value. 2508 2509'-fopenmp' 2510 Enable handling of OpenMP directives '#pragma omp' in C/C++ and 2511 '!$omp' in Fortran. When '-fopenmp' is specified, the compiler 2512 generates parallel code according to the OpenMP Application Program 2513 Interface v4.5 <https://www.openmp.org>. This option implies 2514 '-pthread', and thus is only supported on targets that have support 2515 for '-pthread'. '-fopenmp' implies '-fopenmp-simd'. 2516 2517'-fopenmp-simd' 2518 Enable handling of OpenMP's SIMD directives with '#pragma omp' in 2519 C/C++ and '!$omp' in Fortran. Other OpenMP directives are ignored. 2520 2521'-fgnu-tm' 2522 When the option '-fgnu-tm' is specified, the compiler generates 2523 code for the Linux variant of Intel's current Transactional Memory 2524 ABI specification document (Revision 1.1, May 6 2009). This is an 2525 experimental feature whose interface may change in future versions 2526 of GCC, as the official specification changes. Please note that 2527 not all architectures are supported for this feature. 2528 2529 For more information on GCC's support for transactional memory, 2530 *Note The GNU Transactional Memory Library: (libitm)Enabling 2531 libitm. 2532 2533 Note that the transactional memory feature is not supported with 2534 non-call exceptions ('-fnon-call-exceptions'). 2535 2536'-fms-extensions' 2537 Accept some non-standard constructs used in Microsoft header files. 2538 2539 In C++ code, this allows member names in structures to be similar 2540 to previous types declarations. 2541 2542 typedef int UOW; 2543 struct ABC { 2544 UOW UOW; 2545 }; 2546 2547 Some cases of unnamed fields in structures and unions are only 2548 accepted with this option. *Note Unnamed struct/union fields 2549 within structs/unions: Unnamed Fields, for details. 2550 2551 Note that this option is off for all targets except for x86 targets 2552 using ms-abi. 2553 2554'-fplan9-extensions' 2555 Accept some non-standard constructs used in Plan 9 code. 2556 2557 This enables '-fms-extensions', permits passing pointers to 2558 structures with anonymous fields to functions that expect pointers 2559 to elements of the type of the field, and permits referring to 2560 anonymous fields declared using a typedef. *Note Unnamed 2561 struct/union fields within structs/unions: Unnamed Fields, for 2562 details. This is only supported for C, not C++. 2563 2564'-fcond-mismatch' 2565 Allow conditional expressions with mismatched types in the second 2566 and third arguments. The value of such an expression is void. 2567 This option is not supported for C++. 2568 2569'-flax-vector-conversions' 2570 Allow implicit conversions between vectors with differing numbers 2571 of elements and/or incompatible element types. This option should 2572 not be used for new code. 2573 2574'-funsigned-char' 2575 Let the type 'char' be unsigned, like 'unsigned char'. 2576 2577 Each kind of machine has a default for what 'char' should be. It 2578 is either like 'unsigned char' by default or like 'signed char' by 2579 default. 2580 2581 Ideally, a portable program should always use 'signed char' or 2582 'unsigned char' when it depends on the signedness of an object. 2583 But many programs have been written to use plain 'char' and expect 2584 it to be signed, or expect it to be unsigned, depending on the 2585 machines they were written for. This option, and its inverse, let 2586 you make such a program work with the opposite default. 2587 2588 The type 'char' is always a distinct type from each of 'signed 2589 char' or 'unsigned char', even though its behavior is always just 2590 like one of those two. 2591 2592'-fsigned-char' 2593 Let the type 'char' be signed, like 'signed char'. 2594 2595 Note that this is equivalent to '-fno-unsigned-char', which is the 2596 negative form of '-funsigned-char'. Likewise, the option 2597 '-fno-signed-char' is equivalent to '-funsigned-char'. 2598 2599'-fsigned-bitfields' 2600'-funsigned-bitfields' 2601'-fno-signed-bitfields' 2602'-fno-unsigned-bitfields' 2603 These options control whether a bit-field is signed or unsigned, 2604 when the declaration does not use either 'signed' or 'unsigned'. 2605 By default, such a bit-field is signed, because this is consistent: 2606 the basic integer types such as 'int' are signed types. 2607 2608'-fsso-struct=ENDIANNESS' 2609 Set the default scalar storage order of structures and unions to 2610 the specified endianness. The accepted values are 'big-endian', 2611 'little-endian' and 'native' for the native endianness of the 2612 target (the default). This option is not supported for C++. 2613 2614 *Warning:* the '-fsso-struct' switch causes GCC to generate code 2615 that is not binary compatible with code generated without it if the 2616 specified endianness is not the native endianness of the target. 2617 2618 2619File: gcc.info, Node: C++ Dialect Options, Next: Objective-C and Objective-C++ Dialect Options, Prev: C Dialect Options, Up: Invoking GCC 2620 26213.5 Options Controlling C++ Dialect 2622=================================== 2623 2624This section describes the command-line options that are only meaningful 2625for C++ programs. You can also use most of the GNU compiler options 2626regardless of what language your program is in. For example, you might 2627compile a file 'firstClass.C' like this: 2628 2629 g++ -g -fstrict-enums -O -c firstClass.C 2630 2631In this example, only '-fstrict-enums' is an option meant only for C++ 2632programs; you can use the other options with any language supported by 2633GCC. 2634 2635 Some options for compiling C programs, such as '-std', are also 2636relevant for C++ programs. *Note Options Controlling C Dialect: C 2637Dialect Options. 2638 2639 Here is a list of options that are _only_ for compiling C++ programs: 2640 2641'-fabi-version=N' 2642 Use version N of the C++ ABI. The default is version 0. 2643 2644 Version 0 refers to the version conforming most closely to the C++ 2645 ABI specification. Therefore, the ABI obtained using version 0 2646 will change in different versions of G++ as ABI bugs are fixed. 2647 2648 Version 1 is the version of the C++ ABI that first appeared in G++ 2649 3.2. 2650 2651 Version 2 is the version of the C++ ABI that first appeared in G++ 2652 3.4, and was the default through G++ 4.9. 2653 2654 Version 3 corrects an error in mangling a constant address as a 2655 template argument. 2656 2657 Version 4, which first appeared in G++ 4.5, implements a standard 2658 mangling for vector types. 2659 2660 Version 5, which first appeared in G++ 4.6, corrects the mangling 2661 of attribute const/volatile on function pointer types, decltype of 2662 a plain decl, and use of a function parameter in the declaration of 2663 another parameter. 2664 2665 Version 6, which first appeared in G++ 4.7, corrects the promotion 2666 behavior of C++11 scoped enums and the mangling of template 2667 argument packs, const/static_cast, prefix ++ and -, and a class 2668 scope function used as a template argument. 2669 2670 Version 7, which first appeared in G++ 4.8, that treats nullptr_t 2671 as a builtin type and corrects the mangling of lambdas in default 2672 argument scope. 2673 2674 Version 8, which first appeared in G++ 4.9, corrects the 2675 substitution behavior of function types with 2676 function-cv-qualifiers. 2677 2678 Version 9, which first appeared in G++ 5.2, corrects the alignment 2679 of 'nullptr_t'. 2680 2681 Version 10, which first appeared in G++ 6.1, adds mangling of 2682 attributes that affect type identity, such as ia32 calling 2683 convention attributes (e.g. 'stdcall'). 2684 2685 Version 11, which first appeared in G++ 7, corrects the mangling of 2686 sizeof... expressions and operator names. For multiple entities 2687 with the same name within a function, that are declared in 2688 different scopes, the mangling now changes starting with the 2689 twelfth occurrence. It also implies '-fnew-inheriting-ctors'. 2690 2691 Version 12, which first appeared in G++ 8, corrects the calling 2692 conventions for empty classes on the x86_64 target and for classes 2693 with only deleted copy/move constructors. It accidentally changes 2694 the calling convention for classes with a deleted copy constructor 2695 and a trivial move constructor. 2696 2697 Version 13, which first appeared in G++ 8.2, fixes the accidental 2698 change in version 12. 2699 2700 Version 14, which first appeared in G++ 10, corrects the mangling 2701 of the nullptr expression. 2702 2703 See also '-Wabi'. 2704 2705'-fabi-compat-version=N' 2706 On targets that support strong aliases, G++ works around mangling 2707 changes by creating an alias with the correct mangled name when 2708 defining a symbol with an incorrect mangled name. This switch 2709 specifies which ABI version to use for the alias. 2710 2711 With '-fabi-version=0' (the default), this defaults to 11 (GCC 7 2712 compatibility). If another ABI version is explicitly selected, 2713 this defaults to 0. For compatibility with GCC versions 3.2 2714 through 4.9, use '-fabi-compat-version=2'. 2715 2716 If this option is not provided but '-Wabi=N' is, that version is 2717 used for compatibility aliases. If this option is provided along 2718 with '-Wabi' (without the version), the version from this option is 2719 used for the warning. 2720 2721'-fno-access-control' 2722 Turn off all access checking. This switch is mainly useful for 2723 working around bugs in the access control code. 2724 2725'-faligned-new' 2726 Enable support for C++17 'new' of types that require more alignment 2727 than 'void* ::operator new(std::size_t)' provides. A numeric 2728 argument such as '-faligned-new=32' can be used to specify how much 2729 alignment (in bytes) is provided by that function, but few users 2730 will need to override the default of 'alignof(std::max_align_t)'. 2731 2732 This flag is enabled by default for '-std=c++17'. 2733 2734'-fchar8_t' 2735'-fno-char8_t' 2736 Enable support for 'char8_t' as adopted for C++2a. This includes 2737 the addition of a new 'char8_t' fundamental type, changes to the 2738 types of UTF-8 string and character literals, new signatures for 2739 user-defined literals, associated standard library updates, and new 2740 '__cpp_char8_t' and '__cpp_lib_char8_t' feature test macros. 2741 2742 This option enables functions to be overloaded for ordinary and 2743 UTF-8 strings: 2744 2745 int f(const char *); // #1 2746 int f(const char8_t *); // #2 2747 int v1 = f("text"); // Calls #1 2748 int v2 = f(u8"text"); // Calls #2 2749 2750 and introduces new signatures for user-defined literals: 2751 2752 int operator""_udl1(char8_t); 2753 int v3 = u8'x'_udl1; 2754 int operator""_udl2(const char8_t*, std::size_t); 2755 int v4 = u8"text"_udl2; 2756 template<typename T, T...> int operator""_udl3(); 2757 int v5 = u8"text"_udl3; 2758 2759 The change to the types of UTF-8 string and character literals 2760 introduces incompatibilities with ISO C++11 and later standards. 2761 For example, the following code is well-formed under ISO C++11, but 2762 is ill-formed when '-fchar8_t' is specified. 2763 2764 char ca[] = u8"xx"; // error: char-array initialized from wide 2765 // string 2766 const char *cp = u8"xx";// error: invalid conversion from 2767 // `const char8_t*' to `const char*' 2768 int f(const char*); 2769 auto v = f(u8"xx"); // error: invalid conversion from 2770 // `const char8_t*' to `const char*' 2771 std::string s{u8"xx"}; // error: no matching function for call to 2772 // `std::basic_string<char>::basic_string()' 2773 using namespace std::literals; 2774 s = u8"xx"s; // error: conversion from 2775 // `basic_string<char8_t>' to non-scalar 2776 // type `basic_string<char>' requested 2777 2778'-fcheck-new' 2779 Check that the pointer returned by 'operator new' is non-null 2780 before attempting to modify the storage allocated. This check is 2781 normally unnecessary because the C++ standard specifies that 2782 'operator new' only returns '0' if it is declared 'throw()', in 2783 which case the compiler always checks the return value even without 2784 this option. In all other cases, when 'operator new' has a 2785 non-empty exception specification, memory exhaustion is signalled 2786 by throwing 'std::bad_alloc'. See also 'new (nothrow)'. 2787 2788'-fconcepts' 2789'-fconcepts-ts' 2790 Below '-std=c++2a', '-fconcepts' enables support for the C++ 2791 Extensions for Concepts Technical Specification, ISO 19217 (2015). 2792 2793 With '-std=c++2a' and above, Concepts are part of the language 2794 standard, so '-fconcepts' defaults to on. But the standard 2795 specification of Concepts differs significantly from the TS, so 2796 some constructs that were allowed in the TS but didn't make it into 2797 the standard can still be enabled by '-fconcepts-ts'. 2798 2799'-fconstexpr-depth=N' 2800 Set the maximum nested evaluation depth for C++11 constexpr 2801 functions to N. A limit is needed to detect endless recursion 2802 during constant expression evaluation. The minimum specified by 2803 the standard is 512. 2804 2805'-fconstexpr-cache-depth=N' 2806 Set the maximum level of nested evaluation depth for C++11 2807 constexpr functions that will be cached to N. This is a heuristic 2808 that trades off compilation speed (when the cache avoids repeated 2809 calculations) against memory consumption (when the cache grows very 2810 large from highly recursive evaluations). The default is 8. Very 2811 few users are likely to want to adjust it, but if your code does 2812 heavy constexpr calculations you might want to experiment to find 2813 which value works best for you. 2814 2815'-fconstexpr-loop-limit=N' 2816 Set the maximum number of iterations for a loop in C++14 constexpr 2817 functions to N. A limit is needed to detect infinite loops during 2818 constant expression evaluation. The default is 262144 (1<<18). 2819 2820'-fconstexpr-ops-limit=N' 2821 Set the maximum number of operations during a single constexpr 2822 evaluation. Even when number of iterations of a single loop is 2823 limited with the above limit, if there are several nested loops and 2824 each of them has many iterations but still smaller than the above 2825 limit, or if in a body of some loop or even outside of a loop too 2826 many expressions need to be evaluated, the resulting constexpr 2827 evaluation might take too long. The default is 33554432 (1<<25). 2828 2829'-fcoroutines' 2830 Enable support for the C++ coroutines extension (experimental). 2831 2832'-fno-elide-constructors' 2833 The C++ standard allows an implementation to omit creating a 2834 temporary that is only used to initialize another object of the 2835 same type. Specifying this option disables that optimization, and 2836 forces G++ to call the copy constructor in all cases. This option 2837 also causes G++ to call trivial member functions which otherwise 2838 would be expanded inline. 2839 2840 In C++17, the compiler is required to omit these temporaries, but 2841 this option still affects trivial member functions. 2842 2843'-fno-enforce-eh-specs' 2844 Don't generate code to check for violation of exception 2845 specifications at run time. This option violates the C++ standard, 2846 but may be useful for reducing code size in production builds, much 2847 like defining 'NDEBUG'. This does not give user code permission to 2848 throw exceptions in violation of the exception specifications; the 2849 compiler still optimizes based on the specifications, so throwing 2850 an unexpected exception results in undefined behavior at run time. 2851 2852'-fextern-tls-init' 2853'-fno-extern-tls-init' 2854 The C++11 and OpenMP standards allow 'thread_local' and 2855 'threadprivate' variables to have dynamic (runtime) initialization. 2856 To support this, any use of such a variable goes through a wrapper 2857 function that performs any necessary initialization. When the use 2858 and definition of the variable are in the same translation unit, 2859 this overhead can be optimized away, but when the use is in a 2860 different translation unit there is significant overhead even if 2861 the variable doesn't actually need dynamic initialization. If the 2862 programmer can be sure that no use of the variable in a 2863 non-defining TU needs to trigger dynamic initialization (either 2864 because the variable is statically initialized, or a use of the 2865 variable in the defining TU will be executed before any uses in 2866 another TU), they can avoid this overhead with the 2867 '-fno-extern-tls-init' option. 2868 2869 On targets that support symbol aliases, the default is 2870 '-fextern-tls-init'. On targets that do not support symbol 2871 aliases, the default is '-fno-extern-tls-init'. 2872 2873'-fno-gnu-keywords' 2874 Do not recognize 'typeof' as a keyword, so that code can use this 2875 word as an identifier. You can use the keyword '__typeof__' 2876 instead. This option is implied by the strict ISO C++ dialects: 2877 '-ansi', '-std=c++98', '-std=c++11', etc. 2878 2879'-fno-implicit-templates' 2880 Never emit code for non-inline templates that are instantiated 2881 implicitly (i.e. by use); only emit code for explicit 2882 instantiations. If you use this option, you must take care to 2883 structure your code to include all the necessary explicit 2884 instantiations to avoid getting undefined symbols at link time. 2885 *Note Template Instantiation::, for more information. 2886 2887'-fno-implicit-inline-templates' 2888 Don't emit code for implicit instantiations of inline templates, 2889 either. The default is to handle inlines differently so that 2890 compiles with and without optimization need the same set of 2891 explicit instantiations. 2892 2893'-fno-implement-inlines' 2894 To save space, do not emit out-of-line copies of inline functions 2895 controlled by '#pragma implementation'. This causes linker errors 2896 if these functions are not inlined everywhere they are called. 2897 2898'-fms-extensions' 2899 Disable Wpedantic warnings about constructs used in MFC, such as 2900 implicit int and getting a pointer to member function via 2901 non-standard syntax. 2902 2903'-fnew-inheriting-ctors' 2904 Enable the P0136 adjustment to the semantics of C++11 constructor 2905 inheritance. This is part of C++17 but also considered to be a 2906 Defect Report against C++11 and C++14. This flag is enabled by 2907 default unless '-fabi-version=10' or lower is specified. 2908 2909'-fnew-ttp-matching' 2910 Enable the P0522 resolution to Core issue 150, template template 2911 parameters and default arguments: this allows a template with 2912 default template arguments as an argument for a template template 2913 parameter with fewer template parameters. This flag is enabled by 2914 default for '-std=c++17'. 2915 2916'-fno-nonansi-builtins' 2917 Disable built-in declarations of functions that are not mandated by 2918 ANSI/ISO C. These include 'ffs', 'alloca', '_exit', 'index', 2919 'bzero', 'conjf', and other related functions. 2920 2921'-fnothrow-opt' 2922 Treat a 'throw()' exception specification as if it were a 2923 'noexcept' specification to reduce or eliminate the text size 2924 overhead relative to a function with no exception specification. 2925 If the function has local variables of types with non-trivial 2926 destructors, the exception specification actually makes the 2927 function smaller because the EH cleanups for those variables can be 2928 optimized away. The semantic effect is that an exception thrown 2929 out of a function with such an exception specification results in a 2930 call to 'terminate' rather than 'unexpected'. 2931 2932'-fno-operator-names' 2933 Do not treat the operator name keywords 'and', 'bitand', 'bitor', 2934 'compl', 'not', 'or' and 'xor' as synonyms as keywords. 2935 2936'-fno-optional-diags' 2937 Disable diagnostics that the standard says a compiler does not need 2938 to issue. Currently, the only such diagnostic issued by G++ is the 2939 one for a name having multiple meanings within a class. 2940 2941'-fpermissive' 2942 Downgrade some diagnostics about nonconformant code from errors to 2943 warnings. Thus, using '-fpermissive' allows some nonconforming 2944 code to compile. 2945 2946'-fno-pretty-templates' 2947 When an error message refers to a specialization of a function 2948 template, the compiler normally prints the signature of the 2949 template followed by the template arguments and any typedefs or 2950 typenames in the signature (e.g. 'void f(T) [with T = int]' rather 2951 than 'void f(int)') so that it's clear which template is involved. 2952 When an error message refers to a specialization of a class 2953 template, the compiler omits any template arguments that match the 2954 default template arguments for that template. If either of these 2955 behaviors make it harder to understand the error message rather 2956 than easier, you can use '-fno-pretty-templates' to disable them. 2957 2958'-fno-rtti' 2959 Disable generation of information about every class with virtual 2960 functions for use by the C++ run-time type identification features 2961 ('dynamic_cast' and 'typeid'). If you don't use those parts of the 2962 language, you can save some space by using this flag. Note that 2963 exception handling uses the same information, but G++ generates it 2964 as needed. The 'dynamic_cast' operator can still be used for casts 2965 that do not require run-time type information, i.e. casts to 'void 2966 *' or to unambiguous base classes. 2967 2968 Mixing code compiled with '-frtti' with that compiled with 2969 '-fno-rtti' may not work. For example, programs may fail to link 2970 if a class compiled with '-fno-rtti' is used as a base for a class 2971 compiled with '-frtti'. 2972 2973'-fsized-deallocation' 2974 Enable the built-in global declarations 2975 void operator delete (void *, std::size_t) noexcept; 2976 void operator delete[] (void *, std::size_t) noexcept; 2977 as introduced in C++14. This is useful for user-defined 2978 replacement deallocation functions that, for example, use the size 2979 of the object to make deallocation faster. Enabled by default 2980 under '-std=c++14' and above. The flag '-Wsized-deallocation' 2981 warns about places that might want to add a definition. 2982 2983'-fstrict-enums' 2984 Allow the compiler to optimize using the assumption that a value of 2985 enumerated type can only be one of the values of the enumeration 2986 (as defined in the C++ standard; basically, a value that can be 2987 represented in the minimum number of bits needed to represent all 2988 the enumerators). This assumption may not be valid if the program 2989 uses a cast to convert an arbitrary integer value to the enumerated 2990 type. 2991 2992'-fstrong-eval-order' 2993 Evaluate member access, array subscripting, and shift expressions 2994 in left-to-right order, and evaluate assignment in right-to-left 2995 order, as adopted for C++17. Enabled by default with '-std=c++17'. 2996 '-fstrong-eval-order=some' enables just the ordering of member 2997 access and shift expressions, and is the default without 2998 '-std=c++17'. 2999 3000'-ftemplate-backtrace-limit=N' 3001 Set the maximum number of template instantiation notes for a single 3002 warning or error to N. The default value is 10. 3003 3004'-ftemplate-depth=N' 3005 Set the maximum instantiation depth for template classes to N. A 3006 limit on the template instantiation depth is needed to detect 3007 endless recursions during template class instantiation. ANSI/ISO 3008 C++ conforming programs must not rely on a maximum depth greater 3009 than 17 (changed to 1024 in C++11). The default value is 900, as 3010 the compiler can run out of stack space before hitting 1024 in some 3011 situations. 3012 3013'-fno-threadsafe-statics' 3014 Do not emit the extra code to use the routines specified in the C++ 3015 ABI for thread-safe initialization of local statics. You can use 3016 this option to reduce code size slightly in code that doesn't need 3017 to be thread-safe. 3018 3019'-fuse-cxa-atexit' 3020 Register destructors for objects with static storage duration with 3021 the '__cxa_atexit' function rather than the 'atexit' function. 3022 This option is required for fully standards-compliant handling of 3023 static destructors, but only works if your C library supports 3024 '__cxa_atexit'. 3025 3026'-fno-use-cxa-get-exception-ptr' 3027 Don't use the '__cxa_get_exception_ptr' runtime routine. This 3028 causes 'std::uncaught_exception' to be incorrect, but is necessary 3029 if the runtime routine is not available. 3030 3031'-fvisibility-inlines-hidden' 3032 This switch declares that the user does not attempt to compare 3033 pointers to inline functions or methods where the addresses of the 3034 two functions are taken in different shared objects. 3035 3036 The effect of this is that GCC may, effectively, mark inline 3037 methods with '__attribute__ ((visibility ("hidden")))' so that they 3038 do not appear in the export table of a DSO and do not require a PLT 3039 indirection when used within the DSO. Enabling this option can 3040 have a dramatic effect on load and link times of a DSO as it 3041 massively reduces the size of the dynamic export table when the 3042 library makes heavy use of templates. 3043 3044 The behavior of this switch is not quite the same as marking the 3045 methods as hidden directly, because it does not affect static 3046 variables local to the function or cause the compiler to deduce 3047 that the function is defined in only one shared object. 3048 3049 You may mark a method as having a visibility explicitly to negate 3050 the effect of the switch for that method. For example, if you do 3051 want to compare pointers to a particular inline method, you might 3052 mark it as having default visibility. Marking the enclosing class 3053 with explicit visibility has no effect. 3054 3055 Explicitly instantiated inline methods are unaffected by this 3056 option as their linkage might otherwise cross a shared library 3057 boundary. *Note Template Instantiation::. 3058 3059'-fvisibility-ms-compat' 3060 This flag attempts to use visibility settings to make GCC's C++ 3061 linkage model compatible with that of Microsoft Visual Studio. 3062 3063 The flag makes these changes to GCC's linkage model: 3064 3065 1. It sets the default visibility to 'hidden', like 3066 '-fvisibility=hidden'. 3067 3068 2. Types, but not their members, are not hidden by default. 3069 3070 3. The One Definition Rule is relaxed for types without explicit 3071 visibility specifications that are defined in more than one 3072 shared object: those declarations are permitted if they are 3073 permitted when this option is not used. 3074 3075 In new code it is better to use '-fvisibility=hidden' and export 3076 those classes that are intended to be externally visible. 3077 Unfortunately it is possible for code to rely, perhaps 3078 accidentally, on the Visual Studio behavior. 3079 3080 Among the consequences of these changes are that static data 3081 members of the same type with the same name but defined in 3082 different shared objects are different, so changing one does not 3083 change the other; and that pointers to function members defined in 3084 different shared objects may not compare equal. When this flag is 3085 given, it is a violation of the ODR to define types with the same 3086 name differently. 3087 3088'-fno-weak' 3089 Do not use weak symbol support, even if it is provided by the 3090 linker. By default, G++ uses weak symbols if they are available. 3091 This option exists only for testing, and should not be used by 3092 end-users; it results in inferior code and has no benefits. This 3093 option may be removed in a future release of G++. 3094 3095'-fext-numeric-literals (C++ and Objective-C++ only)' 3096 Accept imaginary, fixed-point, or machine-defined literal number 3097 suffixes as GNU extensions. When this option is turned off these 3098 suffixes are treated as C++11 user-defined literal numeric 3099 suffixes. This is on by default for all pre-C++11 dialects and all 3100 GNU dialects: '-std=c++98', '-std=gnu++98', '-std=gnu++11', 3101 '-std=gnu++14'. This option is off by default for ISO C++11 3102 onwards ('-std=c++11', ...). 3103 3104'-nostdinc++' 3105 Do not search for header files in the standard directories specific 3106 to C++, but do still search the other standard directories. (This 3107 option is used when building the C++ library.) 3108 3109 In addition, these warning options have meanings only for C++ programs: 3110 3111'-Wabi-tag (C++ and Objective-C++ only)' 3112 Warn when a type with an ABI tag is used in a context that does not 3113 have that ABI tag. See *note C++ Attributes:: for more information 3114 about ABI tags. 3115 3116'-Wcomma-subscript (C++ and Objective-C++ only)' 3117 Warn about uses of a comma expression within a subscripting 3118 expression. This usage was deprecated in C++2a. However, a comma 3119 expression wrapped in '( )' is not deprecated. Example: 3120 3121 void f(int *a, int b, int c) { 3122 a[b,c]; // deprecated 3123 a[(b,c)]; // OK 3124 } 3125 3126 Enabled by default with '-std=c++2a'. 3127 3128'-Wctor-dtor-privacy (C++ and Objective-C++ only)' 3129 Warn when a class seems unusable because all the constructors or 3130 destructors in that class are private, and it has neither friends 3131 nor public static member functions. Also warn if there are no 3132 non-private methods, and there's at least one private member 3133 function that isn't a constructor or destructor. 3134 3135'-Wdelete-non-virtual-dtor (C++ and Objective-C++ only)' 3136 Warn when 'delete' is used to destroy an instance of a class that 3137 has virtual functions and non-virtual destructor. It is unsafe to 3138 delete an instance of a derived class through a pointer to a base 3139 class if the base class does not have a virtual destructor. This 3140 warning is enabled by '-Wall'. 3141 3142'-Wdeprecated-copy (C++ and Objective-C++ only)' 3143 Warn that the implicit declaration of a copy constructor or copy 3144 assignment operator is deprecated if the class has a user-provided 3145 copy constructor or copy assignment operator, in C++11 and up. 3146 This warning is enabled by '-Wextra'. With 3147 '-Wdeprecated-copy-dtor', also deprecate if the class has a 3148 user-provided destructor. 3149 3150'-Wno-init-list-lifetime (C++ and Objective-C++ only)' 3151 Do not warn about uses of 'std::initializer_list' that are likely 3152 to result in dangling pointers. Since the underlying array for an 3153 'initializer_list' is handled like a normal C++ temporary object, 3154 it is easy to inadvertently keep a pointer to the array past the 3155 end of the array's lifetime. For example: 3156 3157 * If a function returns a temporary 'initializer_list', or a 3158 local 'initializer_list' variable, the array's lifetime ends 3159 at the end of the return statement, so the value returned has 3160 a dangling pointer. 3161 3162 * If a new-expression creates an 'initializer_list', the array 3163 only lives until the end of the enclosing full-expression, so 3164 the 'initializer_list' in the heap has a dangling pointer. 3165 3166 * When an 'initializer_list' variable is assigned from a 3167 brace-enclosed initializer list, the temporary array created 3168 for the right side of the assignment only lives until the end 3169 of the full-expression, so at the next statement the 3170 'initializer_list' variable has a dangling pointer. 3171 3172 // li's initial underlying array lives as long as li 3173 std::initializer_list<int> li = { 1,2,3 }; 3174 // assignment changes li to point to a temporary array 3175 li = { 4, 5 }; 3176 // now the temporary is gone and li has a dangling pointer 3177 int i = li.begin()[0] // undefined behavior 3178 3179 * When a list constructor stores the 'begin' pointer from the 3180 'initializer_list' argument, this doesn't extend the lifetime 3181 of the array, so if a class variable is constructed from a 3182 temporary 'initializer_list', the pointer is left dangling by 3183 the end of the variable declaration statement. 3184 3185'-Wno-literal-suffix (C++ and Objective-C++ only)' 3186 Do not warn when a string or character literal is followed by a 3187 ud-suffix which does not begin with an underscore. As a conforming 3188 extension, GCC treats such suffixes as separate preprocessing 3189 tokens in order to maintain backwards compatibility with code that 3190 uses formatting macros from '<inttypes.h>'. For example: 3191 3192 #define __STDC_FORMAT_MACROS 3193 #include <inttypes.h> 3194 #include <stdio.h> 3195 3196 int main() { 3197 int64_t i64 = 123; 3198 printf("My int64: %" PRId64"\n", i64); 3199 } 3200 3201 In this case, 'PRId64' is treated as a separate preprocessing 3202 token. 3203 3204 This option also controls warnings when a user-defined literal 3205 operator is declared with a literal suffix identifier that doesn't 3206 begin with an underscore. Literal suffix identifiers that don't 3207 begin with an underscore are reserved for future standardization. 3208 3209 These warnings are enabled by default. 3210 3211'-Wno-narrowing (C++ and Objective-C++ only)' 3212 For C++11 and later standards, narrowing conversions are diagnosed 3213 by default, as required by the standard. A narrowing conversion 3214 from a constant produces an error, and a narrowing conversion from 3215 a non-constant produces a warning, but '-Wno-narrowing' suppresses 3216 the diagnostic. Note that this does not affect the meaning of 3217 well-formed code; narrowing conversions are still considered 3218 ill-formed in SFINAE contexts. 3219 3220 With '-Wnarrowing' in C++98, warn when a narrowing conversion 3221 prohibited by C++11 occurs within '{ }', e.g. 3222 3223 int i = { 2.2 }; // error: narrowing from double to int 3224 3225 This flag is included in '-Wall' and '-Wc++11-compat'. 3226 3227'-Wnoexcept (C++ and Objective-C++ only)' 3228 Warn when a noexcept-expression evaluates to false because of a 3229 call to a function that does not have a non-throwing exception 3230 specification (i.e. 'throw()' or 'noexcept') but is known by the 3231 compiler to never throw an exception. 3232 3233'-Wnoexcept-type (C++ and Objective-C++ only)' 3234 Warn if the C++17 feature making 'noexcept' part of a function type 3235 changes the mangled name of a symbol relative to C++14. Enabled by 3236 '-Wabi' and '-Wc++17-compat'. 3237 3238 As an example: 3239 3240 template <class T> void f(T t) { t(); }; 3241 void g() noexcept; 3242 void h() { f(g); } 3243 3244 In C++14, 'f' calls 'f<void(*)()>', but in C++17 it calls 3245 'f<void(*)()noexcept>'. 3246 3247'-Wclass-memaccess (C++ and Objective-C++ only)' 3248 Warn when the destination of a call to a raw memory function such 3249 as 'memset' or 'memcpy' is an object of class type, and when 3250 writing into such an object might bypass the class non-trivial or 3251 deleted constructor or copy assignment, violate const-correctness 3252 or encapsulation, or corrupt virtual table pointers. Modifying the 3253 representation of such objects may violate invariants maintained by 3254 member functions of the class. For example, the call to 'memset' 3255 below is undefined because it modifies a non-trivial class object 3256 and is, therefore, diagnosed. The safe way to either initialize or 3257 clear the storage of objects of such types is by using the 3258 appropriate constructor or assignment operator, if one is 3259 available. 3260 std::string str = "abc"; 3261 memset (&str, 0, sizeof str); 3262 The '-Wclass-memaccess' option is enabled by '-Wall'. Explicitly 3263 casting the pointer to the class object to 'void *' or to a type 3264 that can be safely accessed by the raw memory function suppresses 3265 the warning. 3266 3267'-Wnon-virtual-dtor (C++ and Objective-C++ only)' 3268 Warn when a class has virtual functions and an accessible 3269 non-virtual destructor itself or in an accessible polymorphic base 3270 class, in which case it is possible but unsafe to delete an 3271 instance of a derived class through a pointer to the class itself 3272 or base class. This warning is automatically enabled if '-Weffc++' 3273 is specified. 3274 3275'-Wregister (C++ and Objective-C++ only)' 3276 Warn on uses of the 'register' storage class specifier, except when 3277 it is part of the GNU *note Explicit Register Variables:: 3278 extension. The use of the 'register' keyword as storage class 3279 specifier has been deprecated in C++11 and removed in C++17. 3280 Enabled by default with '-std=c++17'. 3281 3282'-Wreorder (C++ and Objective-C++ only)' 3283 Warn when the order of member initializers given in the code does 3284 not match the order in which they must be executed. For instance: 3285 3286 struct A { 3287 int i; 3288 int j; 3289 A(): j (0), i (1) { } 3290 }; 3291 3292 The compiler rearranges the member initializers for 'i' and 'j' to 3293 match the declaration order of the members, emitting a warning to 3294 that effect. This warning is enabled by '-Wall'. 3295 3296'-Wno-pessimizing-move (C++ and Objective-C++ only)' 3297 This warning warns when a call to 'std::move' prevents copy 3298 elision. A typical scenario when copy elision can occur is when 3299 returning in a function with a class return type, when the 3300 expression being returned is the name of a non-volatile automatic 3301 object, and is not a function parameter, and has the same type as 3302 the function return type. 3303 3304 struct T { 3305 ... 3306 }; 3307 T fn() 3308 { 3309 T t; 3310 ... 3311 return std::move (t); 3312 } 3313 3314 But in this example, the 'std::move' call prevents copy elision. 3315 3316 This warning is enabled by '-Wall'. 3317 3318'-Wno-redundant-move (C++ and Objective-C++ only)' 3319 This warning warns about redundant calls to 'std::move'; that is, 3320 when a move operation would have been performed even without the 3321 'std::move' call. This happens because the compiler is forced to 3322 treat the object as if it were an rvalue in certain situations such 3323 as returning a local variable, where copy elision isn't applicable. 3324 Consider: 3325 3326 struct T { 3327 ... 3328 }; 3329 T fn(T t) 3330 { 3331 ... 3332 return std::move (t); 3333 } 3334 3335 Here, the 'std::move' call is redundant. Because G++ implements 3336 Core Issue 1579, another example is: 3337 3338 struct T { // convertible to U 3339 ... 3340 }; 3341 struct U { 3342 ... 3343 }; 3344 U fn() 3345 { 3346 T t; 3347 ... 3348 return std::move (t); 3349 } 3350 In this example, copy elision isn't applicable because the type of 3351 the expression being returned and the function return type differ, 3352 yet G++ treats the return value as if it were designated by an 3353 rvalue. 3354 3355 This warning is enabled by '-Wextra'. 3356 3357'-Wredundant-tags (C++ and Objective-C++ only)' 3358 Warn about redundant class-key and enum-key in references to class 3359 types and enumerated types in contexts where the key can be 3360 eliminated without causing an ambiguity. For example: 3361 3362 struct foo; 3363 struct foo *p; // warn that keyword struct can be eliminated 3364 3365 On the other hand, in this example there is no warning: 3366 3367 struct foo; 3368 void foo (); // "hides" struct foo 3369 void bar (struct foo&); // no warning, keyword struct is necessary 3370 3371'-Wno-subobject-linkage (C++ and Objective-C++ only)' 3372 Do not warn if a class type has a base or a field whose type uses 3373 the anonymous namespace or depends on a type with no linkage. If a 3374 type A depends on a type B with no or internal linkage, defining it 3375 in multiple translation units would be an ODR violation because the 3376 meaning of B is different in each translation unit. If A only 3377 appears in a single translation unit, the best way to silence the 3378 warning is to give it internal linkage by putting it in an 3379 anonymous namespace as well. The compiler doesn't give this 3380 warning for types defined in the main .C file, as those are 3381 unlikely to have multiple definitions. '-Wsubobject-linkage' is 3382 enabled by default. 3383 3384'-Weffc++ (C++ and Objective-C++ only)' 3385 Warn about violations of the following style guidelines from Scott 3386 Meyers' 'Effective C++' series of books: 3387 3388 * Define a copy constructor and an assignment operator for 3389 classes with dynamically-allocated memory. 3390 3391 * Prefer initialization to assignment in constructors. 3392 3393 * Have 'operator=' return a reference to '*this'. 3394 3395 * Don't try to return a reference when you must return an 3396 object. 3397 3398 * Distinguish between prefix and postfix forms of increment and 3399 decrement operators. 3400 3401 * Never overload '&&', '||', or ','. 3402 3403 This option also enables '-Wnon-virtual-dtor', which is also one of 3404 the effective C++ recommendations. However, the check is extended 3405 to warn about the lack of virtual destructor in accessible 3406 non-polymorphic bases classes too. 3407 3408 When selecting this option, be aware that the standard library 3409 headers do not obey all of these guidelines; use 'grep -v' to 3410 filter out those warnings. 3411 3412'-Wstrict-null-sentinel (C++ and Objective-C++ only)' 3413 Warn about the use of an uncasted 'NULL' as sentinel. When 3414 compiling only with GCC this is a valid sentinel, as 'NULL' is 3415 defined to '__null'. Although it is a null pointer constant rather 3416 than a null pointer, it is guaranteed to be of the same size as a 3417 pointer. But this use is not portable across different compilers. 3418 3419'-Wno-non-template-friend (C++ and Objective-C++ only)' 3420 Disable warnings when non-template friend functions are declared 3421 within a template. In very old versions of GCC that predate 3422 implementation of the ISO standard, declarations such as 'friend 3423 int foo(int)', where the name of the friend is an unqualified-id, 3424 could be interpreted as a particular specialization of a template 3425 function; the warning exists to diagnose compatibility problems, 3426 and is enabled by default. 3427 3428'-Wold-style-cast (C++ and Objective-C++ only)' 3429 Warn if an old-style (C-style) cast to a non-void type is used 3430 within a C++ program. The new-style casts ('dynamic_cast', 3431 'static_cast', 'reinterpret_cast', and 'const_cast') are less 3432 vulnerable to unintended effects and much easier to search for. 3433 3434'-Woverloaded-virtual (C++ and Objective-C++ only)' 3435 Warn when a function declaration hides virtual functions from a 3436 base class. For example, in: 3437 3438 struct A { 3439 virtual void f(); 3440 }; 3441 3442 struct B: public A { 3443 void f(int); 3444 }; 3445 3446 the 'A' class version of 'f' is hidden in 'B', and code like: 3447 3448 B* b; 3449 b->f(); 3450 3451 fails to compile. 3452 3453'-Wno-pmf-conversions (C++ and Objective-C++ only)' 3454 Disable the diagnostic for converting a bound pointer to member 3455 function to a plain pointer. 3456 3457'-Wsign-promo (C++ and Objective-C++ only)' 3458 Warn when overload resolution chooses a promotion from unsigned or 3459 enumerated type to a signed type, over a conversion to an unsigned 3460 type of the same size. Previous versions of G++ tried to preserve 3461 unsignedness, but the standard mandates the current behavior. 3462 3463'-Wtemplates (C++ and Objective-C++ only)' 3464 Warn when a primary template declaration is encountered. Some 3465 coding rules disallow templates, and this may be used to enforce 3466 that rule. The warning is inactive inside a system header file, 3467 such as the STL, so one can still use the STL. One may also 3468 instantiate or specialize templates. 3469 3470'-Wmismatched-tags (C++ and Objective-C++ only)' 3471 Warn for declarations of structs, classes, and class templates and 3472 their specializations with a class-key that does not match either 3473 the definition or the first declaration if no definition is 3474 provided. 3475 3476 For example, the declaration of 'struct Object' in the argument 3477 list of 'draw' triggers the warning. To avoid it, either remove 3478 the redundant class-key 'struct' or replace it with 'class' to 3479 match its definition. 3480 class Object { 3481 public: 3482 virtual ~Object () = 0; 3483 }; 3484 void draw (struct Object*); 3485 3486 It is not wrong to declare a class with the class-key 'struct' as 3487 the example above shows. The '-Wmismatched-tags' option is 3488 intended to help achieve a consistent style of class declarations. 3489 In code that is intended to be portable to Windows-based compilers 3490 the warning helps prevent unresolved references due to the 3491 difference in the mangling of symbols declared with different 3492 class-keys. The option can be used either on its own or in 3493 conjunction with '-Wredundant-tags'. 3494 3495'-Wmultiple-inheritance (C++ and Objective-C++ only)' 3496 Warn when a class is defined with multiple direct base classes. 3497 Some coding rules disallow multiple inheritance, and this may be 3498 used to enforce that rule. The warning is inactive inside a system 3499 header file, such as the STL, so one can still use the STL. One may 3500 also define classes that indirectly use multiple inheritance. 3501 3502'-Wvirtual-inheritance' 3503 Warn when a class is defined with a virtual direct base class. 3504 Some coding rules disallow multiple inheritance, and this may be 3505 used to enforce that rule. The warning is inactive inside a system 3506 header file, such as the STL, so one can still use the STL. One may 3507 also define classes that indirectly use virtual inheritance. 3508 3509'-Wno-virtual-move-assign' 3510 Suppress warnings about inheriting from a virtual base with a 3511 non-trivial C++11 move assignment operator. This is dangerous 3512 because if the virtual base is reachable along more than one path, 3513 it is moved multiple times, which can mean both objects end up in 3514 the moved-from state. If the move assignment operator is written 3515 to avoid moving from a moved-from object, this warning can be 3516 disabled. 3517 3518'-Wnamespaces' 3519 Warn when a namespace definition is opened. Some coding rules 3520 disallow namespaces, and this may be used to enforce that rule. 3521 The warning is inactive inside a system header file, such as the 3522 STL, so one can still use the STL. One may also use using 3523 directives and qualified names. 3524 3525'-Wno-terminate (C++ and Objective-C++ only)' 3526 Disable the warning about a throw-expression that will immediately 3527 result in a call to 'terminate'. 3528 3529'-Wno-class-conversion (C++ and Objective-C++ only)' 3530 Do not warn when a conversion function converts an object to the 3531 same type, to a base class of that type, or to void; such a 3532 conversion function will never be called. 3533 3534'-Wvolatile (C++ and Objective-C++ only)' 3535 Warn about deprecated uses of the 'volatile' qualifier. This 3536 includes postfix and prefix '++' and '--' expressions of 3537 'volatile'-qualified types, using simple assignments where the left 3538 operand is a 'volatile'-qualified non-class type for their value, 3539 compound assignments where the left operand is a 3540 'volatile'-qualified non-class type, 'volatile'-qualified function 3541 return type, 'volatile'-qualified parameter type, and structured 3542 bindings of a 'volatile'-qualified type. This usage was deprecated 3543 in C++20. 3544 3545 Enabled by default with '-std=c++2a'. 3546 3547'-Wzero-as-null-pointer-constant (C++ and Objective-C++ only)' 3548 Warn when a literal '0' is used as null pointer constant. This can 3549 be useful to facilitate the conversion to 'nullptr' in C++11. 3550 3551'-Waligned-new' 3552 Warn about a new-expression of a type that requires greater 3553 alignment than the 'alignof(std::max_align_t)' but uses an 3554 allocation function without an explicit alignment parameter. This 3555 option is enabled by '-Wall'. 3556 3557 Normally this only warns about global allocation functions, but 3558 '-Waligned-new=all' also warns about class member allocation 3559 functions. 3560 3561'-Wno-placement-new' 3562'-Wplacement-new=N' 3563 Warn about placement new expressions with undefined behavior, such 3564 as constructing an object in a buffer that is smaller than the type 3565 of the object. For example, the placement new expression below is 3566 diagnosed because it attempts to construct an array of 64 integers 3567 in a buffer only 64 bytes large. 3568 char buf [64]; 3569 new (buf) int[64]; 3570 This warning is enabled by default. 3571 3572 '-Wplacement-new=1' 3573 This is the default warning level of '-Wplacement-new'. At 3574 this level the warning is not issued for some strictly 3575 undefined constructs that GCC allows as extensions for 3576 compatibility with legacy code. For example, the following 3577 'new' expression is not diagnosed at this level even though it 3578 has undefined behavior according to the C++ standard because 3579 it writes past the end of the one-element array. 3580 struct S { int n, a[1]; }; 3581 S *s = (S *)malloc (sizeof *s + 31 * sizeof s->a[0]); 3582 new (s->a)int [32](); 3583 3584 '-Wplacement-new=2' 3585 At this level, in addition to diagnosing all the same 3586 constructs as at level 1, a diagnostic is also issued for 3587 placement new expressions that construct an object in the last 3588 member of structure whose type is an array of a single element 3589 and whose size is less than the size of the object being 3590 constructed. While the previous example would be diagnosed, 3591 the following construct makes use of the flexible member array 3592 extension to avoid the warning at level 2. 3593 struct S { int n, a[]; }; 3594 S *s = (S *)malloc (sizeof *s + 32 * sizeof s->a[0]); 3595 new (s->a)int [32](); 3596 3597'-Wcatch-value' 3598'-Wcatch-value=N (C++ and Objective-C++ only)' 3599 Warn about catch handlers that do not catch via reference. With 3600 '-Wcatch-value=1' (or '-Wcatch-value' for short) warn about 3601 polymorphic class types that are caught by value. With 3602 '-Wcatch-value=2' warn about all class types that are caught by 3603 value. With '-Wcatch-value=3' warn about all types that are not 3604 caught by reference. '-Wcatch-value' is enabled by '-Wall'. 3605 3606'-Wconditionally-supported (C++ and Objective-C++ only)' 3607 Warn for conditionally-supported (C++11 [intro.defs]) constructs. 3608 3609'-Wno-delete-incomplete (C++ and Objective-C++ only)' 3610 Do not warn when deleting a pointer to incomplete type, which may 3611 cause undefined behavior at runtime. This warning is enabled by 3612 default. 3613 3614'-Wextra-semi (C++, Objective-C++ only)' 3615 Warn about redundant semicolons after in-class function 3616 definitions. 3617 3618'-Wno-inaccessible-base (C++, Objective-C++ only)' 3619 This option controls warnings when a base class is inaccessible in 3620 a class derived from it due to ambiguity. The warning is enabled 3621 by default. Note that the warning for ambiguous virtual bases is 3622 enabled by the '-Wextra' option. 3623 struct A { int a; }; 3624 3625 struct B : A { }; 3626 3627 struct C : B, A { }; 3628 3629'-Wno-inherited-variadic-ctor' 3630 Suppress warnings about use of C++11 inheriting constructors when 3631 the base class inherited from has a C variadic constructor; the 3632 warning is on by default because the ellipsis is not inherited. 3633 3634'-Wno-invalid-offsetof (C++ and Objective-C++ only)' 3635 Suppress warnings from applying the 'offsetof' macro to a non-POD 3636 type. According to the 2014 ISO C++ standard, applying 'offsetof' 3637 to a non-standard-layout type is undefined. In existing C++ 3638 implementations, however, 'offsetof' typically gives meaningful 3639 results. This flag is for users who are aware that they are 3640 writing nonportable code and who have deliberately chosen to ignore 3641 the warning about it. 3642 3643 The restrictions on 'offsetof' may be relaxed in a future version 3644 of the C++ standard. 3645 3646'-Wsized-deallocation (C++ and Objective-C++ only)' 3647 Warn about a definition of an unsized deallocation function 3648 void operator delete (void *) noexcept; 3649 void operator delete[] (void *) noexcept; 3650 without a definition of the corresponding sized deallocation 3651 function 3652 void operator delete (void *, std::size_t) noexcept; 3653 void operator delete[] (void *, std::size_t) noexcept; 3654 or vice versa. Enabled by '-Wextra' along with 3655 '-fsized-deallocation'. 3656 3657'-Wsuggest-final-types' 3658 Warn about types with virtual methods where code quality would be 3659 improved if the type were declared with the C++11 'final' 3660 specifier, or, if possible, declared in an anonymous namespace. 3661 This allows GCC to more aggressively devirtualize the polymorphic 3662 calls. This warning is more effective with link-time optimization, 3663 where the information about the class hierarchy graph is more 3664 complete. 3665 3666'-Wsuggest-final-methods' 3667 Warn about virtual methods where code quality would be improved if 3668 the method were declared with the C++11 'final' specifier, or, if 3669 possible, its type were declared in an anonymous namespace or with 3670 the 'final' specifier. This warning is more effective with 3671 link-time optimization, where the information about the class 3672 hierarchy graph is more complete. It is recommended to first 3673 consider suggestions of '-Wsuggest-final-types' and then rebuild 3674 with new annotations. 3675 3676'-Wsuggest-override' 3677 Warn about overriding virtual functions that are not marked with 3678 the 'override' keyword. 3679 3680'-Wuseless-cast (C++ and Objective-C++ only)' 3681 Warn when an expression is casted to its own type. 3682 3683'-Wno-conversion-null (C++ and Objective-C++ only)' 3684 Do not warn for conversions between 'NULL' and non-pointer types. 3685 '-Wconversion-null' is enabled by default. 3686 3687 3688File: gcc.info, Node: Objective-C and Objective-C++ Dialect Options, Next: Diagnostic Message Formatting Options, Prev: C++ Dialect Options, Up: Invoking GCC 3689 36903.6 Options Controlling Objective-C and Objective-C++ Dialects 3691============================================================== 3692 3693(NOTE: This manual does not describe the Objective-C and Objective-C++ 3694languages themselves. *Note Language Standards Supported by GCC: 3695Standards, for references.) 3696 3697 This section describes the command-line options that are only 3698meaningful for Objective-C and Objective-C++ programs. You can also use 3699most of the language-independent GNU compiler options. For example, you 3700might compile a file 'some_class.m' like this: 3701 3702 gcc -g -fgnu-runtime -O -c some_class.m 3703 3704In this example, '-fgnu-runtime' is an option meant only for Objective-C 3705and Objective-C++ programs; you can use the other options with any 3706language supported by GCC. 3707 3708 Note that since Objective-C is an extension of the C language, 3709Objective-C compilations may also use options specific to the C 3710front-end (e.g., '-Wtraditional'). Similarly, Objective-C++ 3711compilations may use C++-specific options (e.g., '-Wabi'). 3712 3713 Here is a list of options that are _only_ for compiling Objective-C and 3714Objective-C++ programs: 3715 3716'-fconstant-string-class=CLASS-NAME' 3717 Use CLASS-NAME as the name of the class to instantiate for each 3718 literal string specified with the syntax '@"..."'. The default 3719 class name is 'NXConstantString' if the GNU runtime is being used, 3720 and 'NSConstantString' if the NeXT runtime is being used (see 3721 below). The '-fconstant-cfstrings' option, if also present, 3722 overrides the '-fconstant-string-class' setting and cause '@"..."' 3723 literals to be laid out as constant CoreFoundation strings. 3724 3725'-fgnu-runtime' 3726 Generate object code compatible with the standard GNU Objective-C 3727 runtime. This is the default for most types of systems. 3728 3729'-fnext-runtime' 3730 Generate output compatible with the NeXT runtime. This is the 3731 default for NeXT-based systems, including Darwin and Mac OS X. The 3732 macro '__NEXT_RUNTIME__' is predefined if (and only if) this option 3733 is used. 3734 3735'-fno-nil-receivers' 3736 Assume that all Objective-C message dispatches ('[receiver 3737 message:arg]') in this translation unit ensure that the receiver is 3738 not 'nil'. This allows for more efficient entry points in the 3739 runtime to be used. This option is only available in conjunction 3740 with the NeXT runtime and ABI version 0 or 1. 3741 3742'-fobjc-abi-version=N' 3743 Use version N of the Objective-C ABI for the selected runtime. 3744 This option is currently supported only for the NeXT runtime. In 3745 that case, Version 0 is the traditional (32-bit) ABI without 3746 support for properties and other Objective-C 2.0 additions. 3747 Version 1 is the traditional (32-bit) ABI with support for 3748 properties and other Objective-C 2.0 additions. Version 2 is the 3749 modern (64-bit) ABI. If nothing is specified, the default is 3750 Version 0 on 32-bit target machines, and Version 2 on 64-bit target 3751 machines. 3752 3753'-fobjc-call-cxx-cdtors' 3754 For each Objective-C class, check if any of its instance variables 3755 is a C++ object with a non-trivial default constructor. If so, 3756 synthesize a special '- (id) .cxx_construct' instance method which 3757 runs non-trivial default constructors on any such instance 3758 variables, in order, and then return 'self'. Similarly, check if 3759 any instance variable is a C++ object with a non-trivial 3760 destructor, and if so, synthesize a special '- (void) 3761 .cxx_destruct' method which runs all such default destructors, in 3762 reverse order. 3763 3764 The '- (id) .cxx_construct' and '- (void) .cxx_destruct' methods 3765 thusly generated only operate on instance variables declared in the 3766 current Objective-C class, and not those inherited from 3767 superclasses. It is the responsibility of the Objective-C runtime 3768 to invoke all such methods in an object's inheritance hierarchy. 3769 The '- (id) .cxx_construct' methods are invoked by the runtime 3770 immediately after a new object instance is allocated; the '- (void) 3771 .cxx_destruct' methods are invoked immediately before the runtime 3772 deallocates an object instance. 3773 3774 As of this writing, only the NeXT runtime on Mac OS X 10.4 and 3775 later has support for invoking the '- (id) .cxx_construct' and '- 3776 (void) .cxx_destruct' methods. 3777 3778'-fobjc-direct-dispatch' 3779 Allow fast jumps to the message dispatcher. On Darwin this is 3780 accomplished via the comm page. 3781 3782'-fobjc-exceptions' 3783 Enable syntactic support for structured exception handling in 3784 Objective-C, similar to what is offered by C++. This option is 3785 required to use the Objective-C keywords '@try', '@throw', 3786 '@catch', '@finally' and '@synchronized'. This option is available 3787 with both the GNU runtime and the NeXT runtime (but not available 3788 in conjunction with the NeXT runtime on Mac OS X 10.2 and earlier). 3789 3790'-fobjc-gc' 3791 Enable garbage collection (GC) in Objective-C and Objective-C++ 3792 programs. This option is only available with the NeXT runtime; the 3793 GNU runtime has a different garbage collection implementation that 3794 does not require special compiler flags. 3795 3796'-fobjc-nilcheck' 3797 For the NeXT runtime with version 2 of the ABI, check for a nil 3798 receiver in method invocations before doing the actual method call. 3799 This is the default and can be disabled using '-fno-objc-nilcheck'. 3800 Class methods and super calls are never checked for nil in this way 3801 no matter what this flag is set to. Currently this flag does 3802 nothing when the GNU runtime, or an older version of the NeXT 3803 runtime ABI, is used. 3804 3805'-fobjc-std=objc1' 3806 Conform to the language syntax of Objective-C 1.0, the language 3807 recognized by GCC 4.0. This only affects the Objective-C additions 3808 to the C/C++ language; it does not affect conformance to C/C++ 3809 standards, which is controlled by the separate C/C++ dialect option 3810 flags. When this option is used with the Objective-C or 3811 Objective-C++ compiler, any Objective-C syntax that is not 3812 recognized by GCC 4.0 is rejected. This is useful if you need to 3813 make sure that your Objective-C code can be compiled with older 3814 versions of GCC. 3815 3816'-freplace-objc-classes' 3817 Emit a special marker instructing 'ld(1)' not to statically link in 3818 the resulting object file, and allow 'dyld(1)' to load it in at run 3819 time instead. This is used in conjunction with the 3820 Fix-and-Continue debugging mode, where the object file in question 3821 may be recompiled and dynamically reloaded in the course of program 3822 execution, without the need to restart the program itself. 3823 Currently, Fix-and-Continue functionality is only available in 3824 conjunction with the NeXT runtime on Mac OS X 10.3 and later. 3825 3826'-fzero-link' 3827 When compiling for the NeXT runtime, the compiler ordinarily 3828 replaces calls to 'objc_getClass("...")' (when the name of the 3829 class is known at compile time) with static class references that 3830 get initialized at load time, which improves run-time performance. 3831 Specifying the '-fzero-link' flag suppresses this behavior and 3832 causes calls to 'objc_getClass("...")' to be retained. This is 3833 useful in Zero-Link debugging mode, since it allows for individual 3834 class implementations to be modified during program execution. The 3835 GNU runtime currently always retains calls to 3836 'objc_get_class("...")' regardless of command-line options. 3837 3838'-fno-local-ivars' 3839 By default instance variables in Objective-C can be accessed as if 3840 they were local variables from within the methods of the class 3841 they're declared in. This can lead to shadowing between instance 3842 variables and other variables declared either locally inside a 3843 class method or globally with the same name. Specifying the 3844 '-fno-local-ivars' flag disables this behavior thus avoiding 3845 variable shadowing issues. 3846 3847'-fivar-visibility=[public|protected|private|package]' 3848 Set the default instance variable visibility to the specified 3849 option so that instance variables declared outside the scope of any 3850 access modifier directives default to the specified visibility. 3851 3852'-gen-decls' 3853 Dump interface declarations for all classes seen in the source file 3854 to a file named 'SOURCENAME.decl'. 3855 3856'-Wassign-intercept (Objective-C and Objective-C++ only)' 3857 Warn whenever an Objective-C assignment is being intercepted by the 3858 garbage collector. 3859 3860'-Wno-property-assign-default (Objective-C and Objective-C++ only)' 3861 Do not warn if a property for an Objective-C object has no assign 3862 semantics specified. 3863 3864'-Wno-protocol (Objective-C and Objective-C++ only)' 3865 If a class is declared to implement a protocol, a warning is issued 3866 for every method in the protocol that is not implemented by the 3867 class. The default behavior is to issue a warning for every method 3868 not explicitly implemented in the class, even if a method 3869 implementation is inherited from the superclass. If you use the 3870 '-Wno-protocol' option, then methods inherited from the superclass 3871 are considered to be implemented, and no warning is issued for 3872 them. 3873 3874'-Wselector (Objective-C and Objective-C++ only)' 3875 Warn if multiple methods of different types for the same selector 3876 are found during compilation. The check is performed on the list 3877 of methods in the final stage of compilation. Additionally, a 3878 check is performed for each selector appearing in a 3879 '@selector(...)' expression, and a corresponding method for that 3880 selector has been found during compilation. Because these checks 3881 scan the method table only at the end of compilation, these 3882 warnings are not produced if the final stage of compilation is not 3883 reached, for example because an error is found during compilation, 3884 or because the '-fsyntax-only' option is being used. 3885 3886'-Wstrict-selector-match (Objective-C and Objective-C++ only)' 3887 Warn if multiple methods with differing argument and/or return 3888 types are found for a given selector when attempting to send a 3889 message using this selector to a receiver of type 'id' or 'Class'. 3890 When this flag is off (which is the default behavior), the compiler 3891 omits such warnings if any differences found are confined to types 3892 that share the same size and alignment. 3893 3894'-Wundeclared-selector (Objective-C and Objective-C++ only)' 3895 Warn if a '@selector(...)' expression referring to an undeclared 3896 selector is found. A selector is considered undeclared if no 3897 method with that name has been declared before the '@selector(...)' 3898 expression, either explicitly in an '@interface' or '@protocol' 3899 declaration, or implicitly in an '@implementation' section. This 3900 option always performs its checks as soon as a '@selector(...)' 3901 expression is found, while '-Wselector' only performs its checks in 3902 the final stage of compilation. This also enforces the coding 3903 style convention that methods and selectors must be declared before 3904 being used. 3905 3906'-print-objc-runtime-info' 3907 Generate C header describing the largest structure that is passed 3908 by value, if any. 3909 3910 3911File: gcc.info, Node: Diagnostic Message Formatting Options, Next: Warning Options, Prev: Objective-C and Objective-C++ Dialect Options, Up: Invoking GCC 3912 39133.7 Options to Control Diagnostic Messages Formatting 3914===================================================== 3915 3916Traditionally, diagnostic messages have been formatted irrespective of 3917the output device's aspect (e.g. its width, ...). You can use the 3918options described below to control the formatting algorithm for 3919diagnostic messages, e.g. how many characters per line, how often source 3920location information should be reported. Note that some language front 3921ends may not honor these options. 3922 3923'-fmessage-length=N' 3924 Try to format error messages so that they fit on lines of about N 3925 characters. If N is zero, then no line-wrapping is done; each 3926 error message appears on a single line. This is the default for 3927 all front ends. 3928 3929 Note - this option also affects the display of the '#error' and 3930 '#warning' pre-processor directives, and the 'deprecated' 3931 function/type/variable attribute. It does not however affect the 3932 'pragma GCC warning' and 'pragma GCC error' pragmas. 3933 3934'-fdiagnostics-show-location=once' 3935 Only meaningful in line-wrapping mode. Instructs the diagnostic 3936 messages reporter to emit source location information _once_; that 3937 is, in case the message is too long to fit on a single physical 3938 line and has to be wrapped, the source location won't be emitted 3939 (as prefix) again, over and over, in subsequent continuation lines. 3940 This is the default behavior. 3941 3942'-fdiagnostics-show-location=every-line' 3943 Only meaningful in line-wrapping mode. Instructs the diagnostic 3944 messages reporter to emit the same source location information (as 3945 prefix) for physical lines that result from the process of breaking 3946 a message which is too long to fit on a single line. 3947 3948'-fdiagnostics-color[=WHEN]' 3949'-fno-diagnostics-color' 3950 Use color in diagnostics. WHEN is 'never', 'always', or 'auto'. 3951 The default depends on how the compiler has been configured, it can 3952 be any of the above WHEN options or also 'never' if 'GCC_COLORS' 3953 environment variable isn't present in the environment, and 'auto' 3954 otherwise. 'auto' makes GCC use color only when the standard error 3955 is a terminal, and when not executing in an emacs shell. The forms 3956 '-fdiagnostics-color' and '-fno-diagnostics-color' are aliases for 3957 '-fdiagnostics-color=always' and '-fdiagnostics-color=never', 3958 respectively. 3959 3960 The colors are defined by the environment variable 'GCC_COLORS'. 3961 Its value is a colon-separated list of capabilities and Select 3962 Graphic Rendition (SGR) substrings. SGR commands are interpreted 3963 by the terminal or terminal emulator. (See the section in the 3964 documentation of your text terminal for permitted values and their 3965 meanings as character attributes.) These substring values are 3966 integers in decimal representation and can be concatenated with 3967 semicolons. Common values to concatenate include '1' for bold, '4' 3968 for underline, '5' for blink, '7' for inverse, '39' for default 3969 foreground color, '30' to '37' for foreground colors, '90' to '97' 3970 for 16-color mode foreground colors, '38;5;0' to '38;5;255' for 3971 88-color and 256-color modes foreground colors, '49' for default 3972 background color, '40' to '47' for background colors, '100' to 3973 '107' for 16-color mode background colors, and '48;5;0' to 3974 '48;5;255' for 88-color and 256-color modes background colors. 3975 3976 The default 'GCC_COLORS' is 3977 error=01;31:warning=01;35:note=01;36:range1=32:range2=34:locus=01:\ 3978 quote=01:path=01;36:fixit-insert=32:fixit-delete=31:\ 3979 diff-filename=01:diff-hunk=32:diff-delete=31:diff-insert=32:\ 3980 type-diff=01;32 3981 where '01;31' is bold red, '01;35' is bold magenta, '01;36' is bold 3982 cyan, '32' is green, '34' is blue, '01' is bold, and '31' is red. 3983 Setting 'GCC_COLORS' to the empty string disables colors. 3984 Supported capabilities are as follows. 3985 3986 'error=' 3987 SGR substring for error: markers. 3988 3989 'warning=' 3990 SGR substring for warning: markers. 3991 3992 'note=' 3993 SGR substring for note: markers. 3994 3995 'path=' 3996 SGR substring for colorizing paths of control-flow events as 3997 printed via '-fdiagnostics-path-format=', such as the 3998 identifiers of individual events and lines indicating 3999 interprocedural calls and returns. 4000 4001 'range1=' 4002 SGR substring for first additional range. 4003 4004 'range2=' 4005 SGR substring for second additional range. 4006 4007 'locus=' 4008 SGR substring for location information, 'file:line' or 4009 'file:line:column' etc. 4010 4011 'quote=' 4012 SGR substring for information printed within quotes. 4013 4014 'fixit-insert=' 4015 SGR substring for fix-it hints suggesting text to be inserted 4016 or replaced. 4017 4018 'fixit-delete=' 4019 SGR substring for fix-it hints suggesting text to be deleted. 4020 4021 'diff-filename=' 4022 SGR substring for filename headers within generated patches. 4023 4024 'diff-hunk=' 4025 SGR substring for the starts of hunks within generated 4026 patches. 4027 4028 'diff-delete=' 4029 SGR substring for deleted lines within generated patches. 4030 4031 'diff-insert=' 4032 SGR substring for inserted lines within generated patches. 4033 4034 'type-diff=' 4035 SGR substring for highlighting mismatching types within 4036 template arguments in the C++ frontend. 4037 4038'-fdiagnostics-urls[=WHEN]' 4039 Use escape sequences to embed URLs in diagnostics. For example, 4040 when '-fdiagnostics-show-option' emits text showing the 4041 command-line option controlling a diagnostic, embed a URL for 4042 documentation of that option. 4043 4044 WHEN is 'never', 'always', or 'auto'. 'auto' makes GCC use URL 4045 escape sequences only when the standard error is a terminal, and 4046 when not executing in an emacs shell or any graphical terminal 4047 which is known to be incompatible with this feature, see below. 4048 4049 The default depends on how the compiler has been configured. It 4050 can be any of the above WHEN options. 4051 4052 GCC can also be configured (via the 4053 '--with-diagnostics-urls=auto-if-env' configure-time option) so 4054 that the default is affected by environment variables. Under such 4055 a configuration, GCC defaults to using 'auto' if either 'GCC_URLS' 4056 or 'TERM_URLS' environment variables are present and non-empty in 4057 the environment of the compiler, or 'never' if neither are. 4058 4059 However, even with '-fdiagnostics-urls=always' the behavior is 4060 dependent on those environment variables: If 'GCC_URLS' is set to 4061 empty or 'no', do not embed URLs in diagnostics. If set to 'st', 4062 URLs use ST escape sequences. If set to 'bel', the default, URLs 4063 use BEL escape sequences. Any other non-empty value enables the 4064 feature. If 'GCC_URLS' is not set, use 'TERM_URLS' as a fallback. 4065 Note: ST is an ANSI escape sequence, string terminator 'ESC \', BEL 4066 is an ASCII character, CTRL-G that usually sounds like a beep. 4067 4068 At this time GCC tries to detect also a few terminals that are 4069 known to not implement the URL feature, and have bugs or at least 4070 had bugs in some versions that are still in use, where the URL 4071 escapes are likely to misbehave, i.e. print garbage on the screen. 4072 That list is currently xfce4-terminal, certain known to be buggy 4073 gnome-terminal versions, the linux console, and mingw. This check 4074 can be skipped with the '-fdiagnostics-urls=always'. 4075 4076'-fno-diagnostics-show-option' 4077 By default, each diagnostic emitted includes text indicating the 4078 command-line option that directly controls the diagnostic (if such 4079 an option is known to the diagnostic machinery). Specifying the 4080 '-fno-diagnostics-show-option' flag suppresses that behavior. 4081 4082'-fno-diagnostics-show-caret' 4083 By default, each diagnostic emitted includes the original source 4084 line and a caret '^' indicating the column. This option suppresses 4085 this information. The source line is truncated to N characters, if 4086 the '-fmessage-length=n' option is given. When the output is done 4087 to the terminal, the width is limited to the width given by the 4088 'COLUMNS' environment variable or, if not set, to the terminal 4089 width. 4090 4091'-fno-diagnostics-show-labels' 4092 By default, when printing source code (via 4093 '-fdiagnostics-show-caret'), diagnostics can label ranges of source 4094 code with pertinent information, such as the types of expressions: 4095 4096 printf ("foo %s bar", long_i + long_j); 4097 ~^ ~~~~~~~~~~~~~~~ 4098 | | 4099 char * long int 4100 4101 This option suppresses the printing of these labels (in the example 4102 above, the vertical bars and the "char *" and "long int" text). 4103 4104'-fno-diagnostics-show-cwe' 4105 Diagnostic messages can optionally have an associated CWE 4106 (https://cwe.mitre.org/index.html) identifier. GCC itself only 4107 provides such metadata for some of the '-fanalyzer' diagnostics. 4108 GCC plugins may also provide diagnostics with such metadata. By 4109 default, if this information is present, it will be printed with 4110 the diagnostic. This option suppresses the printing of this 4111 metadata. 4112 4113'-fno-diagnostics-show-line-numbers' 4114 By default, when printing source code (via 4115 '-fdiagnostics-show-caret'), a left margin is printed, showing line 4116 numbers. This option suppresses this left margin. 4117 4118'-fdiagnostics-minimum-margin-width=WIDTH' 4119 This option controls the minimum width of the left margin printed 4120 by '-fdiagnostics-show-line-numbers'. It defaults to 6. 4121 4122'-fdiagnostics-parseable-fixits' 4123 Emit fix-it hints in a machine-parseable format, suitable for 4124 consumption by IDEs. For each fix-it, a line will be printed after 4125 the relevant diagnostic, starting with the string "fix-it:". For 4126 example: 4127 4128 fix-it:"test.c":{45:3-45:21}:"gtk_widget_show_all" 4129 4130 The location is expressed as a half-open range, expressed as a 4131 count of bytes, starting at byte 1 for the initial column. In the 4132 above example, bytes 3 through 20 of line 45 of "test.c" are to be 4133 replaced with the given string: 4134 4135 00000000011111111112222222222 4136 12345678901234567890123456789 4137 gtk_widget_showall (dlg); 4138 ^^^^^^^^^^^^^^^^^^ 4139 gtk_widget_show_all 4140 4141 The filename and replacement string escape backslash as "\\", tab 4142 as "\t", newline as "\n", double quotes as "\"", non-printable 4143 characters as octal (e.g. vertical tab as "\013"). 4144 4145 An empty replacement string indicates that the given range is to be 4146 removed. An empty range (e.g. "45:3-45:3") indicates that the 4147 string is to be inserted at the given position. 4148 4149'-fdiagnostics-generate-patch' 4150 Print fix-it hints to stderr in unified diff format, after any 4151 diagnostics are printed. For example: 4152 4153 --- test.c 4154 +++ test.c 4155 @ -42,5 +42,5 @ 4156 4157 void show_cb(GtkDialog *dlg) 4158 { 4159 - gtk_widget_showall(dlg); 4160 + gtk_widget_show_all(dlg); 4161 } 4162 4163 4164 The diff may or may not be colorized, following the same rules as 4165 for diagnostics (see '-fdiagnostics-color'). 4166 4167'-fdiagnostics-show-template-tree' 4168 4169 In the C++ frontend, when printing diagnostics showing mismatching 4170 template types, such as: 4171 4172 could not convert 'std::map<int, std::vector<double> >()' 4173 from 'map<[...],vector<double>>' to 'map<[...],vector<float>> 4174 4175 the '-fdiagnostics-show-template-tree' flag enables printing a 4176 tree-like structure showing the common and differing parts of the 4177 types, such as: 4178 4179 map< 4180 [...], 4181 vector< 4182 [double != float]>> 4183 4184 The parts that differ are highlighted with color ("double" and 4185 "float" in this case). 4186 4187'-fno-elide-type' 4188 By default when the C++ frontend prints diagnostics showing 4189 mismatching template types, common parts of the types are printed 4190 as "[...]" to simplify the error message. For example: 4191 4192 could not convert 'std::map<int, std::vector<double> >()' 4193 from 'map<[...],vector<double>>' to 'map<[...],vector<float>> 4194 4195 Specifying the '-fno-elide-type' flag suppresses that behavior. 4196 This flag also affects the output of the 4197 '-fdiagnostics-show-template-tree' flag. 4198 4199'-fdiagnostics-path-format=KIND' 4200 Specify how to print paths of control-flow events for diagnostics 4201 that have such a path associated with them. 4202 4203 KIND is 'none', 'separate-events', or 'inline-events', the default. 4204 4205 'none' means to not print diagnostic paths. 4206 4207 'separate-events' means to print a separate "note" diagnostic for 4208 each event within the diagnostic. For example: 4209 4210 test.c:29:5: error: passing NULL as argument 1 to 'PyList_Append' which requires a non-NULL parameter 4211 test.c:25:10: note: (1) when 'PyList_New' fails, returning NULL 4212 test.c:27:3: note: (2) when 'i < count' 4213 test.c:29:5: note: (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 4214 4215 'inline-events' means to print the events "inline" within the 4216 source code. This view attempts to consolidate the events into 4217 runs of sufficiently-close events, printing them as labelled ranges 4218 within the source. 4219 4220 For example, the same events as above might be printed as: 4221 4222 'test': events 1-3 4223 | 4224 | 25 | list = PyList_New(0); 4225 | | ^~~~~~~~~~~~~ 4226 | | | 4227 | | (1) when 'PyList_New' fails, returning NULL 4228 | 26 | 4229 | 27 | for (i = 0; i < count; i++) { 4230 | | ~~~ 4231 | | | 4232 | | (2) when 'i < count' 4233 | 28 | item = PyLong_FromLong(random()); 4234 | 29 | PyList_Append(list, item); 4235 | | ~~~~~~~~~~~~~~~~~~~~~~~~~ 4236 | | | 4237 | | (3) when calling 'PyList_Append', passing NULL from (1) as argument 1 4238 | 4239 4240 Interprocedural control flow is shown by grouping the events by 4241 stack frame, and using indentation to show how stack frames are 4242 nested, pushed, and popped. 4243 4244 For example: 4245 4246 'test': events 1-2 4247 | 4248 | 133 | { 4249 | | ^ 4250 | | | 4251 | | (1) entering 'test' 4252 | 134 | boxed_int *obj = make_boxed_int (i); 4253 | | ~~~~~~~~~~~~~~~~~~ 4254 | | | 4255 | | (2) calling 'make_boxed_int' 4256 | 4257 +--> 'make_boxed_int': events 3-4 4258 | 4259 | 120 | { 4260 | | ^ 4261 | | | 4262 | | (3) entering 'make_boxed_int' 4263 | 121 | boxed_int *result = (boxed_int *)wrapped_malloc (sizeof (boxed_int)); 4264 | | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 4265 | | | 4266 | | (4) calling 'wrapped_malloc' 4267 | 4268 +--> 'wrapped_malloc': events 5-6 4269 | 4270 | 7 | { 4271 | | ^ 4272 | | | 4273 | | (5) entering 'wrapped_malloc' 4274 | 8 | return malloc (size); 4275 | | ~~~~~~~~~~~~~ 4276 | | | 4277 | | (6) calling 'malloc' 4278 | 4279 <-------------+ 4280 | 4281 'test': event 7 4282 | 4283 | 138 | free_boxed_int (obj); 4284 | | ^~~~~~~~~~~~~~~~~~~~ 4285 | | | 4286 | | (7) calling 'free_boxed_int' 4287 | 4288 (etc) 4289 4290'-fdiagnostics-show-path-depths' 4291 This option provides additional information when printing 4292 control-flow paths associated with a diagnostic. 4293 4294 If this is option is provided then the stack depth will be printed 4295 for each run of events within 4296 '-fdiagnostics-path-format=separate-events'. 4297 4298 This is intended for use by GCC developers and plugin developers 4299 when debugging diagnostics that report interprocedural control 4300 flow. 4301 4302'-fno-show-column' 4303 Do not print column numbers in diagnostics. This may be necessary 4304 if diagnostics are being scanned by a program that does not 4305 understand the column numbers, such as 'dejagnu'. 4306 4307'-fdiagnostics-format=FORMAT' 4308 Select a different format for printing diagnostics. FORMAT is 4309 'text' or 'json'. The default is 'text'. 4310 4311 The 'json' format consists of a top-level JSON array containing 4312 JSON objects representing the diagnostics. 4313 4314 The JSON is emitted as one line, without formatting; the examples 4315 below have been formatted for clarity. 4316 4317 Diagnostics can have child diagnostics. For example, this error 4318 and note: 4319 4320 misleading-indentation.c:15:3: warning: this 'if' clause does not 4321 guard... [-Wmisleading-indentation] 4322 15 | if (flag) 4323 | ^~ 4324 misleading-indentation.c:17:5: note: ...this statement, but the latter 4325 is misleadingly indented as if it were guarded by the 'if' 4326 17 | y = 2; 4327 | ^ 4328 4329 might be printed in JSON form (after formatting) like this: 4330 4331 [ 4332 { 4333 "kind": "warning", 4334 "locations": [ 4335 { 4336 "caret": { 4337 "column": 3, 4338 "file": "misleading-indentation.c", 4339 "line": 15 4340 }, 4341 "finish": { 4342 "column": 4, 4343 "file": "misleading-indentation.c", 4344 "line": 15 4345 } 4346 } 4347 ], 4348 "message": "this \u2018if\u2019 clause does not guard...", 4349 "option": "-Wmisleading-indentation", 4350 "option_url": "https://gcc.gnu.org/onlinedocs/gcc/Warning-Options.html#index-Wmisleading-indentation", 4351 "children": [ 4352 { 4353 "kind": "note", 4354 "locations": [ 4355 { 4356 "caret": { 4357 "column": 5, 4358 "file": "misleading-indentation.c", 4359 "line": 17 4360 } 4361 } 4362 ], 4363 "message": "...this statement, but the latter is ..." 4364 } 4365 ] 4366 }, 4367 ... 4368 ] 4369 4370 where the 'note' is a child of the 'warning'. 4371 4372 A diagnostic has a 'kind'. If this is 'warning', then there is an 4373 'option' key describing the command-line option controlling the 4374 warning. 4375 4376 A diagnostic can contain zero or more locations. Each location has 4377 up to three positions within it: a 'caret' position and optional 4378 'start' and 'finish' positions. A location can also have an 4379 optional 'label' string. For example, this error: 4380 4381 bad-binary-ops.c:64:23: error: invalid operands to binary + (have 'S' {aka 4382 'struct s'} and 'T' {aka 'struct t'}) 4383 64 | return callee_4a () + callee_4b (); 4384 | ~~~~~~~~~~~~ ^ ~~~~~~~~~~~~ 4385 | | | 4386 | | T {aka struct t} 4387 | S {aka struct s} 4388 4389 has three locations. Its primary location is at the "+" token at 4390 column 23. It has two secondary locations, describing the left and 4391 right-hand sides of the expression, which have labels. It might be 4392 printed in JSON form as: 4393 4394 { 4395 "children": [], 4396 "kind": "error", 4397 "locations": [ 4398 { 4399 "caret": { 4400 "column": 23, "file": "bad-binary-ops.c", "line": 64 4401 } 4402 }, 4403 { 4404 "caret": { 4405 "column": 10, "file": "bad-binary-ops.c", "line": 64 4406 }, 4407 "finish": { 4408 "column": 21, "file": "bad-binary-ops.c", "line": 64 4409 }, 4410 "label": "S {aka struct s}" 4411 }, 4412 { 4413 "caret": { 4414 "column": 25, "file": "bad-binary-ops.c", "line": 64 4415 }, 4416 "finish": { 4417 "column": 36, "file": "bad-binary-ops.c", "line": 64 4418 }, 4419 "label": "T {aka struct t}" 4420 } 4421 ], 4422 "message": "invalid operands to binary + ..." 4423 } 4424 4425 If a diagnostic contains fix-it hints, it has a 'fixits' array, 4426 consisting of half-open intervals, similar to the output of 4427 '-fdiagnostics-parseable-fixits'. For example, this diagnostic 4428 with a replacement fix-it hint: 4429 4430 demo.c:8:15: error: 'struct s' has no member named 'colour'; did you 4431 mean 'color'? 4432 8 | return ptr->colour; 4433 | ^~~~~~ 4434 | color 4435 4436 might be printed in JSON form as: 4437 4438 { 4439 "children": [], 4440 "fixits": [ 4441 { 4442 "next": { 4443 "column": 21, 4444 "file": "demo.c", 4445 "line": 8 4446 }, 4447 "start": { 4448 "column": 15, 4449 "file": "demo.c", 4450 "line": 8 4451 }, 4452 "string": "color" 4453 } 4454 ], 4455 "kind": "error", 4456 "locations": [ 4457 { 4458 "caret": { 4459 "column": 15, 4460 "file": "demo.c", 4461 "line": 8 4462 }, 4463 "finish": { 4464 "column": 20, 4465 "file": "demo.c", 4466 "line": 8 4467 } 4468 } 4469 ], 4470 "message": "\u2018struct s\u2019 has no member named ..." 4471 } 4472 4473 where the fix-it hint suggests replacing the text from 'start' up 4474 to but not including 'next' with 'string''s value. Deletions are 4475 expressed via an empty value for 'string', insertions by having 4476 'start' equal 'next'. 4477 4478 If the diagnostic has a path of control-flow events associated with 4479 it, it has a 'path' array of objects representing the events. Each 4480 event object has a 'description' string, a 'location' object, along 4481 with a 'function' string and a 'depth' number for representing 4482 interprocedural paths. The 'function' represents the current 4483 function at that event, and the 'depth' represents the stack depth 4484 relative to some baseline: the higher, the more frames are within 4485 the stack. 4486 4487 For example, the intraprocedural example shown for 4488 '-fdiagnostics-path-format=' might have this JSON for its path: 4489 4490 "path": [ 4491 { 4492 "depth": 0, 4493 "description": "when 'PyList_New' fails, returning NULL", 4494 "function": "test", 4495 "location": { 4496 "column": 10, 4497 "file": "test.c", 4498 "line": 25 4499 } 4500 }, 4501 { 4502 "depth": 0, 4503 "description": "when 'i < count'", 4504 "function": "test", 4505 "location": { 4506 "column": 3, 4507 "file": "test.c", 4508 "line": 27 4509 } 4510 }, 4511 { 4512 "depth": 0, 4513 "description": "when calling 'PyList_Append', passing NULL from (1) as argument 1", 4514 "function": "test", 4515 "location": { 4516 "column": 5, 4517 "file": "test.c", 4518 "line": 29 4519 } 4520 } 4521 ] 4522 4523 4524File: gcc.info, Node: Warning Options, Next: Static Analyzer Options, Prev: Diagnostic Message Formatting Options, Up: Invoking GCC 4525 45263.8 Options to Request or Suppress Warnings 4527=========================================== 4528 4529Warnings are diagnostic messages that report constructions that are not 4530inherently erroneous but that are risky or suggest there may have been 4531an error. 4532 4533 The following language-independent options do not enable specific 4534warnings but control the kinds of diagnostics produced by GCC. 4535 4536'-fsyntax-only' 4537 Check the code for syntax errors, but don't do anything beyond 4538 that. 4539 4540'-fmax-errors=N' 4541 Limits the maximum number of error messages to N, at which point 4542 GCC bails out rather than attempting to continue processing the 4543 source code. If N is 0 (the default), there is no limit on the 4544 number of error messages produced. If '-Wfatal-errors' is also 4545 specified, then '-Wfatal-errors' takes precedence over this option. 4546 4547'-w' 4548 Inhibit all warning messages. 4549 4550'-Werror' 4551 Make all warnings into errors. 4552 4553'-Werror=' 4554 Make the specified warning into an error. The specifier for a 4555 warning is appended; for example '-Werror=switch' turns the 4556 warnings controlled by '-Wswitch' into errors. This switch takes a 4557 negative form, to be used to negate '-Werror' for specific 4558 warnings; for example '-Wno-error=switch' makes '-Wswitch' warnings 4559 not be errors, even when '-Werror' is in effect. 4560 4561 The warning message for each controllable warning includes the 4562 option that controls the warning. That option can then be used 4563 with '-Werror=' and '-Wno-error=' as described above. (Printing of 4564 the option in the warning message can be disabled using the 4565 '-fno-diagnostics-show-option' flag.) 4566 4567 Note that specifying '-Werror='FOO automatically implies '-W'FOO. 4568 However, '-Wno-error='FOO does not imply anything. 4569 4570'-Wfatal-errors' 4571 This option causes the compiler to abort compilation on the first 4572 error occurred rather than trying to keep going and printing 4573 further error messages. 4574 4575 You can request many specific warnings with options beginning with 4576'-W', for example '-Wimplicit' to request warnings on implicit 4577declarations. Each of these specific warning options also has a 4578negative form beginning '-Wno-' to turn off warnings; for example, 4579'-Wno-implicit'. This manual lists only one of the two forms, whichever 4580is not the default. For further language-specific options also refer to 4581*note C++ Dialect Options:: and *note Objective-C and Objective-C++ 4582Dialect Options::. Additional warnings can be produced by enabling the 4583static analyzer; *Note Static Analyzer Options::. 4584 4585 Some options, such as '-Wall' and '-Wextra', turn on other options, 4586such as '-Wunused', which may turn on further options, such as 4587'-Wunused-value'. The combined effect of positive and negative forms is 4588that more specific options have priority over less specific ones, 4589independently of their position in the command-line. For options of the 4590same specificity, the last one takes effect. Options enabled or 4591disabled via pragmas (*note Diagnostic Pragmas::) take effect as if they 4592appeared at the end of the command-line. 4593 4594 When an unrecognized warning option is requested (e.g., 4595'-Wunknown-warning'), GCC emits a diagnostic stating that the option is 4596not recognized. However, if the '-Wno-' form is used, the behavior is 4597slightly different: no diagnostic is produced for '-Wno-unknown-warning' 4598unless other diagnostics are being produced. This allows the use of new 4599'-Wno-' options with old compilers, but if something goes wrong, the 4600compiler warns that an unrecognized option is present. 4601 4602 The effectiveness of some warnings depends on optimizations also being 4603enabled. For example '-Wsuggest-final-types' is more effective with 4604link-time optimization and '-Wmaybe-uninitialized' does not warn at all 4605unless optimization is enabled. 4606 4607'-Wpedantic' 4608'-pedantic' 4609 Issue all the warnings demanded by strict ISO C and ISO C++; reject 4610 all programs that use forbidden extensions, and some other programs 4611 that do not follow ISO C and ISO C++. For ISO C, follows the 4612 version of the ISO C standard specified by any '-std' option used. 4613 4614 Valid ISO C and ISO C++ programs should compile properly with or 4615 without this option (though a rare few require '-ansi' or a '-std' 4616 option specifying the required version of ISO C). However, without 4617 this option, certain GNU extensions and traditional C and C++ 4618 features are supported as well. With this option, they are 4619 rejected. 4620 4621 '-Wpedantic' does not cause warning messages for use of the 4622 alternate keywords whose names begin and end with '__'. This 4623 alternate format can also be used to disable warnings for non-ISO 4624 '__intN' types, i.e. '__intN__'. Pedantic warnings are also 4625 disabled in the expression that follows '__extension__'. However, 4626 only system header files should use these escape routes; 4627 application programs should avoid them. *Note Alternate 4628 Keywords::. 4629 4630 Some users try to use '-Wpedantic' to check programs for strict ISO 4631 C conformance. They soon find that it does not do quite what they 4632 want: it finds some non-ISO practices, but not all--only those for 4633 which ISO C _requires_ a diagnostic, and some others for which 4634 diagnostics have been added. 4635 4636 A feature to report any failure to conform to ISO C might be useful 4637 in some instances, but would require considerable additional work 4638 and would be quite different from '-Wpedantic'. We don't have 4639 plans to support such a feature in the near future. 4640 4641 Where the standard specified with '-std' represents a GNU extended 4642 dialect of C, such as 'gnu90' or 'gnu99', there is a corresponding 4643 "base standard", the version of ISO C on which the GNU extended 4644 dialect is based. Warnings from '-Wpedantic' are given where they 4645 are required by the base standard. (It does not make sense for 4646 such warnings to be given only for features not in the specified 4647 GNU C dialect, since by definition the GNU dialects of C include 4648 all features the compiler supports with the given option, and there 4649 would be nothing to warn about.) 4650 4651'-pedantic-errors' 4652 Give an error whenever the "base standard" (see '-Wpedantic') 4653 requires a diagnostic, in some cases where there is undefined 4654 behavior at compile-time and in some other cases that do not 4655 prevent compilation of programs that are valid according to the 4656 standard. This is not equivalent to '-Werror=pedantic', since 4657 there are errors enabled by this option and not enabled by the 4658 latter and vice versa. 4659 4660'-Wall' 4661 This enables all the warnings about constructions that some users 4662 consider questionable, and that are easy to avoid (or modify to 4663 prevent the warning), even in conjunction with macros. This also 4664 enables some language-specific warnings described in *note C++ 4665 Dialect Options:: and *note Objective-C and Objective-C++ Dialect 4666 Options::. 4667 4668 '-Wall' turns on the following warning flags: 4669 4670 -Waddress 4671 -Warray-bounds=1 (only with -O2) 4672 -Wbool-compare 4673 -Wbool-operation 4674 -Wc++11-compat -Wc++14-compat 4675 -Wcatch-value (C++ and Objective-C++ only) 4676 -Wchar-subscripts 4677 -Wcomment 4678 -Wduplicate-decl-specifier (C and Objective-C only) 4679 -Wenum-compare (in C/ObjC; this is on by default in C++) 4680 -Wenum-conversion in C/ObjC; 4681 -Wformat 4682 -Wformat-overflow 4683 -Wformat-truncation 4684 -Wint-in-bool-context 4685 -Wimplicit (C and Objective-C only) 4686 -Wimplicit-int (C and Objective-C only) 4687 -Wimplicit-function-declaration (C and Objective-C only) 4688 -Winit-self (only for C++) 4689 -Wlogical-not-parentheses 4690 -Wmain (only for C/ObjC and unless -ffreestanding) 4691 -Wmaybe-uninitialized 4692 -Wmemset-elt-size 4693 -Wmemset-transposed-args 4694 -Wmisleading-indentation (only for C/C++) 4695 -Wmissing-attributes 4696 -Wmissing-braces (only for C/ObjC) 4697 -Wmultistatement-macros 4698 -Wnarrowing (only for C++) 4699 -Wnonnull 4700 -Wnonnull-compare 4701 -Wopenmp-simd 4702 -Wparentheses 4703 -Wpessimizing-move (only for C++) 4704 -Wpointer-sign 4705 -Wreorder 4706 -Wrestrict 4707 -Wreturn-type 4708 -Wsequence-point 4709 -Wsign-compare (only in C++) 4710 -Wsizeof-pointer-div 4711 -Wsizeof-pointer-memaccess 4712 -Wstrict-aliasing 4713 -Wstrict-overflow=1 4714 -Wswitch 4715 -Wtautological-compare 4716 -Wtrigraphs 4717 -Wuninitialized 4718 -Wunknown-pragmas 4719 -Wunused-function 4720 -Wunused-label 4721 -Wunused-value 4722 -Wunused-variable 4723 -Wvolatile-register-var 4724 -Wzero-length-bounds 4725 4726 Note that some warning flags are not implied by '-Wall'. Some of 4727 them warn about constructions that users generally do not consider 4728 questionable, but which occasionally you might wish to check for; 4729 others warn about constructions that are necessary or hard to avoid 4730 in some cases, and there is no simple way to modify the code to 4731 suppress the warning. Some of them are enabled by '-Wextra' but 4732 many of them must be enabled individually. 4733 4734'-Wextra' 4735 This enables some extra warning flags that are not enabled by 4736 '-Wall'. (This option used to be called '-W'. The older name is 4737 still supported, but the newer name is more descriptive.) 4738 4739 -Wclobbered 4740 -Wcast-function-type 4741 -Wdeprecated-copy (C++ only) 4742 -Wempty-body 4743 -Wignored-qualifiers 4744 -Wimplicit-fallthrough=3 4745 -Wmissing-field-initializers 4746 -Wmissing-parameter-type (C only) 4747 -Wold-style-declaration (C only) 4748 -Woverride-init 4749 -Wsign-compare (C only) 4750 -Wstring-compare 4751 -Wredundant-move (only for C++) 4752 -Wtype-limits 4753 -Wuninitialized 4754 -Wshift-negative-value (in C++11 to C++17 and in C99 and newer) 4755 -Wunused-parameter (only with -Wunused or -Wall) 4756 -Wunused-but-set-parameter (only with -Wunused or -Wall) 4757 4758 The option '-Wextra' also prints warning messages for the following 4759 cases: 4760 4761 * A pointer is compared against integer zero with '<', '<=', 4762 '>', or '>='. 4763 4764 * (C++ only) An enumerator and a non-enumerator both appear in a 4765 conditional expression. 4766 4767 * (C++ only) Ambiguous virtual bases. 4768 4769 * (C++ only) Subscripting an array that has been declared 4770 'register'. 4771 4772 * (C++ only) Taking the address of a variable that has been 4773 declared 'register'. 4774 4775 * (C++ only) A base class is not initialized in the copy 4776 constructor of a derived class. 4777 4778'-Wabi (C, Objective-C, C++ and Objective-C++ only)' 4779 4780 Warn about code affected by ABI changes. This includes code that 4781 may not be compatible with the vendor-neutral C++ ABI as well as 4782 the psABI for the particular target. 4783 4784 Since G++ now defaults to updating the ABI with each major release, 4785 normally '-Wabi' warns only about C++ ABI compatibility problems if 4786 there is a check added later in a release series for an ABI issue 4787 discovered since the initial release. '-Wabi' warns about more 4788 things if an older ABI version is selected (with 4789 '-fabi-version=N'). 4790 4791 '-Wabi' can also be used with an explicit version number to warn 4792 about C++ ABI compatibility with a particular '-fabi-version' 4793 level, e.g. '-Wabi=2' to warn about changes relative to 4794 '-fabi-version=2'. 4795 4796 If an explicit version number is provided and 4797 '-fabi-compat-version' is not specified, the version number from 4798 this option is used for compatibility aliases. If no explicit 4799 version number is provided with this option, but 4800 '-fabi-compat-version' is specified, that version number is used 4801 for C++ ABI warnings. 4802 4803 Although an effort has been made to warn about all such cases, 4804 there are probably some cases that are not warned about, even 4805 though G++ is generating incompatible code. There may also be 4806 cases where warnings are emitted even though the code that is 4807 generated is compatible. 4808 4809 You should rewrite your code to avoid these warnings if you are 4810 concerned about the fact that code generated by G++ may not be 4811 binary compatible with code generated by other compilers. 4812 4813 Known incompatibilities in '-fabi-version=2' (which was the default 4814 from GCC 3.4 to 4.9) include: 4815 4816 * A template with a non-type template parameter of reference 4817 type was mangled incorrectly: 4818 extern int N; 4819 template <int &> struct S {}; 4820 void n (S<N>) {2} 4821 4822 This was fixed in '-fabi-version=3'. 4823 4824 * SIMD vector types declared using '__attribute ((vector_size))' 4825 were mangled in a non-standard way that does not allow for 4826 overloading of functions taking vectors of different sizes. 4827 4828 The mangling was changed in '-fabi-version=4'. 4829 4830 * '__attribute ((const))' and 'noreturn' were mangled as type 4831 qualifiers, and 'decltype' of a plain declaration was folded 4832 away. 4833 4834 These mangling issues were fixed in '-fabi-version=5'. 4835 4836 * Scoped enumerators passed as arguments to a variadic function 4837 are promoted like unscoped enumerators, causing 'va_arg' to 4838 complain. On most targets this does not actually affect the 4839 parameter passing ABI, as there is no way to pass an argument 4840 smaller than 'int'. 4841 4842 Also, the ABI changed the mangling of template argument packs, 4843 'const_cast', 'static_cast', prefix increment/decrement, and a 4844 class scope function used as a template argument. 4845 4846 These issues were corrected in '-fabi-version=6'. 4847 4848 * Lambdas in default argument scope were mangled incorrectly, 4849 and the ABI changed the mangling of 'nullptr_t'. 4850 4851 These issues were corrected in '-fabi-version=7'. 4852 4853 * When mangling a function type with function-cv-qualifiers, the 4854 un-qualified function type was incorrectly treated as a 4855 substitution candidate. 4856 4857 This was fixed in '-fabi-version=8', the default for GCC 5.1. 4858 4859 * 'decltype(nullptr)' incorrectly had an alignment of 1, leading 4860 to unaligned accesses. Note that this did not affect the ABI 4861 of a function with a 'nullptr_t' parameter, as parameters have 4862 a minimum alignment. 4863 4864 This was fixed in '-fabi-version=9', the default for GCC 5.2. 4865 4866 * Target-specific attributes that affect the identity of a type, 4867 such as ia32 calling conventions on a function type (stdcall, 4868 regparm, etc.), did not affect the mangled name, leading to 4869 name collisions when function pointers were used as template 4870 arguments. 4871 4872 This was fixed in '-fabi-version=10', the default for GCC 6.1. 4873 4874 This option also enables warnings about psABI-related changes. The 4875 known psABI changes at this point include: 4876 4877 * For SysV/x86-64, unions with 'long double' members are passed 4878 in memory as specified in psABI. Prior to GCC 4.4, this was 4879 not the case. For example: 4880 4881 union U { 4882 long double ld; 4883 int i; 4884 }; 4885 4886 'union U' is now always passed in memory. 4887 4888'-Wchar-subscripts' 4889 Warn if an array subscript has type 'char'. This is a common cause 4890 of error, as programmers often forget that this type is signed on 4891 some machines. This warning is enabled by '-Wall'. 4892 4893'-Wno-coverage-mismatch' 4894 Warn if feedback profiles do not match when using the 4895 '-fprofile-use' option. If a source file is changed between 4896 compiling with '-fprofile-generate' and with '-fprofile-use', the 4897 files with the profile feedback can fail to match the source file 4898 and GCC cannot use the profile feedback information. By default, 4899 this warning is enabled and is treated as an error. 4900 '-Wno-coverage-mismatch' can be used to disable the warning or 4901 '-Wno-error=coverage-mismatch' can be used to disable the error. 4902 Disabling the error for this warning can result in poorly optimized 4903 code and is useful only in the case of very minor changes such as 4904 bug fixes to an existing code-base. Completely disabling the 4905 warning is not recommended. 4906 4907'-Wno-cpp' 4908 (C, Objective-C, C++, Objective-C++ and Fortran only) Suppress 4909 warning messages emitted by '#warning' directives. 4910 4911'-Wdouble-promotion (C, C++, Objective-C and Objective-C++ only)' 4912 Give a warning when a value of type 'float' is implicitly promoted 4913 to 'double'. CPUs with a 32-bit "single-precision" floating-point 4914 unit implement 'float' in hardware, but emulate 'double' in 4915 software. On such a machine, doing computations using 'double' 4916 values is much more expensive because of the overhead required for 4917 software emulation. 4918 4919 It is easy to accidentally do computations with 'double' because 4920 floating-point literals are implicitly of type 'double'. For 4921 example, in: 4922 float area(float radius) 4923 { 4924 return 3.14159 * radius * radius; 4925 } 4926 the compiler performs the entire computation with 'double' because 4927 the floating-point literal is a 'double'. 4928 4929'-Wduplicate-decl-specifier (C and Objective-C only)' 4930 Warn if a declaration has duplicate 'const', 'volatile', 'restrict' 4931 or '_Atomic' specifier. This warning is enabled by '-Wall'. 4932 4933'-Wformat' 4934'-Wformat=N' 4935 Check calls to 'printf' and 'scanf', etc., to make sure that the 4936 arguments supplied have types appropriate to the format string 4937 specified, and that the conversions specified in the format string 4938 make sense. This includes standard functions, and others specified 4939 by format attributes (*note Function Attributes::), in the 4940 'printf', 'scanf', 'strftime' and 'strfmon' (an X/Open extension, 4941 not in the C standard) families (or other target-specific 4942 families). Which functions are checked without format attributes 4943 having been specified depends on the standard version selected, and 4944 such checks of functions without the attribute specified are 4945 disabled by '-ffreestanding' or '-fno-builtin'. 4946 4947 The formats are checked against the format features supported by 4948 GNU libc version 2.2. These include all ISO C90 and C99 features, 4949 as well as features from the Single Unix Specification and some BSD 4950 and GNU extensions. Other library implementations may not support 4951 all these features; GCC does not support warning about features 4952 that go beyond a particular library's limitations. However, if 4953 '-Wpedantic' is used with '-Wformat', warnings are given about 4954 format features not in the selected standard version (but not for 4955 'strfmon' formats, since those are not in any version of the C 4956 standard). *Note Options Controlling C Dialect: C Dialect Options. 4957 4958 '-Wformat=1' 4959 '-Wformat' 4960 Option '-Wformat' is equivalent to '-Wformat=1', and 4961 '-Wno-format' is equivalent to '-Wformat=0'. Since '-Wformat' 4962 also checks for null format arguments for several functions, 4963 '-Wformat' also implies '-Wnonnull'. Some aspects of this 4964 level of format checking can be disabled by the options: 4965 '-Wno-format-contains-nul', '-Wno-format-extra-args', and 4966 '-Wno-format-zero-length'. '-Wformat' is enabled by '-Wall'. 4967 4968 '-Wformat=2' 4969 Enable '-Wformat' plus additional format checks. Currently 4970 equivalent to '-Wformat -Wformat-nonliteral -Wformat-security 4971 -Wformat-y2k'. 4972 4973'-Wno-format-contains-nul' 4974 If '-Wformat' is specified, do not warn about format strings that 4975 contain NUL bytes. 4976 4977'-Wno-format-extra-args' 4978 If '-Wformat' is specified, do not warn about excess arguments to a 4979 'printf' or 'scanf' format function. The C standard specifies that 4980 such arguments are ignored. 4981 4982 Where the unused arguments lie between used arguments that are 4983 specified with '$' operand number specifications, normally warnings 4984 are still given, since the implementation could not know what type 4985 to pass to 'va_arg' to skip the unused arguments. However, in the 4986 case of 'scanf' formats, this option suppresses the warning if the 4987 unused arguments are all pointers, since the Single Unix 4988 Specification says that such unused arguments are allowed. 4989 4990'-Wformat-overflow' 4991'-Wformat-overflow=LEVEL' 4992 Warn about calls to formatted input/output functions such as 4993 'sprintf' and 'vsprintf' that might overflow the destination 4994 buffer. When the exact number of bytes written by a format 4995 directive cannot be determined at compile-time it is estimated 4996 based on heuristics that depend on the LEVEL argument and on 4997 optimization. While enabling optimization will in most cases 4998 improve the accuracy of the warning, it may also result in false 4999 positives. 5000 5001 '-Wformat-overflow' 5002 '-Wformat-overflow=1' 5003 Level 1 of '-Wformat-overflow' enabled by '-Wformat' employs a 5004 conservative approach that warns only about calls that most 5005 likely overflow the buffer. At this level, numeric arguments 5006 to format directives with unknown values are assumed to have 5007 the value of one, and strings of unknown length to be empty. 5008 Numeric arguments that are known to be bounded to a subrange 5009 of their type, or string arguments whose output is bounded 5010 either by their directive's precision or by a finite set of 5011 string literals, are assumed to take on the value within the 5012 range that results in the most bytes on output. For example, 5013 the call to 'sprintf' below is diagnosed because even with 5014 both A and B equal to zero, the terminating NUL character 5015 (''\0'') appended by the function to the destination buffer 5016 will be written past its end. Increasing the size of the 5017 buffer by a single byte is sufficient to avoid the warning, 5018 though it may not be sufficient to avoid the overflow. 5019 5020 void f (int a, int b) 5021 { 5022 char buf [13]; 5023 sprintf (buf, "a = %i, b = %i\n", a, b); 5024 } 5025 5026 '-Wformat-overflow=2' 5027 Level 2 warns also about calls that might overflow the 5028 destination buffer given an argument of sufficient length or 5029 magnitude. At level 2, unknown numeric arguments are assumed 5030 to have the minimum representable value for signed types with 5031 a precision greater than 1, and the maximum representable 5032 value otherwise. Unknown string arguments whose length cannot 5033 be assumed to be bounded either by the directive's precision, 5034 or by a finite set of string literals they may evaluate to, or 5035 the character array they may point to, are assumed to be 1 5036 character long. 5037 5038 At level 2, the call in the example above is again diagnosed, 5039 but this time because with A equal to a 32-bit 'INT_MIN' the 5040 first '%i' directive will write some of its digits beyond the 5041 end of the destination buffer. To make the call safe 5042 regardless of the values of the two variables, the size of the 5043 destination buffer must be increased to at least 34 bytes. 5044 GCC includes the minimum size of the buffer in an 5045 informational note following the warning. 5046 5047 An alternative to increasing the size of the destination 5048 buffer is to constrain the range of formatted values. The 5049 maximum length of string arguments can be bounded by 5050 specifying the precision in the format directive. When 5051 numeric arguments of format directives can be assumed to be 5052 bounded by less than the precision of their type, choosing an 5053 appropriate length modifier to the format specifier will 5054 reduce the required buffer size. For example, if A and B in 5055 the example above can be assumed to be within the precision of 5056 the 'short int' type then using either the '%hi' format 5057 directive or casting the argument to 'short' reduces the 5058 maximum required size of the buffer to 24 bytes. 5059 5060 void f (int a, int b) 5061 { 5062 char buf [23]; 5063 sprintf (buf, "a = %hi, b = %i\n", a, (short)b); 5064 } 5065 5066'-Wno-format-zero-length' 5067 If '-Wformat' is specified, do not warn about zero-length formats. 5068 The C standard specifies that zero-length formats are allowed. 5069 5070'-Wformat-nonliteral' 5071 If '-Wformat' is specified, also warn if the format string is not a 5072 string literal and so cannot be checked, unless the format function 5073 takes its format arguments as a 'va_list'. 5074 5075'-Wformat-security' 5076 If '-Wformat' is specified, also warn about uses of format 5077 functions that represent possible security problems. At present, 5078 this warns about calls to 'printf' and 'scanf' functions where the 5079 format string is not a string literal and there are no format 5080 arguments, as in 'printf (foo);'. This may be a security hole if 5081 the format string came from untrusted input and contains '%n'. 5082 (This is currently a subset of what '-Wformat-nonliteral' warns 5083 about, but in future warnings may be added to '-Wformat-security' 5084 that are not included in '-Wformat-nonliteral'.) 5085 5086'-Wformat-signedness' 5087 If '-Wformat' is specified, also warn if the format string requires 5088 an unsigned argument and the argument is signed and vice versa. 5089 5090'-Wformat-truncation' 5091'-Wformat-truncation=LEVEL' 5092 Warn about calls to formatted input/output functions such as 5093 'snprintf' and 'vsnprintf' that might result in output truncation. 5094 When the exact number of bytes written by a format directive cannot 5095 be determined at compile-time it is estimated based on heuristics 5096 that depend on the LEVEL argument and on optimization. While 5097 enabling optimization will in most cases improve the accuracy of 5098 the warning, it may also result in false positives. Except as 5099 noted otherwise, the option uses the same logic 5100 '-Wformat-overflow'. 5101 5102 '-Wformat-truncation' 5103 '-Wformat-truncation=1' 5104 Level 1 of '-Wformat-truncation' enabled by '-Wformat' employs 5105 a conservative approach that warns only about calls to bounded 5106 functions whose return value is unused and that will most 5107 likely result in output truncation. 5108 5109 '-Wformat-truncation=2' 5110 Level 2 warns also about calls to bounded functions whose 5111 return value is used and that might result in truncation given 5112 an argument of sufficient length or magnitude. 5113 5114'-Wformat-y2k' 5115 If '-Wformat' is specified, also warn about 'strftime' formats that 5116 may yield only a two-digit year. 5117 5118'-Wnonnull' 5119 Warn about passing a null pointer for arguments marked as requiring 5120 a non-null value by the 'nonnull' function attribute. 5121 5122 '-Wnonnull' is included in '-Wall' and '-Wformat'. It can be 5123 disabled with the '-Wno-nonnull' option. 5124 5125'-Wnonnull-compare' 5126 Warn when comparing an argument marked with the 'nonnull' function 5127 attribute against null inside the function. 5128 5129 '-Wnonnull-compare' is included in '-Wall'. It can be disabled 5130 with the '-Wno-nonnull-compare' option. 5131 5132'-Wnull-dereference' 5133 Warn if the compiler detects paths that trigger erroneous or 5134 undefined behavior due to dereferencing a null pointer. This 5135 option is only active when '-fdelete-null-pointer-checks' is 5136 active, which is enabled by optimizations in most targets. The 5137 precision of the warnings depends on the optimization options used. 5138 5139'-Winit-self (C, C++, Objective-C and Objective-C++ only)' 5140 Warn about uninitialized variables that are initialized with 5141 themselves. Note this option can only be used with the 5142 '-Wuninitialized' option. 5143 5144 For example, GCC warns about 'i' being uninitialized in the 5145 following snippet only when '-Winit-self' has been specified: 5146 int f() 5147 { 5148 int i = i; 5149 return i; 5150 } 5151 5152 This warning is enabled by '-Wall' in C++. 5153 5154'-Wno-implicit-int (C and Objective-C only)' 5155 This option controls warnings when a declaration does not specify a 5156 type. This warning is enabled by default in C99 and later dialects 5157 of C, and also by '-Wall'. 5158 5159'-Wno-implicit-function-declaration (C and Objective-C only)' 5160 This option controls warnings when a function is used before being 5161 declared. This warning is enabled by default in C99 and later 5162 dialects of C, and also by '-Wall'. The warning is made into an 5163 error by '-pedantic-errors'. 5164 5165'-Wimplicit (C and Objective-C only)' 5166 Same as '-Wimplicit-int' and '-Wimplicit-function-declaration'. 5167 This warning is enabled by '-Wall'. 5168 5169'-Wimplicit-fallthrough' 5170 '-Wimplicit-fallthrough' is the same as '-Wimplicit-fallthrough=3' 5171 and '-Wno-implicit-fallthrough' is the same as 5172 '-Wimplicit-fallthrough=0'. 5173 5174'-Wimplicit-fallthrough=N' 5175 Warn when a switch case falls through. For example: 5176 5177 switch (cond) 5178 { 5179 case 1: 5180 a = 1; 5181 break; 5182 case 2: 5183 a = 2; 5184 case 3: 5185 a = 3; 5186 break; 5187 } 5188 5189 This warning does not warn when the last statement of a case cannot 5190 fall through, e.g. when there is a return statement or a call to 5191 function declared with the noreturn attribute. 5192 '-Wimplicit-fallthrough=' also takes into account control flow 5193 statements, such as ifs, and only warns when appropriate. E.g. 5194 5195 switch (cond) 5196 { 5197 case 1: 5198 if (i > 3) { 5199 bar (5); 5200 break; 5201 } else if (i < 1) { 5202 bar (0); 5203 } else 5204 return; 5205 default: 5206 ... 5207 } 5208 5209 Since there are occasions where a switch case fall through is 5210 desirable, GCC provides an attribute, '__attribute__ 5211 ((fallthrough))', that is to be used along with a null statement to 5212 suppress this warning that would normally occur: 5213 5214 switch (cond) 5215 { 5216 case 1: 5217 bar (0); 5218 __attribute__ ((fallthrough)); 5219 default: 5220 ... 5221 } 5222 5223 C++17 provides a standard way to suppress the 5224 '-Wimplicit-fallthrough' warning using '[[fallthrough]];' instead 5225 of the GNU attribute. In C++11 or C++14 users can use 5226 '[[gnu::fallthrough]];', which is a GNU extension. Instead of 5227 these attributes, it is also possible to add a fallthrough comment 5228 to silence the warning. The whole body of the C or C++ style 5229 comment should match the given regular expressions listed below. 5230 The option argument N specifies what kind of comments are accepted: 5231 5232 * '-Wimplicit-fallthrough=0' disables the warning altogether. 5233 5234 * '-Wimplicit-fallthrough=1' matches '.*' regular expression, 5235 any comment is used as fallthrough comment. 5236 5237 * '-Wimplicit-fallthrough=2' case insensitively matches 5238 '.*falls?[ \t-]*thr(ough|u).*' regular expression. 5239 5240 * '-Wimplicit-fallthrough=3' case sensitively matches one of the 5241 following regular expressions: 5242 5243 * '-fallthrough' 5244 5245 * '@fallthrough@' 5246 5247 * 'lint -fallthrough[ \t]*' 5248 5249 * '[ \t.!]*(ELSE,? |INTENTIONAL(LY)? )? 5250 FALL(S | |-)?THR(OUGH|U)[ \t.!]*(-[^\n\r]*)?' 5251 5252 * '[ \t.!]*(Else,? |Intentional(ly)? )? 5253 Fall((s | |-)[Tt]|t)hr(ough|u)[ \t.!]*(-[^\n\r]*)?' 5254 5255 * '[ \t.!]*([Ee]lse,? |[Ii]ntentional(ly)? )? 5256 fall(s | |-)?thr(ough|u)[ \t.!]*(-[^\n\r]*)?' 5257 5258 * '-Wimplicit-fallthrough=4' case sensitively matches one of the 5259 following regular expressions: 5260 5261 * '-fallthrough' 5262 5263 * '@fallthrough@' 5264 5265 * 'lint -fallthrough[ \t]*' 5266 5267 * '[ \t]*FALLTHR(OUGH|U)[ \t]*' 5268 5269 * '-Wimplicit-fallthrough=5' doesn't recognize any comments as 5270 fallthrough comments, only attributes disable the warning. 5271 5272 The comment needs to be followed after optional whitespace and 5273 other comments by 'case' or 'default' keywords or by a user label 5274 that precedes some 'case' or 'default' label. 5275 5276 switch (cond) 5277 { 5278 case 1: 5279 bar (0); 5280 /* FALLTHRU */ 5281 default: 5282 ... 5283 } 5284 5285 The '-Wimplicit-fallthrough=3' warning is enabled by '-Wextra'. 5286 5287'-Wno-if-not-aligned (C, C++, Objective-C and Objective-C++ only)' 5288 Control if warnings triggered by the 'warn_if_not_aligned' 5289 attribute should be issued. These warnings are enabled by default. 5290 5291'-Wignored-qualifiers (C and C++ only)' 5292 Warn if the return type of a function has a type qualifier such as 5293 'const'. For ISO C such a type qualifier has no effect, since the 5294 value returned by a function is not an lvalue. For C++, the 5295 warning is only emitted for scalar types or 'void'. ISO C 5296 prohibits qualified 'void' return types on function definitions, so 5297 such return types always receive a warning even without this 5298 option. 5299 5300 This warning is also enabled by '-Wextra'. 5301 5302'-Wno-ignored-attributes (C and C++ only)' 5303 This option controls warnings when an attribute is ignored. This 5304 is different from the '-Wattributes' option in that it warns 5305 whenever the compiler decides to drop an attribute, not that the 5306 attribute is either unknown, used in a wrong place, etc. This 5307 warning is enabled by default. 5308 5309'-Wmain' 5310 Warn if the type of 'main' is suspicious. 'main' should be a 5311 function with external linkage, returning int, taking either zero 5312 arguments, two, or three arguments of appropriate types. This 5313 warning is enabled by default in C++ and is enabled by either 5314 '-Wall' or '-Wpedantic'. 5315 5316'-Wmisleading-indentation (C and C++ only)' 5317 Warn when the indentation of the code does not reflect the block 5318 structure. Specifically, a warning is issued for 'if', 'else', 5319 'while', and 'for' clauses with a guarded statement that does not 5320 use braces, followed by an unguarded statement with the same 5321 indentation. 5322 5323 In the following example, the call to "bar" is misleadingly 5324 indented as if it were guarded by the "if" conditional. 5325 5326 if (some_condition ()) 5327 foo (); 5328 bar (); /* Gotcha: this is not guarded by the "if". */ 5329 5330 In the case of mixed tabs and spaces, the warning uses the 5331 '-ftabstop=' option to determine if the statements line up 5332 (defaulting to 8). 5333 5334 The warning is not issued for code involving multiline preprocessor 5335 logic such as the following example. 5336 5337 if (flagA) 5338 foo (0); 5339 #if SOME_CONDITION_THAT_DOES_NOT_HOLD 5340 if (flagB) 5341 #endif 5342 foo (1); 5343 5344 The warning is not issued after a '#line' directive, since this 5345 typically indicates autogenerated code, and no assumptions can be 5346 made about the layout of the file that the directive references. 5347 5348 This warning is enabled by '-Wall' in C and C++. 5349 5350'-Wmissing-attributes' 5351 Warn when a declaration of a function is missing one or more 5352 attributes that a related function is declared with and whose 5353 absence may adversely affect the correctness or efficiency of 5354 generated code. For example, the warning is issued for 5355 declarations of aliases that use attributes to specify less 5356 restrictive requirements than those of their targets. This 5357 typically represents a potential optimization opportunity. By 5358 contrast, the '-Wattribute-alias=2' option controls warnings issued 5359 when the alias is more restrictive than the target, which could 5360 lead to incorrect code generation. Attributes considered include 5361 'alloc_align', 'alloc_size', 'cold', 'const', 'hot', 'leaf', 5362 'malloc', 'nonnull', 'noreturn', 'nothrow', 'pure', 5363 'returns_nonnull', and 'returns_twice'. 5364 5365 In C++, the warning is issued when an explicit specialization of a 5366 primary template declared with attribute 'alloc_align', 5367 'alloc_size', 'assume_aligned', 'format', 'format_arg', 'malloc', 5368 or 'nonnull' is declared without it. Attributes 'deprecated', 5369 'error', and 'warning' suppress the warning. (*note Function 5370 Attributes::). 5371 5372 You can use the 'copy' attribute to apply the same set of 5373 attributes to a declaration as that on another declaration without 5374 explicitly enumerating the attributes. This attribute can be 5375 applied to declarations of functions (*note Common Function 5376 Attributes::), variables (*note Common Variable Attributes::), or 5377 types (*note Common Type Attributes::). 5378 5379 '-Wmissing-attributes' is enabled by '-Wall'. 5380 5381 For example, since the declaration of the primary function template 5382 below makes use of both attribute 'malloc' and 'alloc_size' the 5383 declaration of the explicit specialization of the template is 5384 diagnosed because it is missing one of the attributes. 5385 5386 template <class T> 5387 T* __attribute__ ((malloc, alloc_size (1))) 5388 allocate (size_t); 5389 5390 template <> 5391 void* __attribute__ ((malloc)) // missing alloc_size 5392 allocate<void> (size_t); 5393 5394'-Wmissing-braces' 5395 Warn if an aggregate or union initializer is not fully bracketed. 5396 In the following example, the initializer for 'a' is not fully 5397 bracketed, but that for 'b' is fully bracketed. 5398 5399 int a[2][2] = { 0, 1, 2, 3 }; 5400 int b[2][2] = { { 0, 1 }, { 2, 3 } }; 5401 5402 This warning is enabled by '-Wall'. 5403 5404'-Wmissing-include-dirs (C, C++, Objective-C and Objective-C++ only)' 5405 Warn if a user-supplied include directory does not exist. 5406 5407'-Wno-missing-profile' 5408 This option controls warnings if feedback profiles are missing when 5409 using the '-fprofile-use' option. This option diagnoses those 5410 cases where a new function or a new file is added between compiling 5411 with '-fprofile-generate' and with '-fprofile-use', without 5412 regenerating the profiles. In these cases, the profile feedback 5413 data files do not contain any profile feedback information for the 5414 newly added function or file respectively. Also, in the case when 5415 profile count data (.gcda) files are removed, GCC cannot use any 5416 profile feedback information. In all these cases, warnings are 5417 issued to inform you that a profile generation step is due. 5418 Ignoring the warning can result in poorly optimized code. 5419 '-Wno-missing-profile' can be used to disable the warning, but this 5420 is not recommended and should be done only when non-existent 5421 profile data is justified. 5422 5423'-Wmultistatement-macros' 5424 Warn about unsafe multiple statement macros that appear to be 5425 guarded by a clause such as 'if', 'else', 'for', 'switch', or 5426 'while', in which only the first statement is actually guarded 5427 after the macro is expanded. 5428 5429 For example: 5430 5431 #define DOIT x++; y++ 5432 if (c) 5433 DOIT; 5434 5435 will increment 'y' unconditionally, not just when 'c' holds. The 5436 can usually be fixed by wrapping the macro in a do-while loop: 5437 #define DOIT do { x++; y++; } while (0) 5438 if (c) 5439 DOIT; 5440 5441 This warning is enabled by '-Wall' in C and C++. 5442 5443'-Wparentheses' 5444 Warn if parentheses are omitted in certain contexts, such as when 5445 there is an assignment in a context where a truth value is 5446 expected, or when operators are nested whose precedence people 5447 often get confused about. 5448 5449 Also warn if a comparison like 'x<=y<=z' appears; this is 5450 equivalent to '(x<=y ? 1 : 0) <= z', which is a different 5451 interpretation from that of ordinary mathematical notation. 5452 5453 Also warn for dangerous uses of the GNU extension to '?:' with 5454 omitted middle operand. When the condition in the '?': operator is 5455 a boolean expression, the omitted value is always 1. Often 5456 programmers expect it to be a value computed inside the conditional 5457 expression instead. 5458 5459 For C++ this also warns for some cases of unnecessary parentheses 5460 in declarations, which can indicate an attempt at a function call 5461 instead of a declaration: 5462 { 5463 // Declares a local variable called mymutex. 5464 std::unique_lock<std::mutex> (mymutex); 5465 // User meant std::unique_lock<std::mutex> lock (mymutex); 5466 } 5467 5468 This warning is enabled by '-Wall'. 5469 5470'-Wsequence-point' 5471 Warn about code that may have undefined semantics because of 5472 violations of sequence point rules in the C and C++ standards. 5473 5474 The C and C++ standards define the order in which expressions in a 5475 C/C++ program are evaluated in terms of "sequence points", which 5476 represent a partial ordering between the execution of parts of the 5477 program: those executed before the sequence point, and those 5478 executed after it. These occur after the evaluation of a full 5479 expression (one which is not part of a larger expression), after 5480 the evaluation of the first operand of a '&&', '||', '? :' or ',' 5481 (comma) operator, before a function is called (but after the 5482 evaluation of its arguments and the expression denoting the called 5483 function), and in certain other places. Other than as expressed by 5484 the sequence point rules, the order of evaluation of subexpressions 5485 of an expression is not specified. All these rules describe only a 5486 partial order rather than a total order, since, for example, if two 5487 functions are called within one expression with no sequence point 5488 between them, the order in which the functions are called is not 5489 specified. However, the standards committee have ruled that 5490 function calls do not overlap. 5491 5492 It is not specified when between sequence points modifications to 5493 the values of objects take effect. Programs whose behavior depends 5494 on this have undefined behavior; the C and C++ standards specify 5495 that "Between the previous and next sequence point an object shall 5496 have its stored value modified at most once by the evaluation of an 5497 expression. Furthermore, the prior value shall be read only to 5498 determine the value to be stored.". If a program breaks these 5499 rules, the results on any particular implementation are entirely 5500 unpredictable. 5501 5502 Examples of code with undefined behavior are 'a = a++;', 'a[n] = 5503 b[n++]' and 'a[i++] = i;'. Some more complicated cases are not 5504 diagnosed by this option, and it may give an occasional false 5505 positive result, but in general it has been found fairly effective 5506 at detecting this sort of problem in programs. 5507 5508 The C++17 standard will define the order of evaluation of operands 5509 in more cases: in particular it requires that the right-hand side 5510 of an assignment be evaluated before the left-hand side, so the 5511 above examples are no longer undefined. But this option will still 5512 warn about them, to help people avoid writing code that is 5513 undefined in C and earlier revisions of C++. 5514 5515 The standard is worded confusingly, therefore there is some debate 5516 over the precise meaning of the sequence point rules in subtle 5517 cases. Links to discussions of the problem, including proposed 5518 formal definitions, may be found on the GCC readings page, at 5519 <http://gcc.gnu.org/readings.html>. 5520 5521 This warning is enabled by '-Wall' for C and C++. 5522 5523'-Wno-return-local-addr' 5524 Do not warn about returning a pointer (or in C++, a reference) to a 5525 variable that goes out of scope after the function returns. 5526 5527'-Wreturn-type' 5528 Warn whenever a function is defined with a return type that 5529 defaults to 'int'. Also warn about any 'return' statement with no 5530 return value in a function whose return type is not 'void' (falling 5531 off the end of the function body is considered returning without a 5532 value). 5533 5534 For C only, warn about a 'return' statement with an expression in a 5535 function whose return type is 'void', unless the expression type is 5536 also 'void'. As a GNU extension, the latter case is accepted 5537 without a warning unless '-Wpedantic' is used. Attempting to use 5538 the return value of a non-'void' function other than 'main' that 5539 flows off the end by reaching the closing curly brace that 5540 terminates the function is undefined. 5541 5542 Unlike in C, in C++, flowing off the end of a non-'void' function 5543 other than 'main' results in undefined behavior even when the value 5544 of the function is not used. 5545 5546 This warning is enabled by default in C++ and by '-Wall' otherwise. 5547 5548'-Wno-shift-count-negative' 5549 Controls warnings if a shift count is negative. This warning is 5550 enabled by default. 5551 5552'-Wno-shift-count-overflow' 5553 Controls warnings if a shift count is greater than or equal to the 5554 bit width of the type. This warning is enabled by default. 5555 5556'-Wshift-negative-value' 5557 Warn if left shifting a negative value. This warning is enabled by 5558 '-Wextra' in C99 (and newer) and C++11 to C++17 modes. 5559 5560'-Wno-shift-overflow' 5561'-Wshift-overflow=N' 5562 These options control warnings about left shift overflows. 5563 5564 '-Wshift-overflow=1' 5565 This is the warning level of '-Wshift-overflow' and is enabled 5566 by default in C99 and C++11 modes (and newer). This warning 5567 level does not warn about left-shifting 1 into the sign bit. 5568 (However, in C, such an overflow is still rejected in contexts 5569 where an integer constant expression is required.) No warning 5570 is emitted in C++2A mode (and newer), as signed left shifts 5571 always wrap. 5572 5573 '-Wshift-overflow=2' 5574 This warning level also warns about left-shifting 1 into the 5575 sign bit, unless C++14 mode (or newer) is active. 5576 5577'-Wswitch' 5578 Warn whenever a 'switch' statement has an index of enumerated type 5579 and lacks a 'case' for one or more of the named codes of that 5580 enumeration. (The presence of a 'default' label prevents this 5581 warning.) 'case' labels outside the enumeration range also provoke 5582 warnings when this option is used (even if there is a 'default' 5583 label). This warning is enabled by '-Wall'. 5584 5585'-Wswitch-default' 5586 Warn whenever a 'switch' statement does not have a 'default' case. 5587 5588'-Wswitch-enum' 5589 Warn whenever a 'switch' statement has an index of enumerated type 5590 and lacks a 'case' for one or more of the named codes of that 5591 enumeration. 'case' labels outside the enumeration range also 5592 provoke warnings when this option is used. The only difference 5593 between '-Wswitch' and this option is that this option gives a 5594 warning about an omitted enumeration code even if there is a 5595 'default' label. 5596 5597'-Wno-switch-bool' 5598 Do not warn when a 'switch' statement has an index of boolean type 5599 and the case values are outside the range of a boolean type. It is 5600 possible to suppress this warning by casting the controlling 5601 expression to a type other than 'bool'. For example: 5602 switch ((int) (a == 4)) 5603 { 5604 ... 5605 } 5606 This warning is enabled by default for C and C++ programs. 5607 5608'-Wno-switch-outside-range' 5609 This option controls warnings when a 'switch' case has a value that 5610 is outside of its respective type range. This warning is enabled 5611 by default for C and C++ programs. 5612 5613'-Wno-switch-unreachable' 5614 Do not warn when a 'switch' statement contains statements between 5615 the controlling expression and the first case label, which will 5616 never be executed. For example: 5617 switch (cond) 5618 { 5619 i = 15; 5620 ... 5621 case 5: 5622 ... 5623 } 5624 '-Wswitch-unreachable' does not warn if the statement between the 5625 controlling expression and the first case label is just a 5626 declaration: 5627 switch (cond) 5628 { 5629 int i; 5630 ... 5631 case 5: 5632 i = 5; 5633 ... 5634 } 5635 This warning is enabled by default for C and C++ programs. 5636 5637'-Wsync-nand (C and C++ only)' 5638 Warn when '__sync_fetch_and_nand' and '__sync_nand_and_fetch' 5639 built-in functions are used. These functions changed semantics in 5640 GCC 4.4. 5641 5642'-Wunused-but-set-parameter' 5643 Warn whenever a function parameter is assigned to, but otherwise 5644 unused (aside from its declaration). 5645 5646 To suppress this warning use the 'unused' attribute (*note Variable 5647 Attributes::). 5648 5649 This warning is also enabled by '-Wunused' together with '-Wextra'. 5650 5651'-Wunused-but-set-variable' 5652 Warn whenever a local variable is assigned to, but otherwise unused 5653 (aside from its declaration). This warning is enabled by '-Wall'. 5654 5655 To suppress this warning use the 'unused' attribute (*note Variable 5656 Attributes::). 5657 5658 This warning is also enabled by '-Wunused', which is enabled by 5659 '-Wall'. 5660 5661'-Wunused-function' 5662 Warn whenever a static function is declared but not defined or a 5663 non-inline static function is unused. This warning is enabled by 5664 '-Wall'. 5665 5666'-Wunused-label' 5667 Warn whenever a label is declared but not used. This warning is 5668 enabled by '-Wall'. 5669 5670 To suppress this warning use the 'unused' attribute (*note Variable 5671 Attributes::). 5672 5673'-Wunused-local-typedefs (C, Objective-C, C++ and Objective-C++ only)' 5674 Warn when a typedef locally defined in a function is not used. 5675 This warning is enabled by '-Wall'. 5676 5677'-Wunused-parameter' 5678 Warn whenever a function parameter is unused aside from its 5679 declaration. 5680 5681 To suppress this warning use the 'unused' attribute (*note Variable 5682 Attributes::). 5683 5684'-Wno-unused-result' 5685 Do not warn if a caller of a function marked with attribute 5686 'warn_unused_result' (*note Function Attributes::) does not use its 5687 return value. The default is '-Wunused-result'. 5688 5689'-Wunused-variable' 5690 Warn whenever a local or static variable is unused aside from its 5691 declaration. This option implies '-Wunused-const-variable=1' for 5692 C, but not for C++. This warning is enabled by '-Wall'. 5693 5694 To suppress this warning use the 'unused' attribute (*note Variable 5695 Attributes::). 5696 5697'-Wunused-const-variable' 5698'-Wunused-const-variable=N' 5699 Warn whenever a constant static variable is unused aside from its 5700 declaration. '-Wunused-const-variable=1' is enabled by 5701 '-Wunused-variable' for C, but not for C++. In C this declares 5702 variable storage, but in C++ this is not an error since const 5703 variables take the place of '#define's. 5704 5705 To suppress this warning use the 'unused' attribute (*note Variable 5706 Attributes::). 5707 5708 '-Wunused-const-variable=1' 5709 This is the warning level that is enabled by 5710 '-Wunused-variable' for C. It warns only about unused static 5711 const variables defined in the main compilation unit, but not 5712 about static const variables declared in any header included. 5713 5714 '-Wunused-const-variable=2' 5715 This warning level also warns for unused constant static 5716 variables in headers (excluding system headers). This is the 5717 warning level of '-Wunused-const-variable' and must be 5718 explicitly requested since in C++ this isn't an error and in C 5719 it might be harder to clean up all headers included. 5720 5721'-Wunused-value' 5722 Warn whenever a statement computes a result that is explicitly not 5723 used. To suppress this warning cast the unused expression to 5724 'void'. This includes an expression-statement or the left-hand 5725 side of a comma expression that contains no side effects. For 5726 example, an expression such as 'x[i,j]' causes a warning, while 5727 'x[(void)i,j]' does not. 5728 5729 This warning is enabled by '-Wall'. 5730 5731'-Wunused' 5732 All the above '-Wunused' options combined. 5733 5734 In order to get a warning about an unused function parameter, you 5735 must either specify '-Wextra -Wunused' (note that '-Wall' implies 5736 '-Wunused'), or separately specify '-Wunused-parameter'. 5737 5738'-Wuninitialized' 5739 Warn if an automatic variable is used without first being 5740 initialized. In C++, warn if a non-static reference or non-static 5741 'const' member appears in a class without constructors. 5742 5743 If you want to warn about code that uses the uninitialized value of 5744 the variable in its own initializer, use the '-Winit-self' option. 5745 5746 These warnings occur for individual uninitialized elements of 5747 structure, union or array variables as well as for variables that 5748 are uninitialized as a whole. They do not occur for variables or 5749 elements declared 'volatile'. Because these warnings depend on 5750 optimization, the exact variables or elements for which there are 5751 warnings depend on the precise optimization options and version of 5752 GCC used. 5753 5754 Note that there may be no warning about a variable that is used 5755 only to compute a value that itself is never used, because such 5756 computations may be deleted by data flow analysis before the 5757 warnings are printed. 5758 5759'-Wno-invalid-memory-model' 5760 This option controls warnings for invocations of *note __atomic 5761 Builtins::, *note __sync Builtins::, and the C11 atomic generic 5762 functions with a memory consistency argument that is either invalid 5763 for the operation or outside the range of values of the 5764 'memory_order' enumeration. For example, since the 5765 '__atomic_store' and '__atomic_store_n' built-ins are only defined 5766 for the relaxed, release, and sequentially consistent memory orders 5767 the following code is diagnosed: 5768 5769 void store (int *i) 5770 { 5771 __atomic_store_n (i, 0, memory_order_consume); 5772 } 5773 5774 '-Winvalid-memory-model' is enabled by default. 5775 5776'-Wmaybe-uninitialized' 5777 For an automatic (i.e. local) variable, if there exists a path from 5778 the function entry to a use of the variable that is initialized, 5779 but there exist some other paths for which the variable is not 5780 initialized, the compiler emits a warning if it cannot prove the 5781 uninitialized paths are not executed at run time. 5782 5783 These warnings are only possible in optimizing compilation, because 5784 otherwise GCC does not keep track of the state of variables. 5785 5786 These warnings are made optional because GCC may not be able to 5787 determine when the code is correct in spite of appearing to have an 5788 error. Here is one example of how this can happen: 5789 5790 { 5791 int x; 5792 switch (y) 5793 { 5794 case 1: x = 1; 5795 break; 5796 case 2: x = 4; 5797 break; 5798 case 3: x = 5; 5799 } 5800 foo (x); 5801 } 5802 5803 If the value of 'y' is always 1, 2 or 3, then 'x' is always 5804 initialized, but GCC doesn't know this. To suppress the warning, 5805 you need to provide a default case with assert(0) or similar code. 5806 5807 This option also warns when a non-volatile automatic variable might 5808 be changed by a call to 'longjmp'. The compiler sees only the 5809 calls to 'setjmp'. It cannot know where 'longjmp' will be called; 5810 in fact, a signal handler could call it at any point in the code. 5811 As a result, you may get a warning even when there is in fact no 5812 problem because 'longjmp' cannot in fact be called at the place 5813 that would cause a problem. 5814 5815 Some spurious warnings can be avoided if you declare all the 5816 functions you use that never return as 'noreturn'. *Note Function 5817 Attributes::. 5818 5819 This warning is enabled by '-Wall' or '-Wextra'. 5820 5821'-Wunknown-pragmas' 5822 Warn when a '#pragma' directive is encountered that is not 5823 understood by GCC. If this command-line option is used, warnings 5824 are even issued for unknown pragmas in system header files. This 5825 is not the case if the warnings are only enabled by the '-Wall' 5826 command-line option. 5827 5828'-Wno-pragmas' 5829 Do not warn about misuses of pragmas, such as incorrect parameters, 5830 invalid syntax, or conflicts between pragmas. See also 5831 '-Wunknown-pragmas'. 5832 5833'-Wno-prio-ctor-dtor' 5834 Do not warn if a priority from 0 to 100 is used for constructor or 5835 destructor. The use of constructor and destructor attributes allow 5836 you to assign a priority to the constructor/destructor to control 5837 its order of execution before 'main' is called or after it returns. 5838 The priority values must be greater than 100 as the compiler 5839 reserves priority values between 0-100 for the implementation. 5840 5841'-Wstrict-aliasing' 5842 This option is only active when '-fstrict-aliasing' is active. It 5843 warns about code that might break the strict aliasing rules that 5844 the compiler is using for optimization. The warning does not catch 5845 all cases, but does attempt to catch the more common pitfalls. It 5846 is included in '-Wall'. It is equivalent to '-Wstrict-aliasing=3' 5847 5848'-Wstrict-aliasing=n' 5849 This option is only active when '-fstrict-aliasing' is active. It 5850 warns about code that might break the strict aliasing rules that 5851 the compiler is using for optimization. Higher levels correspond 5852 to higher accuracy (fewer false positives). Higher levels also 5853 correspond to more effort, similar to the way '-O' works. 5854 '-Wstrict-aliasing' is equivalent to '-Wstrict-aliasing=3'. 5855 5856 Level 1: Most aggressive, quick, least accurate. Possibly useful 5857 when higher levels do not warn but '-fstrict-aliasing' still breaks 5858 the code, as it has very few false negatives. However, it has many 5859 false positives. Warns for all pointer conversions between 5860 possibly incompatible types, even if never dereferenced. Runs in 5861 the front end only. 5862 5863 Level 2: Aggressive, quick, not too precise. May still have many 5864 false positives (not as many as level 1 though), and few false 5865 negatives (but possibly more than level 1). Unlike level 1, it 5866 only warns when an address is taken. Warns about incomplete types. 5867 Runs in the front end only. 5868 5869 Level 3 (default for '-Wstrict-aliasing'): Should have very few 5870 false positives and few false negatives. Slightly slower than 5871 levels 1 or 2 when optimization is enabled. Takes care of the 5872 common pun+dereference pattern in the front end: 5873 '*(int*)&some_float'. If optimization is enabled, it also runs in 5874 the back end, where it deals with multiple statement cases using 5875 flow-sensitive points-to information. Only warns when the 5876 converted pointer is dereferenced. Does not warn about incomplete 5877 types. 5878 5879'-Wstrict-overflow' 5880'-Wstrict-overflow=N' 5881 This option is only active when signed overflow is undefined. It 5882 warns about cases where the compiler optimizes based on the 5883 assumption that signed overflow does not occur. Note that it does 5884 not warn about all cases where the code might overflow: it only 5885 warns about cases where the compiler implements some optimization. 5886 Thus this warning depends on the optimization level. 5887 5888 An optimization that assumes that signed overflow does not occur is 5889 perfectly safe if the values of the variables involved are such 5890 that overflow never does, in fact, occur. Therefore this warning 5891 can easily give a false positive: a warning about code that is not 5892 actually a problem. To help focus on important issues, several 5893 warning levels are defined. No warnings are issued for the use of 5894 undefined signed overflow when estimating how many iterations a 5895 loop requires, in particular when determining whether a loop will 5896 be executed at all. 5897 5898 '-Wstrict-overflow=1' 5899 Warn about cases that are both questionable and easy to avoid. 5900 For example the compiler simplifies 'x + 1 > x' to '1'. This 5901 level of '-Wstrict-overflow' is enabled by '-Wall'; higher 5902 levels are not, and must be explicitly requested. 5903 5904 '-Wstrict-overflow=2' 5905 Also warn about other cases where a comparison is simplified 5906 to a constant. For example: 'abs (x) >= 0'. This can only be 5907 simplified when signed integer overflow is undefined, because 5908 'abs (INT_MIN)' overflows to 'INT_MIN', which is less than 5909 zero. '-Wstrict-overflow' (with no level) is the same as 5910 '-Wstrict-overflow=2'. 5911 5912 '-Wstrict-overflow=3' 5913 Also warn about other cases where a comparison is simplified. 5914 For example: 'x + 1 > 1' is simplified to 'x > 0'. 5915 5916 '-Wstrict-overflow=4' 5917 Also warn about other simplifications not covered by the above 5918 cases. For example: '(x * 10) / 5' is simplified to 'x * 2'. 5919 5920 '-Wstrict-overflow=5' 5921 Also warn about cases where the compiler reduces the magnitude 5922 of a constant involved in a comparison. For example: 'x + 2 > 5923 y' is simplified to 'x + 1 >= y'. This is reported only at 5924 the highest warning level because this simplification applies 5925 to many comparisons, so this warning level gives a very large 5926 number of false positives. 5927 5928'-Wstring-compare' 5929 Warn for calls to 'strcmp' and 'strncmp' whose result is determined 5930 to be either zero or non-zero in tests for such equality owing to 5931 the length of one argument being greater than the size of the array 5932 the other argument is stored in (or the bound in the case of 5933 'strncmp'). Such calls could be mistakes. For example, the call 5934 to 'strcmp' below is diagnosed because its result is necessarily 5935 non-zero irrespective of the contents of the array 'a'. 5936 5937 extern char a[4]; 5938 void f (char *d) 5939 { 5940 strcpy (d, "string"); 5941 ... 5942 if (0 == strcmp (a, d)) // cannot be true 5943 puts ("a and d are the same"); 5944 } 5945 5946 '-Wstring-compare' is enabled by '-Wextra'. 5947 5948'-Wstringop-overflow' 5949'-Wstringop-overflow=TYPE' 5950 Warn for calls to string manipulation functions such as 'memcpy' 5951 and 'strcpy' that are determined to overflow the destination 5952 buffer. The optional argument is one greater than the type of 5953 Object Size Checking to perform to determine the size of the 5954 destination. *Note Object Size Checking::. The argument is 5955 meaningful only for functions that operate on character arrays but 5956 not for raw memory functions like 'memcpy' which always make use of 5957 Object Size type-0. The option also warns for calls that specify a 5958 size in excess of the largest possible object or at most 'SIZE_MAX 5959 / 2' bytes. The option produces the best results with optimization 5960 enabled but can detect a small subset of simple buffer overflows 5961 even without optimization in calls to the GCC built-in functions 5962 like '__builtin_memcpy' that correspond to the standard functions. 5963 In any case, the option warns about just a subset of buffer 5964 overflows detected by the corresponding overflow checking 5965 built-ins. For example, the option issues a warning for the 5966 'strcpy' call below because it copies at least 5 characters (the 5967 string '"blue"' including the terminating NUL) into the buffer of 5968 size 4. 5969 5970 enum Color { blue, purple, yellow }; 5971 const char* f (enum Color clr) 5972 { 5973 static char buf [4]; 5974 const char *str; 5975 switch (clr) 5976 { 5977 case blue: str = "blue"; break; 5978 case purple: str = "purple"; break; 5979 case yellow: str = "yellow"; break; 5980 } 5981 5982 return strcpy (buf, str); // warning here 5983 } 5984 5985 Option '-Wstringop-overflow=2' is enabled by default. 5986 5987 '-Wstringop-overflow' 5988 '-Wstringop-overflow=1' 5989 The '-Wstringop-overflow=1' option uses type-zero Object Size 5990 Checking to determine the sizes of destination objects. This 5991 is the default setting of the option. At this setting the 5992 option does not warn for writes past the end of subobjects of 5993 larger objects accessed by pointers unless the size of the 5994 largest surrounding object is known. When the destination may 5995 be one of several objects it is assumed to be the largest one 5996 of them. On Linux systems, when optimization is enabled at 5997 this setting the option warns for the same code as when the 5998 '_FORTIFY_SOURCE' macro is defined to a non-zero value. 5999 6000 '-Wstringop-overflow=2' 6001 The '-Wstringop-overflow=2' option uses type-one Object Size 6002 Checking to determine the sizes of destination objects. At 6003 this setting the option warna about overflows when writing to 6004 members of the largest complete objects whose exact size is 6005 known. However, it does not warn for excessive writes to the 6006 same members of unknown objects referenced by pointers since 6007 they may point to arrays containing unknown numbers of 6008 elements. 6009 6010 '-Wstringop-overflow=3' 6011 The '-Wstringop-overflow=3' option uses type-two Object Size 6012 Checking to determine the sizes of destination objects. At 6013 this setting the option warns about overflowing the smallest 6014 object or data member. This is the most restrictive setting 6015 of the option that may result in warnings for safe code. 6016 6017 '-Wstringop-overflow=4' 6018 The '-Wstringop-overflow=4' option uses type-three Object Size 6019 Checking to determine the sizes of destination objects. At 6020 this setting the option warns about overflowing any data 6021 members, and when the destination is one of several objects it 6022 uses the size of the largest of them to decide whether to 6023 issue a warning. Similarly to '-Wstringop-overflow=3' this 6024 setting of the option may result in warnings for benign code. 6025 6026'-Wno-stringop-truncation' 6027 Do not warn for calls to bounded string manipulation functions such 6028 as 'strncat', 'strncpy', and 'stpncpy' that may either truncate the 6029 copied string or leave the destination unchanged. 6030 6031 In the following example, the call to 'strncat' specifies a bound 6032 that is less than the length of the source string. As a result, 6033 the copy of the source will be truncated and so the call is 6034 diagnosed. To avoid the warning use 'bufsize - strlen (buf) - 1)' 6035 as the bound. 6036 6037 void append (char *buf, size_t bufsize) 6038 { 6039 strncat (buf, ".txt", 3); 6040 } 6041 6042 As another example, the following call to 'strncpy' results in 6043 copying to 'd' just the characters preceding the terminating NUL, 6044 without appending the NUL to the end. Assuming the result of 6045 'strncpy' is necessarily a NUL-terminated string is a common 6046 mistake, and so the call is diagnosed. To avoid the warning when 6047 the result is not expected to be NUL-terminated, call 'memcpy' 6048 instead. 6049 6050 void copy (char *d, const char *s) 6051 { 6052 strncpy (d, s, strlen (s)); 6053 } 6054 6055 In the following example, the call to 'strncpy' specifies the size 6056 of the destination buffer as the bound. If the length of the 6057 source string is equal to or greater than this size the result of 6058 the copy will not be NUL-terminated. Therefore, the call is also 6059 diagnosed. To avoid the warning, specify 'sizeof buf - 1' as the 6060 bound and set the last element of the buffer to 'NUL'. 6061 6062 void copy (const char *s) 6063 { 6064 char buf[80]; 6065 strncpy (buf, s, sizeof buf); 6066 ... 6067 } 6068 6069 In situations where a character array is intended to store a 6070 sequence of bytes with no terminating 'NUL' such an array may be 6071 annotated with attribute 'nonstring' to avoid this warning. Such 6072 arrays, however, are not suitable arguments to functions that 6073 expect 'NUL'-terminated strings. To help detect accidental misuses 6074 of such arrays GCC issues warnings unless it can prove that the use 6075 is safe. *Note Common Variable Attributes::. 6076 6077'-Wsuggest-attribute=[pure|const|noreturn|format|cold|malloc]' 6078 Warn for cases where adding an attribute may be beneficial. The 6079 attributes currently supported are listed below. 6080 6081 '-Wsuggest-attribute=pure' 6082 '-Wsuggest-attribute=const' 6083 '-Wsuggest-attribute=noreturn' 6084 '-Wmissing-noreturn' 6085 '-Wsuggest-attribute=malloc' 6086 6087 Warn about functions that might be candidates for attributes 6088 'pure', 'const' or 'noreturn' or 'malloc'. The compiler only 6089 warns for functions visible in other compilation units or (in 6090 the case of 'pure' and 'const') if it cannot prove that the 6091 function returns normally. A function returns normally if it 6092 doesn't contain an infinite loop or return abnormally by 6093 throwing, calling 'abort' or trapping. This analysis requires 6094 option '-fipa-pure-const', which is enabled by default at '-O' 6095 and higher. Higher optimization levels improve the accuracy 6096 of the analysis. 6097 6098 '-Wsuggest-attribute=format' 6099 '-Wmissing-format-attribute' 6100 6101 Warn about function pointers that might be candidates for 6102 'format' attributes. Note these are only possible candidates, 6103 not absolute ones. GCC guesses that function pointers with 6104 'format' attributes that are used in assignment, 6105 initialization, parameter passing or return statements should 6106 have a corresponding 'format' attribute in the resulting type. 6107 I.e. the left-hand side of the assignment or initialization, 6108 the type of the parameter variable, or the return type of the 6109 containing function respectively should also have a 'format' 6110 attribute to avoid the warning. 6111 6112 GCC also warns about function definitions that might be 6113 candidates for 'format' attributes. Again, these are only 6114 possible candidates. GCC guesses that 'format' attributes 6115 might be appropriate for any function that calls a function 6116 like 'vprintf' or 'vscanf', but this might not always be the 6117 case, and some functions for which 'format' attributes are 6118 appropriate may not be detected. 6119 6120 '-Wsuggest-attribute=cold' 6121 6122 Warn about functions that might be candidates for 'cold' 6123 attribute. This is based on static detection and generally 6124 only warns about functions which always leads to a call to 6125 another 'cold' function such as wrappers of C++ 'throw' or 6126 fatal error reporting functions leading to 'abort'. 6127 6128'-Walloc-zero' 6129 Warn about calls to allocation functions decorated with attribute 6130 'alloc_size' that specify zero bytes, including those to the 6131 built-in forms of the functions 'aligned_alloc', 'alloca', 6132 'calloc', 'malloc', and 'realloc'. Because the behavior of these 6133 functions when called with a zero size differs among 6134 implementations (and in the case of 'realloc' has been deprecated) 6135 relying on it may result in subtle portability bugs and should be 6136 avoided. 6137 6138'-Walloc-size-larger-than=BYTE-SIZE' 6139 Warn about calls to functions decorated with attribute 'alloc_size' 6140 that attempt to allocate objects larger than the specified number 6141 of bytes, or where the result of the size computation in an integer 6142 type with infinite precision would exceed the value of 6143 'PTRDIFF_MAX' on the target. 6144 '-Walloc-size-larger-than=''PTRDIFF_MAX' is enabled by default. 6145 Warnings controlled by the option can be disabled either by 6146 specifying BYTE-SIZE of 'SIZE_MAX' or more or by 6147 '-Wno-alloc-size-larger-than'. *Note Function Attributes::. 6148 6149'-Wno-alloc-size-larger-than' 6150 Disable '-Walloc-size-larger-than=' warnings. The option is 6151 equivalent to '-Walloc-size-larger-than=''SIZE_MAX' or larger. 6152 6153'-Walloca' 6154 This option warns on all uses of 'alloca' in the source. 6155 6156'-Walloca-larger-than=BYTE-SIZE' 6157 This option warns on calls to 'alloca' with an integer argument 6158 whose value is either zero, or that is not bounded by a controlling 6159 predicate that limits its value to at most BYTE-SIZE. It also 6160 warns for calls to 'alloca' where the bound value is unknown. 6161 Arguments of non-integer types are considered unbounded even if 6162 they appear to be constrained to the expected range. 6163 6164 For example, a bounded case of 'alloca' could be: 6165 6166 void func (size_t n) 6167 { 6168 void *p; 6169 if (n <= 1000) 6170 p = alloca (n); 6171 else 6172 p = malloc (n); 6173 f (p); 6174 } 6175 6176 In the above example, passing '-Walloca-larger-than=1000' would not 6177 issue a warning because the call to 'alloca' is known to be at most 6178 1000 bytes. However, if '-Walloca-larger-than=500' were passed, 6179 the compiler would emit a warning. 6180 6181 Unbounded uses, on the other hand, are uses of 'alloca' with no 6182 controlling predicate constraining its integer argument. For 6183 example: 6184 6185 void func () 6186 { 6187 void *p = alloca (n); 6188 f (p); 6189 } 6190 6191 If '-Walloca-larger-than=500' were passed, the above would trigger 6192 a warning, but this time because of the lack of bounds checking. 6193 6194 Note, that even seemingly correct code involving signed integers 6195 could cause a warning: 6196 6197 void func (signed int n) 6198 { 6199 if (n < 500) 6200 { 6201 p = alloca (n); 6202 f (p); 6203 } 6204 } 6205 6206 In the above example, N could be negative, causing a larger than 6207 expected argument to be implicitly cast into the 'alloca' call. 6208 6209 This option also warns when 'alloca' is used in a loop. 6210 6211 '-Walloca-larger-than=''PTRDIFF_MAX' is enabled by default but is 6212 usually only effective when '-ftree-vrp' is active (default for 6213 '-O2' and above). 6214 6215 See also '-Wvla-larger-than=''byte-size'. 6216 6217'-Wno-alloca-larger-than' 6218 Disable '-Walloca-larger-than=' warnings. The option is equivalent 6219 to '-Walloca-larger-than=''SIZE_MAX' or larger. 6220 6221'-Warith-conversion' 6222 Do warn about implicit conversions from arithmetic operations even 6223 when conversion of the operands to the same type cannot change 6224 their values. This affects warnings from '-Wconversion', 6225 '-Wfloat-conversion', and '-Wsign-conversion'. 6226 6227 void f (char c, int i) 6228 { 6229 c = c + i; // warns with -Wconversion 6230 c = c + 1; // only warns with -Warith-conversion 6231 } 6232 6233'-Warray-bounds' 6234'-Warray-bounds=N' 6235 This option is only active when '-ftree-vrp' is active (default for 6236 '-O2' and above). It warns about subscripts to arrays that are 6237 always out of bounds. This warning is enabled by '-Wall'. 6238 6239 '-Warray-bounds=1' 6240 This is the warning level of '-Warray-bounds' and is enabled 6241 by '-Wall'; higher levels are not, and must be explicitly 6242 requested. 6243 6244 '-Warray-bounds=2' 6245 This warning level also warns about out of bounds access for 6246 arrays at the end of a struct and for arrays accessed through 6247 pointers. This warning level may give a larger number of 6248 false positives and is deactivated by default. 6249 6250'-Wattribute-alias=N' 6251'-Wno-attribute-alias' 6252 Warn about declarations using the 'alias' and similar attributes 6253 whose target is incompatible with the type of the alias. *Note 6254 Declaring Attributes of Functions: Function Attributes. 6255 6256 '-Wattribute-alias=1' 6257 The default warning level of the '-Wattribute-alias' option 6258 diagnoses incompatibilities between the type of the alias 6259 declaration and that of its target. Such incompatibilities 6260 are typically indicative of bugs. 6261 6262 '-Wattribute-alias=2' 6263 6264 At this level '-Wattribute-alias' also diagnoses cases where 6265 the attributes of the alias declaration are more restrictive 6266 than the attributes applied to its target. These mismatches 6267 can potentially result in incorrect code generation. In other 6268 cases they may be benign and could be resolved simply by 6269 adding the missing attribute to the target. For comparison, 6270 see the '-Wmissing-attributes' option, which controls 6271 diagnostics when the alias declaration is less restrictive 6272 than the target, rather than more restrictive. 6273 6274 Attributes considered include 'alloc_align', 'alloc_size', 6275 'cold', 'const', 'hot', 'leaf', 'malloc', 'nonnull', 6276 'noreturn', 'nothrow', 'pure', 'returns_nonnull', and 6277 'returns_twice'. 6278 6279 '-Wattribute-alias' is equivalent to '-Wattribute-alias=1'. This 6280 is the default. You can disable these warnings with either 6281 '-Wno-attribute-alias' or '-Wattribute-alias=0'. 6282 6283'-Wbool-compare' 6284 Warn about boolean expression compared with an integer value 6285 different from 'true'/'false'. For instance, the following 6286 comparison is always false: 6287 int n = 5; 6288 ... 6289 if ((n > 1) == 2) { ... } 6290 This warning is enabled by '-Wall'. 6291 6292'-Wbool-operation' 6293 Warn about suspicious operations on expressions of a boolean type. 6294 For instance, bitwise negation of a boolean is very likely a bug in 6295 the program. For C, this warning also warns about incrementing or 6296 decrementing a boolean, which rarely makes sense. (In C++, 6297 decrementing a boolean is always invalid. Incrementing a boolean 6298 is invalid in C++17, and deprecated otherwise.) 6299 6300 This warning is enabled by '-Wall'. 6301 6302'-Wduplicated-branches' 6303 Warn when an if-else has identical branches. This warning detects 6304 cases like 6305 if (p != NULL) 6306 return 0; 6307 else 6308 return 0; 6309 It doesn't warn when both branches contain just a null statement. 6310 This warning also warn for conditional operators: 6311 int i = x ? *p : *p; 6312 6313'-Wduplicated-cond' 6314 Warn about duplicated conditions in an if-else-if chain. For 6315 instance, warn for the following code: 6316 if (p->q != NULL) { ... } 6317 else if (p->q != NULL) { ... } 6318 6319'-Wframe-address' 6320 Warn when the '__builtin_frame_address' or 6321 '__builtin_return_address' is called with an argument greater than 6322 0. Such calls may return indeterminate values or crash the 6323 program. The warning is included in '-Wall'. 6324 6325'-Wno-discarded-qualifiers (C and Objective-C only)' 6326 Do not warn if type qualifiers on pointers are being discarded. 6327 Typically, the compiler warns if a 'const char *' variable is 6328 passed to a function that takes a 'char *' parameter. This option 6329 can be used to suppress such a warning. 6330 6331'-Wno-discarded-array-qualifiers (C and Objective-C only)' 6332 Do not warn if type qualifiers on arrays which are pointer targets 6333 are being discarded. Typically, the compiler warns if a 'const int 6334 (*)[]' variable is passed to a function that takes a 'int (*)[]' 6335 parameter. This option can be used to suppress such a warning. 6336 6337'-Wno-incompatible-pointer-types (C and Objective-C only)' 6338 Do not warn when there is a conversion between pointers that have 6339 incompatible types. This warning is for cases not covered by 6340 '-Wno-pointer-sign', which warns for pointer argument passing or 6341 assignment with different signedness. 6342 6343'-Wno-int-conversion (C and Objective-C only)' 6344 Do not warn about incompatible integer to pointer and pointer to 6345 integer conversions. This warning is about implicit conversions; 6346 for explicit conversions the warnings '-Wno-int-to-pointer-cast' 6347 and '-Wno-pointer-to-int-cast' may be used. 6348 6349'-Wzero-length-bounds' 6350 Warn about accesses to elements of zero-length array members that 6351 might overlap other members of the same object. Declaring interior 6352 zero-length arrays is discouraged because accesses to them are 6353 undefined. See *Note Zero Length::. 6354 6355 For example, the first two stores in function 'bad' are diagnosed 6356 because the array elements overlap the subsequent members 'b' and 6357 'c'. The third store is diagnosed by '-Warray-bounds' because it 6358 is beyond the bounds of the enclosing object. 6359 6360 struct X { int a[0]; int b, c; }; 6361 struct X x; 6362 6363 void bad (void) 6364 { 6365 x.a[0] = 0; // -Wzero-length-bounds 6366 x.a[1] = 1; // -Wzero-length-bounds 6367 x.a[2] = 2; // -Warray-bounds 6368 } 6369 6370 Option '-Wzero-length-bounds' is enabled by '-Warray-bounds'. 6371 6372'-Wno-div-by-zero' 6373 Do not warn about compile-time integer division by zero. 6374 Floating-point division by zero is not warned about, as it can be a 6375 legitimate way of obtaining infinities and NaNs. 6376 6377'-Wsystem-headers' 6378 Print warning messages for constructs found in system header files. 6379 Warnings from system headers are normally suppressed, on the 6380 assumption that they usually do not indicate real problems and 6381 would only make the compiler output harder to read. Using this 6382 command-line option tells GCC to emit warnings from system headers 6383 as if they occurred in user code. However, note that using '-Wall' 6384 in conjunction with this option does _not_ warn about unknown 6385 pragmas in system headers--for that, '-Wunknown-pragmas' must also 6386 be used. 6387 6388'-Wtautological-compare' 6389 Warn if a self-comparison always evaluates to true or false. This 6390 warning detects various mistakes such as: 6391 int i = 1; 6392 ... 6393 if (i > i) { ... } 6394 6395 This warning also warns about bitwise comparisons that always 6396 evaluate to true or false, for instance: 6397 if ((a & 16) == 10) { ... } 6398 will always be false. 6399 6400 This warning is enabled by '-Wall'. 6401 6402'-Wtrampolines' 6403 Warn about trampolines generated for pointers to nested functions. 6404 A trampoline is a small piece of data or code that is created at 6405 run time on the stack when the address of a nested function is 6406 taken, and is used to call the nested function indirectly. For 6407 some targets, it is made up of data only and thus requires no 6408 special treatment. But, for most targets, it is made up of code 6409 and thus requires the stack to be made executable in order for the 6410 program to work properly. 6411 6412'-Wfloat-equal' 6413 Warn if floating-point values are used in equality comparisons. 6414 6415 The idea behind this is that sometimes it is convenient (for the 6416 programmer) to consider floating-point values as approximations to 6417 infinitely precise real numbers. If you are doing this, then you 6418 need to compute (by analyzing the code, or in some other way) the 6419 maximum or likely maximum error that the computation introduces, 6420 and allow for it when performing comparisons (and when producing 6421 output, but that's a different problem). In particular, instead of 6422 testing for equality, you should check to see whether the two 6423 values have ranges that overlap; and this is done with the 6424 relational operators, so equality comparisons are probably 6425 mistaken. 6426 6427'-Wtraditional (C and Objective-C only)' 6428 Warn about certain constructs that behave differently in 6429 traditional and ISO C. Also warn about ISO C constructs that have 6430 no traditional C equivalent, and/or problematic constructs that 6431 should be avoided. 6432 6433 * Macro parameters that appear within string literals in the 6434 macro body. In traditional C macro replacement takes place 6435 within string literals, but in ISO C it does not. 6436 6437 * In traditional C, some preprocessor directives did not exist. 6438 Traditional preprocessors only considered a line to be a 6439 directive if the '#' appeared in column 1 on the line. 6440 Therefore '-Wtraditional' warns about directives that 6441 traditional C understands but ignores because the '#' does not 6442 appear as the first character on the line. It also suggests 6443 you hide directives like '#pragma' not understood by 6444 traditional C by indenting them. Some traditional 6445 implementations do not recognize '#elif', so this option 6446 suggests avoiding it altogether. 6447 6448 * A function-like macro that appears without arguments. 6449 6450 * The unary plus operator. 6451 6452 * The 'U' integer constant suffix, or the 'F' or 'L' 6453 floating-point constant suffixes. (Traditional C does support 6454 the 'L' suffix on integer constants.) Note, these suffixes 6455 appear in macros defined in the system headers of most modern 6456 systems, e.g. the '_MIN'/'_MAX' macros in '<limits.h>'. Use 6457 of these macros in user code might normally lead to spurious 6458 warnings, however GCC's integrated preprocessor has enough 6459 context to avoid warning in these cases. 6460 6461 * A function declared external in one block and then used after 6462 the end of the block. 6463 6464 * A 'switch' statement has an operand of type 'long'. 6465 6466 * A non-'static' function declaration follows a 'static' one. 6467 This construct is not accepted by some traditional C 6468 compilers. 6469 6470 * The ISO type of an integer constant has a different width or 6471 signedness from its traditional type. This warning is only 6472 issued if the base of the constant is ten. I.e. hexadecimal 6473 or octal values, which typically represent bit patterns, are 6474 not warned about. 6475 6476 * Usage of ISO string concatenation is detected. 6477 6478 * Initialization of automatic aggregates. 6479 6480 * Identifier conflicts with labels. Traditional C lacks a 6481 separate namespace for labels. 6482 6483 * Initialization of unions. If the initializer is zero, the 6484 warning is omitted. This is done under the assumption that 6485 the zero initializer in user code appears conditioned on e.g. 6486 '__STDC__' to avoid missing initializer warnings and relies on 6487 default initialization to zero in the traditional C case. 6488 6489 * Conversions by prototypes between fixed/floating-point values 6490 and vice versa. The absence of these prototypes when 6491 compiling with traditional C causes serious problems. This is 6492 a subset of the possible conversion warnings; for the full set 6493 use '-Wtraditional-conversion'. 6494 6495 * Use of ISO C style function definitions. This warning 6496 intentionally is _not_ issued for prototype declarations or 6497 variadic functions because these ISO C features appear in your 6498 code when using libiberty's traditional C compatibility 6499 macros, 'PARAMS' and 'VPARAMS'. This warning is also bypassed 6500 for nested functions because that feature is already a GCC 6501 extension and thus not relevant to traditional C 6502 compatibility. 6503 6504'-Wtraditional-conversion (C and Objective-C only)' 6505 Warn if a prototype causes a type conversion that is different from 6506 what would happen to the same argument in the absence of a 6507 prototype. This includes conversions of fixed point to floating 6508 and vice versa, and conversions changing the width or signedness of 6509 a fixed-point argument except when the same as the default 6510 promotion. 6511 6512'-Wdeclaration-after-statement (C and Objective-C only)' 6513 Warn when a declaration is found after a statement in a block. 6514 This construct, known from C++, was introduced with ISO C99 and is 6515 by default allowed in GCC. It is not supported by ISO C90. *Note 6516 Mixed Declarations::. 6517 6518'-Wshadow' 6519 Warn whenever a local variable or type declaration shadows another 6520 variable, parameter, type, class member (in C++), or instance 6521 variable (in Objective-C) or whenever a built-in function is 6522 shadowed. Note that in C++, the compiler warns if a local variable 6523 shadows an explicit typedef, but not if it shadows a 6524 struct/class/enum. If this warning is enabled, it includes also 6525 all instances of local shadowing. This means that 6526 '-Wno-shadow=local' and '-Wno-shadow=compatible-local' are ignored 6527 when '-Wshadow' is used. Same as '-Wshadow=global'. 6528 6529'-Wno-shadow-ivar (Objective-C only)' 6530 Do not warn whenever a local variable shadows an instance variable 6531 in an Objective-C method. 6532 6533'-Wshadow=global' 6534 Warn for any shadowing. Same as '-Wshadow'. 6535 6536'-Wshadow=local' 6537 Warn when a local variable shadows another local variable or 6538 parameter. 6539 6540'-Wshadow=compatible-local' 6541 Warn when a local variable shadows another local variable or 6542 parameter whose type is compatible with that of the shadowing 6543 variable. In C++, type compatibility here means the type of the 6544 shadowing variable can be converted to that of the shadowed 6545 variable. The creation of this flag (in addition to 6546 '-Wshadow=local') is based on the idea that when a local variable 6547 shadows another one of incompatible type, it is most likely 6548 intentional, not a bug or typo, as shown in the following example: 6549 6550 for (SomeIterator i = SomeObj.begin(); i != SomeObj.end(); ++i) 6551 { 6552 for (int i = 0; i < N; ++i) 6553 { 6554 ... 6555 } 6556 ... 6557 } 6558 6559 Since the two variable 'i' in the example above have incompatible 6560 types, enabling only '-Wshadow=compatible-local' does not emit a 6561 warning. Because their types are incompatible, if a programmer 6562 accidentally uses one in place of the other, type checking is 6563 expected to catch that and emit an error or warning. Use of this 6564 flag instead of '-Wshadow=local' can possibly reduce the number of 6565 warnings triggered by intentional shadowing. Note that this also 6566 means that shadowing 'const char *i' by 'char *i' does not emit a 6567 warning. 6568 6569 This warning is also enabled by '-Wshadow=local'. 6570 6571'-Wlarger-than=BYTE-SIZE' 6572 Warn whenever an object is defined whose size exceeds BYTE-SIZE. 6573 '-Wlarger-than=''PTRDIFF_MAX' is enabled by default. Warnings 6574 controlled by the option can be disabled either by specifying 6575 BYTE-SIZE of 'SIZE_MAX' or more or by '-Wno-larger-than'. 6576 6577'-Wno-larger-than' 6578 Disable '-Wlarger-than=' warnings. The option is equivalent to 6579 '-Wlarger-than=''SIZE_MAX' or larger. 6580 6581'-Wframe-larger-than=BYTE-SIZE' 6582 Warn if the size of a function frame exceeds BYTE-SIZE. The 6583 computation done to determine the stack frame size is approximate 6584 and not conservative. The actual requirements may be somewhat 6585 greater than BYTE-SIZE even if you do not get a warning. In 6586 addition, any space allocated via 'alloca', variable-length arrays, 6587 or related constructs is not included by the compiler when 6588 determining whether or not to issue a warning. 6589 '-Wframe-larger-than=''PTRDIFF_MAX' is enabled by default. 6590 Warnings controlled by the option can be disabled either by 6591 specifying BYTE-SIZE of 'SIZE_MAX' or more or by 6592 '-Wno-frame-larger-than'. 6593 6594'-Wno-frame-larger-than' 6595 Disable '-Wframe-larger-than=' warnings. The option is equivalent 6596 to '-Wframe-larger-than=''SIZE_MAX' or larger. 6597 6598'-Wno-free-nonheap-object' 6599 Do not warn when attempting to free an object that was not 6600 allocated on the heap. 6601 6602'-Wstack-usage=BYTE-SIZE' 6603 Warn if the stack usage of a function might exceed BYTE-SIZE. The 6604 computation done to determine the stack usage is conservative. Any 6605 space allocated via 'alloca', variable-length arrays, or related 6606 constructs is included by the compiler when determining whether or 6607 not to issue a warning. 6608 6609 The message is in keeping with the output of '-fstack-usage'. 6610 6611 * If the stack usage is fully static but exceeds the specified 6612 amount, it's: 6613 6614 warning: stack usage is 1120 bytes 6615 * If the stack usage is (partly) dynamic but bounded, it's: 6616 6617 warning: stack usage might be 1648 bytes 6618 * If the stack usage is (partly) dynamic and not bounded, it's: 6619 6620 warning: stack usage might be unbounded 6621 6622 '-Wstack-usage=''PTRDIFF_MAX' is enabled by default. Warnings 6623 controlled by the option can be disabled either by specifying 6624 BYTE-SIZE of 'SIZE_MAX' or more or by '-Wno-stack-usage'. 6625 6626'-Wno-stack-usage' 6627 Disable '-Wstack-usage=' warnings. The option is equivalent to 6628 '-Wstack-usage=''SIZE_MAX' or larger. 6629 6630'-Wunsafe-loop-optimizations' 6631 Warn if the loop cannot be optimized because the compiler cannot 6632 assume anything on the bounds of the loop indices. With 6633 '-funsafe-loop-optimizations' warn if the compiler makes such 6634 assumptions. 6635 6636'-Wno-pedantic-ms-format (MinGW targets only)' 6637 When used in combination with '-Wformat' and '-pedantic' without 6638 GNU extensions, this option disables the warnings about non-ISO 6639 'printf' / 'scanf' format width specifiers 'I32', 'I64', and 'I' 6640 used on Windows targets, which depend on the MS runtime. 6641 6642'-Wpointer-arith' 6643 Warn about anything that depends on the "size of" a function type 6644 or of 'void'. GNU C assigns these types a size of 1, for 6645 convenience in calculations with 'void *' pointers and pointers to 6646 functions. In C++, warn also when an arithmetic operation involves 6647 'NULL'. This warning is also enabled by '-Wpedantic'. 6648 6649'-Wno-pointer-compare' 6650 Do not warn if a pointer is compared with a zero character 6651 constant. This usually means that the pointer was meant to be 6652 dereferenced. For example: 6653 6654 const char *p = foo (); 6655 if (p == '\0') 6656 return 42; 6657 6658 Note that the code above is invalid in C++11. 6659 6660 This warning is enabled by default. 6661 6662'-Wtype-limits' 6663 Warn if a comparison is always true or always false due to the 6664 limited range of the data type, but do not warn for constant 6665 expressions. For example, warn if an unsigned variable is compared 6666 against zero with '<' or '>='. This warning is also enabled by 6667 '-Wextra'. 6668 6669'-Wabsolute-value (C and Objective-C only)' 6670 Warn for calls to standard functions that compute the absolute 6671 value of an argument when a more appropriate standard function is 6672 available. For example, calling 'abs(3.14)' triggers the warning 6673 because the appropriate function to call to compute the absolute 6674 value of a double argument is 'fabs'. The option also triggers 6675 warnings when the argument in a call to such a function has an 6676 unsigned type. This warning can be suppressed with an explicit 6677 type cast and it is also enabled by '-Wextra'. 6678 6679'-Wcomment' 6680'-Wcomments' 6681 Warn whenever a comment-start sequence '/*' appears in a '/*' 6682 comment, or whenever a backslash-newline appears in a '//' comment. 6683 This warning is enabled by '-Wall'. 6684 6685'-Wtrigraphs' 6686 Warn if any trigraphs are encountered that might change the meaning 6687 of the program. Trigraphs within comments are not warned about, 6688 except those that would form escaped newlines. 6689 6690 This option is implied by '-Wall'. If '-Wall' is not given, this 6691 option is still enabled unless trigraphs are enabled. To get 6692 trigraph conversion without warnings, but get the other '-Wall' 6693 warnings, use '-trigraphs -Wall -Wno-trigraphs'. 6694 6695'-Wundef' 6696 Warn if an undefined identifier is evaluated in an '#if' directive. 6697 Such identifiers are replaced with zero. 6698 6699'-Wexpansion-to-defined' 6700 Warn whenever 'defined' is encountered in the expansion of a macro 6701 (including the case where the macro is expanded by an '#if' 6702 directive). Such usage is not portable. This warning is also 6703 enabled by '-Wpedantic' and '-Wextra'. 6704 6705'-Wunused-macros' 6706 Warn about macros defined in the main file that are unused. A 6707 macro is "used" if it is expanded or tested for existence at least 6708 once. The preprocessor also warns if the macro has not been used 6709 at the time it is redefined or undefined. 6710 6711 Built-in macros, macros defined on the command line, and macros 6712 defined in include files are not warned about. 6713 6714 _Note:_ If a macro is actually used, but only used in skipped 6715 conditional blocks, then the preprocessor reports it as unused. To 6716 avoid the warning in such a case, you might improve the scope of 6717 the macro's definition by, for example, moving it into the first 6718 skipped block. Alternatively, you could provide a dummy use with 6719 something like: 6720 6721 #if defined the_macro_causing_the_warning 6722 #endif 6723 6724'-Wno-endif-labels' 6725 Do not warn whenever an '#else' or an '#endif' are followed by 6726 text. This sometimes happens in older programs with code of the 6727 form 6728 6729 #if FOO 6730 ... 6731 #else FOO 6732 ... 6733 #endif FOO 6734 6735 The second and third 'FOO' should be in comments. This warning is 6736 on by default. 6737 6738'-Wbad-function-cast (C and Objective-C only)' 6739 Warn when a function call is cast to a non-matching type. For 6740 example, warn if a call to a function returning an integer type is 6741 cast to a pointer type. 6742 6743'-Wc90-c99-compat (C and Objective-C only)' 6744 Warn about features not present in ISO C90, but present in ISO C99. 6745 For instance, warn about use of variable length arrays, 'long long' 6746 type, 'bool' type, compound literals, designated initializers, and 6747 so on. This option is independent of the standards mode. Warnings 6748 are disabled in the expression that follows '__extension__'. 6749 6750'-Wc99-c11-compat (C and Objective-C only)' 6751 Warn about features not present in ISO C99, but present in ISO C11. 6752 For instance, warn about use of anonymous structures and unions, 6753 '_Atomic' type qualifier, '_Thread_local' storage-class specifier, 6754 '_Alignas' specifier, 'Alignof' operator, '_Generic' keyword, and 6755 so on. This option is independent of the standards mode. Warnings 6756 are disabled in the expression that follows '__extension__'. 6757 6758'-Wc11-c2x-compat (C and Objective-C only)' 6759 Warn about features not present in ISO C11, but present in ISO C2X. 6760 For instance, warn about omitting the string in '_Static_assert', 6761 use of '[[]]' syntax for attributes, use of decimal floating-point 6762 types, and so on. This option is independent of the standards 6763 mode. Warnings are disabled in the expression that follows 6764 '__extension__'. 6765 6766'-Wc++-compat (C and Objective-C only)' 6767 Warn about ISO C constructs that are outside of the common subset 6768 of ISO C and ISO C++, e.g. request for implicit conversion from 6769 'void *' to a pointer to non-'void' type. 6770 6771'-Wc++11-compat (C++ and Objective-C++ only)' 6772 Warn about C++ constructs whose meaning differs between ISO C++ 6773 1998 and ISO C++ 2011, e.g., identifiers in ISO C++ 1998 that are 6774 keywords in ISO C++ 2011. This warning turns on '-Wnarrowing' and 6775 is enabled by '-Wall'. 6776 6777'-Wc++14-compat (C++ and Objective-C++ only)' 6778 Warn about C++ constructs whose meaning differs between ISO C++ 6779 2011 and ISO C++ 2014. This warning is enabled by '-Wall'. 6780 6781'-Wc++17-compat (C++ and Objective-C++ only)' 6782 Warn about C++ constructs whose meaning differs between ISO C++ 6783 2014 and ISO C++ 2017. This warning is enabled by '-Wall'. 6784 6785'-Wc++20-compat (C++ and Objective-C++ only)' 6786 Warn about C++ constructs whose meaning differs between ISO C++ 6787 2017 and ISO C++ 2020. This warning is enabled by '-Wall'. 6788 6789'-Wcast-qual' 6790 Warn whenever a pointer is cast so as to remove a type qualifier 6791 from the target type. For example, warn if a 'const char *' is 6792 cast to an ordinary 'char *'. 6793 6794 Also warn when making a cast that introduces a type qualifier in an 6795 unsafe way. For example, casting 'char **' to 'const char **' is 6796 unsafe, as in this example: 6797 6798 /* p is char ** value. */ 6799 const char **q = (const char **) p; 6800 /* Assignment of readonly string to const char * is OK. */ 6801 *q = "string"; 6802 /* Now char** pointer points to read-only memory. */ 6803 **p = 'b'; 6804 6805'-Wcast-align' 6806 Warn whenever a pointer is cast such that the required alignment of 6807 the target is increased. For example, warn if a 'char *' is cast 6808 to an 'int *' on machines where integers can only be accessed at 6809 two- or four-byte boundaries. 6810 6811'-Wcast-align=strict' 6812 Warn whenever a pointer is cast such that the required alignment of 6813 the target is increased. For example, warn if a 'char *' is cast 6814 to an 'int *' regardless of the target machine. 6815 6816'-Wcast-function-type' 6817 Warn when a function pointer is cast to an incompatible function 6818 pointer. In a cast involving function types with a variable 6819 argument list only the types of initial arguments that are provided 6820 are considered. Any parameter of pointer-type matches any other 6821 pointer-type. Any benign differences in integral types are 6822 ignored, like 'int' vs. 'long' on ILP32 targets. Likewise type 6823 qualifiers are ignored. The function type 'void (*) (void)' is 6824 special and matches everything, which can be used to suppress this 6825 warning. In a cast involving pointer to member types this warning 6826 warns whenever the type cast is changing the pointer to member 6827 type. This warning is enabled by '-Wextra'. 6828 6829'-Wwrite-strings' 6830 When compiling C, give string constants the type 'const 6831 char[LENGTH]' so that copying the address of one into a non-'const' 6832 'char *' pointer produces a warning. These warnings help you find 6833 at compile time code that can try to write into a string constant, 6834 but only if you have been very careful about using 'const' in 6835 declarations and prototypes. Otherwise, it is just a nuisance. 6836 This is why we did not make '-Wall' request these warnings. 6837 6838 When compiling C++, warn about the deprecated conversion from 6839 string literals to 'char *'. This warning is enabled by default 6840 for C++ programs. 6841 6842'-Wclobbered' 6843 Warn for variables that might be changed by 'longjmp' or 'vfork'. 6844 This warning is also enabled by '-Wextra'. 6845 6846'-Wconversion' 6847 Warn for implicit conversions that may alter a value. This 6848 includes conversions between real and integer, like 'abs (x)' when 6849 'x' is 'double'; conversions between signed and unsigned, like 6850 'unsigned ui = -1'; and conversions to smaller types, like 'sqrtf 6851 (M_PI)'. Do not warn for explicit casts like 'abs ((int) x)' and 6852 'ui = (unsigned) -1', or if the value is not changed by the 6853 conversion like in 'abs (2.0)'. Warnings about conversions between 6854 signed and unsigned integers can be disabled by using 6855 '-Wno-sign-conversion'. 6856 6857 For C++, also warn for confusing overload resolution for 6858 user-defined conversions; and conversions that never use a type 6859 conversion operator: conversions to 'void', the same type, a base 6860 class or a reference to them. Warnings about conversions between 6861 signed and unsigned integers are disabled by default in C++ unless 6862 '-Wsign-conversion' is explicitly enabled. 6863 6864 Warnings about conversion from arithmetic on a small type back to 6865 that type are only given with '-Warith-conversion'. 6866 6867'-Wdangling-else' 6868 Warn about constructions where there may be confusion to which 'if' 6869 statement an 'else' branch belongs. Here is an example of such a 6870 case: 6871 6872 { 6873 if (a) 6874 if (b) 6875 foo (); 6876 else 6877 bar (); 6878 } 6879 6880 In C/C++, every 'else' branch belongs to the innermost possible 6881 'if' statement, which in this example is 'if (b)'. This is often 6882 not what the programmer expected, as illustrated in the above 6883 example by indentation the programmer chose. When there is the 6884 potential for this confusion, GCC issues a warning when this flag 6885 is specified. To eliminate the warning, add explicit braces around 6886 the innermost 'if' statement so there is no way the 'else' can 6887 belong to the enclosing 'if'. The resulting code looks like this: 6888 6889 { 6890 if (a) 6891 { 6892 if (b) 6893 foo (); 6894 else 6895 bar (); 6896 } 6897 } 6898 6899 This warning is enabled by '-Wparentheses'. 6900 6901'-Wdate-time' 6902 Warn when macros '__TIME__', '__DATE__' or '__TIMESTAMP__' are 6903 encountered as they might prevent bit-wise-identical reproducible 6904 compilations. 6905 6906'-Wempty-body' 6907 Warn if an empty body occurs in an 'if', 'else' or 'do while' 6908 statement. This warning is also enabled by '-Wextra'. 6909 6910'-Wno-endif-labels' 6911 Do not warn about stray tokens after '#else' and '#endif'. 6912 6913'-Wenum-compare' 6914 Warn about a comparison between values of different enumerated 6915 types. In C++ enumerated type mismatches in conditional 6916 expressions are also diagnosed and the warning is enabled by 6917 default. In C this warning is enabled by '-Wall'. 6918 6919'-Wenum-conversion (C, Objective-C only)' 6920 Warn when a value of enumerated type is implicitly converted to a 6921 different enumerated type. This warning is enabled by '-Wextra'. 6922 6923'-Wjump-misses-init (C, Objective-C only)' 6924 Warn if a 'goto' statement or a 'switch' statement jumps forward 6925 across the initialization of a variable, or jumps backward to a 6926 label after the variable has been initialized. This only warns 6927 about variables that are initialized when they are declared. This 6928 warning is only supported for C and Objective-C; in C++ this sort 6929 of branch is an error in any case. 6930 6931 '-Wjump-misses-init' is included in '-Wc++-compat'. It can be 6932 disabled with the '-Wno-jump-misses-init' option. 6933 6934'-Wsign-compare' 6935 Warn when a comparison between signed and unsigned values could 6936 produce an incorrect result when the signed value is converted to 6937 unsigned. In C++, this warning is also enabled by '-Wall'. In C, 6938 it is also enabled by '-Wextra'. 6939 6940'-Wsign-conversion' 6941 Warn for implicit conversions that may change the sign of an 6942 integer value, like assigning a signed integer expression to an 6943 unsigned integer variable. An explicit cast silences the warning. 6944 In C, this option is enabled also by '-Wconversion'. 6945 6946'-Wfloat-conversion' 6947 Warn for implicit conversions that reduce the precision of a real 6948 value. This includes conversions from real to integer, and from 6949 higher precision real to lower precision real values. This option 6950 is also enabled by '-Wconversion'. 6951 6952'-Wno-scalar-storage-order' 6953 Do not warn on suspicious constructs involving reverse scalar 6954 storage order. 6955 6956'-Wsizeof-pointer-div' 6957 Warn for suspicious divisions of two sizeof expressions that divide 6958 the pointer size by the element size, which is the usual way to 6959 compute the array size but won't work out correctly with pointers. 6960 This warning warns e.g. about 'sizeof (ptr) / sizeof (ptr[0])' if 6961 'ptr' is not an array, but a pointer. This warning is enabled by 6962 '-Wall'. 6963 6964'-Wsizeof-pointer-memaccess' 6965 Warn for suspicious length parameters to certain string and memory 6966 built-in functions if the argument uses 'sizeof'. This warning 6967 triggers for example for 'memset (ptr, 0, sizeof (ptr));' if 'ptr' 6968 is not an array, but a pointer, and suggests a possible fix, or 6969 about 'memcpy (&foo, ptr, sizeof (&foo));'. 6970 '-Wsizeof-pointer-memaccess' also warns about calls to bounded 6971 string copy functions like 'strncat' or 'strncpy' that specify as 6972 the bound a 'sizeof' expression of the source array. For example, 6973 in the following function the call to 'strncat' specifies the size 6974 of the source string as the bound. That is almost certainly a 6975 mistake and so the call is diagnosed. 6976 void make_file (const char *name) 6977 { 6978 char path[PATH_MAX]; 6979 strncpy (path, name, sizeof path - 1); 6980 strncat (path, ".text", sizeof ".text"); 6981 ... 6982 } 6983 6984 The '-Wsizeof-pointer-memaccess' option is enabled by '-Wall'. 6985 6986'-Wno-sizeof-array-argument' 6987 Do not warn when the 'sizeof' operator is applied to a parameter 6988 that is declared as an array in a function definition. This 6989 warning is enabled by default for C and C++ programs. 6990 6991'-Wmemset-elt-size' 6992 Warn for suspicious calls to the 'memset' built-in function, if the 6993 first argument references an array, and the third argument is a 6994 number equal to the number of elements, but not equal to the size 6995 of the array in memory. This indicates that the user has omitted a 6996 multiplication by the element size. This warning is enabled by 6997 '-Wall'. 6998 6999'-Wmemset-transposed-args' 7000 Warn for suspicious calls to the 'memset' built-in function where 7001 the second argument is not zero and the third argument is zero. 7002 For example, the call 'memset (buf, sizeof buf, 0)' is diagnosed 7003 because 'memset (buf, 0, sizeof buf)' was meant instead. The 7004 diagnostic is only emitted if the third argument is a literal zero. 7005 Otherwise, if it is an expression that is folded to zero, or a cast 7006 of zero to some type, it is far less likely that the arguments have 7007 been mistakenly transposed and no warning is emitted. This warning 7008 is enabled by '-Wall'. 7009 7010'-Waddress' 7011 Warn about suspicious uses of memory addresses. These include 7012 using the address of a function in a conditional expression, such 7013 as 'void func(void); if (func)', and comparisons against the memory 7014 address of a string literal, such as 'if (x == "abc")'. Such uses 7015 typically indicate a programmer error: the address of a function 7016 always evaluates to true, so their use in a conditional usually 7017 indicate that the programmer forgot the parentheses in a function 7018 call; and comparisons against string literals result in unspecified 7019 behavior and are not portable in C, so they usually indicate that 7020 the programmer intended to use 'strcmp'. This warning is enabled 7021 by '-Wall'. 7022 7023'-Wno-address-of-packed-member' 7024 Do not warn when the address of packed member of struct or union is 7025 taken, which usually results in an unaligned pointer value. This 7026 is enabled by default. 7027 7028'-Wlogical-op' 7029 Warn about suspicious uses of logical operators in expressions. 7030 This includes using logical operators in contexts where a bit-wise 7031 operator is likely to be expected. Also warns when the operands of 7032 a logical operator are the same: 7033 extern int a; 7034 if (a < 0 && a < 0) { ... } 7035 7036'-Wlogical-not-parentheses' 7037 Warn about logical not used on the left hand side operand of a 7038 comparison. This option does not warn if the right operand is 7039 considered to be a boolean expression. Its purpose is to detect 7040 suspicious code like the following: 7041 int a; 7042 ... 7043 if (!a > 1) { ... } 7044 7045 It is possible to suppress the warning by wrapping the LHS into 7046 parentheses: 7047 if ((!a) > 1) { ... } 7048 7049 This warning is enabled by '-Wall'. 7050 7051'-Waggregate-return' 7052 Warn if any functions that return structures or unions are defined 7053 or called. (In languages where you can return an array, this also 7054 elicits a warning.) 7055 7056'-Wno-aggressive-loop-optimizations' 7057 Warn if in a loop with constant number of iterations the compiler 7058 detects undefined behavior in some statement during one or more of 7059 the iterations. 7060 7061'-Wno-attributes' 7062 Do not warn if an unexpected '__attribute__' is used, such as 7063 unrecognized attributes, function attributes applied to variables, 7064 etc. This does not stop errors for incorrect use of supported 7065 attributes. 7066 7067'-Wno-builtin-declaration-mismatch' 7068 Warn if a built-in function is declared with an incompatible 7069 signature or as a non-function, or when a built-in function 7070 declared with a type that does not include a prototype is called 7071 with arguments whose promoted types do not match those expected by 7072 the function. When '-Wextra' is specified, also warn when a 7073 built-in function that takes arguments is declared without a 7074 prototype. The '-Wbuiltin-declaration-mismatch' warning is enabled 7075 by default. To avoid the warning include the appropriate header to 7076 bring the prototypes of built-in functions into scope. 7077 7078 For example, the call to 'memset' below is diagnosed by the warning 7079 because the function expects a value of type 'size_t' as its 7080 argument but the type of '32' is 'int'. With '-Wextra', the 7081 declaration of the function is diagnosed as well. 7082 extern void* memset (); 7083 void f (void *d) 7084 { 7085 memset (d, '\0', 32); 7086 } 7087 7088'-Wno-builtin-macro-redefined' 7089 Do not warn if certain built-in macros are redefined. This 7090 suppresses warnings for redefinition of '__TIMESTAMP__', 7091 '__TIME__', '__DATE__', '__FILE__', and '__BASE_FILE__'. 7092 7093'-Wstrict-prototypes (C and Objective-C only)' 7094 Warn if a function is declared or defined without specifying the 7095 argument types. (An old-style function definition is permitted 7096 without a warning if preceded by a declaration that specifies the 7097 argument types.) 7098 7099'-Wold-style-declaration (C and Objective-C only)' 7100 Warn for obsolescent usages, according to the C Standard, in a 7101 declaration. For example, warn if storage-class specifiers like 7102 'static' are not the first things in a declaration. This warning 7103 is also enabled by '-Wextra'. 7104 7105'-Wold-style-definition (C and Objective-C only)' 7106 Warn if an old-style function definition is used. A warning is 7107 given even if there is a previous prototype. A definition using 7108 '()' is not considered an old-style definition in C2X mode, because 7109 it is equivalent to '(void)' in that case, but is considered an 7110 old-style definition for older standards. 7111 7112'-Wmissing-parameter-type (C and Objective-C only)' 7113 A function parameter is declared without a type specifier in 7114 K&R-style functions: 7115 7116 void foo(bar) { } 7117 7118 This warning is also enabled by '-Wextra'. 7119 7120'-Wmissing-prototypes (C and Objective-C only)' 7121 Warn if a global function is defined without a previous prototype 7122 declaration. This warning is issued even if the definition itself 7123 provides a prototype. Use this option to detect global functions 7124 that do not have a matching prototype declaration in a header file. 7125 This option is not valid for C++ because all function declarations 7126 provide prototypes and a non-matching declaration declares an 7127 overload rather than conflict with an earlier declaration. Use 7128 '-Wmissing-declarations' to detect missing declarations in C++. 7129 7130'-Wmissing-declarations' 7131 Warn if a global function is defined without a previous 7132 declaration. Do so even if the definition itself provides a 7133 prototype. Use this option to detect global functions that are not 7134 declared in header files. In C, no warnings are issued for 7135 functions with previous non-prototype declarations; use 7136 '-Wmissing-prototypes' to detect missing prototypes. In C++, no 7137 warnings are issued for function templates, or for inline 7138 functions, or for functions in anonymous namespaces. 7139 7140'-Wmissing-field-initializers' 7141 Warn if a structure's initializer has some fields missing. For 7142 example, the following code causes such a warning, because 'x.h' is 7143 implicitly zero: 7144 7145 struct s { int f, g, h; }; 7146 struct s x = { 3, 4 }; 7147 7148 This option does not warn about designated initializers, so the 7149 following modification does not trigger a warning: 7150 7151 struct s { int f, g, h; }; 7152 struct s x = { .f = 3, .g = 4 }; 7153 7154 In C this option does not warn about the universal zero initializer 7155 '{ 0 }': 7156 7157 struct s { int f, g, h; }; 7158 struct s x = { 0 }; 7159 7160 Likewise, in C++ this option does not warn about the empty { } 7161 initializer, for example: 7162 7163 struct s { int f, g, h; }; 7164 s x = { }; 7165 7166 This warning is included in '-Wextra'. To get other '-Wextra' 7167 warnings without this one, use '-Wextra 7168 -Wno-missing-field-initializers'. 7169 7170'-Wno-multichar' 7171 Do not warn if a multicharacter constant (''FOOF'') is used. 7172 Usually they indicate a typo in the user's code, as they have 7173 implementation-defined values, and should not be used in portable 7174 code. 7175 7176'-Wnormalized=[none|id|nfc|nfkc]' 7177 In ISO C and ISO C++, two identifiers are different if they are 7178 different sequences of characters. However, sometimes when 7179 characters outside the basic ASCII character set are used, you can 7180 have two different character sequences that look the same. To 7181 avoid confusion, the ISO 10646 standard sets out some 7182 "normalization rules" which when applied ensure that two sequences 7183 that look the same are turned into the same sequence. GCC can warn 7184 you if you are using identifiers that have not been normalized; 7185 this option controls that warning. 7186 7187 There are four levels of warning supported by GCC. The default is 7188 '-Wnormalized=nfc', which warns about any identifier that is not in 7189 the ISO 10646 "C" normalized form, "NFC". NFC is the recommended 7190 form for most uses. It is equivalent to '-Wnormalized'. 7191 7192 Unfortunately, there are some characters allowed in identifiers by 7193 ISO C and ISO C++ that, when turned into NFC, are not allowed in 7194 identifiers. That is, there's no way to use these symbols in 7195 portable ISO C or C++ and have all your identifiers in NFC. 7196 '-Wnormalized=id' suppresses the warning for these characters. It 7197 is hoped that future versions of the standards involved will 7198 correct this, which is why this option is not the default. 7199 7200 You can switch the warning off for all characters by writing 7201 '-Wnormalized=none' or '-Wno-normalized'. You should only do this 7202 if you are using some other normalization scheme (like "D"), 7203 because otherwise you can easily create bugs that are literally 7204 impossible to see. 7205 7206 Some characters in ISO 10646 have distinct meanings but look 7207 identical in some fonts or display methodologies, especially once 7208 formatting has been applied. For instance '\u207F', "SUPERSCRIPT 7209 LATIN SMALL LETTER N", displays just like a regular 'n' that has 7210 been placed in a superscript. ISO 10646 defines the "NFKC" 7211 normalization scheme to convert all these into a standard form as 7212 well, and GCC warns if your code is not in NFKC if you use 7213 '-Wnormalized=nfkc'. This warning is comparable to warning about 7214 every identifier that contains the letter O because it might be 7215 confused with the digit 0, and so is not the default, but may be 7216 useful as a local coding convention if the programming environment 7217 cannot be fixed to display these characters distinctly. 7218 7219'-Wno-attribute-warning' 7220 Do not warn about usage of functions (*note Function Attributes::) 7221 declared with 'warning' attribute. By default, this warning is 7222 enabled. '-Wno-attribute-warning' can be used to disable the 7223 warning or '-Wno-error=attribute-warning' can be used to disable 7224 the error when compiled with '-Werror' flag. 7225 7226'-Wno-deprecated' 7227 Do not warn about usage of deprecated features. *Note Deprecated 7228 Features::. 7229 7230'-Wno-deprecated-declarations' 7231 Do not warn about uses of functions (*note Function Attributes::), 7232 variables (*note Variable Attributes::), and types (*note Type 7233 Attributes::) marked as deprecated by using the 'deprecated' 7234 attribute. 7235 7236'-Wno-overflow' 7237 Do not warn about compile-time overflow in constant expressions. 7238 7239'-Wno-odr' 7240 Warn about One Definition Rule violations during link-time 7241 optimization. Enabled by default. 7242 7243'-Wopenmp-simd' 7244 Warn if the vectorizer cost model overrides the OpenMP simd 7245 directive set by user. The '-fsimd-cost-model=unlimited' option 7246 can be used to relax the cost model. 7247 7248'-Woverride-init (C and Objective-C only)' 7249 Warn if an initialized field without side effects is overridden 7250 when using designated initializers (*note Designated Initializers: 7251 Designated Inits.). 7252 7253 This warning is included in '-Wextra'. To get other '-Wextra' 7254 warnings without this one, use '-Wextra -Wno-override-init'. 7255 7256'-Wno-override-init-side-effects (C and Objective-C only)' 7257 Do not warn if an initialized field with side effects is overridden 7258 when using designated initializers (*note Designated Initializers: 7259 Designated Inits.). This warning is enabled by default. 7260 7261'-Wpacked' 7262 Warn if a structure is given the packed attribute, but the packed 7263 attribute has no effect on the layout or size of the structure. 7264 Such structures may be mis-aligned for little benefit. For 7265 instance, in this code, the variable 'f.x' in 'struct bar' is 7266 misaligned even though 'struct bar' does not itself have the packed 7267 attribute: 7268 7269 struct foo { 7270 int x; 7271 char a, b, c, d; 7272 } __attribute__((packed)); 7273 struct bar { 7274 char z; 7275 struct foo f; 7276 }; 7277 7278'-Wnopacked-bitfield-compat' 7279 The 4.1, 4.2 and 4.3 series of GCC ignore the 'packed' attribute on 7280 bit-fields of type 'char'. This was fixed in GCC 4.4 but the 7281 change can lead to differences in the structure layout. GCC 7282 informs you when the offset of such a field has changed in GCC 4.4. 7283 For example there is no longer a 4-bit padding between field 'a' 7284 and 'b' in this structure: 7285 7286 struct foo 7287 { 7288 char a:4; 7289 char b:8; 7290 } __attribute__ ((packed)); 7291 7292 This warning is enabled by default. Use 7293 '-Wno-packed-bitfield-compat' to disable this warning. 7294 7295'-Wpacked-not-aligned (C, C++, Objective-C and Objective-C++ only)' 7296 Warn if a structure field with explicitly specified alignment in a 7297 packed struct or union is misaligned. For example, a warning will 7298 be issued on 'struct S', like, 'warning: alignment 1 of 'struct S' 7299 is less than 8', in this code: 7300 7301 struct __attribute__ ((aligned (8))) S8 { char a[8]; }; 7302 struct __attribute__ ((packed)) S { 7303 struct S8 s8; 7304 }; 7305 7306 This warning is enabled by '-Wall'. 7307 7308'-Wpadded' 7309 Warn if padding is included in a structure, either to align an 7310 element of the structure or to align the whole structure. 7311 Sometimes when this happens it is possible to rearrange the fields 7312 of the structure to reduce the padding and so make the structure 7313 smaller. 7314 7315'-Wredundant-decls' 7316 Warn if anything is declared more than once in the same scope, even 7317 in cases where multiple declaration is valid and changes nothing. 7318 7319'-Wrestrict' 7320 Warn when an object referenced by a 'restrict'-qualified parameter 7321 (or, in C++, a '__restrict'-qualified parameter) is aliased by 7322 another argument, or when copies between such objects overlap. For 7323 example, the call to the 'strcpy' function below attempts to 7324 truncate the string by replacing its initial characters with the 7325 last four. However, because the call writes the terminating NUL 7326 into 'a[4]', the copies overlap and the call is diagnosed. 7327 7328 void foo (void) 7329 { 7330 char a[] = "abcd1234"; 7331 strcpy (a, a + 4); 7332 ... 7333 } 7334 The '-Wrestrict' option detects some instances of simple overlap 7335 even without optimization but works best at '-O2' and above. It is 7336 included in '-Wall'. 7337 7338'-Wnested-externs (C and Objective-C only)' 7339 Warn if an 'extern' declaration is encountered within a function. 7340 7341'-Winline' 7342 Warn if a function that is declared as inline cannot be inlined. 7343 Even with this option, the compiler does not warn about failures to 7344 inline functions declared in system headers. 7345 7346 The compiler uses a variety of heuristics to determine whether or 7347 not to inline a function. For example, the compiler takes into 7348 account the size of the function being inlined and the amount of 7349 inlining that has already been done in the current function. 7350 Therefore, seemingly insignificant changes in the source program 7351 can cause the warnings produced by '-Winline' to appear or 7352 disappear. 7353 7354'-Wint-in-bool-context' 7355 Warn for suspicious use of integer values where boolean values are 7356 expected, such as conditional expressions (?:) using non-boolean 7357 integer constants in boolean context, like 'if (a <= b ? 2 : 3)'. 7358 Or left shifting of signed integers in boolean context, like 'for 7359 (a = 0; 1 << a; a++);'. Likewise for all kinds of multiplications 7360 regardless of the data type. This warning is enabled by '-Wall'. 7361 7362'-Wno-int-to-pointer-cast' 7363 Suppress warnings from casts to pointer type of an integer of a 7364 different size. In C++, casting to a pointer type of smaller size 7365 is an error. 'Wint-to-pointer-cast' is enabled by default. 7366 7367'-Wno-pointer-to-int-cast (C and Objective-C only)' 7368 Suppress warnings from casts from a pointer to an integer type of a 7369 different size. 7370 7371'-Winvalid-pch' 7372 Warn if a precompiled header (*note Precompiled Headers::) is found 7373 in the search path but cannot be used. 7374 7375'-Wlong-long' 7376 Warn if 'long long' type is used. This is enabled by either 7377 '-Wpedantic' or '-Wtraditional' in ISO C90 and C++98 modes. To 7378 inhibit the warning messages, use '-Wno-long-long'. 7379 7380'-Wvariadic-macros' 7381 Warn if variadic macros are used in ISO C90 mode, or if the GNU 7382 alternate syntax is used in ISO C99 mode. This is enabled by 7383 either '-Wpedantic' or '-Wtraditional'. To inhibit the warning 7384 messages, use '-Wno-variadic-macros'. 7385 7386'-Wno-varargs' 7387 Do not warn upon questionable usage of the macros used to handle 7388 variable arguments like 'va_start'. These warnings are enabled by 7389 default. 7390 7391'-Wvector-operation-performance' 7392 Warn if vector operation is not implemented via SIMD capabilities 7393 of the architecture. Mainly useful for the performance tuning. 7394 Vector operation can be implemented 'piecewise', which means that 7395 the scalar operation is performed on every vector element; 'in 7396 parallel', which means that the vector operation is implemented 7397 using scalars of wider type, which normally is more performance 7398 efficient; and 'as a single scalar', which means that vector fits 7399 into a scalar type. 7400 7401'-Wvla' 7402 Warn if a variable-length array is used in the code. '-Wno-vla' 7403 prevents the '-Wpedantic' warning of the variable-length array. 7404 7405'-Wvla-larger-than=BYTE-SIZE' 7406 If this option is used, the compiler warns for declarations of 7407 variable-length arrays whose size is either unbounded, or bounded 7408 by an argument that allows the array size to exceed BYTE-SIZE 7409 bytes. This is similar to how '-Walloca-larger-than='BYTE-SIZE 7410 works, but with variable-length arrays. 7411 7412 Note that GCC may optimize small variable-length arrays of a known 7413 value into plain arrays, so this warning may not get triggered for 7414 such arrays. 7415 7416 '-Wvla-larger-than=''PTRDIFF_MAX' is enabled by default but is 7417 typically only effective when '-ftree-vrp' is active (default for 7418 '-O2' and above). 7419 7420 See also '-Walloca-larger-than=BYTE-SIZE'. 7421 7422'-Wno-vla-larger-than' 7423 Disable '-Wvla-larger-than=' warnings. The option is equivalent to 7424 '-Wvla-larger-than=''SIZE_MAX' or larger. 7425 7426'-Wvolatile-register-var' 7427 Warn if a register variable is declared volatile. The volatile 7428 modifier does not inhibit all optimizations that may eliminate 7429 reads and/or writes to register variables. This warning is enabled 7430 by '-Wall'. 7431 7432'-Wdisabled-optimization' 7433 Warn if a requested optimization pass is disabled. This warning 7434 does not generally indicate that there is anything wrong with your 7435 code; it merely indicates that GCC's optimizers are unable to 7436 handle the code effectively. Often, the problem is that your code 7437 is too big or too complex; GCC refuses to optimize programs when 7438 the optimization itself is likely to take inordinate amounts of 7439 time. 7440 7441'-Wpointer-sign (C and Objective-C only)' 7442 Warn for pointer argument passing or assignment with different 7443 signedness. This option is only supported for C and Objective-C. 7444 It is implied by '-Wall' and by '-Wpedantic', which can be disabled 7445 with '-Wno-pointer-sign'. 7446 7447'-Wstack-protector' 7448 This option is only active when '-fstack-protector' is active. It 7449 warns about functions that are not protected against stack 7450 smashing. 7451 7452'-Woverlength-strings' 7453 Warn about string constants that are longer than the "minimum 7454 maximum" length specified in the C standard. Modern compilers 7455 generally allow string constants that are much longer than the 7456 standard's minimum limit, but very portable programs should avoid 7457 using longer strings. 7458 7459 The limit applies _after_ string constant concatenation, and does 7460 not count the trailing NUL. In C90, the limit was 509 characters; 7461 in C99, it was raised to 4095. C++98 does not specify a normative 7462 minimum maximum, so we do not diagnose overlength strings in C++. 7463 7464 This option is implied by '-Wpedantic', and can be disabled with 7465 '-Wno-overlength-strings'. 7466 7467'-Wunsuffixed-float-constants (C and Objective-C only)' 7468 7469 Issue a warning for any floating constant that does not have a 7470 suffix. When used together with '-Wsystem-headers' it warns about 7471 such constants in system header files. This can be useful when 7472 preparing code to use with the 'FLOAT_CONST_DECIMAL64' pragma from 7473 the decimal floating-point extension to C99. 7474 7475'-Wno-lto-type-mismatch' 7476 7477 During the link-time optimization, do not warn about type 7478 mismatches in global declarations from different compilation units. 7479 Requires '-flto' to be enabled. Enabled by default. 7480 7481'-Wno-designated-init (C and Objective-C only)' 7482 Suppress warnings when a positional initializer is used to 7483 initialize a structure that has been marked with the 7484 'designated_init' attribute. 7485 7486'-Wno-hsa' 7487 Do not warn when HSAIL cannot be emitted for the compiled function 7488 or OpenMP construct. These warnings are enabled by default. 7489 7490 7491File: gcc.info, Node: Static Analyzer Options, Next: Debugging Options, Prev: Warning Options, Up: Invoking GCC 7492 74933.9 Options That Control Static Analysis 7494======================================== 7495 7496'-fanalyzer' 7497 This option enables an static analysis of program flow which looks 7498 for "interesting" interprocedural paths through the code, and 7499 issues warnings for problems found on them. 7500 7501 This analysis is much more expensive than other GCC warnings. 7502 7503 Enabling this option effectively enables the following warnings: 7504 7505 7506 -Wanalyzer-double-fclose 7507 -Wanalyzer-double-free 7508 -Wanalyzer-exposure-through-output-file 7509 -Wanalyzer-file-leak 7510 -Wanalyzer-free-of-non-heap 7511 -Wanalyzer-malloc-leak 7512 -Wanalyzer-possible-null-argument 7513 -Wanalyzer-possible-null-dereference 7514 -Wanalyzer-null-argument 7515 -Wanalyzer-null-dereference 7516 -Wanalyzer-stale-setjmp-buffer 7517 -Wanalyzer-tainted-array-index 7518 -Wanalyzer-unsafe-call-within-signal-handler 7519 -Wanalyzer-use-after-free 7520 -Wanalyzer-use-of-pointer-in-stale-stack-frame 7521 7522 7523 This option is only available if GCC was configured with analyzer 7524 support enabled. 7525 7526'-Wanalyzer-too-complex' 7527 If '-fanalyzer' is enabled, the analyzer uses various heuristics to 7528 attempt to explore the control flow and data flow in the program, 7529 but these can be defeated by sufficiently complicated code. 7530 7531 By default, the analysis silently stops if the code is too 7532 complicated for the analyzer to fully explore and it reaches an 7533 internal limit. The '-Wanalyzer-too-complex' option warns if this 7534 occurs. 7535 7536'-Wno-analyzer-double-fclose' 7537 This warning requires '-fanalyzer', which enables it; use 7538 '-Wno-analyzer-double-fclose' to disable it. 7539 7540 This diagnostic warns for paths through the code in which a 'FILE 7541 *' can have 'fclose' called on it more than once. 7542 7543'-Wno-analyzer-double-free' 7544 This warning requires '-fanalyzer', which enables it; use 7545 '-Wno-analyzer-double-free' to disable it. 7546 7547 This diagnostic warns for paths through the code in which a pointer 7548 can have 'free' called on it more than once. 7549 7550'-Wno-analyzer-exposure-through-output-file' 7551 This warning requires '-fanalyzer', which enables it; use 7552 '-Wno-analyzer-exposure-through-output-file' to disable it. 7553 7554 This diagnostic warns for paths through the code in which a 7555 security-sensitive value is written to an output file (such as 7556 writing a password to a log file). 7557 7558'-Wno-analyzer-file-leak' 7559 This warning requires '-fanalyzer', which enables it; use 7560 '-Wno-analyzer-file-leak' to disable it. 7561 7562 This diagnostic warns for paths through the code in which a 7563 '<stdio.h>' 'FILE *' stream object is leaked. 7564 7565'-Wno-analyzer-free-of-non-heap' 7566 This warning requires '-fanalyzer', which enables it; use 7567 '-Wno-analyzer-free-of-non-heap' to disable it. 7568 7569 This diagnostic warns for paths through the code in which 'free' is 7570 called on a non-heap pointer (e.g. an on-stack buffer, or a 7571 global). 7572 7573'-Wno-analyzer-malloc-leak' 7574 This warning requires '-fanalyzer', which enables it; use 7575 '-Wno-analyzer-malloc-leak' to disable it. 7576 7577 This diagnostic warns for paths through the code in which a pointer 7578 allocated via 'malloc' is leaked. 7579 7580'-Wno-analyzer-possible-null-argument' 7581 This warning requires '-fanalyzer', which enables it; use 7582 '-Wno-analyzer-possible-null-argument' to disable it. 7583 7584 This diagnostic warns for paths through the code in which a 7585 possibly-NULL value is passed to a function argument marked with 7586 '__attribute__((nonnull))' as requiring a non-NULL value. 7587 7588'-Wno-analyzer-possible-null-dereference' 7589 This warning requires '-fanalyzer', which enables it; use 7590 '-Wno-analyzer-possible-null-dereference' to disable it. 7591 7592 This diagnostic warns for paths through the code in which a 7593 possibly-NULL value is dereferenced. 7594 7595'-Wno-analyzer-null-argument' 7596 This warning requires '-fanalyzer', which enables it; use 7597 '-Wno-analyzer-null-argument' to disable it. 7598 7599 This diagnostic warns for paths through the code in which a value 7600 known to be NULL is passed to a function argument marked with 7601 '__attribute__((nonnull))' as requiring a non-NULL value. 7602 7603'-Wno-analyzer-null-dereference' 7604 This warning requires '-fanalyzer', which enables it; use 7605 '-Wno-analyzer-null-dereference' to disable it. 7606 7607 This diagnostic warns for paths through the code in which a value 7608 known to be NULL is dereferenced. 7609 7610'-Wno-analyzer-stale-setjmp-buffer' 7611 This warning requires '-fanalyzer', which enables it; use 7612 '-Wno-analyzer-stale-setjmp-buffer' to disable it. 7613 7614 This diagnostic warns for paths through the code in which 'longjmp' 7615 is called to rewind to a 'jmp_buf' relating to a 'setjmp' call in a 7616 function that has returned. 7617 7618 When 'setjmp' is called on a 'jmp_buf' to record a rewind location, 7619 it records the stack frame. The stack frame becomes invalid when 7620 the function containing the 'setjmp' call returns. Attempting to 7621 rewind to it via 'longjmp' would reference a stack frame that no 7622 longer exists, and likely lead to a crash (or worse). 7623 7624'-Wno-analyzer-tainted-array-index' 7625 This warning requires both '-fanalyzer' and 7626 '-fanalyzer-checker=taint' to enable it; use 7627 '-Wno-analyzer-tainted-array-index' to disable it. 7628 7629 This diagnostic warns for paths through the code in which a value 7630 that could be under an attacker's control is used as the index of 7631 an array access without being sanitized. 7632 7633'-Wno-analyzer-unsafe-call-within-signal-handler' 7634 This warning requires '-fanalyzer', which enables it; use 7635 '-Wno-analyzer-unsafe-call-within-signal-handler' to disable it. 7636 7637 This diagnostic warns for paths through the code in which a 7638 function known to be async-signal-unsafe (such as 'fprintf') is 7639 called from a signal handler. 7640 7641'-Wno-analyzer-use-after-free' 7642 This warning requires '-fanalyzer', which enables it; use 7643 '-Wno-analyzer-use-after-free' to disable it. 7644 7645 This diagnostic warns for paths through the code in which a pointer 7646 is used after 'free' is called on it. 7647 7648'-Wno-analyzer-use-of-pointer-in-stale-stack-frame' 7649 This warning requires '-fanalyzer', which enables it; use 7650 '-Wno-analyzer-use-of-pointer-in-stale-stack-frame' to disable it. 7651 7652 This diagnostic warns for paths through the code in which a pointer 7653 is dereferenced that points to a variable in a stale stack frame. 7654 7655 Pertinent parameters for controlling the exploration are: '--param 7656analyzer-bb-explosion-factor=VALUE', '--param 7657analyzer-max-enodes-per-program-point=VALUE', '--param 7658analyzer-max-recursion-depth=VALUE', and '--param 7659analyzer-min-snodes-for-call-summary=VALUE'. 7660 7661 The following options control the analyzer. 7662 7663'-fanalyzer-call-summaries' 7664 Simplify interprocedural analysis by computing the effect of 7665 certain calls, rather than exploring all paths through the function 7666 from callsite to each possible return. 7667 7668 If enabled, call summaries are only used for functions with more 7669 than one call site, and that are sufficiently complicated (as per 7670 '--param analyzer-min-snodes-for-call-summary=VALUE'). 7671 7672'-fanalyzer-checker=NAME' 7673 Restrict the analyzer to run just the named checker, and enable it. 7674 7675 Some checkers are disabled by default (even with '-fanalyzer'), 7676 such as the 'taint' checker that implements 7677 '-Wanalyzer-tainted-array-index', and this option is required to 7678 enable them. 7679 7680'-fanalyzer-fine-grained' 7681 This option is intended for analyzer developers. 7682 7683 Internally the analyzer builds an "exploded graph" that combines 7684 control flow graphs with data flow information. 7685 7686 By default, an edge in this graph can contain the effects of a run 7687 of multiple statements within a basic block. With 7688 '-fanalyzer-fine-grained', each statement gets its own edge. 7689 7690'-fanalyzer-show-duplicate-count' 7691 This option is intended for analyzer developers: if multiple 7692 diagnostics have been detected as being duplicates of each other, 7693 it emits a note when reporting the best diagnostic, giving the 7694 number of additional diagnostics that were suppressed by the 7695 deduplication logic. 7696 7697'-fno-analyzer-state-merge' 7698 This option is intended for analyzer developers. 7699 7700 By default the analyzer attempts to simplify analysis by merging 7701 sufficiently similar states at each program point as it builds its 7702 "exploded graph". With '-fno-analyzer-state-merge' this merging 7703 can be suppressed, for debugging state-handling issues. 7704 7705'-fno-analyzer-state-purge' 7706 This option is intended for analyzer developers. 7707 7708 By default the analyzer attempts to simplify analysis by purging 7709 aspects of state at a program point that appear to no longer be 7710 relevant e.g. the values of locals that aren't accessed later in 7711 the function and which aren't relevant to leak analysis. 7712 7713 With '-fno-analyzer-state-purge' this purging of state can be 7714 suppressed, for debugging state-handling issues. 7715 7716'-fanalyzer-transitivity' 7717 This option enables transitivity of constraints within the 7718 analyzer. 7719 7720'-fanalyzer-verbose-edges' 7721 This option is intended for analyzer developers. It enables more 7722 verbose, lower-level detail in the descriptions of control flow 7723 within diagnostic paths. 7724 7725'-fanalyzer-verbose-state-changes' 7726 This option is intended for analyzer developers. It enables more 7727 verbose, lower-level detail in the descriptions of events relating 7728 to state machines within diagnostic paths. 7729 7730'-fanalyzer-verbosity=LEVEL' 7731 This option controls the complexity of the control flow paths that 7732 are emitted for analyzer diagnostics. 7733 7734 The LEVEL can be one of: 7735 7736 '0' 7737 At this level, interprocedural call and return events are 7738 displayed, along with the most pertinent state-change events 7739 relating to a diagnostic. For example, for a double-'free' 7740 diagnostic, both calls to 'free' will be shown. 7741 7742 '1' 7743 As per the previous level, but also show events for the entry 7744 to each function. 7745 7746 '2' 7747 As per the previous level, but also show events relating to 7748 control flow that are significant to triggering the issue 7749 (e.g. "true path taken" at a conditional). 7750 7751 This level is the default. 7752 7753 '3' 7754 As per the previous level, but show all control flow events, 7755 not just significant ones. 7756 7757 '4' 7758 This level is intended for analyzer developers; it adds 7759 various other events intended for debugging the analyzer. 7760 7761'-fdump-analyzer' 7762 Dump internal details about what the analyzer is doing to 7763 'FILE.analyzer.txt'. This option is overridden by 7764 '-fdump-analyzer-stderr'. 7765 7766'-fdump-analyzer-stderr' 7767 Dump internal details about what the analyzer is doing to stderr. 7768 This option overrides '-fdump-analyzer'. 7769 7770'-fdump-analyzer-callgraph' 7771 Dump a representation of the call graph suitable for viewing with 7772 GraphViz to 'FILE.callgraph.dot'. 7773 7774'-fdump-analyzer-exploded-graph' 7775 Dump a representation of the "exploded graph" suitable for viewing 7776 with GraphViz to 'FILE.eg.dot'. Nodes are color-coded based on 7777 state-machine states to emphasize state changes. 7778 7779'-fdump-analyzer-exploded-nodes' 7780 Emit diagnostics showing where nodes in the "exploded graph" are in 7781 relation to the program source. 7782 7783'-fdump-analyzer-exploded-nodes-2' 7784 Dump a textual representation of the "exploded graph" to 7785 'FILE.eg.txt'. 7786 7787'-fdump-analyzer-exploded-nodes-3' 7788 Dump a textual representation of the "exploded graph" to one dump 7789 file per node, to 'FILE.eg-ID.txt'. This is typically a large 7790 number of dump files. 7791 7792'-fdump-analyzer-state-purge' 7793 As per '-fdump-analyzer-supergraph', dump a representation of the 7794 "supergraph" suitable for viewing with GraphViz, but annotate the 7795 graph with information on what state will be purged at each node. 7796 The graph is written to 'FILE.state-purge.dot'. 7797 7798'-fdump-analyzer-supergraph' 7799 Dump representations of the "supergraph" suitable for viewing with 7800 GraphViz to 'FILE.supergraph.dot' and to 'FILE.supergraph-eg.dot'. 7801 These show all of the control flow graphs in the program, with 7802 interprocedural edges for calls and returns. The second dump 7803 contains annotations showing nodes in the "exploded graph" and 7804 diagnostics associated with them. 7805 7806 7807File: gcc.info, Node: Debugging Options, Next: Optimize Options, Prev: Static Analyzer Options, Up: Invoking GCC 7808 78093.10 Options for Debugging Your Program 7810======================================= 7811 7812To tell GCC to emit extra information for use by a debugger, in almost 7813all cases you need only to add '-g' to your other options. 7814 7815 GCC allows you to use '-g' with '-O'. The shortcuts taken by optimized 7816code may occasionally be surprising: some variables you declared may not 7817exist at all; flow of control may briefly move where you did not expect 7818it; some statements may not be executed because they compute constant 7819results or their values are already at hand; some statements may execute 7820in different places because they have been moved out of loops. 7821Nevertheless it is possible to debug optimized output. This makes it 7822reasonable to use the optimizer for programs that might have bugs. 7823 7824 If you are not using some other optimization option, consider using 7825'-Og' (*note Optimize Options::) with '-g'. With no '-O' option at all, 7826some compiler passes that collect information useful for debugging do 7827not run at all, so that '-Og' may result in a better debugging 7828experience. 7829 7830'-g' 7831 Produce debugging information in the operating system's native 7832 format (stabs, COFF, XCOFF, or DWARF). GDB can work with this 7833 debugging information. 7834 7835 On most systems that use stabs format, '-g' enables use of extra 7836 debugging information that only GDB can use; this extra information 7837 makes debugging work better in GDB but probably makes other 7838 debuggers crash or refuse to read the program. If you want to 7839 control for certain whether to generate the extra information, use 7840 '-gstabs+', '-gstabs', '-gxcoff+', '-gxcoff', or '-gvms' (see 7841 below). 7842 7843'-ggdb' 7844 Produce debugging information for use by GDB. This means to use 7845 the most expressive format available (DWARF, stabs, or the native 7846 format if neither of those are supported), including GDB extensions 7847 if at all possible. 7848 7849'-gdwarf' 7850'-gdwarf-VERSION' 7851 Produce debugging information in DWARF format (if that is 7852 supported). The value of VERSION may be either 2, 3, 4 or 5; the 7853 default version for most targets is 4. DWARF Version 5 is only 7854 experimental. 7855 7856 Note that with DWARF Version 2, some ports require and always use 7857 some non-conflicting DWARF 3 extensions in the unwind tables. 7858 7859 Version 4 may require GDB 7.0 and '-fvar-tracking-assignments' for 7860 maximum benefit. 7861 7862 GCC no longer supports DWARF Version 1, which is substantially 7863 different than Version 2 and later. For historical reasons, some 7864 other DWARF-related options such as '-fno-dwarf2-cfi-asm') retain a 7865 reference to DWARF Version 2 in their names, but apply to all 7866 currently-supported versions of DWARF. 7867 7868'-gstabs' 7869 Produce debugging information in stabs format (if that is 7870 supported), without GDB extensions. This is the format used by DBX 7871 on most BSD systems. On MIPS, Alpha and System V Release 4 systems 7872 this option produces stabs debugging output that is not understood 7873 by DBX. On System V Release 4 systems this option requires the GNU 7874 assembler. 7875 7876'-gstabs+' 7877 Produce debugging information in stabs format (if that is 7878 supported), using GNU extensions understood only by the GNU 7879 debugger (GDB). The use of these extensions is likely to make 7880 other debuggers crash or refuse to read the program. 7881 7882'-gxcoff' 7883 Produce debugging information in XCOFF format (if that is 7884 supported). This is the format used by the DBX debugger on IBM 7885 RS/6000 systems. 7886 7887'-gxcoff+' 7888 Produce debugging information in XCOFF format (if that is 7889 supported), using GNU extensions understood only by the GNU 7890 debugger (GDB). The use of these extensions is likely to make 7891 other debuggers crash or refuse to read the program, and may cause 7892 assemblers other than the GNU assembler (GAS) to fail with an 7893 error. 7894 7895'-gvms' 7896 Produce debugging information in Alpha/VMS debug format (if that is 7897 supported). This is the format used by DEBUG on Alpha/VMS systems. 7898 7899'-gLEVEL' 7900'-ggdbLEVEL' 7901'-gstabsLEVEL' 7902'-gxcoffLEVEL' 7903'-gvmsLEVEL' 7904 Request debugging information and also use LEVEL to specify how 7905 much information. The default level is 2. 7906 7907 Level 0 produces no debug information at all. Thus, '-g0' negates 7908 '-g'. 7909 7910 Level 1 produces minimal information, enough for making backtraces 7911 in parts of the program that you don't plan to debug. This 7912 includes descriptions of functions and external variables, and line 7913 number tables, but no information about local variables. 7914 7915 Level 3 includes extra information, such as all the macro 7916 definitions present in the program. Some debuggers support macro 7917 expansion when you use '-g3'. 7918 7919 If you use multiple '-g' options, with or without level numbers, 7920 the last such option is the one that is effective. 7921 7922 '-gdwarf' does not accept a concatenated debug level, to avoid 7923 confusion with '-gdwarf-LEVEL'. Instead use an additional 7924 '-gLEVEL' option to change the debug level for DWARF. 7925 7926'-fno-eliminate-unused-debug-symbols' 7927 By default, no debug information is produced for symbols that are 7928 not actually used. Use this option if you want debug information 7929 for all symbols. 7930 7931'-femit-class-debug-always' 7932 Instead of emitting debugging information for a C++ class in only 7933 one object file, emit it in all object files using the class. This 7934 option should be used only with debuggers that are unable to handle 7935 the way GCC normally emits debugging information for classes 7936 because using this option increases the size of debugging 7937 information by as much as a factor of two. 7938 7939'-fno-merge-debug-strings' 7940 Direct the linker to not merge together strings in the debugging 7941 information that are identical in different object files. Merging 7942 is not supported by all assemblers or linkers. Merging decreases 7943 the size of the debug information in the output file at the cost of 7944 increasing link processing time. Merging is enabled by default. 7945 7946'-fdebug-prefix-map=OLD=NEW' 7947 When compiling files residing in directory 'OLD', record debugging 7948 information describing them as if the files resided in directory 7949 'NEW' instead. This can be used to replace a build-time path with 7950 an install-time path in the debug info. It can also be used to 7951 change an absolute path to a relative path by using '.' for NEW. 7952 This can give more reproducible builds, which are location 7953 independent, but may require an extra command to tell GDB where to 7954 find the source files. See also '-ffile-prefix-map'. 7955 7956'-fvar-tracking' 7957 Run variable tracking pass. It computes where variables are stored 7958 at each position in code. Better debugging information is then 7959 generated (if the debugging information format supports this 7960 information). 7961 7962 It is enabled by default when compiling with optimization ('-Os', 7963 '-O', '-O2', ...), debugging information ('-g') and the debug info 7964 format supports it. 7965 7966'-fvar-tracking-assignments' 7967 Annotate assignments to user variables early in the compilation and 7968 attempt to carry the annotations over throughout the compilation 7969 all the way to the end, in an attempt to improve debug information 7970 while optimizing. Use of '-gdwarf-4' is recommended along with it. 7971 7972 It can be enabled even if var-tracking is disabled, in which case 7973 annotations are created and maintained, but discarded at the end. 7974 By default, this flag is enabled together with '-fvar-tracking', 7975 except when selective scheduling is enabled. 7976 7977'-gsplit-dwarf' 7978 Separate as much DWARF debugging information as possible into a 7979 separate output file with the extension '.dwo'. This option allows 7980 the build system to avoid linking files with debug information. To 7981 be useful, this option requires a debugger capable of reading 7982 '.dwo' files. 7983 7984'-gdescribe-dies' 7985 Add description attributes to some DWARF DIEs that have no name 7986 attribute, such as artificial variables, external references and 7987 call site parameter DIEs. 7988 7989'-gpubnames' 7990 Generate DWARF '.debug_pubnames' and '.debug_pubtypes' sections. 7991 7992'-ggnu-pubnames' 7993 Generate '.debug_pubnames' and '.debug_pubtypes' sections in a 7994 format suitable for conversion into a GDB index. This option is 7995 only useful with a linker that can produce GDB index version 7. 7996 7997'-fdebug-types-section' 7998 When using DWARF Version 4 or higher, type DIEs can be put into 7999 their own '.debug_types' section instead of making them part of the 8000 '.debug_info' section. It is more efficient to put them in a 8001 separate comdat section since the linker can then remove 8002 duplicates. But not all DWARF consumers support '.debug_types' 8003 sections yet and on some objects '.debug_types' produces larger 8004 instead of smaller debugging information. 8005 8006'-grecord-gcc-switches' 8007'-gno-record-gcc-switches' 8008 This switch causes the command-line options used to invoke the 8009 compiler that may affect code generation to be appended to the 8010 DW_AT_producer attribute in DWARF debugging information. The 8011 options are concatenated with spaces separating them from each 8012 other and from the compiler version. It is enabled by default. 8013 See also '-frecord-gcc-switches' for another way of storing 8014 compiler options into the object file. 8015 8016'-gstrict-dwarf' 8017 Disallow using extensions of later DWARF standard version than 8018 selected with '-gdwarf-VERSION'. On most targets using 8019 non-conflicting DWARF extensions from later standard versions is 8020 allowed. 8021 8022'-gno-strict-dwarf' 8023 Allow using extensions of later DWARF standard version than 8024 selected with '-gdwarf-VERSION'. 8025 8026'-gas-loc-support' 8027 Inform the compiler that the assembler supports '.loc' directives. 8028 It may then use them for the assembler to generate DWARF2+ line 8029 number tables. 8030 8031 This is generally desirable, because assembler-generated 8032 line-number tables are a lot more compact than those the compiler 8033 can generate itself. 8034 8035 This option will be enabled by default if, at GCC configure time, 8036 the assembler was found to support such directives. 8037 8038'-gno-as-loc-support' 8039 Force GCC to generate DWARF2+ line number tables internally, if 8040 DWARF2+ line number tables are to be generated. 8041 8042'-gas-locview-support' 8043 Inform the compiler that the assembler supports 'view' assignment 8044 and reset assertion checking in '.loc' directives. 8045 8046 This option will be enabled by default if, at GCC configure time, 8047 the assembler was found to support them. 8048 8049'-gno-as-locview-support' 8050 Force GCC to assign view numbers internally, if 8051 '-gvariable-location-views' are explicitly requested. 8052 8053'-gcolumn-info' 8054'-gno-column-info' 8055 Emit location column information into DWARF debugging information, 8056 rather than just file and line. This option is enabled by default. 8057 8058'-gstatement-frontiers' 8059'-gno-statement-frontiers' 8060 This option causes GCC to create markers in the internal 8061 representation at the beginning of statements, and to keep them 8062 roughly in place throughout compilation, using them to guide the 8063 output of 'is_stmt' markers in the line number table. This is 8064 enabled by default when compiling with optimization ('-Os', '-O', 8065 '-O2', ...), and outputting DWARF 2 debug information at the normal 8066 level. 8067 8068'-gvariable-location-views' 8069'-gvariable-location-views=incompat5' 8070'-gno-variable-location-views' 8071 Augment variable location lists with progressive view numbers 8072 implied from the line number table. This enables debug information 8073 consumers to inspect state at certain points of the program, even 8074 if no instructions associated with the corresponding source 8075 locations are present at that point. If the assembler lacks 8076 support for view numbers in line number tables, this will cause the 8077 compiler to emit the line number table, which generally makes them 8078 somewhat less compact. The augmented line number tables and 8079 location lists are fully backward-compatible, so they can be 8080 consumed by debug information consumers that are not aware of these 8081 augmentations, but they won't derive any benefit from them either. 8082 8083 This is enabled by default when outputting DWARF 2 debug 8084 information at the normal level, as long as there is assembler 8085 support, '-fvar-tracking-assignments' is enabled and 8086 '-gstrict-dwarf' is not. When assembler support is not available, 8087 this may still be enabled, but it will force GCC to output internal 8088 line number tables, and if '-ginternal-reset-location-views' is not 8089 enabled, that will most certainly lead to silently mismatching 8090 location views. 8091 8092 There is a proposed representation for view numbers that is not 8093 backward compatible with the location list format introduced in 8094 DWARF 5, that can be enabled with 8095 '-gvariable-location-views=incompat5'. This option may be removed 8096 in the future, is only provided as a reference implementation of 8097 the proposed representation. Debug information consumers are not 8098 expected to support this extended format, and they would be 8099 rendered unable to decode location lists using it. 8100 8101'-ginternal-reset-location-views' 8102'-gno-internal-reset-location-views' 8103 Attempt to determine location views that can be omitted from 8104 location view lists. This requires the compiler to have very 8105 accurate insn length estimates, which isn't always the case, and it 8106 may cause incorrect view lists to be generated silently when using 8107 an assembler that does not support location view lists. The GNU 8108 assembler will flag any such error as a 'view number mismatch'. 8109 This is only enabled on ports that define a reliable estimation 8110 function. 8111 8112'-ginline-points' 8113'-gno-inline-points' 8114 Generate extended debug information for inlined functions. 8115 Location view tracking markers are inserted at inlined entry 8116 points, so that address and view numbers can be computed and output 8117 in debug information. This can be enabled independently of 8118 location views, in which case the view numbers won't be output, but 8119 it can only be enabled along with statement frontiers, and it is 8120 only enabled by default if location views are enabled. 8121 8122'-gz[=TYPE]' 8123 Produce compressed debug sections in DWARF format, if that is 8124 supported. If TYPE is not given, the default type depends on the 8125 capabilities of the assembler and linker used. TYPE may be one of 8126 'none' (don't compress debug sections), 'zlib' (use zlib 8127 compression in ELF gABI format), or 'zlib-gnu' (use zlib 8128 compression in traditional GNU format). If the linker doesn't 8129 support writing compressed debug sections, the option is rejected. 8130 Otherwise, if the assembler does not support them, '-gz' is 8131 silently ignored when producing object files. 8132 8133'-femit-struct-debug-baseonly' 8134 Emit debug information for struct-like types only when the base 8135 name of the compilation source file matches the base name of file 8136 in which the struct is defined. 8137 8138 This option substantially reduces the size of debugging 8139 information, but at significant potential loss in type information 8140 to the debugger. See '-femit-struct-debug-reduced' for a less 8141 aggressive option. See '-femit-struct-debug-detailed' for more 8142 detailed control. 8143 8144 This option works only with DWARF debug output. 8145 8146'-femit-struct-debug-reduced' 8147 Emit debug information for struct-like types only when the base 8148 name of the compilation source file matches the base name of file 8149 in which the type is defined, unless the struct is a template or 8150 defined in a system header. 8151 8152 This option significantly reduces the size of debugging 8153 information, with some potential loss in type information to the 8154 debugger. See '-femit-struct-debug-baseonly' for a more aggressive 8155 option. See '-femit-struct-debug-detailed' for more detailed 8156 control. 8157 8158 This option works only with DWARF debug output. 8159 8160'-femit-struct-debug-detailed[=SPEC-LIST]' 8161 Specify the struct-like types for which the compiler generates 8162 debug information. The intent is to reduce duplicate struct debug 8163 information between different object files within the same program. 8164 8165 This option is a detailed version of '-femit-struct-debug-reduced' 8166 and '-femit-struct-debug-baseonly', which serves for most needs. 8167 8168 A specification has the syntax 8169 ['dir:'|'ind:']['ord:'|'gen:']('any'|'sys'|'base'|'none') 8170 8171 The optional first word limits the specification to structs that 8172 are used directly ('dir:') or used indirectly ('ind:'). A struct 8173 type is used directly when it is the type of a variable, member. 8174 Indirect uses arise through pointers to structs. That is, when use 8175 of an incomplete struct is valid, the use is indirect. An example 8176 is 'struct one direct; struct two * indirect;'. 8177 8178 The optional second word limits the specification to ordinary 8179 structs ('ord:') or generic structs ('gen:'). Generic structs are 8180 a bit complicated to explain. For C++, these are non-explicit 8181 specializations of template classes, or non-template classes within 8182 the above. Other programming languages have generics, but 8183 '-femit-struct-debug-detailed' does not yet implement them. 8184 8185 The third word specifies the source files for those structs for 8186 which the compiler should emit debug information. The values 8187 'none' and 'any' have the normal meaning. The value 'base' means 8188 that the base of name of the file in which the type declaration 8189 appears must match the base of the name of the main compilation 8190 file. In practice, this means that when compiling 'foo.c', debug 8191 information is generated for types declared in that file and 8192 'foo.h', but not other header files. The value 'sys' means those 8193 types satisfying 'base' or declared in system or compiler headers. 8194 8195 You may need to experiment to determine the best settings for your 8196 application. 8197 8198 The default is '-femit-struct-debug-detailed=all'. 8199 8200 This option works only with DWARF debug output. 8201 8202'-fno-dwarf2-cfi-asm' 8203 Emit DWARF unwind info as compiler generated '.eh_frame' section 8204 instead of using GAS '.cfi_*' directives. 8205 8206'-fno-eliminate-unused-debug-types' 8207 Normally, when producing DWARF output, GCC avoids producing debug 8208 symbol output for types that are nowhere used in the source file 8209 being compiled. Sometimes it is useful to have GCC emit debugging 8210 information for all types declared in a compilation unit, 8211 regardless of whether or not they are actually used in that 8212 compilation unit, for example if, in the debugger, you want to cast 8213 a value to a type that is not actually used in your program (but is 8214 declared). More often, however, this results in a significant 8215 amount of wasted space. 8216 8217 8218File: gcc.info, Node: Optimize Options, Next: Instrumentation Options, Prev: Debugging Options, Up: Invoking GCC 8219 82203.11 Options That Control Optimization 8221====================================== 8222 8223These options control various sorts of optimizations. 8224 8225 Without any optimization option, the compiler's goal is to reduce the 8226cost of compilation and to make debugging produce the expected results. 8227Statements are independent: if you stop the program with a breakpoint 8228between statements, you can then assign a new value to any variable or 8229change the program counter to any other statement in the function and 8230get exactly the results you expect from the source code. 8231 8232 Turning on optimization flags makes the compiler attempt to improve the 8233performance and/or code size at the expense of compilation time and 8234possibly the ability to debug the program. 8235 8236 The compiler performs optimization based on the knowledge it has of the 8237program. Compiling multiple files at once to a single output file mode 8238allows the compiler to use information gained from all of the files when 8239compiling each of them. 8240 8241 Not all optimizations are controlled directly by a flag. Only 8242optimizations that have a flag are listed in this section. 8243 8244 Most optimizations are completely disabled at '-O0' or if an '-O' level 8245is not set on the command line, even if individual optimization flags 8246are specified. Similarly, '-Og' suppresses many optimization passes. 8247 8248 Depending on the target and how GCC was configured, a slightly 8249different set of optimizations may be enabled at each '-O' level than 8250those listed here. You can invoke GCC with '-Q --help=optimizers' to 8251find out the exact set of optimizations that are enabled at each level. 8252*Note Overall Options::, for examples. 8253 8254'-O' 8255'-O1' 8256 Optimize. Optimizing compilation takes somewhat more time, and a 8257 lot more memory for a large function. 8258 8259 With '-O', the compiler tries to reduce code size and execution 8260 time, without performing any optimizations that take a great deal 8261 of compilation time. 8262 8263 '-O' turns on the following optimization flags: 8264 8265 -fauto-inc-dec 8266 -fbranch-count-reg 8267 -fcombine-stack-adjustments 8268 -fcompare-elim 8269 -fcprop-registers 8270 -fdce 8271 -fdefer-pop 8272 -fdelayed-branch 8273 -fdse 8274 -fforward-propagate 8275 -fguess-branch-probability 8276 -fif-conversion 8277 -fif-conversion2 8278 -finline-functions-called-once 8279 -fipa-profile 8280 -fipa-pure-const 8281 -fipa-reference 8282 -fipa-reference-addressable 8283 -fmerge-constants 8284 -fmove-loop-invariants 8285 -fomit-frame-pointer 8286 -freorder-blocks 8287 -fshrink-wrap 8288 -fshrink-wrap-separate 8289 -fsplit-wide-types 8290 -fssa-backprop 8291 -fssa-phiopt 8292 -ftree-bit-ccp 8293 -ftree-ccp 8294 -ftree-ch 8295 -ftree-coalesce-vars 8296 -ftree-copy-prop 8297 -ftree-dce 8298 -ftree-dominator-opts 8299 -ftree-dse 8300 -ftree-forwprop 8301 -ftree-fre 8302 -ftree-phiprop 8303 -ftree-pta 8304 -ftree-scev-cprop 8305 -ftree-sink 8306 -ftree-slsr 8307 -ftree-sra 8308 -ftree-ter 8309 -funit-at-a-time 8310 8311'-O2' 8312 Optimize even more. GCC performs nearly all supported 8313 optimizations that do not involve a space-speed tradeoff. As 8314 compared to '-O', this option increases both compilation time and 8315 the performance of the generated code. 8316 8317 '-O2' turns on all optimization flags specified by '-O'. It also 8318 turns on the following optimization flags: 8319 8320 -falign-functions -falign-jumps 8321 -falign-labels -falign-loops 8322 -fcaller-saves 8323 -fcode-hoisting 8324 -fcrossjumping 8325 -fcse-follow-jumps -fcse-skip-blocks 8326 -fdelete-null-pointer-checks 8327 -fdevirtualize -fdevirtualize-speculatively 8328 -fexpensive-optimizations 8329 -ffinite-loops 8330 -fgcse -fgcse-lm 8331 -fhoist-adjacent-loads 8332 -finline-functions 8333 -finline-small-functions 8334 -findirect-inlining 8335 -fipa-bit-cp -fipa-cp -fipa-icf 8336 -fipa-ra -fipa-sra -fipa-vrp 8337 -fisolate-erroneous-paths-dereference 8338 -flra-remat 8339 -foptimize-sibling-calls 8340 -foptimize-strlen 8341 -fpartial-inlining 8342 -fpeephole2 8343 -freorder-blocks-algorithm=stc 8344 -freorder-blocks-and-partition -freorder-functions 8345 -frerun-cse-after-loop 8346 -fschedule-insns -fschedule-insns2 8347 -fsched-interblock -fsched-spec 8348 -fstore-merging 8349 -fstrict-aliasing 8350 -fthread-jumps 8351 -ftree-builtin-call-dce 8352 -ftree-pre 8353 -ftree-switch-conversion -ftree-tail-merge 8354 -ftree-vrp 8355 8356 Please note the warning under '-fgcse' about invoking '-O2' on 8357 programs that use computed gotos. 8358 8359'-O3' 8360 Optimize yet more. '-O3' turns on all optimizations specified by 8361 '-O2' and also turns on the following optimization flags: 8362 8363 -fgcse-after-reload 8364 -fipa-cp-clone 8365 -floop-interchange 8366 -floop-unroll-and-jam 8367 -fpeel-loops 8368 -fpredictive-commoning 8369 -fsplit-loops 8370 -fsplit-paths 8371 -ftree-loop-distribution 8372 -ftree-loop-vectorize 8373 -ftree-partial-pre 8374 -ftree-slp-vectorize 8375 -funswitch-loops 8376 -fvect-cost-model 8377 -fvect-cost-model=dynamic 8378 -fversion-loops-for-strides 8379 8380'-O0' 8381 Reduce compilation time and make debugging produce the expected 8382 results. This is the default. 8383 8384'-Os' 8385 Optimize for size. '-Os' enables all '-O2' optimizations except 8386 those that often increase code size: 8387 8388 -falign-functions -falign-jumps 8389 -falign-labels -falign-loops 8390 -fprefetch-loop-arrays -freorder-blocks-algorithm=stc 8391 8392 It also enables '-finline-functions', causes the compiler to tune 8393 for code size rather than execution speed, and performs further 8394 optimizations designed to reduce code size. 8395 8396'-Ofast' 8397 Disregard strict standards compliance. '-Ofast' enables all '-O3' 8398 optimizations. It also enables optimizations that are not valid 8399 for all standard-compliant programs. It turns on '-ffast-math', 8400 '-fallow-store-data-races' and the Fortran-specific 8401 '-fstack-arrays', unless '-fmax-stack-var-size' is specified, and 8402 '-fno-protect-parens'. 8403 8404'-Og' 8405 Optimize debugging experience. '-Og' should be the optimization 8406 level of choice for the standard edit-compile-debug cycle, offering 8407 a reasonable level of optimization while maintaining fast 8408 compilation and a good debugging experience. It is a better choice 8409 than '-O0' for producing debuggable code because some compiler 8410 passes that collect debug information are disabled at '-O0'. 8411 8412 Like '-O0', '-Og' completely disables a number of optimization 8413 passes so that individual options controlling them have no effect. 8414 Otherwise '-Og' enables all '-O1' optimization flags except for 8415 those that may interfere with debugging: 8416 8417 -fbranch-count-reg -fdelayed-branch 8418 -fdse -fif-conversion -fif-conversion2 8419 -finline-functions-called-once 8420 -fmove-loop-invariants -fssa-phiopt 8421 -ftree-bit-ccp -ftree-dse -ftree-pta -ftree-sra 8422 8423 If you use multiple '-O' options, with or without level numbers, the 8424last such option is the one that is effective. 8425 8426 Options of the form '-fFLAG' specify machine-independent flags. Most 8427flags have both positive and negative forms; the negative form of 8428'-ffoo' is '-fno-foo'. In the table below, only one of the forms is 8429listed--the one you typically use. You can figure out the other form by 8430either removing 'no-' or adding it. 8431 8432 The following options control specific optimizations. They are either 8433activated by '-O' options or are related to ones that are. You can use 8434the following flags in the rare cases when "fine-tuning" of 8435optimizations to be performed is desired. 8436 8437'-fno-defer-pop' 8438 For machines that must pop arguments after a function call, always 8439 pop the arguments as soon as each function returns. At levels 8440 '-O1' and higher, '-fdefer-pop' is the default; this allows the 8441 compiler to let arguments accumulate on the stack for several 8442 function calls and pop them all at once. 8443 8444'-fforward-propagate' 8445 Perform a forward propagation pass on RTL. The pass tries to 8446 combine two instructions and checks if the result can be 8447 simplified. If loop unrolling is active, two passes are performed 8448 and the second is scheduled after loop unrolling. 8449 8450 This option is enabled by default at optimization levels '-O', 8451 '-O2', '-O3', '-Os'. 8452 8453'-ffp-contract=STYLE' 8454 '-ffp-contract=off' disables floating-point expression contraction. 8455 '-ffp-contract=fast' enables floating-point expression contraction 8456 such as forming of fused multiply-add operations if the target has 8457 native support for them. '-ffp-contract=on' enables floating-point 8458 expression contraction if allowed by the language standard. This 8459 is currently not implemented and treated equal to 8460 '-ffp-contract=off'. 8461 8462 The default is '-ffp-contract=fast'. 8463 8464'-fomit-frame-pointer' 8465 Omit the frame pointer in functions that don't need one. This 8466 avoids the instructions to save, set up and restore the frame 8467 pointer; on many targets it also makes an extra register available. 8468 8469 On some targets this flag has no effect because the standard 8470 calling sequence always uses a frame pointer, so it cannot be 8471 omitted. 8472 8473 Note that '-fno-omit-frame-pointer' doesn't guarantee the frame 8474 pointer is used in all functions. Several targets always omit the 8475 frame pointer in leaf functions. 8476 8477 Enabled by default at '-O' and higher. 8478 8479'-foptimize-sibling-calls' 8480 Optimize sibling and tail recursive calls. 8481 8482 Enabled at levels '-O2', '-O3', '-Os'. 8483 8484'-foptimize-strlen' 8485 Optimize various standard C string functions (e.g. 'strlen', 8486 'strchr' or 'strcpy') and their '_FORTIFY_SOURCE' counterparts into 8487 faster alternatives. 8488 8489 Enabled at levels '-O2', '-O3'. 8490 8491'-fno-inline' 8492 Do not expand any functions inline apart from those marked with the 8493 'always_inline' attribute. This is the default when not 8494 optimizing. 8495 8496 Single functions can be exempted from inlining by marking them with 8497 the 'noinline' attribute. 8498 8499'-finline-small-functions' 8500 Integrate functions into their callers when their body is smaller 8501 than expected function call code (so overall size of program gets 8502 smaller). The compiler heuristically decides which functions are 8503 simple enough to be worth integrating in this way. This inlining 8504 applies to all functions, even those not declared inline. 8505 8506 Enabled at levels '-O2', '-O3', '-Os'. 8507 8508'-findirect-inlining' 8509 Inline also indirect calls that are discovered to be known at 8510 compile time thanks to previous inlining. This option has any 8511 effect only when inlining itself is turned on by the 8512 '-finline-functions' or '-finline-small-functions' options. 8513 8514 Enabled at levels '-O2', '-O3', '-Os'. 8515 8516'-finline-functions' 8517 Consider all functions for inlining, even if they are not declared 8518 inline. The compiler heuristically decides which functions are 8519 worth integrating in this way. 8520 8521 If all calls to a given function are integrated, and the function 8522 is declared 'static', then the function is normally not output as 8523 assembler code in its own right. 8524 8525 Enabled at levels '-O2', '-O3', '-Os'. Also enabled by 8526 '-fprofile-use' and '-fauto-profile'. 8527 8528'-finline-functions-called-once' 8529 Consider all 'static' functions called once for inlining into their 8530 caller even if they are not marked 'inline'. If a call to a given 8531 function is integrated, then the function is not output as 8532 assembler code in its own right. 8533 8534 Enabled at levels '-O1', '-O2', '-O3' and '-Os', but not '-Og'. 8535 8536'-fearly-inlining' 8537 Inline functions marked by 'always_inline' and functions whose body 8538 seems smaller than the function call overhead early before doing 8539 '-fprofile-generate' instrumentation and real inlining pass. Doing 8540 so makes profiling significantly cheaper and usually inlining 8541 faster on programs having large chains of nested wrapper functions. 8542 8543 Enabled by default. 8544 8545'-fipa-sra' 8546 Perform interprocedural scalar replacement of aggregates, removal 8547 of unused parameters and replacement of parameters passed by 8548 reference by parameters passed by value. 8549 8550 Enabled at levels '-O2', '-O3' and '-Os'. 8551 8552'-finline-limit=N' 8553 By default, GCC limits the size of functions that can be inlined. 8554 This flag allows coarse control of this limit. N is the size of 8555 functions that can be inlined in number of pseudo instructions. 8556 8557 Inlining is actually controlled by a number of parameters, which 8558 may be specified individually by using '--param NAME=VALUE'. The 8559 '-finline-limit=N' option sets some of these parameters as follows: 8560 8561 'max-inline-insns-single' 8562 is set to N/2. 8563 'max-inline-insns-auto' 8564 is set to N/2. 8565 8566 See below for a documentation of the individual parameters 8567 controlling inlining and for the defaults of these parameters. 8568 8569 _Note:_ there may be no value to '-finline-limit' that results in 8570 default behavior. 8571 8572 _Note:_ pseudo instruction represents, in this particular context, 8573 an abstract measurement of function's size. In no way does it 8574 represent a count of assembly instructions and as such its exact 8575 meaning might change from one release to an another. 8576 8577'-fno-keep-inline-dllexport' 8578 This is a more fine-grained version of '-fkeep-inline-functions', 8579 which applies only to functions that are declared using the 8580 'dllexport' attribute or declspec. *Note Declaring Attributes of 8581 Functions: Function Attributes. 8582 8583'-fkeep-inline-functions' 8584 In C, emit 'static' functions that are declared 'inline' into the 8585 object file, even if the function has been inlined into all of its 8586 callers. This switch does not affect functions using the 'extern 8587 inline' extension in GNU C90. In C++, emit any and all inline 8588 functions into the object file. 8589 8590'-fkeep-static-functions' 8591 Emit 'static' functions into the object file, even if the function 8592 is never used. 8593 8594'-fkeep-static-consts' 8595 Emit variables declared 'static const' when optimization isn't 8596 turned on, even if the variables aren't referenced. 8597 8598 GCC enables this option by default. If you want to force the 8599 compiler to check if a variable is referenced, regardless of 8600 whether or not optimization is turned on, use the 8601 '-fno-keep-static-consts' option. 8602 8603'-fmerge-constants' 8604 Attempt to merge identical constants (string constants and 8605 floating-point constants) across compilation units. 8606 8607 This option is the default for optimized compilation if the 8608 assembler and linker support it. Use '-fno-merge-constants' to 8609 inhibit this behavior. 8610 8611 Enabled at levels '-O', '-O2', '-O3', '-Os'. 8612 8613'-fmerge-all-constants' 8614 Attempt to merge identical constants and identical variables. 8615 8616 This option implies '-fmerge-constants'. In addition to 8617 '-fmerge-constants' this considers e.g. even constant initialized 8618 arrays or initialized constant variables with integral or 8619 floating-point types. Languages like C or C++ require each 8620 variable, including multiple instances of the same variable in 8621 recursive calls, to have distinct locations, so using this option 8622 results in non-conforming behavior. 8623 8624'-fmodulo-sched' 8625 Perform swing modulo scheduling immediately before the first 8626 scheduling pass. This pass looks at innermost loops and reorders 8627 their instructions by overlapping different iterations. 8628 8629'-fmodulo-sched-allow-regmoves' 8630 Perform more aggressive SMS-based modulo scheduling with register 8631 moves allowed. By setting this flag certain anti-dependences edges 8632 are deleted, which triggers the generation of reg-moves based on 8633 the life-range analysis. This option is effective only with 8634 '-fmodulo-sched' enabled. 8635 8636'-fno-branch-count-reg' 8637 Disable the optimization pass that scans for opportunities to use 8638 "decrement and branch" instructions on a count register instead of 8639 instruction sequences that decrement a register, compare it against 8640 zero, and then branch based upon the result. This option is only 8641 meaningful on architectures that support such instructions, which 8642 include x86, PowerPC, IA-64 and S/390. Note that the 8643 '-fno-branch-count-reg' option doesn't remove the decrement and 8644 branch instructions from the generated instruction stream 8645 introduced by other optimization passes. 8646 8647 The default is '-fbranch-count-reg' at '-O1' and higher, except for 8648 '-Og'. 8649 8650'-fno-function-cse' 8651 Do not put function addresses in registers; make each instruction 8652 that calls a constant function contain the function's address 8653 explicitly. 8654 8655 This option results in less efficient code, but some strange hacks 8656 that alter the assembler output may be confused by the 8657 optimizations performed when this option is not used. 8658 8659 The default is '-ffunction-cse' 8660 8661'-fno-zero-initialized-in-bss' 8662 If the target supports a BSS section, GCC by default puts variables 8663 that are initialized to zero into BSS. This can save space in the 8664 resulting code. 8665 8666 This option turns off this behavior because some programs 8667 explicitly rely on variables going to the data section--e.g., so 8668 that the resulting executable can find the beginning of that 8669 section and/or make assumptions based on that. 8670 8671 The default is '-fzero-initialized-in-bss'. 8672 8673'-fthread-jumps' 8674 Perform optimizations that check to see if a jump branches to a 8675 location where another comparison subsumed by the first is found. 8676 If so, the first branch is redirected to either the destination of 8677 the second branch or a point immediately following it, depending on 8678 whether the condition is known to be true or false. 8679 8680 Enabled at levels '-O2', '-O3', '-Os'. 8681 8682'-fsplit-wide-types' 8683 When using a type that occupies multiple registers, such as 'long 8684 long' on a 32-bit system, split the registers apart and allocate 8685 them independently. This normally generates better code for those 8686 types, but may make debugging more difficult. 8687 8688 Enabled at levels '-O', '-O2', '-O3', '-Os'. 8689 8690'-fsplit-wide-types-early' 8691 Fully split wide types early, instead of very late. This option 8692 has no effect unless '-fsplit-wide-types' is turned on. 8693 8694 This is the default on some targets. 8695 8696'-fcse-follow-jumps' 8697 In common subexpression elimination (CSE), scan through jump 8698 instructions when the target of the jump is not reached by any 8699 other path. For example, when CSE encounters an 'if' statement 8700 with an 'else' clause, CSE follows the jump when the condition 8701 tested is false. 8702 8703 Enabled at levels '-O2', '-O3', '-Os'. 8704 8705'-fcse-skip-blocks' 8706 This is similar to '-fcse-follow-jumps', but causes CSE to follow 8707 jumps that conditionally skip over blocks. When CSE encounters a 8708 simple 'if' statement with no else clause, '-fcse-skip-blocks' 8709 causes CSE to follow the jump around the body of the 'if'. 8710 8711 Enabled at levels '-O2', '-O3', '-Os'. 8712 8713'-frerun-cse-after-loop' 8714 Re-run common subexpression elimination after loop optimizations 8715 are performed. 8716 8717 Enabled at levels '-O2', '-O3', '-Os'. 8718 8719'-fgcse' 8720 Perform a global common subexpression elimination pass. This pass 8721 also performs global constant and copy propagation. 8722 8723 _Note:_ When compiling a program using computed gotos, a GCC 8724 extension, you may get better run-time performance if you disable 8725 the global common subexpression elimination pass by adding 8726 '-fno-gcse' to the command line. 8727 8728 Enabled at levels '-O2', '-O3', '-Os'. 8729 8730'-fgcse-lm' 8731 When '-fgcse-lm' is enabled, global common subexpression 8732 elimination attempts to move loads that are only killed by stores 8733 into themselves. This allows a loop containing a load/store 8734 sequence to be changed to a load outside the loop, and a copy/store 8735 within the loop. 8736 8737 Enabled by default when '-fgcse' is enabled. 8738 8739'-fgcse-sm' 8740 When '-fgcse-sm' is enabled, a store motion pass is run after 8741 global common subexpression elimination. This pass attempts to 8742 move stores out of loops. When used in conjunction with 8743 '-fgcse-lm', loops containing a load/store sequence can be changed 8744 to a load before the loop and a store after the loop. 8745 8746 Not enabled at any optimization level. 8747 8748'-fgcse-las' 8749 When '-fgcse-las' is enabled, the global common subexpression 8750 elimination pass eliminates redundant loads that come after stores 8751 to the same memory location (both partial and full redundancies). 8752 8753 Not enabled at any optimization level. 8754 8755'-fgcse-after-reload' 8756 When '-fgcse-after-reload' is enabled, a redundant load elimination 8757 pass is performed after reload. The purpose of this pass is to 8758 clean up redundant spilling. 8759 8760 Enabled by '-fprofile-use' and '-fauto-profile'. 8761 8762'-faggressive-loop-optimizations' 8763 This option tells the loop optimizer to use language constraints to 8764 derive bounds for the number of iterations of a loop. This assumes 8765 that loop code does not invoke undefined behavior by for example 8766 causing signed integer overflows or out-of-bound array accesses. 8767 The bounds for the number of iterations of a loop are used to guide 8768 loop unrolling and peeling and loop exit test optimizations. This 8769 option is enabled by default. 8770 8771'-funconstrained-commons' 8772 This option tells the compiler that variables declared in common 8773 blocks (e.g. Fortran) may later be overridden with longer trailing 8774 arrays. This prevents certain optimizations that depend on knowing 8775 the array bounds. 8776 8777'-fcrossjumping' 8778 Perform cross-jumping transformation. This transformation unifies 8779 equivalent code and saves code size. The resulting code may or may 8780 not perform better than without cross-jumping. 8781 8782 Enabled at levels '-O2', '-O3', '-Os'. 8783 8784'-fauto-inc-dec' 8785 Combine increments or decrements of addresses with memory accesses. 8786 This pass is always skipped on architectures that do not have 8787 instructions to support this. Enabled by default at '-O' and 8788 higher on architectures that support this. 8789 8790'-fdce' 8791 Perform dead code elimination (DCE) on RTL. Enabled by default at 8792 '-O' and higher. 8793 8794'-fdse' 8795 Perform dead store elimination (DSE) on RTL. Enabled by default at 8796 '-O' and higher. 8797 8798'-fif-conversion' 8799 Attempt to transform conditional jumps into branch-less 8800 equivalents. This includes use of conditional moves, min, max, set 8801 flags and abs instructions, and some tricks doable by standard 8802 arithmetics. The use of conditional execution on chips where it is 8803 available is controlled by '-fif-conversion2'. 8804 8805 Enabled at levels '-O', '-O2', '-O3', '-Os', but not with '-Og'. 8806 8807'-fif-conversion2' 8808 Use conditional execution (where available) to transform 8809 conditional jumps into branch-less equivalents. 8810 8811 Enabled at levels '-O', '-O2', '-O3', '-Os', but not with '-Og'. 8812 8813'-fdeclone-ctor-dtor' 8814 The C++ ABI requires multiple entry points for constructors and 8815 destructors: one for a base subobject, one for a complete object, 8816 and one for a virtual destructor that calls operator delete 8817 afterwards. For a hierarchy with virtual bases, the base and 8818 complete variants are clones, which means two copies of the 8819 function. With this option, the base and complete variants are 8820 changed to be thunks that call a common implementation. 8821 8822 Enabled by '-Os'. 8823 8824'-fdelete-null-pointer-checks' 8825 Assume that programs cannot safely dereference null pointers, and 8826 that no code or data element resides at address zero. This option 8827 enables simple constant folding optimizations at all optimization 8828 levels. In addition, other optimization passes in GCC use this 8829 flag to control global dataflow analyses that eliminate useless 8830 checks for null pointers; these assume that a memory access to 8831 address zero always results in a trap, so that if a pointer is 8832 checked after it has already been dereferenced, it cannot be null. 8833 8834 Note however that in some environments this assumption is not true. 8835 Use '-fno-delete-null-pointer-checks' to disable this optimization 8836 for programs that depend on that behavior. 8837 8838 This option is enabled by default on most targets. On Nios II ELF, 8839 it defaults to off. On AVR, CR16, and MSP430, this option is 8840 completely disabled. 8841 8842 Passes that use the dataflow information are enabled independently 8843 at different optimization levels. 8844 8845'-fdevirtualize' 8846 Attempt to convert calls to virtual functions to direct calls. 8847 This is done both within a procedure and interprocedurally as part 8848 of indirect inlining ('-findirect-inlining') and interprocedural 8849 constant propagation ('-fipa-cp'). Enabled at levels '-O2', '-O3', 8850 '-Os'. 8851 8852'-fdevirtualize-speculatively' 8853 Attempt to convert calls to virtual functions to speculative direct 8854 calls. Based on the analysis of the type inheritance graph, 8855 determine for a given call the set of likely targets. If the set 8856 is small, preferably of size 1, change the call into a conditional 8857 deciding between direct and indirect calls. The speculative calls 8858 enable more optimizations, such as inlining. When they seem 8859 useless after further optimization, they are converted back into 8860 original form. 8861 8862'-fdevirtualize-at-ltrans' 8863 Stream extra information needed for aggressive devirtualization 8864 when running the link-time optimizer in local transformation mode. 8865 This option enables more devirtualization but significantly 8866 increases the size of streamed data. For this reason it is 8867 disabled by default. 8868 8869'-fexpensive-optimizations' 8870 Perform a number of minor optimizations that are relatively 8871 expensive. 8872 8873 Enabled at levels '-O2', '-O3', '-Os'. 8874 8875'-free' 8876 Attempt to remove redundant extension instructions. This is 8877 especially helpful for the x86-64 architecture, which implicitly 8878 zero-extends in 64-bit registers after writing to their lower 8879 32-bit half. 8880 8881 Enabled for Alpha, AArch64 and x86 at levels '-O2', '-O3', '-Os'. 8882 8883'-fno-lifetime-dse' 8884 In C++ the value of an object is only affected by changes within 8885 its lifetime: when the constructor begins, the object has an 8886 indeterminate value, and any changes during the lifetime of the 8887 object are dead when the object is destroyed. Normally dead store 8888 elimination will take advantage of this; if your code relies on the 8889 value of the object storage persisting beyond the lifetime of the 8890 object, you can use this flag to disable this optimization. To 8891 preserve stores before the constructor starts (e.g. because your 8892 operator new clears the object storage) but still treat the object 8893 as dead after the destructor, you can use '-flifetime-dse=1'. The 8894 default behavior can be explicitly selected with 8895 '-flifetime-dse=2'. '-flifetime-dse=0' is equivalent to 8896 '-fno-lifetime-dse'. 8897 8898'-flive-range-shrinkage' 8899 Attempt to decrease register pressure through register live range 8900 shrinkage. This is helpful for fast processors with small or 8901 moderate size register sets. 8902 8903'-fira-algorithm=ALGORITHM' 8904 Use the specified coloring algorithm for the integrated register 8905 allocator. The ALGORITHM argument can be 'priority', which 8906 specifies Chow's priority coloring, or 'CB', which specifies 8907 Chaitin-Briggs coloring. Chaitin-Briggs coloring is not 8908 implemented for all architectures, but for those targets that do 8909 support it, it is the default because it generates better code. 8910 8911'-fira-region=REGION' 8912 Use specified regions for the integrated register allocator. The 8913 REGION argument should be one of the following: 8914 8915 'all' 8916 Use all loops as register allocation regions. This can give 8917 the best results for machines with a small and/or irregular 8918 register set. 8919 8920 'mixed' 8921 Use all loops except for loops with small register pressure as 8922 the regions. This value usually gives the best results in 8923 most cases and for most architectures, and is enabled by 8924 default when compiling with optimization for speed ('-O', 8925 '-O2', ...). 8926 8927 'one' 8928 Use all functions as a single region. This typically results 8929 in the smallest code size, and is enabled by default for '-Os' 8930 or '-O0'. 8931 8932'-fira-hoist-pressure' 8933 Use IRA to evaluate register pressure in the code hoisting pass for 8934 decisions to hoist expressions. This option usually results in 8935 smaller code, but it can slow the compiler down. 8936 8937 This option is enabled at level '-Os' for all targets. 8938 8939'-fira-loop-pressure' 8940 Use IRA to evaluate register pressure in loops for decisions to 8941 move loop invariants. This option usually results in generation of 8942 faster and smaller code on machines with large register files (>= 8943 32 registers), but it can slow the compiler down. 8944 8945 This option is enabled at level '-O3' for some targets. 8946 8947'-fno-ira-share-save-slots' 8948 Disable sharing of stack slots used for saving call-used hard 8949 registers living through a call. Each hard register gets a 8950 separate stack slot, and as a result function stack frames are 8951 larger. 8952 8953'-fno-ira-share-spill-slots' 8954 Disable sharing of stack slots allocated for pseudo-registers. 8955 Each pseudo-register that does not get a hard register gets a 8956 separate stack slot, and as a result function stack frames are 8957 larger. 8958 8959'-flra-remat' 8960 Enable CFG-sensitive rematerialization in LRA. Instead of loading 8961 values of spilled pseudos, LRA tries to rematerialize (recalculate) 8962 values if it is profitable. 8963 8964 Enabled at levels '-O2', '-O3', '-Os'. 8965 8966'-fdelayed-branch' 8967 If supported for the target machine, attempt to reorder 8968 instructions to exploit instruction slots available after delayed 8969 branch instructions. 8970 8971 Enabled at levels '-O', '-O2', '-O3', '-Os', but not at '-Og'. 8972 8973'-fschedule-insns' 8974 If supported for the target machine, attempt to reorder 8975 instructions to eliminate execution stalls due to required data 8976 being unavailable. This helps machines that have slow floating 8977 point or memory load instructions by allowing other instructions to 8978 be issued until the result of the load or floating-point 8979 instruction is required. 8980 8981 Enabled at levels '-O2', '-O3'. 8982 8983'-fschedule-insns2' 8984 Similar to '-fschedule-insns', but requests an additional pass of 8985 instruction scheduling after register allocation has been done. 8986 This is especially useful on machines with a relatively small 8987 number of registers and where memory load instructions take more 8988 than one cycle. 8989 8990 Enabled at levels '-O2', '-O3', '-Os'. 8991 8992'-fno-sched-interblock' 8993 Disable instruction scheduling across basic blocks, which is 8994 normally enabled when scheduling before register allocation, i.e. 8995 with '-fschedule-insns' or at '-O2' or higher. 8996 8997'-fno-sched-spec' 8998 Disable speculative motion of non-load instructions, which is 8999 normally enabled when scheduling before register allocation, i.e. 9000 with '-fschedule-insns' or at '-O2' or higher. 9001 9002'-fsched-pressure' 9003 Enable register pressure sensitive insn scheduling before register 9004 allocation. This only makes sense when scheduling before register 9005 allocation is enabled, i.e. with '-fschedule-insns' or at '-O2' or 9006 higher. Usage of this option can improve the generated code and 9007 decrease its size by preventing register pressure increase above 9008 the number of available hard registers and subsequent spills in 9009 register allocation. 9010 9011'-fsched-spec-load' 9012 Allow speculative motion of some load instructions. This only 9013 makes sense when scheduling before register allocation, i.e. with 9014 '-fschedule-insns' or at '-O2' or higher. 9015 9016'-fsched-spec-load-dangerous' 9017 Allow speculative motion of more load instructions. This only 9018 makes sense when scheduling before register allocation, i.e. with 9019 '-fschedule-insns' or at '-O2' or higher. 9020 9021'-fsched-stalled-insns' 9022'-fsched-stalled-insns=N' 9023 Define how many insns (if any) can be moved prematurely from the 9024 queue of stalled insns into the ready list during the second 9025 scheduling pass. '-fno-sched-stalled-insns' means that no insns 9026 are moved prematurely, '-fsched-stalled-insns=0' means there is no 9027 limit on how many queued insns can be moved prematurely. 9028 '-fsched-stalled-insns' without a value is equivalent to 9029 '-fsched-stalled-insns=1'. 9030 9031'-fsched-stalled-insns-dep' 9032'-fsched-stalled-insns-dep=N' 9033 Define how many insn groups (cycles) are examined for a dependency 9034 on a stalled insn that is a candidate for premature removal from 9035 the queue of stalled insns. This has an effect only during the 9036 second scheduling pass, and only if '-fsched-stalled-insns' is 9037 used. '-fno-sched-stalled-insns-dep' is equivalent to 9038 '-fsched-stalled-insns-dep=0'. '-fsched-stalled-insns-dep' without 9039 a value is equivalent to '-fsched-stalled-insns-dep=1'. 9040 9041'-fsched2-use-superblocks' 9042 When scheduling after register allocation, use superblock 9043 scheduling. This allows motion across basic block boundaries, 9044 resulting in faster schedules. This option is experimental, as not 9045 all machine descriptions used by GCC model the CPU closely enough 9046 to avoid unreliable results from the algorithm. 9047 9048 This only makes sense when scheduling after register allocation, 9049 i.e. with '-fschedule-insns2' or at '-O2' or higher. 9050 9051'-fsched-group-heuristic' 9052 Enable the group heuristic in the scheduler. This heuristic favors 9053 the instruction that belongs to a schedule group. This is enabled 9054 by default when scheduling is enabled, i.e. with '-fschedule-insns' 9055 or '-fschedule-insns2' or at '-O2' or higher. 9056 9057'-fsched-critical-path-heuristic' 9058 Enable the critical-path heuristic in the scheduler. This 9059 heuristic favors instructions on the critical path. This is 9060 enabled by default when scheduling is enabled, i.e. with 9061 '-fschedule-insns' or '-fschedule-insns2' or at '-O2' or higher. 9062 9063'-fsched-spec-insn-heuristic' 9064 Enable the speculative instruction heuristic in the scheduler. 9065 This heuristic favors speculative instructions with greater 9066 dependency weakness. This is enabled by default when scheduling is 9067 enabled, i.e. with '-fschedule-insns' or '-fschedule-insns2' or at 9068 '-O2' or higher. 9069 9070'-fsched-rank-heuristic' 9071 Enable the rank heuristic in the scheduler. This heuristic favors 9072 the instruction belonging to a basic block with greater size or 9073 frequency. This is enabled by default when scheduling is enabled, 9074 i.e. with '-fschedule-insns' or '-fschedule-insns2' or at '-O2' or 9075 higher. 9076 9077'-fsched-last-insn-heuristic' 9078 Enable the last-instruction heuristic in the scheduler. This 9079 heuristic favors the instruction that is less dependent on the last 9080 instruction scheduled. This is enabled by default when scheduling 9081 is enabled, i.e. with '-fschedule-insns' or '-fschedule-insns2' or 9082 at '-O2' or higher. 9083 9084'-fsched-dep-count-heuristic' 9085 Enable the dependent-count heuristic in the scheduler. This 9086 heuristic favors the instruction that has more instructions 9087 depending on it. This is enabled by default when scheduling is 9088 enabled, i.e. with '-fschedule-insns' or '-fschedule-insns2' or at 9089 '-O2' or higher. 9090 9091'-freschedule-modulo-scheduled-loops' 9092 Modulo scheduling is performed before traditional scheduling. If a 9093 loop is modulo scheduled, later scheduling passes may change its 9094 schedule. Use this option to control that behavior. 9095 9096'-fselective-scheduling' 9097 Schedule instructions using selective scheduling algorithm. 9098 Selective scheduling runs instead of the first scheduler pass. 9099 9100'-fselective-scheduling2' 9101 Schedule instructions using selective scheduling algorithm. 9102 Selective scheduling runs instead of the second scheduler pass. 9103 9104'-fsel-sched-pipelining' 9105 Enable software pipelining of innermost loops during selective 9106 scheduling. This option has no effect unless one of 9107 '-fselective-scheduling' or '-fselective-scheduling2' is turned on. 9108 9109'-fsel-sched-pipelining-outer-loops' 9110 When pipelining loops during selective scheduling, also pipeline 9111 outer loops. This option has no effect unless 9112 '-fsel-sched-pipelining' is turned on. 9113 9114'-fsemantic-interposition' 9115 Some object formats, like ELF, allow interposing of symbols by the 9116 dynamic linker. This means that for symbols exported from the DSO, 9117 the compiler cannot perform interprocedural propagation, inlining 9118 and other optimizations in anticipation that the function or 9119 variable in question may change. While this feature is useful, for 9120 example, to rewrite memory allocation functions by a debugging 9121 implementation, it is expensive in the terms of code quality. With 9122 '-fno-semantic-interposition' the compiler assumes that if 9123 interposition happens for functions the overwriting function will 9124 have precisely the same semantics (and side effects). Similarly if 9125 interposition happens for variables, the constructor of the 9126 variable will be the same. The flag has no effect for functions 9127 explicitly declared inline (where it is never allowed for 9128 interposition to change semantics) and for symbols explicitly 9129 declared weak. 9130 9131'-fshrink-wrap' 9132 Emit function prologues only before parts of the function that need 9133 it, rather than at the top of the function. This flag is enabled 9134 by default at '-O' and higher. 9135 9136'-fshrink-wrap-separate' 9137 Shrink-wrap separate parts of the prologue and epilogue separately, 9138 so that those parts are only executed when needed. This option is 9139 on by default, but has no effect unless '-fshrink-wrap' is also 9140 turned on and the target supports this. 9141 9142'-fcaller-saves' 9143 Enable allocation of values to registers that are clobbered by 9144 function calls, by emitting extra instructions to save and restore 9145 the registers around such calls. Such allocation is done only when 9146 it seems to result in better code. 9147 9148 This option is always enabled by default on certain machines, 9149 usually those which have no call-preserved registers to use 9150 instead. 9151 9152 Enabled at levels '-O2', '-O3', '-Os'. 9153 9154'-fcombine-stack-adjustments' 9155 Tracks stack adjustments (pushes and pops) and stack memory 9156 references and then tries to find ways to combine them. 9157 9158 Enabled by default at '-O1' and higher. 9159 9160'-fipa-ra' 9161 Use caller save registers for allocation if those registers are not 9162 used by any called function. In that case it is not necessary to 9163 save and restore them around calls. This is only possible if 9164 called functions are part of same compilation unit as current 9165 function and they are compiled before it. 9166 9167 Enabled at levels '-O2', '-O3', '-Os', however the option is 9168 disabled if generated code will be instrumented for profiling 9169 ('-p', or '-pg') or if callee's register usage cannot be known 9170 exactly (this happens on targets that do not expose prologues and 9171 epilogues in RTL). 9172 9173'-fconserve-stack' 9174 Attempt to minimize stack usage. The compiler attempts to use less 9175 stack space, even if that makes the program slower. This option 9176 implies setting the 'large-stack-frame' parameter to 100 and the 9177 'large-stack-frame-growth' parameter to 400. 9178 9179'-ftree-reassoc' 9180 Perform reassociation on trees. This flag is enabled by default at 9181 '-O' and higher. 9182 9183'-fcode-hoisting' 9184 Perform code hoisting. Code hoisting tries to move the evaluation 9185 of expressions executed on all paths to the function exit as early 9186 as possible. This is especially useful as a code size 9187 optimization, but it often helps for code speed as well. This flag 9188 is enabled by default at '-O2' and higher. 9189 9190'-ftree-pre' 9191 Perform partial redundancy elimination (PRE) on trees. This flag 9192 is enabled by default at '-O2' and '-O3'. 9193 9194'-ftree-partial-pre' 9195 Make partial redundancy elimination (PRE) more aggressive. This 9196 flag is enabled by default at '-O3'. 9197 9198'-ftree-forwprop' 9199 Perform forward propagation on trees. This flag is enabled by 9200 default at '-O' and higher. 9201 9202'-ftree-fre' 9203 Perform full redundancy elimination (FRE) on trees. The difference 9204 between FRE and PRE is that FRE only considers expressions that are 9205 computed on all paths leading to the redundant computation. This 9206 analysis is faster than PRE, though it exposes fewer redundancies. 9207 This flag is enabled by default at '-O' and higher. 9208 9209'-ftree-phiprop' 9210 Perform hoisting of loads from conditional pointers on trees. This 9211 pass is enabled by default at '-O' and higher. 9212 9213'-fhoist-adjacent-loads' 9214 Speculatively hoist loads from both branches of an if-then-else if 9215 the loads are from adjacent locations in the same structure and the 9216 target architecture has a conditional move instruction. This flag 9217 is enabled by default at '-O2' and higher. 9218 9219'-ftree-copy-prop' 9220 Perform copy propagation on trees. This pass eliminates 9221 unnecessary copy operations. This flag is enabled by default at 9222 '-O' and higher. 9223 9224'-fipa-pure-const' 9225 Discover which functions are pure or constant. Enabled by default 9226 at '-O' and higher. 9227 9228'-fipa-reference' 9229 Discover which static variables do not escape the compilation unit. 9230 Enabled by default at '-O' and higher. 9231 9232'-fipa-reference-addressable' 9233 Discover read-only, write-only and non-addressable static 9234 variables. Enabled by default at '-O' and higher. 9235 9236'-fipa-stack-alignment' 9237 Reduce stack alignment on call sites if possible. Enabled by 9238 default. 9239 9240'-fipa-pta' 9241 Perform interprocedural pointer analysis and interprocedural 9242 modification and reference analysis. This option can cause 9243 excessive memory and compile-time usage on large compilation units. 9244 It is not enabled by default at any optimization level. 9245 9246'-fipa-profile' 9247 Perform interprocedural profile propagation. The functions called 9248 only from cold functions are marked as cold. Also functions 9249 executed once (such as 'cold', 'noreturn', static constructors or 9250 destructors) are identified. Cold functions and loop less parts of 9251 functions executed once are then optimized for size. Enabled by 9252 default at '-O' and higher. 9253 9254'-fipa-cp' 9255 Perform interprocedural constant propagation. This optimization 9256 analyzes the program to determine when values passed to functions 9257 are constants and then optimizes accordingly. This optimization 9258 can substantially increase performance if the application has 9259 constants passed to functions. This flag is enabled by default at 9260 '-O2', '-Os' and '-O3'. It is also enabled by '-fprofile-use' and 9261 '-fauto-profile'. 9262 9263'-fipa-cp-clone' 9264 Perform function cloning to make interprocedural constant 9265 propagation stronger. When enabled, interprocedural constant 9266 propagation performs function cloning when externally visible 9267 function can be called with constant arguments. Because this 9268 optimization can create multiple copies of functions, it may 9269 significantly increase code size (see '--param 9270 ipa-cp-unit-growth=VALUE'). This flag is enabled by default at 9271 '-O3'. It is also enabled by '-fprofile-use' and '-fauto-profile'. 9272 9273'-fipa-bit-cp' 9274 When enabled, perform interprocedural bitwise constant propagation. 9275 This flag is enabled by default at '-O2' and by '-fprofile-use' and 9276 '-fauto-profile'. It requires that '-fipa-cp' is enabled. 9277 9278'-fipa-vrp' 9279 When enabled, perform interprocedural propagation of value ranges. 9280 This flag is enabled by default at '-O2'. It requires that 9281 '-fipa-cp' is enabled. 9282 9283'-fipa-icf' 9284 Perform Identical Code Folding for functions and read-only 9285 variables. The optimization reduces code size and may disturb 9286 unwind stacks by replacing a function by equivalent one with a 9287 different name. The optimization works more effectively with 9288 link-time optimization enabled. 9289 9290 Although the behavior is similar to the Gold Linker's ICF 9291 optimization, GCC ICF works on different levels and thus the 9292 optimizations are not same - there are equivalences that are found 9293 only by GCC and equivalences found only by Gold. 9294 9295 This flag is enabled by default at '-O2' and '-Os'. 9296 9297'-flive-patching=LEVEL' 9298 Control GCC's optimizations to produce output suitable for 9299 live-patching. 9300 9301 If the compiler's optimization uses a function's body or 9302 information extracted from its body to optimize/change another 9303 function, the latter is called an impacted function of the former. 9304 If a function is patched, its impacted functions should be patched 9305 too. 9306 9307 The impacted functions are determined by the compiler's 9308 interprocedural optimizations. For example, a caller is impacted 9309 when inlining a function into its caller, cloning a function and 9310 changing its caller to call this new clone, or extracting a 9311 function's pureness/constness information to optimize its direct or 9312 indirect callers, etc. 9313 9314 Usually, the more IPA optimizations enabled, the larger the number 9315 of impacted functions for each function. In order to control the 9316 number of impacted functions and more easily compute the list of 9317 impacted function, IPA optimizations can be partially enabled at 9318 two different levels. 9319 9320 The LEVEL argument should be one of the following: 9321 9322 'inline-clone' 9323 9324 Only enable inlining and cloning optimizations, which includes 9325 inlining, cloning, interprocedural scalar replacement of 9326 aggregates and partial inlining. As a result, when patching a 9327 function, all its callers and its clones' callers are 9328 impacted, therefore need to be patched as well. 9329 9330 '-flive-patching=inline-clone' disables the following 9331 optimization flags: 9332 -fwhole-program -fipa-pta -fipa-reference -fipa-ra 9333 -fipa-icf -fipa-icf-functions -fipa-icf-variables 9334 -fipa-bit-cp -fipa-vrp -fipa-pure-const -fipa-reference-addressable 9335 -fipa-stack-alignment 9336 9337 'inline-only-static' 9338 9339 Only enable inlining of static functions. As a result, when 9340 patching a static function, all its callers are impacted and 9341 so need to be patched as well. 9342 9343 In addition to all the flags that 9344 '-flive-patching=inline-clone' disables, 9345 '-flive-patching=inline-only-static' disables the following 9346 additional optimization flags: 9347 -fipa-cp-clone -fipa-sra -fpartial-inlining -fipa-cp 9348 9349 When '-flive-patching' is specified without any value, the default 9350 value is INLINE-CLONE. 9351 9352 This flag is disabled by default. 9353 9354 Note that '-flive-patching' is not supported with link-time 9355 optimization ('-flto'). 9356 9357'-fisolate-erroneous-paths-dereference' 9358 Detect paths that trigger erroneous or undefined behavior due to 9359 dereferencing a null pointer. Isolate those paths from the main 9360 control flow and turn the statement with erroneous or undefined 9361 behavior into a trap. This flag is enabled by default at '-O2' and 9362 higher and depends on '-fdelete-null-pointer-checks' also being 9363 enabled. 9364 9365'-fisolate-erroneous-paths-attribute' 9366 Detect paths that trigger erroneous or undefined behavior due to a 9367 null value being used in a way forbidden by a 'returns_nonnull' or 9368 'nonnull' attribute. Isolate those paths from the main control 9369 flow and turn the statement with erroneous or undefined behavior 9370 into a trap. This is not currently enabled, but may be enabled by 9371 '-O2' in the future. 9372 9373'-ftree-sink' 9374 Perform forward store motion on trees. This flag is enabled by 9375 default at '-O' and higher. 9376 9377'-ftree-bit-ccp' 9378 Perform sparse conditional bit constant propagation on trees and 9379 propagate pointer alignment information. This pass only operates 9380 on local scalar variables and is enabled by default at '-O1' and 9381 higher, except for '-Og'. It requires that '-ftree-ccp' is 9382 enabled. 9383 9384'-ftree-ccp' 9385 Perform sparse conditional constant propagation (CCP) on trees. 9386 This pass only operates on local scalar variables and is enabled by 9387 default at '-O' and higher. 9388 9389'-fssa-backprop' 9390 Propagate information about uses of a value up the definition chain 9391 in order to simplify the definitions. For example, this pass 9392 strips sign operations if the sign of a value never matters. The 9393 flag is enabled by default at '-O' and higher. 9394 9395'-fssa-phiopt' 9396 Perform pattern matching on SSA PHI nodes to optimize conditional 9397 code. This pass is enabled by default at '-O1' and higher, except 9398 for '-Og'. 9399 9400'-ftree-switch-conversion' 9401 Perform conversion of simple initializations in a switch to 9402 initializations from a scalar array. This flag is enabled by 9403 default at '-O2' and higher. 9404 9405'-ftree-tail-merge' 9406 Look for identical code sequences. When found, replace one with a 9407 jump to the other. This optimization is known as tail merging or 9408 cross jumping. This flag is enabled by default at '-O2' and 9409 higher. The compilation time in this pass can be limited using 9410 'max-tail-merge-comparisons' parameter and 9411 'max-tail-merge-iterations' parameter. 9412 9413'-ftree-dce' 9414 Perform dead code elimination (DCE) on trees. This flag is enabled 9415 by default at '-O' and higher. 9416 9417'-ftree-builtin-call-dce' 9418 Perform conditional dead code elimination (DCE) for calls to 9419 built-in functions that may set 'errno' but are otherwise free of 9420 side effects. This flag is enabled by default at '-O2' and higher 9421 if '-Os' is not also specified. 9422 9423'-ffinite-loops' 9424 Assume that a loop with an exit will eventually take the exit and 9425 not loop indefinitely. This allows the compiler to remove loops 9426 that otherwise have no side-effects, not considering eventual 9427 endless looping as such. 9428 9429 This option is enabled by default at '-O2' for C++ with -std=c++11 9430 or higher. 9431 9432'-ftree-dominator-opts' 9433 Perform a variety of simple scalar cleanups (constant/copy 9434 propagation, redundancy elimination, range propagation and 9435 expression simplification) based on a dominator tree traversal. 9436 This also performs jump threading (to reduce jumps to jumps). This 9437 flag is enabled by default at '-O' and higher. 9438 9439'-ftree-dse' 9440 Perform dead store elimination (DSE) on trees. A dead store is a 9441 store into a memory location that is later overwritten by another 9442 store without any intervening loads. In this case the earlier 9443 store can be deleted. This flag is enabled by default at '-O' and 9444 higher. 9445 9446'-ftree-ch' 9447 Perform loop header copying on trees. This is beneficial since it 9448 increases effectiveness of code motion optimizations. It also 9449 saves one jump. This flag is enabled by default at '-O' and 9450 higher. It is not enabled for '-Os', since it usually increases 9451 code size. 9452 9453'-ftree-loop-optimize' 9454 Perform loop optimizations on trees. This flag is enabled by 9455 default at '-O' and higher. 9456 9457'-ftree-loop-linear' 9458'-floop-strip-mine' 9459'-floop-block' 9460 Perform loop nest optimizations. Same as '-floop-nest-optimize'. 9461 To use this code transformation, GCC has to be configured with 9462 '--with-isl' to enable the Graphite loop transformation 9463 infrastructure. 9464 9465'-fgraphite-identity' 9466 Enable the identity transformation for graphite. For every SCoP we 9467 generate the polyhedral representation and transform it back to 9468 gimple. Using '-fgraphite-identity' we can check the costs or 9469 benefits of the GIMPLE -> GRAPHITE -> GIMPLE transformation. Some 9470 minimal optimizations are also performed by the code generator isl, 9471 like index splitting and dead code elimination in loops. 9472 9473'-floop-nest-optimize' 9474 Enable the isl based loop nest optimizer. This is a generic loop 9475 nest optimizer based on the Pluto optimization algorithms. It 9476 calculates a loop structure optimized for data-locality and 9477 parallelism. This option is experimental. 9478 9479'-floop-parallelize-all' 9480 Use the Graphite data dependence analysis to identify loops that 9481 can be parallelized. Parallelize all the loops that can be 9482 analyzed to not contain loop carried dependences without checking 9483 that it is profitable to parallelize the loops. 9484 9485'-ftree-coalesce-vars' 9486 While transforming the program out of the SSA representation, 9487 attempt to reduce copying by coalescing versions of different 9488 user-defined variables, instead of just compiler temporaries. This 9489 may severely limit the ability to debug an optimized program 9490 compiled with '-fno-var-tracking-assignments'. In the negated 9491 form, this flag prevents SSA coalescing of user variables. This 9492 option is enabled by default if optimization is enabled, and it 9493 does very little otherwise. 9494 9495'-ftree-loop-if-convert' 9496 Attempt to transform conditional jumps in the innermost loops to 9497 branch-less equivalents. The intent is to remove control-flow from 9498 the innermost loops in order to improve the ability of the 9499 vectorization pass to handle these loops. This is enabled by 9500 default if vectorization is enabled. 9501 9502'-ftree-loop-distribution' 9503 Perform loop distribution. This flag can improve cache performance 9504 on big loop bodies and allow further loop optimizations, like 9505 parallelization or vectorization, to take place. For example, the 9506 loop 9507 DO I = 1, N 9508 A(I) = B(I) + C 9509 D(I) = E(I) * F 9510 ENDDO 9511 is transformed to 9512 DO I = 1, N 9513 A(I) = B(I) + C 9514 ENDDO 9515 DO I = 1, N 9516 D(I) = E(I) * F 9517 ENDDO 9518 This flag is enabled by default at '-O3'. It is also enabled by 9519 '-fprofile-use' and '-fauto-profile'. 9520 9521'-ftree-loop-distribute-patterns' 9522 Perform loop distribution of patterns that can be code generated 9523 with calls to a library. This flag is enabled by default at '-O2' 9524 and higher, and by '-fprofile-use' and '-fauto-profile'. 9525 9526 This pass distributes the initialization loops and generates a call 9527 to memset zero. For example, the loop 9528 DO I = 1, N 9529 A(I) = 0 9530 B(I) = A(I) + I 9531 ENDDO 9532 is transformed to 9533 DO I = 1, N 9534 A(I) = 0 9535 ENDDO 9536 DO I = 1, N 9537 B(I) = A(I) + I 9538 ENDDO 9539 and the initialization loop is transformed into a call to memset 9540 zero. This flag is enabled by default at '-O3'. It is also 9541 enabled by '-fprofile-use' and '-fauto-profile'. 9542 9543'-floop-interchange' 9544 Perform loop interchange outside of graphite. This flag can 9545 improve cache performance on loop nest and allow further loop 9546 optimizations, like vectorization, to take place. For example, the 9547 loop 9548 for (int i = 0; i < N; i++) 9549 for (int j = 0; j < N; j++) 9550 for (int k = 0; k < N; k++) 9551 c[i][j] = c[i][j] + a[i][k]*b[k][j]; 9552 is transformed to 9553 for (int i = 0; i < N; i++) 9554 for (int k = 0; k < N; k++) 9555 for (int j = 0; j < N; j++) 9556 c[i][j] = c[i][j] + a[i][k]*b[k][j]; 9557 This flag is enabled by default at '-O3'. It is also enabled by 9558 '-fprofile-use' and '-fauto-profile'. 9559 9560'-floop-unroll-and-jam' 9561 Apply unroll and jam transformations on feasible loops. In a loop 9562 nest this unrolls the outer loop by some factor and fuses the 9563 resulting multiple inner loops. This flag is enabled by default at 9564 '-O3'. It is also enabled by '-fprofile-use' and '-fauto-profile'. 9565 9566'-ftree-loop-im' 9567 Perform loop invariant motion on trees. This pass moves only 9568 invariants that are hard to handle at RTL level (function calls, 9569 operations that expand to nontrivial sequences of insns). With 9570 '-funswitch-loops' it also moves operands of conditions that are 9571 invariant out of the loop, so that we can use just trivial 9572 invariantness analysis in loop unswitching. The pass also includes 9573 store motion. 9574 9575'-ftree-loop-ivcanon' 9576 Create a canonical counter for number of iterations in loops for 9577 which determining number of iterations requires complicated 9578 analysis. Later optimizations then may determine the number 9579 easily. Useful especially in connection with unrolling. 9580 9581'-ftree-scev-cprop' 9582 Perform final value replacement. If a variable is modified in a 9583 loop in such a way that its value when exiting the loop can be 9584 determined using only its initial value and the number of loop 9585 iterations, replace uses of the final value by such a computation, 9586 provided it is sufficiently cheap. This reduces data dependencies 9587 and may allow further simplifications. Enabled by default at '-O' 9588 and higher. 9589 9590'-fivopts' 9591 Perform induction variable optimizations (strength reduction, 9592 induction variable merging and induction variable elimination) on 9593 trees. 9594 9595'-ftree-parallelize-loops=n' 9596 Parallelize loops, i.e., split their iteration space to run in n 9597 threads. This is only possible for loops whose iterations are 9598 independent and can be arbitrarily reordered. The optimization is 9599 only profitable on multiprocessor machines, for loops that are 9600 CPU-intensive, rather than constrained e.g. by memory bandwidth. 9601 This option implies '-pthread', and thus is only supported on 9602 targets that have support for '-pthread'. 9603 9604'-ftree-pta' 9605 Perform function-local points-to analysis on trees. This flag is 9606 enabled by default at '-O1' and higher, except for '-Og'. 9607 9608'-ftree-sra' 9609 Perform scalar replacement of aggregates. This pass replaces 9610 structure references with scalars to prevent committing structures 9611 to memory too early. This flag is enabled by default at '-O1' and 9612 higher, except for '-Og'. 9613 9614'-fstore-merging' 9615 Perform merging of narrow stores to consecutive memory addresses. 9616 This pass merges contiguous stores of immediate values narrower 9617 than a word into fewer wider stores to reduce the number of 9618 instructions. This is enabled by default at '-O2' and higher as 9619 well as '-Os'. 9620 9621'-ftree-ter' 9622 Perform temporary expression replacement during the SSA->normal 9623 phase. Single use/single def temporaries are replaced at their use 9624 location with their defining expression. This results in 9625 non-GIMPLE code, but gives the expanders much more complex trees to 9626 work on resulting in better RTL generation. This is enabled by 9627 default at '-O' and higher. 9628 9629'-ftree-slsr' 9630 Perform straight-line strength reduction on trees. This recognizes 9631 related expressions involving multiplications and replaces them by 9632 less expensive calculations when possible. This is enabled by 9633 default at '-O' and higher. 9634 9635'-ftree-vectorize' 9636 Perform vectorization on trees. This flag enables 9637 '-ftree-loop-vectorize' and '-ftree-slp-vectorize' if not 9638 explicitly specified. 9639 9640'-ftree-loop-vectorize' 9641 Perform loop vectorization on trees. This flag is enabled by 9642 default at '-O3' and by '-ftree-vectorize', '-fprofile-use', and 9643 '-fauto-profile'. 9644 9645'-ftree-slp-vectorize' 9646 Perform basic block vectorization on trees. This flag is enabled 9647 by default at '-O3' and by '-ftree-vectorize', '-fprofile-use', and 9648 '-fauto-profile'. 9649 9650'-fvect-cost-model=MODEL' 9651 Alter the cost model used for vectorization. The MODEL argument 9652 should be one of 'unlimited', 'dynamic' or 'cheap'. With the 9653 'unlimited' model the vectorized code-path is assumed to be 9654 profitable while with the 'dynamic' model a runtime check guards 9655 the vectorized code-path to enable it only for iteration counts 9656 that will likely execute faster than when executing the original 9657 scalar loop. The 'cheap' model disables vectorization of loops 9658 where doing so would be cost prohibitive for example due to 9659 required runtime checks for data dependence or alignment but 9660 otherwise is equal to the 'dynamic' model. The default cost model 9661 depends on other optimization flags and is either 'dynamic' or 9662 'cheap'. 9663 9664'-fsimd-cost-model=MODEL' 9665 Alter the cost model used for vectorization of loops marked with 9666 the OpenMP simd directive. The MODEL argument should be one of 9667 'unlimited', 'dynamic', 'cheap'. All values of MODEL have the same 9668 meaning as described in '-fvect-cost-model' and by default a cost 9669 model defined with '-fvect-cost-model' is used. 9670 9671'-ftree-vrp' 9672 Perform Value Range Propagation on trees. This is similar to the 9673 constant propagation pass, but instead of values, ranges of values 9674 are propagated. This allows the optimizers to remove unnecessary 9675 range checks like array bound checks and null pointer checks. This 9676 is enabled by default at '-O2' and higher. Null pointer check 9677 elimination is only done if '-fdelete-null-pointer-checks' is 9678 enabled. 9679 9680'-fsplit-paths' 9681 Split paths leading to loop backedges. This can improve dead code 9682 elimination and common subexpression elimination. This is enabled 9683 by default at '-O3' and above. 9684 9685'-fsplit-ivs-in-unroller' 9686 Enables expression of values of induction variables in later 9687 iterations of the unrolled loop using the value in the first 9688 iteration. This breaks long dependency chains, thus improving 9689 efficiency of the scheduling passes. 9690 9691 A combination of '-fweb' and CSE is often sufficient to obtain the 9692 same effect. However, that is not reliable in cases where the loop 9693 body is more complicated than a single basic block. It also does 9694 not work at all on some architectures due to restrictions in the 9695 CSE pass. 9696 9697 This optimization is enabled by default. 9698 9699'-fvariable-expansion-in-unroller' 9700 With this option, the compiler creates multiple copies of some 9701 local variables when unrolling a loop, which can result in superior 9702 code. 9703 9704 This optimization is enabled by default for PowerPC targets, but 9705 disabled by default otherwise. 9706 9707'-fpartial-inlining' 9708 Inline parts of functions. This option has any effect only when 9709 inlining itself is turned on by the '-finline-functions' or 9710 '-finline-small-functions' options. 9711 9712 Enabled at levels '-O2', '-O3', '-Os'. 9713 9714'-fpredictive-commoning' 9715 Perform predictive commoning optimization, i.e., reusing 9716 computations (especially memory loads and stores) performed in 9717 previous iterations of loops. 9718 9719 This option is enabled at level '-O3'. It is also enabled by 9720 '-fprofile-use' and '-fauto-profile'. 9721 9722'-fprefetch-loop-arrays' 9723 If supported by the target machine, generate instructions to 9724 prefetch memory to improve the performance of loops that access 9725 large arrays. 9726 9727 This option may generate better or worse code; results are highly 9728 dependent on the structure of loops within the source code. 9729 9730 Disabled at level '-Os'. 9731 9732'-fno-printf-return-value' 9733 Do not substitute constants for known return value of formatted 9734 output functions such as 'sprintf', 'snprintf', 'vsprintf', and 9735 'vsnprintf' (but not 'printf' of 'fprintf'). This transformation 9736 allows GCC to optimize or even eliminate branches based on the 9737 known return value of these functions called with arguments that 9738 are either constant, or whose values are known to be in a range 9739 that makes determining the exact return value possible. For 9740 example, when '-fprintf-return-value' is in effect, both the branch 9741 and the body of the 'if' statement (but not the call to 'snprint') 9742 can be optimized away when 'i' is a 32-bit or smaller integer 9743 because the return value is guaranteed to be at most 8. 9744 9745 char buf[9]; 9746 if (snprintf (buf, "%08x", i) >= sizeof buf) 9747 ... 9748 9749 The '-fprintf-return-value' option relies on other optimizations 9750 and yields best results with '-O2' and above. It works in tandem 9751 with the '-Wformat-overflow' and '-Wformat-truncation' options. 9752 The '-fprintf-return-value' option is enabled by default. 9753 9754'-fno-peephole' 9755'-fno-peephole2' 9756 Disable any machine-specific peephole optimizations. The 9757 difference between '-fno-peephole' and '-fno-peephole2' is in how 9758 they are implemented in the compiler; some targets use one, some 9759 use the other, a few use both. 9760 9761 '-fpeephole' is enabled by default. '-fpeephole2' enabled at 9762 levels '-O2', '-O3', '-Os'. 9763 9764'-fno-guess-branch-probability' 9765 Do not guess branch probabilities using heuristics. 9766 9767 GCC uses heuristics to guess branch probabilities if they are not 9768 provided by profiling feedback ('-fprofile-arcs'). These 9769 heuristics are based on the control flow graph. If some branch 9770 probabilities are specified by '__builtin_expect', then the 9771 heuristics are used to guess branch probabilities for the rest of 9772 the control flow graph, taking the '__builtin_expect' info into 9773 account. The interactions between the heuristics and 9774 '__builtin_expect' can be complex, and in some cases, it may be 9775 useful to disable the heuristics so that the effects of 9776 '__builtin_expect' are easier to understand. 9777 9778 It is also possible to specify expected probability of the 9779 expression with '__builtin_expect_with_probability' built-in 9780 function. 9781 9782 The default is '-fguess-branch-probability' at levels '-O', '-O2', 9783 '-O3', '-Os'. 9784 9785'-freorder-blocks' 9786 Reorder basic blocks in the compiled function in order to reduce 9787 number of taken branches and improve code locality. 9788 9789 Enabled at levels '-O', '-O2', '-O3', '-Os'. 9790 9791'-freorder-blocks-algorithm=ALGORITHM' 9792 Use the specified algorithm for basic block reordering. The 9793 ALGORITHM argument can be 'simple', which does not increase code 9794 size (except sometimes due to secondary effects like alignment), or 9795 'stc', the "software trace cache" algorithm, which tries to put all 9796 often executed code together, minimizing the number of branches 9797 executed by making extra copies of code. 9798 9799 The default is 'simple' at levels '-O', '-Os', and 'stc' at levels 9800 '-O2', '-O3'. 9801 9802'-freorder-blocks-and-partition' 9803 In addition to reordering basic blocks in the compiled function, in 9804 order to reduce number of taken branches, partitions hot and cold 9805 basic blocks into separate sections of the assembly and '.o' files, 9806 to improve paging and cache locality performance. 9807 9808 This optimization is automatically turned off in the presence of 9809 exception handling or unwind tables (on targets using 9810 setjump/longjump or target specific scheme), for linkonce sections, 9811 for functions with a user-defined section attribute and on any 9812 architecture that does not support named sections. When 9813 '-fsplit-stack' is used this option is not enabled by default (to 9814 avoid linker errors), but may be enabled explicitly (if using a 9815 working linker). 9816 9817 Enabled for x86 at levels '-O2', '-O3', '-Os'. 9818 9819'-freorder-functions' 9820 Reorder functions in the object file in order to improve code 9821 locality. This is implemented by using special subsections 9822 '.text.hot' for most frequently executed functions and 9823 '.text.unlikely' for unlikely executed functions. Reordering is 9824 done by the linker so object file format must support named 9825 sections and linker must place them in a reasonable way. 9826 9827 This option isn't effective unless you either provide profile 9828 feedback (see '-fprofile-arcs' for details) or manually annotate 9829 functions with 'hot' or 'cold' attributes (*note Common Function 9830 Attributes::). 9831 9832 Enabled at levels '-O2', '-O3', '-Os'. 9833 9834'-fstrict-aliasing' 9835 Allow the compiler to assume the strictest aliasing rules 9836 applicable to the language being compiled. For C (and C++), this 9837 activates optimizations based on the type of expressions. In 9838 particular, an object of one type is assumed never to reside at the 9839 same address as an object of a different type, unless the types are 9840 almost the same. For example, an 'unsigned int' can alias an 9841 'int', but not a 'void*' or a 'double'. A character type may alias 9842 any other type. 9843 9844 Pay special attention to code like this: 9845 union a_union { 9846 int i; 9847 double d; 9848 }; 9849 9850 int f() { 9851 union a_union t; 9852 t.d = 3.0; 9853 return t.i; 9854 } 9855 The practice of reading from a different union member than the one 9856 most recently written to (called "type-punning") is common. Even 9857 with '-fstrict-aliasing', type-punning is allowed, provided the 9858 memory is accessed through the union type. So, the code above 9859 works as expected. *Note Structures unions enumerations and 9860 bit-fields implementation::. However, this code might not: 9861 int f() { 9862 union a_union t; 9863 int* ip; 9864 t.d = 3.0; 9865 ip = &t.i; 9866 return *ip; 9867 } 9868 9869 Similarly, access by taking the address, casting the resulting 9870 pointer and dereferencing the result has undefined behavior, even 9871 if the cast uses a union type, e.g.: 9872 int f() { 9873 double d = 3.0; 9874 return ((union a_union *) &d)->i; 9875 } 9876 9877 The '-fstrict-aliasing' option is enabled at levels '-O2', '-O3', 9878 '-Os'. 9879 9880'-falign-functions' 9881'-falign-functions=N' 9882'-falign-functions=N:M' 9883'-falign-functions=N:M:N2' 9884'-falign-functions=N:M:N2:M2' 9885 Align the start of functions to the next power-of-two greater than 9886 or equal to N, skipping up to M-1 bytes. This ensures that at 9887 least the first M bytes of the function can be fetched by the CPU 9888 without crossing an N-byte alignment boundary. 9889 9890 If M is not specified, it defaults to N. 9891 9892 Examples: '-falign-functions=32' aligns functions to the next 9893 32-byte boundary, '-falign-functions=24' aligns to the next 32-byte 9894 boundary only if this can be done by skipping 23 bytes or less, 9895 '-falign-functions=32:7' aligns to the next 32-byte boundary only 9896 if this can be done by skipping 6 bytes or less. 9897 9898 The second pair of N2:M2 values allows you to specify a secondary 9899 alignment: '-falign-functions=64:7:32:3' aligns to the next 64-byte 9900 boundary if this can be done by skipping 6 bytes or less, otherwise 9901 aligns to the next 32-byte boundary if this can be done by skipping 9902 2 bytes or less. If M2 is not specified, it defaults to N2. 9903 9904 Some assemblers only support this flag when N is a power of two; in 9905 that case, it is rounded up. 9906 9907 '-fno-align-functions' and '-falign-functions=1' are equivalent and 9908 mean that functions are not aligned. 9909 9910 If N is not specified or is zero, use a machine-dependent default. 9911 The maximum allowed N option value is 65536. 9912 9913 Enabled at levels '-O2', '-O3'. 9914 9915'-flimit-function-alignment' 9916 If this option is enabled, the compiler tries to avoid 9917 unnecessarily overaligning functions. It attempts to instruct the 9918 assembler to align by the amount specified by '-falign-functions', 9919 but not to skip more bytes than the size of the function. 9920 9921'-falign-labels' 9922'-falign-labels=N' 9923'-falign-labels=N:M' 9924'-falign-labels=N:M:N2' 9925'-falign-labels=N:M:N2:M2' 9926 Align all branch targets to a power-of-two boundary. 9927 9928 Parameters of this option are analogous to the '-falign-functions' 9929 option. '-fno-align-labels' and '-falign-labels=1' are equivalent 9930 and mean that labels are not aligned. 9931 9932 If '-falign-loops' or '-falign-jumps' are applicable and are 9933 greater than this value, then their values are used instead. 9934 9935 If N is not specified or is zero, use a machine-dependent default 9936 which is very likely to be '1', meaning no alignment. The maximum 9937 allowed N option value is 65536. 9938 9939 Enabled at levels '-O2', '-O3'. 9940 9941'-falign-loops' 9942'-falign-loops=N' 9943'-falign-loops=N:M' 9944'-falign-loops=N:M:N2' 9945'-falign-loops=N:M:N2:M2' 9946 Align loops to a power-of-two boundary. If the loops are executed 9947 many times, this makes up for any execution of the dummy padding 9948 instructions. 9949 9950 If '-falign-labels' is greater than this value, then its value is 9951 used instead. 9952 9953 Parameters of this option are analogous to the '-falign-functions' 9954 option. '-fno-align-loops' and '-falign-loops=1' are equivalent 9955 and mean that loops are not aligned. The maximum allowed N option 9956 value is 65536. 9957 9958 If N is not specified or is zero, use a machine-dependent default. 9959 9960 Enabled at levels '-O2', '-O3'. 9961 9962'-falign-jumps' 9963'-falign-jumps=N' 9964'-falign-jumps=N:M' 9965'-falign-jumps=N:M:N2' 9966'-falign-jumps=N:M:N2:M2' 9967 Align branch targets to a power-of-two boundary, for branch targets 9968 where the targets can only be reached by jumping. In this case, no 9969 dummy operations need be executed. 9970 9971 If '-falign-labels' is greater than this value, then its value is 9972 used instead. 9973 9974 Parameters of this option are analogous to the '-falign-functions' 9975 option. '-fno-align-jumps' and '-falign-jumps=1' are equivalent 9976 and mean that loops are not aligned. 9977 9978 If N is not specified or is zero, use a machine-dependent default. 9979 The maximum allowed N option value is 65536. 9980 9981 Enabled at levels '-O2', '-O3'. 9982 9983'-fno-allocation-dce' 9984 Do not remove unused C++ allocations in dead code elimination. 9985 9986'-fallow-store-data-races' 9987 Allow the compiler to introduce new data races on stores. 9988 9989 Enabled at level '-Ofast'. 9990 9991'-funit-at-a-time' 9992 This option is left for compatibility reasons. '-funit-at-a-time' 9993 has no effect, while '-fno-unit-at-a-time' implies 9994 '-fno-toplevel-reorder' and '-fno-section-anchors'. 9995 9996 Enabled by default. 9997 9998'-fno-toplevel-reorder' 9999 Do not reorder top-level functions, variables, and 'asm' 10000 statements. Output them in the same order that they appear in the 10001 input file. When this option is used, unreferenced static 10002 variables are not removed. This option is intended to support 10003 existing code that relies on a particular ordering. For new code, 10004 it is better to use attributes when possible. 10005 10006 '-ftoplevel-reorder' is the default at '-O1' and higher, and also 10007 at '-O0' if '-fsection-anchors' is explicitly requested. 10008 Additionally '-fno-toplevel-reorder' implies 10009 '-fno-section-anchors'. 10010 10011'-fweb' 10012 Constructs webs as commonly used for register allocation purposes 10013 and assign each web individual pseudo register. This allows the 10014 register allocation pass to operate on pseudos directly, but also 10015 strengthens several other optimization passes, such as CSE, loop 10016 optimizer and trivial dead code remover. It can, however, make 10017 debugging impossible, since variables no longer stay in a "home 10018 register". 10019 10020 Enabled by default with '-funroll-loops'. 10021 10022'-fwhole-program' 10023 Assume that the current compilation unit represents the whole 10024 program being compiled. All public functions and variables with 10025 the exception of 'main' and those merged by attribute 10026 'externally_visible' become static functions and in effect are 10027 optimized more aggressively by interprocedural optimizers. 10028 10029 This option should not be used in combination with '-flto'. 10030 Instead relying on a linker plugin should provide safer and more 10031 precise information. 10032 10033'-flto[=N]' 10034 This option runs the standard link-time optimizer. When invoked 10035 with source code, it generates GIMPLE (one of GCC's internal 10036 representations) and writes it to special ELF sections in the 10037 object file. When the object files are linked together, all the 10038 function bodies are read from these ELF sections and instantiated 10039 as if they had been part of the same translation unit. 10040 10041 To use the link-time optimizer, '-flto' and optimization options 10042 should be specified at compile time and during the final link. It 10043 is recommended that you compile all the files participating in the 10044 same link with the same options and also specify those options at 10045 link time. For example: 10046 10047 gcc -c -O2 -flto foo.c 10048 gcc -c -O2 -flto bar.c 10049 gcc -o myprog -flto -O2 foo.o bar.o 10050 10051 The first two invocations to GCC save a bytecode representation of 10052 GIMPLE into special ELF sections inside 'foo.o' and 'bar.o'. The 10053 final invocation reads the GIMPLE bytecode from 'foo.o' and 10054 'bar.o', merges the two files into a single internal image, and 10055 compiles the result as usual. Since both 'foo.o' and 'bar.o' are 10056 merged into a single image, this causes all the interprocedural 10057 analyses and optimizations in GCC to work across the two files as 10058 if they were a single one. This means, for example, that the 10059 inliner is able to inline functions in 'bar.o' into functions in 10060 'foo.o' and vice-versa. 10061 10062 Another (simpler) way to enable link-time optimization is: 10063 10064 gcc -o myprog -flto -O2 foo.c bar.c 10065 10066 The above generates bytecode for 'foo.c' and 'bar.c', merges them 10067 together into a single GIMPLE representation and optimizes them as 10068 usual to produce 'myprog'. 10069 10070 The important thing to keep in mind is that to enable link-time 10071 optimizations you need to use the GCC driver to perform the link 10072 step. GCC automatically performs link-time optimization if any of 10073 the objects involved were compiled with the '-flto' command-line 10074 option. You can always override the automatic decision to do 10075 link-time optimization by passing '-fno-lto' to the link command. 10076 10077 To make whole program optimization effective, it is necessary to 10078 make certain whole program assumptions. The compiler needs to know 10079 what functions and variables can be accessed by libraries and 10080 runtime outside of the link-time optimized unit. When supported by 10081 the linker, the linker plugin (see '-fuse-linker-plugin') passes 10082 information to the compiler about used and externally visible 10083 symbols. When the linker plugin is not available, 10084 '-fwhole-program' should be used to allow the compiler to make 10085 these assumptions, which leads to more aggressive optimization 10086 decisions. 10087 10088 When a file is compiled with '-flto' without '-fuse-linker-plugin', 10089 the generated object file is larger than a regular object file 10090 because it contains GIMPLE bytecodes and the usual final code (see 10091 '-ffat-lto-objects'. This means that object files with LTO 10092 information can be linked as normal object files; if '-fno-lto' is 10093 passed to the linker, no interprocedural optimizations are applied. 10094 Note that when '-fno-fat-lto-objects' is enabled the compile stage 10095 is faster but you cannot perform a regular, non-LTO link on them. 10096 10097 When producing the final binary, GCC only applies link-time 10098 optimizations to those files that contain bytecode. Therefore, you 10099 can mix and match object files and libraries with GIMPLE bytecodes 10100 and final object code. GCC automatically selects which files to 10101 optimize in LTO mode and which files to link without further 10102 processing. 10103 10104 Generally, options specified at link time override those specified 10105 at compile time, although in some cases GCC attempts to infer 10106 link-time options from the settings used to compile the input 10107 files. 10108 10109 If you do not specify an optimization level option '-O' at link 10110 time, then GCC uses the highest optimization level used when 10111 compiling the object files. Note that it is generally ineffective 10112 to specify an optimization level option only at link time and not 10113 at compile time, for two reasons. First, compiling without 10114 optimization suppresses compiler passes that gather information 10115 needed for effective optimization at link time. Second, some early 10116 optimization passes can be performed only at compile time and not 10117 at link time. 10118 10119 There are some code generation flags preserved by GCC when 10120 generating bytecodes, as they need to be used during the final 10121 link. Currently, the following options and their settings are 10122 taken from the first object file that explicitly specifies them: 10123 '-fPIC', '-fpic', '-fpie', '-fcommon', '-fexceptions', 10124 '-fnon-call-exceptions', '-fgnu-tm' and all the '-m' target flags. 10125 10126 Certain ABI-changing flags are required to match in all compilation 10127 units, and trying to override this at link time with a conflicting 10128 value is ignored. This includes options such as 10129 '-freg-struct-return' and '-fpcc-struct-return'. 10130 10131 Other options such as '-ffp-contract', '-fno-strict-overflow', 10132 '-fwrapv', '-fno-trapv' or '-fno-strict-aliasing' are passed 10133 through to the link stage and merged conservatively for conflicting 10134 translation units. Specifically '-fno-strict-overflow', '-fwrapv' 10135 and '-fno-trapv' take precedence; and for example 10136 '-ffp-contract=off' takes precedence over '-ffp-contract=fast'. 10137 You can override them at link time. 10138 10139 Diagnostic options such as '-Wstringop-overflow' are passed through 10140 to the link stage and their setting matches that of the 10141 compile-step at function granularity. Note that this matters only 10142 for diagnostics emitted during optimization. Note that code 10143 transforms such as inlining can lead to warnings being enabled or 10144 disabled for regions if code not consistent with the setting at 10145 compile time. 10146 10147 When you need to pass options to the assembler via '-Wa' or 10148 '-Xassembler' make sure to either compile such translation units 10149 with '-fno-lto' or consistently use the same assembler options on 10150 all translation units. You can alternatively also specify 10151 assembler options at LTO link time. 10152 10153 To enable debug info generation you need to supply '-g' at compile 10154 time. If any of the input files at link time were built with debug 10155 info generation enabled the link will enable debug info generation 10156 as well. Any elaborate debug info settings like the dwarf level 10157 '-gdwarf-5' need to be explicitly repeated at the linker command 10158 line and mixing different settings in different translation units 10159 is discouraged. 10160 10161 If LTO encounters objects with C linkage declared with incompatible 10162 types in separate translation units to be linked together 10163 (undefined behavior according to ISO C99 6.2.7), a non-fatal 10164 diagnostic may be issued. The behavior is still undefined at run 10165 time. Similar diagnostics may be raised for other languages. 10166 10167 Another feature of LTO is that it is possible to apply 10168 interprocedural optimizations on files written in different 10169 languages: 10170 10171 gcc -c -flto foo.c 10172 g++ -c -flto bar.cc 10173 gfortran -c -flto baz.f90 10174 g++ -o myprog -flto -O3 foo.o bar.o baz.o -lgfortran 10175 10176 Notice that the final link is done with 'g++' to get the C++ 10177 runtime libraries and '-lgfortran' is added to get the Fortran 10178 runtime libraries. In general, when mixing languages in LTO mode, 10179 you should use the same link command options as when mixing 10180 languages in a regular (non-LTO) compilation. 10181 10182 If object files containing GIMPLE bytecode are stored in a library 10183 archive, say 'libfoo.a', it is possible to extract and use them in 10184 an LTO link if you are using a linker with plugin support. To 10185 create static libraries suitable for LTO, use 'gcc-ar' and 10186 'gcc-ranlib' instead of 'ar' and 'ranlib'; to show the symbols of 10187 object files with GIMPLE bytecode, use 'gcc-nm'. Those commands 10188 require that 'ar', 'ranlib' and 'nm' have been compiled with plugin 10189 support. At link time, use the flag '-fuse-linker-plugin' to 10190 ensure that the library participates in the LTO optimization 10191 process: 10192 10193 gcc -o myprog -O2 -flto -fuse-linker-plugin a.o b.o -lfoo 10194 10195 With the linker plugin enabled, the linker extracts the needed 10196 GIMPLE files from 'libfoo.a' and passes them on to the running GCC 10197 to make them part of the aggregated GIMPLE image to be optimized. 10198 10199 If you are not using a linker with plugin support and/or do not 10200 enable the linker plugin, then the objects inside 'libfoo.a' are 10201 extracted and linked as usual, but they do not participate in the 10202 LTO optimization process. In order to make a static library 10203 suitable for both LTO optimization and usual linkage, compile its 10204 object files with '-flto' '-ffat-lto-objects'. 10205 10206 Link-time optimizations do not require the presence of the whole 10207 program to operate. If the program does not require any symbols to 10208 be exported, it is possible to combine '-flto' and 10209 '-fwhole-program' to allow the interprocedural optimizers to use 10210 more aggressive assumptions which may lead to improved optimization 10211 opportunities. Use of '-fwhole-program' is not needed when linker 10212 plugin is active (see '-fuse-linker-plugin'). 10213 10214 The current implementation of LTO makes no attempt to generate 10215 bytecode that is portable between different types of hosts. The 10216 bytecode files are versioned and there is a strict version check, 10217 so bytecode files generated in one version of GCC do not work with 10218 an older or newer version of GCC. 10219 10220 Link-time optimization does not work well with generation of 10221 debugging information on systems other than those using a 10222 combination of ELF and DWARF. 10223 10224 If you specify the optional N, the optimization and code generation 10225 done at link time is executed in parallel using N parallel jobs by 10226 utilizing an installed 'make' program. The environment variable 10227 'MAKE' may be used to override the program used. 10228 10229 You can also specify '-flto=jobserver' to use GNU make's job server 10230 mode to determine the number of parallel jobs. This is useful when 10231 the Makefile calling GCC is already executing in parallel. You 10232 must prepend a '+' to the command recipe in the parent Makefile for 10233 this to work. This option likely only works if 'MAKE' is GNU make. 10234 Even without the option value, GCC tries to automatically detect a 10235 running GNU make's job server. 10236 10237 Use '-flto=auto' to use GNU make's job server, if available, or 10238 otherwise fall back to autodetection of the number of CPU threads 10239 present in your system. 10240 10241'-flto-partition=ALG' 10242 Specify the partitioning algorithm used by the link-time optimizer. 10243 The value is either '1to1' to specify a partitioning mirroring the 10244 original source files or 'balanced' to specify partitioning into 10245 equally sized chunks (whenever possible) or 'max' to create new 10246 partition for every symbol where possible. Specifying 'none' as an 10247 algorithm disables partitioning and streaming completely. The 10248 default value is 'balanced'. While '1to1' can be used as an 10249 workaround for various code ordering issues, the 'max' partitioning 10250 is intended for internal testing only. The value 'one' specifies 10251 that exactly one partition should be used while the value 'none' 10252 bypasses partitioning and executes the link-time optimization step 10253 directly from the WPA phase. 10254 10255'-flto-compression-level=N' 10256 This option specifies the level of compression used for 10257 intermediate language written to LTO object files, and is only 10258 meaningful in conjunction with LTO mode ('-flto'). Valid values 10259 are 0 (no compression) to 9 (maximum compression). Values outside 10260 this range are clamped to either 0 or 9. If the option is not 10261 given, a default balanced compression setting is used. 10262 10263'-fuse-linker-plugin' 10264 Enables the use of a linker plugin during link-time optimization. 10265 This option relies on plugin support in the linker, which is 10266 available in gold or in GNU ld 2.21 or newer. 10267 10268 This option enables the extraction of object files with GIMPLE 10269 bytecode out of library archives. This improves the quality of 10270 optimization by exposing more code to the link-time optimizer. 10271 This information specifies what symbols can be accessed externally 10272 (by non-LTO object or during dynamic linking). Resulting code 10273 quality improvements on binaries (and shared libraries that use 10274 hidden visibility) are similar to '-fwhole-program'. See '-flto' 10275 for a description of the effect of this flag and how to use it. 10276 10277 This option is enabled by default when LTO support in GCC is 10278 enabled and GCC was configured for use with a linker supporting 10279 plugins (GNU ld 2.21 or newer or gold). 10280 10281'-ffat-lto-objects' 10282 Fat LTO objects are object files that contain both the intermediate 10283 language and the object code. This makes them usable for both LTO 10284 linking and normal linking. This option is effective only when 10285 compiling with '-flto' and is ignored at link time. 10286 10287 '-fno-fat-lto-objects' improves compilation time over plain LTO, 10288 but requires the complete toolchain to be aware of LTO. It requires 10289 a linker with linker plugin support for basic functionality. 10290 Additionally, 'nm', 'ar' and 'ranlib' need to support linker 10291 plugins to allow a full-featured build environment (capable of 10292 building static libraries etc). GCC provides the 'gcc-ar', 10293 'gcc-nm', 'gcc-ranlib' wrappers to pass the right options to these 10294 tools. With non fat LTO makefiles need to be modified to use them. 10295 10296 Note that modern binutils provide plugin auto-load mechanism. 10297 Installing the linker plugin into '$libdir/bfd-plugins' has the 10298 same effect as usage of the command wrappers ('gcc-ar', 'gcc-nm' 10299 and 'gcc-ranlib'). 10300 10301 The default is '-fno-fat-lto-objects' on targets with linker plugin 10302 support. 10303 10304'-fcompare-elim' 10305 After register allocation and post-register allocation instruction 10306 splitting, identify arithmetic instructions that compute processor 10307 flags similar to a comparison operation based on that arithmetic. 10308 If possible, eliminate the explicit comparison operation. 10309 10310 This pass only applies to certain targets that cannot explicitly 10311 represent the comparison operation before register allocation is 10312 complete. 10313 10314 Enabled at levels '-O', '-O2', '-O3', '-Os'. 10315 10316'-fcprop-registers' 10317 After register allocation and post-register allocation instruction 10318 splitting, perform a copy-propagation pass to try to reduce 10319 scheduling dependencies and occasionally eliminate the copy. 10320 10321 Enabled at levels '-O', '-O2', '-O3', '-Os'. 10322 10323'-fprofile-correction' 10324 Profiles collected using an instrumented binary for multi-threaded 10325 programs may be inconsistent due to missed counter updates. When 10326 this option is specified, GCC uses heuristics to correct or smooth 10327 out such inconsistencies. By default, GCC emits an error message 10328 when an inconsistent profile is detected. 10329 10330 This option is enabled by '-fauto-profile'. 10331 10332'-fprofile-partial-training' 10333 With '-fprofile-use' all portions of programs not executed during 10334 train run are optimized agressively for size rather than speed. In 10335 some cases it is not practical to train all possible hot paths in 10336 the program. (For example, program may contain functions specific 10337 for a given hardware and trianing may not cover all hardware 10338 configurations program is run on.) With 10339 '-fprofile-partial-training' profile feedback will be ignored for 10340 all functions not executed during the train run leading them to be 10341 optimized as if they were compiled without profile feedback. This 10342 leads to better performance when train run is not representative 10343 but also leads to significantly bigger code. 10344 10345'-fprofile-use' 10346'-fprofile-use=PATH' 10347 Enable profile feedback-directed optimizations, and the following 10348 optimizations, many of which are generally profitable only with 10349 profile feedback available: 10350 10351 -fbranch-probabilities -fprofile-values 10352 -funroll-loops -fpeel-loops -ftracer -fvpt 10353 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp 10354 -fpredictive-commoning -fsplit-loops -funswitch-loops 10355 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize 10356 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns 10357 -fprofile-reorder-functions 10358 10359 Before you can use this option, you must first generate profiling 10360 information. *Note Instrumentation Options::, for information 10361 about the '-fprofile-generate' option. 10362 10363 By default, GCC emits an error message if the feedback profiles do 10364 not match the source code. This error can be turned into a warning 10365 by using '-Wno-error=coverage-mismatch'. Note this may result in 10366 poorly optimized code. Additionally, by default, GCC also emits a 10367 warning message if the feedback profiles do not exist (see 10368 '-Wmissing-profile'). 10369 10370 If PATH is specified, GCC looks at the PATH to find the profile 10371 feedback data files. See '-fprofile-dir'. 10372 10373'-fauto-profile' 10374'-fauto-profile=PATH' 10375 Enable sampling-based feedback-directed optimizations, and the 10376 following optimizations, many of which are generally profitable 10377 only with profile feedback available: 10378 10379 -fbranch-probabilities -fprofile-values 10380 -funroll-loops -fpeel-loops -ftracer -fvpt 10381 -finline-functions -fipa-cp -fipa-cp-clone -fipa-bit-cp 10382 -fpredictive-commoning -fsplit-loops -funswitch-loops 10383 -fgcse-after-reload -ftree-loop-vectorize -ftree-slp-vectorize 10384 -fvect-cost-model=dynamic -ftree-loop-distribute-patterns 10385 -fprofile-correction 10386 10387 PATH is the name of a file containing AutoFDO profile information. 10388 If omitted, it defaults to 'fbdata.afdo' in the current directory. 10389 10390 Producing an AutoFDO profile data file requires running your 10391 program with the 'perf' utility on a supported GNU/Linux target 10392 system. For more information, see <https://perf.wiki.kernel.org/>. 10393 10394 E.g. 10395 perf record -e br_inst_retired:near_taken -b -o perf.data \ 10396 -- your_program 10397 10398 Then use the 'create_gcov' tool to convert the raw profile data to 10399 a format that can be used by GCC. You must also supply the 10400 unstripped binary for your program to this tool. See 10401 <https://github.com/google/autofdo>. 10402 10403 E.g. 10404 create_gcov --binary=your_program.unstripped --profile=perf.data \ 10405 --gcov=profile.afdo 10406 10407 The following options control compiler behavior regarding 10408floating-point arithmetic. These options trade off between speed and 10409correctness. All must be specifically enabled. 10410 10411'-ffloat-store' 10412 Do not store floating-point variables in registers, and inhibit 10413 other options that might change whether a floating-point value is 10414 taken from a register or memory. 10415 10416 This option prevents undesirable excess precision on machines such 10417 as the 68000 where the floating registers (of the 68881) keep more 10418 precision than a 'double' is supposed to have. Similarly for the 10419 x86 architecture. For most programs, the excess precision does 10420 only good, but a few programs rely on the precise definition of 10421 IEEE floating point. Use '-ffloat-store' for such programs, after 10422 modifying them to store all pertinent intermediate computations 10423 into variables. 10424 10425'-fexcess-precision=STYLE' 10426 This option allows further control over excess precision on 10427 machines where floating-point operations occur in a format with 10428 more precision or range than the IEEE standard and interchange 10429 floating-point types. By default, '-fexcess-precision=fast' is in 10430 effect; this means that operations may be carried out in a wider 10431 precision than the types specified in the source if that would 10432 result in faster code, and it is unpredictable when rounding to the 10433 types specified in the source code takes place. When compiling C, 10434 if '-fexcess-precision=standard' is specified then excess precision 10435 follows the rules specified in ISO C99; in particular, both casts 10436 and assignments cause values to be rounded to their semantic types 10437 (whereas '-ffloat-store' only affects assignments). This option is 10438 enabled by default for C if a strict conformance option such as 10439 '-std=c99' is used. '-ffast-math' enables 10440 '-fexcess-precision=fast' by default regardless of whether a strict 10441 conformance option is used. 10442 10443 '-fexcess-precision=standard' is not implemented for languages 10444 other than C. On the x86, it has no effect if '-mfpmath=sse' or 10445 '-mfpmath=sse+387' is specified; in the former case, IEEE semantics 10446 apply without excess precision, and in the latter, rounding is 10447 unpredictable. 10448 10449'-ffast-math' 10450 Sets the options '-fno-math-errno', '-funsafe-math-optimizations', 10451 '-ffinite-math-only', '-fno-rounding-math', '-fno-signaling-nans', 10452 '-fcx-limited-range' and '-fexcess-precision=fast'. 10453 10454 This option causes the preprocessor macro '__FAST_MATH__' to be 10455 defined. 10456 10457 This option is not turned on by any '-O' option besides '-Ofast' 10458 since it can result in incorrect output for programs that depend on 10459 an exact implementation of IEEE or ISO rules/specifications for 10460 math functions. It may, however, yield faster code for programs 10461 that do not require the guarantees of these specifications. 10462 10463'-fno-math-errno' 10464 Do not set 'errno' after calling math functions that are executed 10465 with a single instruction, e.g., 'sqrt'. A program that relies on 10466 IEEE exceptions for math error handling may want to use this flag 10467 for speed while maintaining IEEE arithmetic compatibility. 10468 10469 This option is not turned on by any '-O' option since it can result 10470 in incorrect output for programs that depend on an exact 10471 implementation of IEEE or ISO rules/specifications for math 10472 functions. It may, however, yield faster code for programs that do 10473 not require the guarantees of these specifications. 10474 10475 The default is '-fmath-errno'. 10476 10477 On Darwin systems, the math library never sets 'errno'. There is 10478 therefore no reason for the compiler to consider the possibility 10479 that it might, and '-fno-math-errno' is the default. 10480 10481'-funsafe-math-optimizations' 10482 10483 Allow optimizations for floating-point arithmetic that (a) assume 10484 that arguments and results are valid and (b) may violate IEEE or 10485 ANSI standards. When used at link time, it may include libraries 10486 or startup files that change the default FPU control word or other 10487 similar optimizations. 10488 10489 This option is not turned on by any '-O' option since it can result 10490 in incorrect output for programs that depend on an exact 10491 implementation of IEEE or ISO rules/specifications for math 10492 functions. It may, however, yield faster code for programs that do 10493 not require the guarantees of these specifications. Enables 10494 '-fno-signed-zeros', '-fno-trapping-math', '-fassociative-math' and 10495 '-freciprocal-math'. 10496 10497 The default is '-fno-unsafe-math-optimizations'. 10498 10499'-fassociative-math' 10500 10501 Allow re-association of operands in series of floating-point 10502 operations. This violates the ISO C and C++ language standard by 10503 possibly changing computation result. NOTE: re-ordering may change 10504 the sign of zero as well as ignore NaNs and inhibit or create 10505 underflow or overflow (and thus cannot be used on code that relies 10506 on rounding behavior like '(x + 2**52) - 2**52'. May also reorder 10507 floating-point comparisons and thus may not be used when ordered 10508 comparisons are required. This option requires that both 10509 '-fno-signed-zeros' and '-fno-trapping-math' be in effect. 10510 Moreover, it doesn't make much sense with '-frounding-math'. For 10511 Fortran the option is automatically enabled when both 10512 '-fno-signed-zeros' and '-fno-trapping-math' are in effect. 10513 10514 The default is '-fno-associative-math'. 10515 10516'-freciprocal-math' 10517 10518 Allow the reciprocal of a value to be used instead of dividing by 10519 the value if this enables optimizations. For example 'x / y' can 10520 be replaced with 'x * (1/y)', which is useful if '(1/y)' is subject 10521 to common subexpression elimination. Note that this loses 10522 precision and increases the number of flops operating on the value. 10523 10524 The default is '-fno-reciprocal-math'. 10525 10526'-ffinite-math-only' 10527 Allow optimizations for floating-point arithmetic that assume that 10528 arguments and results are not NaNs or +-Infs. 10529 10530 This option is not turned on by any '-O' option since it can result 10531 in incorrect output for programs that depend on an exact 10532 implementation of IEEE or ISO rules/specifications for math 10533 functions. It may, however, yield faster code for programs that do 10534 not require the guarantees of these specifications. 10535 10536 The default is '-fno-finite-math-only'. 10537 10538'-fno-signed-zeros' 10539 Allow optimizations for floating-point arithmetic that ignore the 10540 signedness of zero. IEEE arithmetic specifies the behavior of 10541 distinct +0.0 and -0.0 values, which then prohibits simplification 10542 of expressions such as x+0.0 or 0.0*x (even with 10543 '-ffinite-math-only'). This option implies that the sign of a zero 10544 result isn't significant. 10545 10546 The default is '-fsigned-zeros'. 10547 10548'-fno-trapping-math' 10549 Compile code assuming that floating-point operations cannot 10550 generate user-visible traps. These traps include division by zero, 10551 overflow, underflow, inexact result and invalid operation. This 10552 option requires that '-fno-signaling-nans' be in effect. Setting 10553 this option may allow faster code if one relies on "non-stop" IEEE 10554 arithmetic, for example. 10555 10556 This option should never be turned on by any '-O' option since it 10557 can result in incorrect output for programs that depend on an exact 10558 implementation of IEEE or ISO rules/specifications for math 10559 functions. 10560 10561 The default is '-ftrapping-math'. 10562 10563'-frounding-math' 10564 Disable transformations and optimizations that assume default 10565 floating-point rounding behavior. This is round-to-zero for all 10566 floating point to integer conversions, and round-to-nearest for all 10567 other arithmetic truncations. This option should be specified for 10568 programs that change the FP rounding mode dynamically, or that may 10569 be executed with a non-default rounding mode. This option disables 10570 constant folding of floating-point expressions at compile time 10571 (which may be affected by rounding mode) and arithmetic 10572 transformations that are unsafe in the presence of sign-dependent 10573 rounding modes. 10574 10575 The default is '-fno-rounding-math'. 10576 10577 This option is experimental and does not currently guarantee to 10578 disable all GCC optimizations that are affected by rounding mode. 10579 Future versions of GCC may provide finer control of this setting 10580 using C99's 'FENV_ACCESS' pragma. This command-line option will be 10581 used to specify the default state for 'FENV_ACCESS'. 10582 10583'-fsignaling-nans' 10584 Compile code assuming that IEEE signaling NaNs may generate 10585 user-visible traps during floating-point operations. Setting this 10586 option disables optimizations that may change the number of 10587 exceptions visible with signaling NaNs. This option implies 10588 '-ftrapping-math'. 10589 10590 This option causes the preprocessor macro '__SUPPORT_SNAN__' to be 10591 defined. 10592 10593 The default is '-fno-signaling-nans'. 10594 10595 This option is experimental and does not currently guarantee to 10596 disable all GCC optimizations that affect signaling NaN behavior. 10597 10598'-fno-fp-int-builtin-inexact' 10599 Do not allow the built-in functions 'ceil', 'floor', 'round' and 10600 'trunc', and their 'float' and 'long double' variants, to generate 10601 code that raises the "inexact" floating-point exception for 10602 noninteger arguments. ISO C99 and C11 allow these functions to 10603 raise the "inexact" exception, but ISO/IEC TS 18661-1:2014, the C 10604 bindings to IEEE 754-2008, as integrated into ISO C2X, does not 10605 allow these functions to do so. 10606 10607 The default is '-ffp-int-builtin-inexact', allowing the exception 10608 to be raised, unless C2X or a later C standard is selected. This 10609 option does nothing unless '-ftrapping-math' is in effect. 10610 10611 Even if '-fno-fp-int-builtin-inexact' is used, if the functions 10612 generate a call to a library function then the "inexact" exception 10613 may be raised if the library implementation does not follow TS 10614 18661. 10615 10616'-fsingle-precision-constant' 10617 Treat floating-point constants as single precision instead of 10618 implicitly converting them to double-precision constants. 10619 10620'-fcx-limited-range' 10621 When enabled, this option states that a range reduction step is not 10622 needed when performing complex division. Also, there is no 10623 checking whether the result of a complex multiplication or division 10624 is 'NaN + I*NaN', with an attempt to rescue the situation in that 10625 case. The default is '-fno-cx-limited-range', but is enabled by 10626 '-ffast-math'. 10627 10628 This option controls the default setting of the ISO C99 10629 'CX_LIMITED_RANGE' pragma. Nevertheless, the option applies to all 10630 languages. 10631 10632'-fcx-fortran-rules' 10633 Complex multiplication and division follow Fortran rules. Range 10634 reduction is done as part of complex division, but there is no 10635 checking whether the result of a complex multiplication or division 10636 is 'NaN + I*NaN', with an attempt to rescue the situation in that 10637 case. 10638 10639 The default is '-fno-cx-fortran-rules'. 10640 10641 The following options control optimizations that may improve 10642performance, but are not enabled by any '-O' options. This section 10643includes experimental options that may produce broken code. 10644 10645'-fbranch-probabilities' 10646 After running a program compiled with '-fprofile-arcs' (*note 10647 Instrumentation Options::), you can compile it a second time using 10648 '-fbranch-probabilities', to improve optimizations based on the 10649 number of times each branch was taken. When a program compiled 10650 with '-fprofile-arcs' exits, it saves arc execution counts to a 10651 file called 'SOURCENAME.gcda' for each source file. The 10652 information in this data file is very dependent on the structure of 10653 the generated code, so you must use the same source code and the 10654 same optimization options for both compilations. 10655 10656 With '-fbranch-probabilities', GCC puts a 'REG_BR_PROB' note on 10657 each 'JUMP_INSN' and 'CALL_INSN'. These can be used to improve 10658 optimization. Currently, they are only used in one place: in 10659 'reorg.c', instead of guessing which path a branch is most likely 10660 to take, the 'REG_BR_PROB' values are used to exactly determine 10661 which path is taken more often. 10662 10663 Enabled by '-fprofile-use' and '-fauto-profile'. 10664 10665'-fprofile-values' 10666 If combined with '-fprofile-arcs', it adds code so that some data 10667 about values of expressions in the program is gathered. 10668 10669 With '-fbranch-probabilities', it reads back the data gathered from 10670 profiling values of expressions for usage in optimizations. 10671 10672 Enabled by '-fprofile-generate', '-fprofile-use', and 10673 '-fauto-profile'. 10674 10675'-fprofile-reorder-functions' 10676 Function reordering based on profile instrumentation collects first 10677 time of execution of a function and orders these functions in 10678 ascending order. 10679 10680 Enabled with '-fprofile-use'. 10681 10682'-fvpt' 10683 If combined with '-fprofile-arcs', this option instructs the 10684 compiler to add code to gather information about values of 10685 expressions. 10686 10687 With '-fbranch-probabilities', it reads back the data gathered and 10688 actually performs the optimizations based on them. Currently the 10689 optimizations include specialization of division operations using 10690 the knowledge about the value of the denominator. 10691 10692 Enabled with '-fprofile-use' and '-fauto-profile'. 10693 10694'-frename-registers' 10695 Attempt to avoid false dependencies in scheduled code by making use 10696 of registers left over after register allocation. This 10697 optimization most benefits processors with lots of registers. 10698 Depending on the debug information format adopted by the target, 10699 however, it can make debugging impossible, since variables no 10700 longer stay in a "home register". 10701 10702 Enabled by default with '-funroll-loops'. 10703 10704'-fschedule-fusion' 10705 Performs a target dependent pass over the instruction stream to 10706 schedule instructions of same type together because target machine 10707 can execute them more efficiently if they are adjacent to each 10708 other in the instruction flow. 10709 10710 Enabled at levels '-O2', '-O3', '-Os'. 10711 10712'-ftracer' 10713 Perform tail duplication to enlarge superblock size. This 10714 transformation simplifies the control flow of the function allowing 10715 other optimizations to do a better job. 10716 10717 Enabled by '-fprofile-use' and '-fauto-profile'. 10718 10719'-funroll-loops' 10720 Unroll loops whose number of iterations can be determined at 10721 compile time or upon entry to the loop. '-funroll-loops' implies 10722 '-frerun-cse-after-loop', '-fweb' and '-frename-registers'. It 10723 also turns on complete loop peeling (i.e. complete removal of loops 10724 with a small constant number of iterations). This option makes 10725 code larger, and may or may not make it run faster. 10726 10727 Enabled by '-fprofile-use' and '-fauto-profile'. 10728 10729'-funroll-all-loops' 10730 Unroll all loops, even if their number of iterations is uncertain 10731 when the loop is entered. This usually makes programs run more 10732 slowly. '-funroll-all-loops' implies the same options as 10733 '-funroll-loops'. 10734 10735'-fpeel-loops' 10736 Peels loops for which there is enough information that they do not 10737 roll much (from profile feedback or static analysis). It also 10738 turns on complete loop peeling (i.e. complete removal of loops with 10739 small constant number of iterations). 10740 10741 Enabled by '-O3', '-fprofile-use', and '-fauto-profile'. 10742 10743'-fmove-loop-invariants' 10744 Enables the loop invariant motion pass in the RTL loop optimizer. 10745 Enabled at level '-O1' and higher, except for '-Og'. 10746 10747'-fsplit-loops' 10748 Split a loop into two if it contains a condition that's always true 10749 for one side of the iteration space and false for the other. 10750 10751 Enabled by '-fprofile-use' and '-fauto-profile'. 10752 10753'-funswitch-loops' 10754 Move branches with loop invariant conditions out of the loop, with 10755 duplicates of the loop on both branches (modified according to 10756 result of the condition). 10757 10758 Enabled by '-fprofile-use' and '-fauto-profile'. 10759 10760'-fversion-loops-for-strides' 10761 If a loop iterates over an array with a variable stride, create 10762 another version of the loop that assumes the stride is always one. 10763 For example: 10764 10765 for (int i = 0; i < n; ++i) 10766 x[i * stride] = ...; 10767 10768 becomes: 10769 10770 if (stride == 1) 10771 for (int i = 0; i < n; ++i) 10772 x[i] = ...; 10773 else 10774 for (int i = 0; i < n; ++i) 10775 x[i * stride] = ...; 10776 10777 This is particularly useful for assumed-shape arrays in Fortran 10778 where (for example) it allows better vectorization assuming 10779 contiguous accesses. This flag is enabled by default at '-O3'. It 10780 is also enabled by '-fprofile-use' and '-fauto-profile'. 10781 10782'-ffunction-sections' 10783'-fdata-sections' 10784 Place each function or data item into its own section in the output 10785 file if the target supports arbitrary sections. The name of the 10786 function or the name of the data item determines the section's name 10787 in the output file. 10788 10789 Use these options on systems where the linker can perform 10790 optimizations to improve locality of reference in the instruction 10791 space. Most systems using the ELF object format have linkers with 10792 such optimizations. On AIX, the linker rearranges sections 10793 (CSECTs) based on the call graph. The performance impact varies. 10794 10795 Together with a linker garbage collection (linker '--gc-sections' 10796 option) these options may lead to smaller statically-linked 10797 executables (after stripping). 10798 10799 On ELF/DWARF systems these options do not degenerate the quality of 10800 the debug information. There could be issues with other object 10801 files/debug info formats. 10802 10803 Only use these options when there are significant benefits from 10804 doing so. When you specify these options, the assembler and linker 10805 create larger object and executable files and are also slower. 10806 These options affect code generation. They prevent optimizations 10807 by the compiler and assembler using relative locations inside a 10808 translation unit since the locations are unknown until link time. 10809 An example of such an optimization is relaxing calls to short call 10810 instructions. 10811 10812'-fstdarg-opt' 10813 Optimize the prologue of variadic argument functions with respect 10814 to usage of those arguments. 10815 10816'-fsection-anchors' 10817 Try to reduce the number of symbolic address calculations by using 10818 shared "anchor" symbols to address nearby objects. This 10819 transformation can help to reduce the number of GOT entries and GOT 10820 accesses on some targets. 10821 10822 For example, the implementation of the following function 'foo': 10823 10824 static int a, b, c; 10825 int foo (void) { return a + b + c; } 10826 10827 usually calculates the addresses of all three variables, but if you 10828 compile it with '-fsection-anchors', it accesses the variables from 10829 a common anchor point instead. The effect is similar to the 10830 following pseudocode (which isn't valid C): 10831 10832 int foo (void) 10833 { 10834 register int *xr = &x; 10835 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 10836 } 10837 10838 Not all targets support this option. 10839 10840'--param NAME=VALUE' 10841 In some places, GCC uses various constants to control the amount of 10842 optimization that is done. For example, GCC does not inline 10843 functions that contain more than a certain number of instructions. 10844 You can control some of these constants on the command line using 10845 the '--param' option. 10846 10847 The names of specific parameters, and the meaning of the values, 10848 are tied to the internals of the compiler, and are subject to 10849 change without notice in future releases. 10850 10851 In order to get minimal, maximal and default value of a parameter, 10852 one can use '--help=param -Q' options. 10853 10854 In each case, the VALUE is an integer. The following choices of 10855 NAME are recognized for all targets: 10856 10857 'predictable-branch-outcome' 10858 When branch is predicted to be taken with probability lower 10859 than this threshold (in percent), then it is considered well 10860 predictable. 10861 10862 'max-rtl-if-conversion-insns' 10863 RTL if-conversion tries to remove conditional branches around 10864 a block and replace them with conditionally executed 10865 instructions. This parameter gives the maximum number of 10866 instructions in a block which should be considered for 10867 if-conversion. The compiler will also use other heuristics to 10868 decide whether if-conversion is likely to be profitable. 10869 10870 'max-rtl-if-conversion-predictable-cost' 10871 'max-rtl-if-conversion-unpredictable-cost' 10872 RTL if-conversion will try to remove conditional branches 10873 around a block and replace them with conditionally executed 10874 instructions. These parameters give the maximum permissible 10875 cost for the sequence that would be generated by if-conversion 10876 depending on whether the branch is statically determined to be 10877 predictable or not. The units for this parameter are the same 10878 as those for the GCC internal seq_cost metric. The compiler 10879 will try to provide a reasonable default for this parameter 10880 using the BRANCH_COST target macro. 10881 10882 'max-crossjump-edges' 10883 The maximum number of incoming edges to consider for 10884 cross-jumping. The algorithm used by '-fcrossjumping' is 10885 O(N^2) in the number of edges incoming to each block. 10886 Increasing values mean more aggressive optimization, making 10887 the compilation time increase with probably small improvement 10888 in executable size. 10889 10890 'min-crossjump-insns' 10891 The minimum number of instructions that must be matched at the 10892 end of two blocks before cross-jumping is performed on them. 10893 This value is ignored in the case where all instructions in 10894 the block being cross-jumped from are matched. 10895 10896 'max-grow-copy-bb-insns' 10897 The maximum code size expansion factor when copying basic 10898 blocks instead of jumping. The expansion is relative to a 10899 jump instruction. 10900 10901 'max-goto-duplication-insns' 10902 The maximum number of instructions to duplicate to a block 10903 that jumps to a computed goto. To avoid O(N^2) behavior in a 10904 number of passes, GCC factors computed gotos early in the 10905 compilation process, and unfactors them as late as possible. 10906 Only computed jumps at the end of a basic blocks with no more 10907 than max-goto-duplication-insns are unfactored. 10908 10909 'max-delay-slot-insn-search' 10910 The maximum number of instructions to consider when looking 10911 for an instruction to fill a delay slot. If more than this 10912 arbitrary number of instructions are searched, the time 10913 savings from filling the delay slot are minimal, so stop 10914 searching. Increasing values mean more aggressive 10915 optimization, making the compilation time increase with 10916 probably small improvement in execution time. 10917 10918 'max-delay-slot-live-search' 10919 When trying to fill delay slots, the maximum number of 10920 instructions to consider when searching for a block with valid 10921 live register information. Increasing this arbitrarily chosen 10922 value means more aggressive optimization, increasing the 10923 compilation time. This parameter should be removed when the 10924 delay slot code is rewritten to maintain the control-flow 10925 graph. 10926 10927 'max-gcse-memory' 10928 The approximate maximum amount of memory that can be allocated 10929 in order to perform the global common subexpression 10930 elimination optimization. If more memory than specified is 10931 required, the optimization is not done. 10932 10933 'max-gcse-insertion-ratio' 10934 If the ratio of expression insertions to deletions is larger 10935 than this value for any expression, then RTL PRE inserts or 10936 removes the expression and thus leaves partially redundant 10937 computations in the instruction stream. 10938 10939 'max-pending-list-length' 10940 The maximum number of pending dependencies scheduling allows 10941 before flushing the current state and starting over. Large 10942 functions with few branches or calls can create excessively 10943 large lists which needlessly consume memory and resources. 10944 10945 'max-modulo-backtrack-attempts' 10946 The maximum number of backtrack attempts the scheduler should 10947 make when modulo scheduling a loop. Larger values can 10948 exponentially increase compilation time. 10949 10950 'max-inline-insns-single' 10951 Several parameters control the tree inliner used in GCC. This 10952 number sets the maximum number of instructions (counted in 10953 GCC's internal representation) in a single function that the 10954 tree inliner considers for inlining. This only affects 10955 functions declared inline and methods implemented in a class 10956 declaration (C++). 10957 10958 'max-inline-insns-auto' 10959 When you use '-finline-functions' (included in '-O3'), a lot 10960 of functions that would otherwise not be considered for 10961 inlining by the compiler are investigated. To those 10962 functions, a different (more restrictive) limit compared to 10963 functions declared inline can be applied ('--param 10964 max-inline-insns-auto'). 10965 10966 'max-inline-insns-small' 10967 This is bound applied to calls which are considered relevant 10968 with '-finline-small-functions'. 10969 10970 'max-inline-insns-size' 10971 This is bound applied to calls which are optimized for size. 10972 Small growth may be desirable to anticipate optimization 10973 oppurtunities exposed by inlining. 10974 10975 'uninlined-function-insns' 10976 Number of instructions accounted by inliner for function 10977 overhead such as function prologue and epilogue. 10978 10979 'uninlined-function-time' 10980 Extra time accounted by inliner for function overhead such as 10981 time needed to execute function prologue and epilogue 10982 10983 'inline-heuristics-hint-percent' 10984 The scale (in percents) applied to 'inline-insns-single', 10985 'inline-insns-single-O2', 'inline-insns-auto' when inline 10986 heuristics hints that inlining is very profitable (will enable 10987 later optimizations). 10988 10989 'uninlined-thunk-insns' 10990 'uninlined-thunk-time' 10991 Same as '--param uninlined-function-insns' and '--param 10992 uninlined-function-time' but applied to function thunks 10993 10994 'inline-min-speedup' 10995 When estimated performance improvement of caller + callee 10996 runtime exceeds this threshold (in percent), the function can 10997 be inlined regardless of the limit on '--param 10998 max-inline-insns-single' and '--param max-inline-insns-auto'. 10999 11000 'large-function-insns' 11001 The limit specifying really large functions. For functions 11002 larger than this limit after inlining, inlining is constrained 11003 by '--param large-function-growth'. This parameter is useful 11004 primarily to avoid extreme compilation time caused by 11005 non-linear algorithms used by the back end. 11006 11007 'large-function-growth' 11008 Specifies maximal growth of large function caused by inlining 11009 in percents. For example, parameter value 100 limits large 11010 function growth to 2.0 times the original size. 11011 11012 'large-unit-insns' 11013 The limit specifying large translation unit. Growth caused by 11014 inlining of units larger than this limit is limited by 11015 '--param inline-unit-growth'. For small units this might be 11016 too tight. For example, consider a unit consisting of 11017 function A that is inline and B that just calls A three times. 11018 If B is small relative to A, the growth of unit is 300\% and 11019 yet such inlining is very sane. For very large units 11020 consisting of small inlineable functions, however, the overall 11021 unit growth limit is needed to avoid exponential explosion of 11022 code size. Thus for smaller units, the size is increased to 11023 '--param large-unit-insns' before applying '--param 11024 inline-unit-growth'. 11025 11026 'inline-unit-growth' 11027 Specifies maximal overall growth of the compilation unit 11028 caused by inlining. For example, parameter value 20 limits 11029 unit growth to 1.2 times the original size. Cold functions 11030 (either marked cold via an attribute or by profile feedback) 11031 are not accounted into the unit size. 11032 11033 'ipa-cp-unit-growth' 11034 Specifies maximal overall growth of the compilation unit 11035 caused by interprocedural constant propagation. For example, 11036 parameter value 10 limits unit growth to 1.1 times the 11037 original size. 11038 11039 'large-stack-frame' 11040 The limit specifying large stack frames. While inlining the 11041 algorithm is trying to not grow past this limit too much. 11042 11043 'large-stack-frame-growth' 11044 Specifies maximal growth of large stack frames caused by 11045 inlining in percents. For example, parameter value 1000 11046 limits large stack frame growth to 11 times the original size. 11047 11048 'max-inline-insns-recursive' 11049 'max-inline-insns-recursive-auto' 11050 Specifies the maximum number of instructions an out-of-line 11051 copy of a self-recursive inline function can grow into by 11052 performing recursive inlining. 11053 11054 '--param max-inline-insns-recursive' applies to functions 11055 declared inline. For functions not declared inline, recursive 11056 inlining happens only when '-finline-functions' (included in 11057 '-O3') is enabled; '--param max-inline-insns-recursive-auto' 11058 applies instead. 11059 11060 'max-inline-recursive-depth' 11061 'max-inline-recursive-depth-auto' 11062 Specifies the maximum recursion depth used for recursive 11063 inlining. 11064 11065 '--param max-inline-recursive-depth' applies to functions 11066 declared inline. For functions not declared inline, recursive 11067 inlining happens only when '-finline-functions' (included in 11068 '-O3') is enabled; '--param max-inline-recursive-depth-auto' 11069 applies instead. 11070 11071 'min-inline-recursive-probability' 11072 Recursive inlining is profitable only for function having deep 11073 recursion in average and can hurt for function having little 11074 recursion depth by increasing the prologue size or complexity 11075 of function body to other optimizers. 11076 11077 When profile feedback is available (see '-fprofile-generate') 11078 the actual recursion depth can be guessed from the probability 11079 that function recurses via a given call expression. This 11080 parameter limits inlining only to call expressions whose 11081 probability exceeds the given threshold (in percents). 11082 11083 'early-inlining-insns' 11084 Specify growth that the early inliner can make. In effect it 11085 increases the amount of inlining for code having a large 11086 abstraction penalty. 11087 11088 'max-early-inliner-iterations' 11089 Limit of iterations of the early inliner. This basically 11090 bounds the number of nested indirect calls the early inliner 11091 can resolve. Deeper chains are still handled by late 11092 inlining. 11093 11094 'comdat-sharing-probability' 11095 Probability (in percent) that C++ inline function with comdat 11096 visibility are shared across multiple compilation units. 11097 11098 'profile-func-internal-id' 11099 A parameter to control whether to use function internal id in 11100 profile database lookup. If the value is 0, the compiler uses 11101 an id that is based on function assembler name and filename, 11102 which makes old profile data more tolerant to source changes 11103 such as function reordering etc. 11104 11105 'min-vect-loop-bound' 11106 The minimum number of iterations under which loops are not 11107 vectorized when '-ftree-vectorize' is used. The number of 11108 iterations after vectorization needs to be greater than the 11109 value specified by this option to allow vectorization. 11110 11111 'gcse-cost-distance-ratio' 11112 Scaling factor in calculation of maximum distance an 11113 expression can be moved by GCSE optimizations. This is 11114 currently supported only in the code hoisting pass. The 11115 bigger the ratio, the more aggressive code hoisting is with 11116 simple expressions, i.e., the expressions that have cost less 11117 than 'gcse-unrestricted-cost'. Specifying 0 disables hoisting 11118 of simple expressions. 11119 11120 'gcse-unrestricted-cost' 11121 Cost, roughly measured as the cost of a single typical machine 11122 instruction, at which GCSE optimizations do not constrain the 11123 distance an expression can travel. This is currently 11124 supported only in the code hoisting pass. The lesser the 11125 cost, the more aggressive code hoisting is. Specifying 0 11126 allows all expressions to travel unrestricted distances. 11127 11128 'max-hoist-depth' 11129 The depth of search in the dominator tree for expressions to 11130 hoist. This is used to avoid quadratic behavior in hoisting 11131 algorithm. The value of 0 does not limit on the search, but 11132 may slow down compilation of huge functions. 11133 11134 'max-tail-merge-comparisons' 11135 The maximum amount of similar bbs to compare a bb with. This 11136 is used to avoid quadratic behavior in tree tail merging. 11137 11138 'max-tail-merge-iterations' 11139 The maximum amount of iterations of the pass over the 11140 function. This is used to limit compilation time in tree tail 11141 merging. 11142 11143 'store-merging-allow-unaligned' 11144 Allow the store merging pass to introduce unaligned stores if 11145 it is legal to do so. 11146 11147 'max-stores-to-merge' 11148 The maximum number of stores to attempt to merge into wider 11149 stores in the store merging pass. 11150 11151 'max-unrolled-insns' 11152 The maximum number of instructions that a loop may have to be 11153 unrolled. If a loop is unrolled, this parameter also 11154 determines how many times the loop code is unrolled. 11155 11156 'max-average-unrolled-insns' 11157 The maximum number of instructions biased by probabilities of 11158 their execution that a loop may have to be unrolled. If a 11159 loop is unrolled, this parameter also determines how many 11160 times the loop code is unrolled. 11161 11162 'max-unroll-times' 11163 The maximum number of unrollings of a single loop. 11164 11165 'max-peeled-insns' 11166 The maximum number of instructions that a loop may have to be 11167 peeled. If a loop is peeled, this parameter also determines 11168 how many times the loop code is peeled. 11169 11170 'max-peel-times' 11171 The maximum number of peelings of a single loop. 11172 11173 'max-peel-branches' 11174 The maximum number of branches on the hot path through the 11175 peeled sequence. 11176 11177 'max-completely-peeled-insns' 11178 The maximum number of insns of a completely peeled loop. 11179 11180 'max-completely-peel-times' 11181 The maximum number of iterations of a loop to be suitable for 11182 complete peeling. 11183 11184 'max-completely-peel-loop-nest-depth' 11185 The maximum depth of a loop nest suitable for complete 11186 peeling. 11187 11188 'max-unswitch-insns' 11189 The maximum number of insns of an unswitched loop. 11190 11191 'max-unswitch-level' 11192 The maximum number of branches unswitched in a single loop. 11193 11194 'lim-expensive' 11195 The minimum cost of an expensive expression in the loop 11196 invariant motion. 11197 11198 'min-loop-cond-split-prob' 11199 When FDO profile information is available, 11200 'min-loop-cond-split-prob' specifies minimum threshold for 11201 probability of semi-invariant condition statement to trigger 11202 loop split. 11203 11204 'iv-consider-all-candidates-bound' 11205 Bound on number of candidates for induction variables, below 11206 which all candidates are considered for each use in induction 11207 variable optimizations. If there are more candidates than 11208 this, only the most relevant ones are considered to avoid 11209 quadratic time complexity. 11210 11211 'iv-max-considered-uses' 11212 The induction variable optimizations give up on loops that 11213 contain more induction variable uses. 11214 11215 'iv-always-prune-cand-set-bound' 11216 If the number of candidates in the set is smaller than this 11217 value, always try to remove unnecessary ivs from the set when 11218 adding a new one. 11219 11220 'avg-loop-niter' 11221 Average number of iterations of a loop. 11222 11223 'dse-max-object-size' 11224 Maximum size (in bytes) of objects tracked bytewise by dead 11225 store elimination. Larger values may result in larger 11226 compilation times. 11227 11228 'dse-max-alias-queries-per-store' 11229 Maximum number of queries into the alias oracle per store. 11230 Larger values result in larger compilation times and may 11231 result in more removed dead stores. 11232 11233 'scev-max-expr-size' 11234 Bound on size of expressions used in the scalar evolutions 11235 analyzer. Large expressions slow the analyzer. 11236 11237 'scev-max-expr-complexity' 11238 Bound on the complexity of the expressions in the scalar 11239 evolutions analyzer. Complex expressions slow the analyzer. 11240 11241 'max-tree-if-conversion-phi-args' 11242 Maximum number of arguments in a PHI supported by TREE if 11243 conversion unless the loop is marked with simd pragma. 11244 11245 'vect-max-version-for-alignment-checks' 11246 The maximum number of run-time checks that can be performed 11247 when doing loop versioning for alignment in the vectorizer. 11248 11249 'vect-max-version-for-alias-checks' 11250 The maximum number of run-time checks that can be performed 11251 when doing loop versioning for alias in the vectorizer. 11252 11253 'vect-max-peeling-for-alignment' 11254 The maximum number of loop peels to enhance access alignment 11255 for vectorizer. Value -1 means no limit. 11256 11257 'max-iterations-to-track' 11258 The maximum number of iterations of a loop the brute-force 11259 algorithm for analysis of the number of iterations of the loop 11260 tries to evaluate. 11261 11262 'hot-bb-count-fraction' 11263 The denominator n of fraction 1/n of the maximal execution 11264 count of a basic block in the entire program that a basic 11265 block needs to at least have in order to be considered hot. 11266 The default is 10000, which means that a basic block is 11267 considered hot if its execution count is greater than 1/10000 11268 of the maximal execution count. 0 means that it is never 11269 considered hot. Used in non-LTO mode. 11270 11271 'hot-bb-count-ws-permille' 11272 The number of most executed permilles, ranging from 0 to 1000, 11273 of the profiled execution of the entire program to which the 11274 execution count of a basic block must be part of in order to 11275 be considered hot. The default is 990, which means that a 11276 basic block is considered hot if its execution count 11277 contributes to the upper 990 permilles, or 99.0%, of the 11278 profiled execution of the entire program. 0 means that it is 11279 never considered hot. Used in LTO mode. 11280 11281 'hot-bb-frequency-fraction' 11282 The denominator n of fraction 1/n of the execution frequency 11283 of the entry block of a function that a basic block of this 11284 function needs to at least have in order to be considered hot. 11285 The default is 1000, which means that a basic block is 11286 considered hot in a function if it is executed more frequently 11287 than 1/1000 of the frequency of the entry block of the 11288 function. 0 means that it is never considered hot. 11289 11290 'unlikely-bb-count-fraction' 11291 The denominator n of fraction 1/n of the number of profiled 11292 runs of the entire program below which the execution count of 11293 a basic block must be in order for the basic block to be 11294 considered unlikely executed. The default is 20, which means 11295 that a basic block is considered unlikely executed if it is 11296 executed in fewer than 1/20, or 5%, of the runs of the 11297 program. 0 means that it is always considered unlikely 11298 executed. 11299 11300 'max-predicted-iterations' 11301 The maximum number of loop iterations we predict statically. 11302 This is useful in cases where a function contains a single 11303 loop with known bound and another loop with unknown bound. 11304 The known number of iterations is predicted correctly, while 11305 the unknown number of iterations average to roughly 10. This 11306 means that the loop without bounds appears artificially cold 11307 relative to the other one. 11308 11309 'builtin-expect-probability' 11310 Control the probability of the expression having the specified 11311 value. This parameter takes a percentage (i.e. 0 ... 100) as 11312 input. 11313 11314 'builtin-string-cmp-inline-length' 11315 The maximum length of a constant string for a builtin string 11316 cmp call eligible for inlining. 11317 11318 'align-threshold' 11319 11320 Select fraction of the maximal frequency of executions of a 11321 basic block in a function to align the basic block. 11322 11323 'align-loop-iterations' 11324 11325 A loop expected to iterate at least the selected number of 11326 iterations is aligned. 11327 11328 'tracer-dynamic-coverage' 11329 'tracer-dynamic-coverage-feedback' 11330 11331 This value is used to limit superblock formation once the 11332 given percentage of executed instructions is covered. This 11333 limits unnecessary code size expansion. 11334 11335 The 'tracer-dynamic-coverage-feedback' parameter is used only 11336 when profile feedback is available. The real profiles (as 11337 opposed to statically estimated ones) are much less balanced 11338 allowing the threshold to be larger value. 11339 11340 'tracer-max-code-growth' 11341 Stop tail duplication once code growth has reached given 11342 percentage. This is a rather artificial limit, as most of the 11343 duplicates are eliminated later in cross jumping, so it may be 11344 set to much higher values than is the desired code growth. 11345 11346 'tracer-min-branch-ratio' 11347 11348 Stop reverse growth when the reverse probability of best edge 11349 is less than this threshold (in percent). 11350 11351 'tracer-min-branch-probability' 11352 'tracer-min-branch-probability-feedback' 11353 11354 Stop forward growth if the best edge has probability lower 11355 than this threshold. 11356 11357 Similarly to 'tracer-dynamic-coverage' two parameters are 11358 provided. 'tracer-min-branch-probability-feedback' is used 11359 for compilation with profile feedback and 11360 'tracer-min-branch-probability' compilation without. The 11361 value for compilation with profile feedback needs to be more 11362 conservative (higher) in order to make tracer effective. 11363 11364 'stack-clash-protection-guard-size' 11365 Specify the size of the operating system provided stack guard 11366 as 2 raised to NUM bytes. Higher values may reduce the number 11367 of explicit probes, but a value larger than the operating 11368 system provided guard will leave code vulnerable to stack 11369 clash style attacks. 11370 11371 'stack-clash-protection-probe-interval' 11372 Stack clash protection involves probing stack space as it is 11373 allocated. This param controls the maximum distance between 11374 probes into the stack as 2 raised to NUM bytes. Higher values 11375 may reduce the number of explicit probes, but a value larger 11376 than the operating system provided guard will leave code 11377 vulnerable to stack clash style attacks. 11378 11379 'max-cse-path-length' 11380 11381 The maximum number of basic blocks on path that CSE considers. 11382 11383 'max-cse-insns' 11384 The maximum number of instructions CSE processes before 11385 flushing. 11386 11387 'ggc-min-expand' 11388 11389 GCC uses a garbage collector to manage its own memory 11390 allocation. This parameter specifies the minimum percentage 11391 by which the garbage collector's heap should be allowed to 11392 expand between collections. Tuning this may improve 11393 compilation speed; it has no effect on code generation. 11394 11395 The default is 30% + 70% * (RAM/1GB) with an upper bound of 11396 100% when RAM >= 1GB. If 'getrlimit' is available, the notion 11397 of "RAM" is the smallest of actual RAM and 'RLIMIT_DATA' or 11398 'RLIMIT_AS'. If GCC is not able to calculate RAM on a 11399 particular platform, the lower bound of 30% is used. Setting 11400 this parameter and 'ggc-min-heapsize' to zero causes a full 11401 collection to occur at every opportunity. This is extremely 11402 slow, but can be useful for debugging. 11403 11404 'ggc-min-heapsize' 11405 11406 Minimum size of the garbage collector's heap before it begins 11407 bothering to collect garbage. The first collection occurs 11408 after the heap expands by 'ggc-min-expand'% beyond 11409 'ggc-min-heapsize'. Again, tuning this may improve 11410 compilation speed, and has no effect on code generation. 11411 11412 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit 11413 that tries to ensure that RLIMIT_DATA or RLIMIT_AS are not 11414 exceeded, but with a lower bound of 4096 (four megabytes) and 11415 an upper bound of 131072 (128 megabytes). If GCC is not able 11416 to calculate RAM on a particular platform, the lower bound is 11417 used. Setting this parameter very large effectively disables 11418 garbage collection. Setting this parameter and 11419 'ggc-min-expand' to zero causes a full collection to occur at 11420 every opportunity. 11421 11422 'max-reload-search-insns' 11423 The maximum number of instruction reload should look backward 11424 for equivalent register. Increasing values mean more 11425 aggressive optimization, making the compilation time increase 11426 with probably slightly better performance. 11427 11428 'max-cselib-memory-locations' 11429 The maximum number of memory locations cselib should take into 11430 account. Increasing values mean more aggressive optimization, 11431 making the compilation time increase with probably slightly 11432 better performance. 11433 11434 'max-sched-ready-insns' 11435 The maximum number of instructions ready to be issued the 11436 scheduler should consider at any given time during the first 11437 scheduling pass. Increasing values mean more thorough 11438 searches, making the compilation time increase with probably 11439 little benefit. 11440 11441 'max-sched-region-blocks' 11442 The maximum number of blocks in a region to be considered for 11443 interblock scheduling. 11444 11445 'max-pipeline-region-blocks' 11446 The maximum number of blocks in a region to be considered for 11447 pipelining in the selective scheduler. 11448 11449 'max-sched-region-insns' 11450 The maximum number of insns in a region to be considered for 11451 interblock scheduling. 11452 11453 'max-pipeline-region-insns' 11454 The maximum number of insns in a region to be considered for 11455 pipelining in the selective scheduler. 11456 11457 'min-spec-prob' 11458 The minimum probability (in percents) of reaching a source 11459 block for interblock speculative scheduling. 11460 11461 'max-sched-extend-regions-iters' 11462 The maximum number of iterations through CFG to extend 11463 regions. A value of 0 disables region extensions. 11464 11465 'max-sched-insn-conflict-delay' 11466 The maximum conflict delay for an insn to be considered for 11467 speculative motion. 11468 11469 'sched-spec-prob-cutoff' 11470 The minimal probability of speculation success (in percents), 11471 so that speculative insns are scheduled. 11472 11473 'sched-state-edge-prob-cutoff' 11474 The minimum probability an edge must have for the scheduler to 11475 save its state across it. 11476 11477 'sched-mem-true-dep-cost' 11478 Minimal distance (in CPU cycles) between store and load 11479 targeting same memory locations. 11480 11481 'selsched-max-lookahead' 11482 The maximum size of the lookahead window of selective 11483 scheduling. It is a depth of search for available 11484 instructions. 11485 11486 'selsched-max-sched-times' 11487 The maximum number of times that an instruction is scheduled 11488 during selective scheduling. This is the limit on the number 11489 of iterations through which the instruction may be pipelined. 11490 11491 'selsched-insns-to-rename' 11492 The maximum number of best instructions in the ready list that 11493 are considered for renaming in the selective scheduler. 11494 11495 'sms-min-sc' 11496 The minimum value of stage count that swing modulo scheduler 11497 generates. 11498 11499 'max-last-value-rtl' 11500 The maximum size measured as number of RTLs that can be 11501 recorded in an expression in combiner for a pseudo register as 11502 last known value of that register. 11503 11504 'max-combine-insns' 11505 The maximum number of instructions the RTL combiner tries to 11506 combine. 11507 11508 'integer-share-limit' 11509 Small integer constants can use a shared data structure, 11510 reducing the compiler's memory usage and increasing its speed. 11511 This sets the maximum value of a shared integer constant. 11512 11513 'ssp-buffer-size' 11514 The minimum size of buffers (i.e. arrays) that receive stack 11515 smashing protection when '-fstack-protection' is used. 11516 11517 'min-size-for-stack-sharing' 11518 The minimum size of variables taking part in stack slot 11519 sharing when not optimizing. 11520 11521 'max-jump-thread-duplication-stmts' 11522 Maximum number of statements allowed in a block that needs to 11523 be duplicated when threading jumps. 11524 11525 'max-fields-for-field-sensitive' 11526 Maximum number of fields in a structure treated in a field 11527 sensitive manner during pointer analysis. 11528 11529 'prefetch-latency' 11530 Estimate on average number of instructions that are executed 11531 before prefetch finishes. The distance prefetched ahead is 11532 proportional to this constant. Increasing this number may 11533 also lead to less streams being prefetched (see 11534 'simultaneous-prefetches'). 11535 11536 'simultaneous-prefetches' 11537 Maximum number of prefetches that can run at the same time. 11538 11539 'l1-cache-line-size' 11540 The size of cache line in L1 data cache, in bytes. 11541 11542 'l1-cache-size' 11543 The size of L1 data cache, in kilobytes. 11544 11545 'l2-cache-size' 11546 The size of L2 data cache, in kilobytes. 11547 11548 'prefetch-dynamic-strides' 11549 Whether the loop array prefetch pass should issue software 11550 prefetch hints for strides that are non-constant. In some 11551 cases this may be beneficial, though the fact the stride is 11552 non-constant may make it hard to predict when there is clear 11553 benefit to issuing these hints. 11554 11555 Set to 1 if the prefetch hints should be issued for 11556 non-constant strides. Set to 0 if prefetch hints should be 11557 issued only for strides that are known to be constant and 11558 below 'prefetch-minimum-stride'. 11559 11560 'prefetch-minimum-stride' 11561 Minimum constant stride, in bytes, to start using prefetch 11562 hints for. If the stride is less than this threshold, 11563 prefetch hints will not be issued. 11564 11565 This setting is useful for processors that have hardware 11566 prefetchers, in which case there may be conflicts between the 11567 hardware prefetchers and the software prefetchers. If the 11568 hardware prefetchers have a maximum stride they can handle, it 11569 should be used here to improve the use of software 11570 prefetchers. 11571 11572 A value of -1 means we don't have a threshold and therefore 11573 prefetch hints can be issued for any constant stride. 11574 11575 This setting is only useful for strides that are known and 11576 constant. 11577 11578 'loop-interchange-max-num-stmts' 11579 The maximum number of stmts in a loop to be interchanged. 11580 11581 'loop-interchange-stride-ratio' 11582 The minimum ratio between stride of two loops for interchange 11583 to be profitable. 11584 11585 'min-insn-to-prefetch-ratio' 11586 The minimum ratio between the number of instructions and the 11587 number of prefetches to enable prefetching in a loop. 11588 11589 'prefetch-min-insn-to-mem-ratio' 11590 The minimum ratio between the number of instructions and the 11591 number of memory references to enable prefetching in a loop. 11592 11593 'use-canonical-types' 11594 Whether the compiler should use the "canonical" type system. 11595 Should always be 1, which uses a more efficient internal 11596 mechanism for comparing types in C++ and Objective-C++. 11597 However, if bugs in the canonical type system are causing 11598 compilation failures, set this value to 0 to disable canonical 11599 types. 11600 11601 'switch-conversion-max-branch-ratio' 11602 Switch initialization conversion refuses to create arrays that 11603 are bigger than 'switch-conversion-max-branch-ratio' times the 11604 number of branches in the switch. 11605 11606 'max-partial-antic-length' 11607 Maximum length of the partial antic set computed during the 11608 tree partial redundancy elimination optimization 11609 ('-ftree-pre') when optimizing at '-O3' and above. For some 11610 sorts of source code the enhanced partial redundancy 11611 elimination optimization can run away, consuming all of the 11612 memory available on the host machine. This parameter sets a 11613 limit on the length of the sets that are computed, which 11614 prevents the runaway behavior. Setting a value of 0 for this 11615 parameter allows an unlimited set length. 11616 11617 'rpo-vn-max-loop-depth' 11618 Maximum loop depth that is value-numbered optimistically. 11619 When the limit hits the innermost RPO-VN-MAX-LOOP-DEPTH loops 11620 and the outermost loop in the loop nest are value-numbered 11621 optimistically and the remaining ones not. 11622 11623 'sccvn-max-alias-queries-per-access' 11624 Maximum number of alias-oracle queries we perform when looking 11625 for redundancies for loads and stores. If this limit is hit 11626 the search is aborted and the load or store is not considered 11627 redundant. The number of queries is algorithmically limited 11628 to the number of stores on all paths from the load to the 11629 function entry. 11630 11631 'ira-max-loops-num' 11632 IRA uses regional register allocation by default. If a 11633 function contains more loops than the number given by this 11634 parameter, only at most the given number of the most 11635 frequently-executed loops form regions for regional register 11636 allocation. 11637 11638 'ira-max-conflict-table-size' 11639 Although IRA uses a sophisticated algorithm to compress the 11640 conflict table, the table can still require excessive amounts 11641 of memory for huge functions. If the conflict table for a 11642 function could be more than the size in MB given by this 11643 parameter, the register allocator instead uses a faster, 11644 simpler, and lower-quality algorithm that does not require 11645 building a pseudo-register conflict table. 11646 11647 'ira-loop-reserved-regs' 11648 IRA can be used to evaluate more accurate register pressure in 11649 loops for decisions to move loop invariants (see '-O3'). The 11650 number of available registers reserved for some other purposes 11651 is given by this parameter. Default of the parameter is the 11652 best found from numerous experiments. 11653 11654 'lra-inheritance-ebb-probability-cutoff' 11655 LRA tries to reuse values reloaded in registers in subsequent 11656 insns. This optimization is called inheritance. EBB is used 11657 as a region to do this optimization. The parameter defines a 11658 minimal fall-through edge probability in percentage used to 11659 add BB to inheritance EBB in LRA. The default value was chosen 11660 from numerous runs of SPEC2000 on x86-64. 11661 11662 'loop-invariant-max-bbs-in-loop' 11663 Loop invariant motion can be very expensive, both in 11664 compilation time and in amount of needed compile-time memory, 11665 with very large loops. Loops with more basic blocks than this 11666 parameter won't have loop invariant motion optimization 11667 performed on them. 11668 11669 'loop-max-datarefs-for-datadeps' 11670 Building data dependencies is expensive for very large loops. 11671 This parameter limits the number of data references in loops 11672 that are considered for data dependence analysis. These large 11673 loops are no handled by the optimizations using loop data 11674 dependencies. 11675 11676 'max-vartrack-size' 11677 Sets a maximum number of hash table slots to use during 11678 variable tracking dataflow analysis of any function. If this 11679 limit is exceeded with variable tracking at assignments 11680 enabled, analysis for that function is retried without it, 11681 after removing all debug insns from the function. If the 11682 limit is exceeded even without debug insns, var tracking 11683 analysis is completely disabled for the function. Setting the 11684 parameter to zero makes it unlimited. 11685 11686 'max-vartrack-expr-depth' 11687 Sets a maximum number of recursion levels when attempting to 11688 map variable names or debug temporaries to value expressions. 11689 This trades compilation time for more complete debug 11690 information. If this is set too low, value expressions that 11691 are available and could be represented in debug information 11692 may end up not being used; setting this higher may enable the 11693 compiler to find more complex debug expressions, but compile 11694 time and memory use may grow. 11695 11696 'max-debug-marker-count' 11697 Sets a threshold on the number of debug markers (e.g. begin 11698 stmt markers) to avoid complexity explosion at inlining or 11699 expanding to RTL. If a function has more such gimple stmts 11700 than the set limit, such stmts will be dropped from the 11701 inlined copy of a function, and from its RTL expansion. 11702 11703 'min-nondebug-insn-uid' 11704 Use uids starting at this parameter for nondebug insns. The 11705 range below the parameter is reserved exclusively for debug 11706 insns created by '-fvar-tracking-assignments', but debug insns 11707 may get (non-overlapping) uids above it if the reserved range 11708 is exhausted. 11709 11710 'ipa-sra-ptr-growth-factor' 11711 IPA-SRA replaces a pointer to an aggregate with one or more 11712 new parameters only when their cumulative size is less or 11713 equal to 'ipa-sra-ptr-growth-factor' times the size of the 11714 original pointer parameter. 11715 11716 'ipa-sra-max-replacements' 11717 Maximum pieces of an aggregate that IPA-SRA tracks. As a 11718 consequence, it is also the maximum number of replacements of 11719 a formal parameter. 11720 11721 'sra-max-scalarization-size-Ospeed' 11722 'sra-max-scalarization-size-Osize' 11723 The two Scalar Reduction of Aggregates passes (SRA and 11724 IPA-SRA) aim to replace scalar parts of aggregates with uses 11725 of independent scalar variables. These parameters control the 11726 maximum size, in storage units, of aggregate which is 11727 considered for replacement when compiling for speed 11728 ('sra-max-scalarization-size-Ospeed') or size 11729 ('sra-max-scalarization-size-Osize') respectively. 11730 11731 'sra-max-propagations' 11732 The maximum number of artificial accesses that Scalar 11733 Replacement of Aggregates (SRA) will track, per one local 11734 variable, in order to facilitate copy propagation. 11735 11736 'tm-max-aggregate-size' 11737 When making copies of thread-local variables in a transaction, 11738 this parameter specifies the size in bytes after which 11739 variables are saved with the logging functions as opposed to 11740 save/restore code sequence pairs. This option only applies 11741 when using '-fgnu-tm'. 11742 11743 'graphite-max-nb-scop-params' 11744 To avoid exponential effects in the Graphite loop transforms, 11745 the number of parameters in a Static Control Part (SCoP) is 11746 bounded. A value of zero can be used to lift the bound. A 11747 variable whose value is unknown at compilation time and 11748 defined outside a SCoP is a parameter of the SCoP. 11749 11750 'loop-block-tile-size' 11751 Loop blocking or strip mining transforms, enabled with 11752 '-floop-block' or '-floop-strip-mine', strip mine each loop in 11753 the loop nest by a given number of iterations. The strip 11754 length can be changed using the 'loop-block-tile-size' 11755 parameter. 11756 11757 'ipa-cp-value-list-size' 11758 IPA-CP attempts to track all possible values and types passed 11759 to a function's parameter in order to propagate them and 11760 perform devirtualization. 'ipa-cp-value-list-size' is the 11761 maximum number of values and types it stores per one formal 11762 parameter of a function. 11763 11764 'ipa-cp-eval-threshold' 11765 IPA-CP calculates its own score of cloning profitability 11766 heuristics and performs those cloning opportunities with 11767 scores that exceed 'ipa-cp-eval-threshold'. 11768 11769 'ipa-cp-max-recursive-depth' 11770 Maximum depth of recursive cloning for self-recursive 11771 function. 11772 11773 'ipa-cp-min-recursive-probability' 11774 Recursive cloning only when the probability of call being 11775 executed exceeds the parameter. 11776 11777 'ipa-cp-recursion-penalty' 11778 Percentage penalty the recursive functions will receive when 11779 they are evaluated for cloning. 11780 11781 'ipa-cp-single-call-penalty' 11782 Percentage penalty functions containing a single call to 11783 another function will receive when they are evaluated for 11784 cloning. 11785 11786 'ipa-max-agg-items' 11787 IPA-CP is also capable to propagate a number of scalar values 11788 passed in an aggregate. 'ipa-max-agg-items' controls the 11789 maximum number of such values per one parameter. 11790 11791 'ipa-cp-loop-hint-bonus' 11792 When IPA-CP determines that a cloning candidate would make the 11793 number of iterations of a loop known, it adds a bonus of 11794 'ipa-cp-loop-hint-bonus' to the profitability score of the 11795 candidate. 11796 11797 'ipa-max-aa-steps' 11798 During its analysis of function bodies, IPA-CP employs alias 11799 analysis in order to track values pointed to by function 11800 parameters. In order not spend too much time analyzing huge 11801 functions, it gives up and consider all memory clobbered after 11802 examining 'ipa-max-aa-steps' statements modifying memory. 11803 11804 'ipa-max-switch-predicate-bounds' 11805 Maximal number of boundary endpoints of case ranges of switch 11806 statement. For switch exceeding this limit, IPA-CP will not 11807 construct cloning cost predicate, which is used to estimate 11808 cloning benefit, for default case of the switch statement. 11809 11810 'ipa-max-param-expr-ops' 11811 IPA-CP will analyze conditional statement that references some 11812 function parameter to estimate benefit for cloning upon 11813 certain constant value. But if number of operations in a 11814 parameter expression exceeds 'ipa-max-param-expr-ops', the 11815 expression is treated as complicated one, and is not handled 11816 by IPA analysis. 11817 11818 'lto-partitions' 11819 Specify desired number of partitions produced during WHOPR 11820 compilation. The number of partitions should exceed the 11821 number of CPUs used for compilation. 11822 11823 'lto-min-partition' 11824 Size of minimal partition for WHOPR (in estimated 11825 instructions). This prevents expenses of splitting very small 11826 programs into too many partitions. 11827 11828 'lto-max-partition' 11829 Size of max partition for WHOPR (in estimated instructions). 11830 to provide an upper bound for individual size of partition. 11831 Meant to be used only with balanced partitioning. 11832 11833 'lto-max-streaming-parallelism' 11834 Maximal number of parallel processes used for LTO streaming. 11835 11836 'cxx-max-namespaces-for-diagnostic-help' 11837 The maximum number of namespaces to consult for suggestions 11838 when C++ name lookup fails for an identifier. 11839 11840 'sink-frequency-threshold' 11841 The maximum relative execution frequency (in percents) of the 11842 target block relative to a statement's original block to allow 11843 statement sinking of a statement. Larger numbers result in 11844 more aggressive statement sinking. A small positive 11845 adjustment is applied for statements with memory operands as 11846 those are even more profitable so sink. 11847 11848 'max-stores-to-sink' 11849 The maximum number of conditional store pairs that can be 11850 sunk. Set to 0 if either vectorization ('-ftree-vectorize') 11851 or if-conversion ('-ftree-loop-if-convert') is disabled. 11852 11853 'case-values-threshold' 11854 The smallest number of different values for which it is best 11855 to use a jump-table instead of a tree of conditional branches. 11856 If the value is 0, use the default for the machine. 11857 11858 'jump-table-max-growth-ratio-for-size' 11859 The maximum code size growth ratio when expanding into a jump 11860 table (in percent). The parameter is used when optimizing for 11861 size. 11862 11863 'jump-table-max-growth-ratio-for-speed' 11864 The maximum code size growth ratio when expanding into a jump 11865 table (in percent). The parameter is used when optimizing for 11866 speed. 11867 11868 'tree-reassoc-width' 11869 Set the maximum number of instructions executed in parallel in 11870 reassociated tree. This parameter overrides target dependent 11871 heuristics used by default if has non zero value. 11872 11873 'sched-pressure-algorithm' 11874 Choose between the two available implementations of 11875 '-fsched-pressure'. Algorithm 1 is the original 11876 implementation and is the more likely to prevent instructions 11877 from being reordered. Algorithm 2 was designed to be a 11878 compromise between the relatively conservative approach taken 11879 by algorithm 1 and the rather aggressive approach taken by the 11880 default scheduler. It relies more heavily on having a regular 11881 register file and accurate register pressure classes. See 11882 'haifa-sched.c' in the GCC sources for more details. 11883 11884 The default choice depends on the target. 11885 11886 'max-slsr-cand-scan' 11887 Set the maximum number of existing candidates that are 11888 considered when seeking a basis for a new straight-line 11889 strength reduction candidate. 11890 11891 'asan-globals' 11892 Enable buffer overflow detection for global objects. This 11893 kind of protection is enabled by default if you are using 11894 '-fsanitize=address' option. To disable global objects 11895 protection use '--param asan-globals=0'. 11896 11897 'asan-stack' 11898 Enable buffer overflow detection for stack objects. This kind 11899 of protection is enabled by default when using 11900 '-fsanitize=address'. To disable stack protection use 11901 '--param asan-stack=0' option. 11902 11903 'asan-instrument-reads' 11904 Enable buffer overflow detection for memory reads. This kind 11905 of protection is enabled by default when using 11906 '-fsanitize=address'. To disable memory reads protection use 11907 '--param asan-instrument-reads=0'. 11908 11909 'asan-instrument-writes' 11910 Enable buffer overflow detection for memory writes. This kind 11911 of protection is enabled by default when using 11912 '-fsanitize=address'. To disable memory writes protection use 11913 '--param asan-instrument-writes=0' option. 11914 11915 'asan-memintrin' 11916 Enable detection for built-in functions. This kind of 11917 protection is enabled by default when using 11918 '-fsanitize=address'. To disable built-in functions 11919 protection use '--param asan-memintrin=0'. 11920 11921 'asan-use-after-return' 11922 Enable detection of use-after-return. This kind of protection 11923 is enabled by default when using the '-fsanitize=address' 11924 option. To disable it use '--param asan-use-after-return=0'. 11925 11926 Note: By default the check is disabled at run time. To enable 11927 it, add 'detect_stack_use_after_return=1' to the environment 11928 variable 'ASAN_OPTIONS'. 11929 11930 'asan-instrumentation-with-call-threshold' 11931 If number of memory accesses in function being instrumented is 11932 greater or equal to this number, use callbacks instead of 11933 inline checks. E.g. to disable inline code use '--param 11934 asan-instrumentation-with-call-threshold=0'. 11935 11936 'use-after-scope-direct-emission-threshold' 11937 If the size of a local variable in bytes is smaller or equal 11938 to this number, directly poison (or unpoison) shadow memory 11939 instead of using run-time callbacks. 11940 11941 'max-fsm-thread-path-insns' 11942 Maximum number of instructions to copy when duplicating blocks 11943 on a finite state automaton jump thread path. 11944 11945 'max-fsm-thread-length' 11946 Maximum number of basic blocks on a finite state automaton 11947 jump thread path. 11948 11949 'max-fsm-thread-paths' 11950 Maximum number of new jump thread paths to create for a finite 11951 state automaton. 11952 11953 'parloops-chunk-size' 11954 Chunk size of omp schedule for loops parallelized by parloops. 11955 11956 'parloops-schedule' 11957 Schedule type of omp schedule for loops parallelized by 11958 parloops (static, dynamic, guided, auto, runtime). 11959 11960 'parloops-min-per-thread' 11961 The minimum number of iterations per thread of an innermost 11962 parallelized loop for which the parallelized variant is 11963 preferred over the single threaded one. Note that for a 11964 parallelized loop nest the minimum number of iterations of the 11965 outermost loop per thread is two. 11966 11967 'max-ssa-name-query-depth' 11968 Maximum depth of recursion when querying properties of SSA 11969 names in things like fold routines. One level of recursion 11970 corresponds to following a use-def chain. 11971 11972 'hsa-gen-debug-stores' 11973 Enable emission of special debug stores within HSA kernels 11974 which are then read and reported by libgomp plugin. 11975 Generation of these stores is disabled by default, use 11976 '--param hsa-gen-debug-stores=1' to enable it. 11977 11978 'max-speculative-devirt-maydefs' 11979 The maximum number of may-defs we analyze when looking for a 11980 must-def specifying the dynamic type of an object that invokes 11981 a virtual call we may be able to devirtualize speculatively. 11982 11983 'max-vrp-switch-assertions' 11984 The maximum number of assertions to add along the default edge 11985 of a switch statement during VRP. 11986 11987 'unroll-jam-min-percent' 11988 The minimum percentage of memory references that must be 11989 optimized away for the unroll-and-jam transformation to be 11990 considered profitable. 11991 11992 'unroll-jam-max-unroll' 11993 The maximum number of times the outer loop should be unrolled 11994 by the unroll-and-jam transformation. 11995 11996 'max-rtl-if-conversion-unpredictable-cost' 11997 Maximum permissible cost for the sequence that would be 11998 generated by the RTL if-conversion pass for a branch that is 11999 considered unpredictable. 12000 12001 'max-variable-expansions-in-unroller' 12002 If '-fvariable-expansion-in-unroller' is used, the maximum 12003 number of times that an individual variable will be expanded 12004 during loop unrolling. 12005 12006 'tracer-min-branch-probability-feedback' 12007 Stop forward growth if the probability of best edge is less 12008 than this threshold (in percent). Used when profile feedback 12009 is available. 12010 12011 'partial-inlining-entry-probability' 12012 Maximum probability of the entry BB of split region (in 12013 percent relative to entry BB of the function) to make partial 12014 inlining happen. 12015 12016 'max-tracked-strlens' 12017 Maximum number of strings for which strlen optimization pass 12018 will track string lengths. 12019 12020 'gcse-after-reload-partial-fraction' 12021 The threshold ratio for performing partial redundancy 12022 elimination after reload. 12023 12024 'gcse-after-reload-critical-fraction' 12025 The threshold ratio of critical edges execution count that 12026 permit performing redundancy elimination after reload. 12027 12028 'max-loop-header-insns' 12029 The maximum number of insns in loop header duplicated by the 12030 copy loop headers pass. 12031 12032 'vect-epilogues-nomask' 12033 Enable loop epilogue vectorization using smaller vector size. 12034 12035 'slp-max-insns-in-bb' 12036 Maximum number of instructions in basic block to be considered 12037 for SLP vectorization. 12038 12039 'avoid-fma-max-bits' 12040 Maximum number of bits for which we avoid creating FMAs. 12041 12042 'sms-loop-average-count-threshold' 12043 A threshold on the average loop count considered by the swing 12044 modulo scheduler. 12045 12046 'sms-dfa-history' 12047 The number of cycles the swing modulo scheduler considers when 12048 checking conflicts using DFA. 12049 12050 'max-inline-insns-recursive-auto' 12051 The maximum number of instructions non-inline function can 12052 grow to via recursive inlining. 12053 12054 'graphite-allow-codegen-errors' 12055 Whether codegen errors should be ICEs when '-fchecking'. 12056 12057 'sms-max-ii-factor' 12058 A factor for tuning the upper bound that swing modulo 12059 scheduler uses for scheduling a loop. 12060 12061 'lra-max-considered-reload-pseudos' 12062 The max number of reload pseudos which are considered during 12063 spilling a non-reload pseudo. 12064 12065 'max-pow-sqrt-depth' 12066 Maximum depth of sqrt chains to use when synthesizing 12067 exponentiation by a real constant. 12068 12069 'max-dse-active-local-stores' 12070 Maximum number of active local stores in RTL dead store 12071 elimination. 12072 12073 'asan-instrument-allocas' 12074 Enable asan allocas/VLAs protection. 12075 12076 'max-iterations-computation-cost' 12077 Bound on the cost of an expression to compute the number of 12078 iterations. 12079 12080 'max-isl-operations' 12081 Maximum number of isl operations, 0 means unlimited. 12082 12083 'graphite-max-arrays-per-scop' 12084 Maximum number of arrays per scop. 12085 12086 'max-vartrack-reverse-op-size' 12087 Max. size of loc list for which reverse ops should be added. 12088 12089 'tracer-dynamic-coverage-feedback' 12090 The percentage of function, weighted by execution frequency, 12091 that must be covered by trace formation. Used when profile 12092 feedback is available. 12093 12094 'max-inline-recursive-depth-auto' 12095 The maximum depth of recursive inlining for non-inline 12096 functions. 12097 12098 'fsm-scale-path-stmts' 12099 Scale factor to apply to the number of statements in a 12100 threading path when comparing to the number of (scaled) 12101 blocks. 12102 12103 'fsm-maximum-phi-arguments' 12104 Maximum number of arguments a PHI may have before the FSM 12105 threader will not try to thread through its block. 12106 12107 'uninit-control-dep-attempts' 12108 Maximum number of nested calls to search for control 12109 dependencies during uninitialized variable analysis. 12110 12111 'sra-max-scalarization-size-Osize' 12112 Maximum size, in storage units, of an aggregate which should 12113 be considered for scalarization when compiling for size. 12114 12115 'fsm-scale-path-blocks' 12116 Scale factor to apply to the number of blocks in a threading 12117 path when comparing to the number of (scaled) statements. 12118 12119 'sched-autopref-queue-depth' 12120 Hardware autoprefetcher scheduler model control flag. Number 12121 of lookahead cycles the model looks into; at ' ' only enable 12122 instruction sorting heuristic. 12123 12124 'loop-versioning-max-inner-insns' 12125 The maximum number of instructions that an inner loop can have 12126 before the loop versioning pass considers it too big to copy. 12127 12128 'loop-versioning-max-outer-insns' 12129 The maximum number of instructions that an outer loop can have 12130 before the loop versioning pass considers it too big to copy, 12131 discounting any instructions in inner loops that directly 12132 benefit from versioning. 12133 12134 'ssa-name-def-chain-limit' 12135 The maximum number of SSA_NAME assignments to follow in 12136 determining a property of a variable such as its value. This 12137 limits the number of iterations or recursive calls GCC 12138 performs when optimizing certain statements or when 12139 determining their validity prior to issuing diagnostics. 12140 12141 'store-merging-max-size' 12142 Maximum size of a single store merging region in bytes. 12143 12144 'hash-table-verification-limit' 12145 The number of elements for which hash table verification is 12146 done for each searched element. 12147 12148 'max-find-base-term-values' 12149 Maximum number of VALUEs handled during a single 12150 find_base_term call. 12151 12152 'analyzer-max-enodes-per-program-point' 12153 The maximum number of exploded nodes per program point within 12154 the analyzer, before terminating analysis of that point. 12155 12156 'analyzer-min-snodes-for-call-summary' 12157 The minimum number of supernodes within a function for the 12158 analyzer to consider summarizing its effects at call sites. 12159 12160 'analyzer-max-recursion-depth' 12161 The maximum number of times a callsite can appear in a call 12162 stack within the analyzer, before terminating analysis of a 12163 call that would recurse deeper. 12164 12165 'gimple-fe-computed-hot-bb-threshold' 12166 The number of executions of a basic block which is considered 12167 hot. The parameter is used only in GIMPLE FE. 12168 12169 'analyzer-bb-explosion-factor' 12170 The maximum number of 'after supernode' exploded nodes within 12171 the analyzer per supernode, before terminating analysis. 12172 12173 The following choices of NAME are available on AArch64 targets: 12174 12175 'aarch64-sve-compare-costs' 12176 When vectorizing for SVE, consider using "unpacked" vectors 12177 for smaller elements and use the cost model to pick the 12178 cheapest approach. Also use the cost model to choose between 12179 SVE and Advanced SIMD vectorization. 12180 12181 Using unpacked vectors includes storing smaller elements in 12182 larger containers and accessing elements with extending loads 12183 and truncating stores. 12184 12185 'aarch64-float-recp-precision' 12186 The number of Newton iterations for calculating the reciprocal 12187 for float type. The precision of division is proportional to 12188 this param when division approximation is enabled. The 12189 default value is 1. 12190 12191 'aarch64-double-recp-precision' 12192 The number of Newton iterations for calculating the reciprocal 12193 for double type. The precision of division is propotional to 12194 this param when division approximation is enabled. The 12195 default value is 2. 12196 12197 'aarch64-autovec-preference' 12198 Force an ISA selection strategy for auto-vectorization. 12199 Accepts values from 0 to 4, inclusive. 12200 '0' 12201 Use the default heuristics. 12202 '1' 12203 Use only Advanced SIMD for auto-vectorization. 12204 '2' 12205 Use only SVE for auto-vectorization. 12206 '3' 12207 Use both Advanced SIMD and SVE. Prefer Advanced SIMD when 12208 the costs are deemed equal. 12209 '4' 12210 Use both Advanced SIMD and SVE. Prefer SVE when the costs 12211 are deemed equal. 12212 The default value is 0. 12213 12214 12215File: gcc.info, Node: Instrumentation Options, Next: Preprocessor Options, Prev: Optimize Options, Up: Invoking GCC 12216 122173.12 Program Instrumentation Options 12218==================================== 12219 12220GCC supports a number of command-line options that control adding 12221run-time instrumentation to the code it normally generates. For 12222example, one purpose of instrumentation is collect profiling statistics 12223for use in finding program hot spots, code coverage analysis, or 12224profile-guided optimizations. Another class of program instrumentation 12225is adding run-time checking to detect programming errors like invalid 12226pointer dereferences or out-of-bounds array accesses, as well as 12227deliberately hostile attacks such as stack smashing or C++ vtable 12228hijacking. There is also a general hook which can be used to implement 12229other forms of tracing or function-level instrumentation for debug or 12230program analysis purposes. 12231 12232'-p' 12233'-pg' 12234 Generate extra code to write profile information suitable for the 12235 analysis program 'prof' (for '-p') or 'gprof' (for '-pg'). You 12236 must use this option when compiling the source files you want data 12237 about, and you must also use it when linking. 12238 12239 You can use the function attribute 'no_instrument_function' to 12240 suppress profiling of individual functions when compiling with 12241 these options. *Note Common Function Attributes::. 12242 12243'-fprofile-arcs' 12244 Add code so that program flow "arcs" are instrumented. During 12245 execution the program records how many times each branch and call 12246 is executed and how many times it is taken or returns. On targets 12247 that support constructors with priority support, profiling properly 12248 handles constructors, destructors and C++ constructors (and 12249 destructors) of classes which are used as a type of a global 12250 variable. 12251 12252 When the compiled program exits it saves this data to a file called 12253 'AUXNAME.gcda' for each source file. The data may be used for 12254 profile-directed optimizations ('-fbranch-probabilities'), or for 12255 test coverage analysis ('-ftest-coverage'). Each object file's 12256 AUXNAME is generated from the name of the output file, if 12257 explicitly specified and it is not the final executable, otherwise 12258 it is the basename of the source file. In both cases any suffix is 12259 removed (e.g. 'foo.gcda' for input file 'dir/foo.c', or 12260 'dir/foo.gcda' for output file specified as '-o dir/foo.o'). *Note 12261 Cross-profiling::. 12262 12263'--coverage' 12264 12265 This option is used to compile and link code instrumented for 12266 coverage analysis. The option is a synonym for '-fprofile-arcs' 12267 '-ftest-coverage' (when compiling) and '-lgcov' (when linking). 12268 See the documentation for those options for more details. 12269 12270 * Compile the source files with '-fprofile-arcs' plus 12271 optimization and code generation options. For test coverage 12272 analysis, use the additional '-ftest-coverage' option. You do 12273 not need to profile every source file in a program. 12274 12275 * Compile the source files additionally with 12276 '-fprofile-abs-path' to create absolute path names in the 12277 '.gcno' files. This allows 'gcov' to find the correct sources 12278 in projects where compilations occur with different working 12279 directories. 12280 12281 * Link your object files with '-lgcov' or '-fprofile-arcs' (the 12282 latter implies the former). 12283 12284 * Run the program on a representative workload to generate the 12285 arc profile information. This may be repeated any number of 12286 times. You can run concurrent instances of your program, and 12287 provided that the file system supports locking, the data files 12288 will be correctly updated. Unless a strict ISO C dialect 12289 option is in effect, 'fork' calls are detected and correctly 12290 handled without double counting. 12291 12292 * For profile-directed optimizations, compile the source files 12293 again with the same optimization and code generation options 12294 plus '-fbranch-probabilities' (*note Options that Control 12295 Optimization: Optimize Options.). 12296 12297 * For test coverage analysis, use 'gcov' to produce human 12298 readable information from the '.gcno' and '.gcda' files. 12299 Refer to the 'gcov' documentation for further information. 12300 12301 With '-fprofile-arcs', for each function of your program GCC 12302 creates a program flow graph, then finds a spanning tree for the 12303 graph. Only arcs that are not on the spanning tree have to be 12304 instrumented: the compiler adds code to count the number of times 12305 that these arcs are executed. When an arc is the only exit or only 12306 entrance to a block, the instrumentation code can be added to the 12307 block; otherwise, a new basic block must be created to hold the 12308 instrumentation code. 12309 12310'-ftest-coverage' 12311 Produce a notes file that the 'gcov' code-coverage utility (*note 12312 'gcov'--a Test Coverage Program: Gcov.) can use to show program 12313 coverage. Each source file's note file is called 'AUXNAME.gcno'. 12314 Refer to the '-fprofile-arcs' option above for a description of 12315 AUXNAME and instructions on how to generate test coverage data. 12316 Coverage data matches the source files more closely if you do not 12317 optimize. 12318 12319'-fprofile-abs-path' 12320 Automatically convert relative source file names to absolute path 12321 names in the '.gcno' files. This allows 'gcov' to find the correct 12322 sources in projects where compilations occur with different working 12323 directories. 12324 12325'-fprofile-dir=PATH' 12326 12327 Set the directory to search for the profile data files in to PATH. 12328 This option affects only the profile data generated by 12329 '-fprofile-generate', '-ftest-coverage', '-fprofile-arcs' and used 12330 by '-fprofile-use' and '-fbranch-probabilities' and its related 12331 options. Both absolute and relative paths can be used. By 12332 default, GCC uses the current directory as PATH, thus the profile 12333 data file appears in the same directory as the object file. In 12334 order to prevent the file name clashing, if the object file name is 12335 not an absolute path, we mangle the absolute path of the 12336 'SOURCENAME.gcda' file and use it as the file name of a '.gcda' 12337 file. See similar option '-fprofile-note'. 12338 12339 When an executable is run in a massive parallel environment, it is 12340 recommended to save profile to different folders. That can be done 12341 with variables in PATH that are exported during run-time: 12342 12343 '%p' 12344 process ID. 12345 12346 '%q{VAR}' 12347 value of environment variable VAR 12348 12349'-fprofile-generate' 12350'-fprofile-generate=PATH' 12351 12352 Enable options usually used for instrumenting application to 12353 produce profile useful for later recompilation with profile 12354 feedback based optimization. You must use '-fprofile-generate' 12355 both when compiling and when linking your program. 12356 12357 The following options are enabled: '-fprofile-arcs', 12358 '-fprofile-values', '-finline-functions', and '-fipa-bit-cp'. 12359 12360 If PATH is specified, GCC looks at the PATH to find the profile 12361 feedback data files. See '-fprofile-dir'. 12362 12363 To optimize the program based on the collected profile information, 12364 use '-fprofile-use'. *Note Optimize Options::, for more 12365 information. 12366 12367'-fprofile-note=PATH' 12368 12369 If PATH is specified, GCC saves '.gcno' file into PATH location. 12370 If you combine the option with multiple source files, the '.gcno' 12371 file will be overwritten. 12372 12373'-fprofile-prefix-path=PATH' 12374 12375 This option can be used in combination with 12376 'profile-generate='PROFILE_DIR and 'profile-use='PROFILE_DIR to 12377 inform GCC where is the base directory of built source tree. By 12378 default PROFILE_DIR will contain files with mangled absolute paths 12379 of all object files in the built project. This is not desirable 12380 when directory used to build the instrumented binary differs from 12381 the directory used to build the binary optimized with profile 12382 feedback because the profile data will not be found during the 12383 optimized build. In such setups '-fprofile-prefix-path='PATH with 12384 PATH pointing to the base directory of the build can be used to 12385 strip the irrelevant part of the path and keep all file names 12386 relative to the main build directory. 12387 12388'-fprofile-update=METHOD' 12389 12390 Alter the update method for an application instrumented for profile 12391 feedback based optimization. The METHOD argument should be one of 12392 'single', 'atomic' or 'prefer-atomic'. The first one is useful for 12393 single-threaded applications, while the second one prevents profile 12394 corruption by emitting thread-safe code. 12395 12396 *Warning:* When an application does not properly join all threads 12397 (or creates an detached thread), a profile file can be still 12398 corrupted. 12399 12400 Using 'prefer-atomic' would be transformed either to 'atomic', when 12401 supported by a target, or to 'single' otherwise. The GCC driver 12402 automatically selects 'prefer-atomic' when '-pthread' is present in 12403 the command line. 12404 12405'-fprofile-filter-files=REGEX' 12406 12407 Instrument only functions from files where names match any regular 12408 expression (separated by a semi-colon). 12409 12410 For example, '-fprofile-filter-files=main.c;module.*.c' will 12411 instrument only 'main.c' and all C files starting with 'module'. 12412 12413'-fprofile-exclude-files=REGEX' 12414 12415 Instrument only functions from files where names do not match all 12416 the regular expressions (separated by a semi-colon). 12417 12418 For example, '-fprofile-exclude-files=/usr/*' will prevent 12419 instrumentation of all files that are located in '/usr/' folder. 12420 12421'-fprofile-reproducible=[multithreaded|parallel-runs|serial]' 12422 Control level of reproducibility of profile gathered by 12423 '-fprofile-generate'. This makes it possible to rebuild program 12424 with same outcome which is useful, for example, for distribution 12425 packages. 12426 12427 With '-fprofile-reproducible=serial' the profile gathered by 12428 '-fprofile-generate' is reproducible provided the trained program 12429 behaves the same at each invocation of the train run, it is not 12430 multi-threaded and profile data streaming is always done in the 12431 same order. Note that profile streaming happens at the end of 12432 program run but also before 'fork' function is invoked. 12433 12434 Note that it is quite common that execution counts of some part of 12435 programs depends, for example, on length of temporary file names or 12436 memory space randomization (that may affect hash-table collision 12437 rate). Such non-reproducible part of programs may be annotated by 12438 'no_instrument_function' function attribute. 'gcov-dump' with '-l' 12439 can be used to dump gathered data and verify that they are indeed 12440 reproducible. 12441 12442 With '-fprofile-reproducible=parallel-runs' collected profile stays 12443 reproducible regardless the order of streaming of the data into 12444 gcda files. This setting makes it possible to run multiple 12445 instances of instrumented program in parallel (such as with 'make 12446 -j'). This reduces quality of gathered data, in particular of 12447 indirect call profiling. 12448 12449'-fsanitize=address' 12450 Enable AddressSanitizer, a fast memory error detector. Memory 12451 access instructions are instrumented to detect out-of-bounds and 12452 use-after-free bugs. The option enables 12453 '-fsanitize-address-use-after-scope'. See 12454 <https://github.com/google/sanitizers/wiki/AddressSanitizer> for 12455 more details. The run-time behavior can be influenced using the 12456 'ASAN_OPTIONS' environment variable. When set to 'help=1', the 12457 available options are shown at startup of the instrumented program. 12458 See 12459 <https://github.com/google/sanitizers/wiki/AddressSanitizerFlags#run-time-flags> 12460 for a list of supported options. The option cannot be combined 12461 with '-fsanitize=thread'. 12462 12463'-fsanitize=kernel-address' 12464 Enable AddressSanitizer for Linux kernel. See 12465 <https://github.com/google/kasan/wiki> for more details. 12466 12467'-fsanitize=pointer-compare' 12468 Instrument comparison operation (<, <=, >, >=) with pointer 12469 operands. The option must be combined with either 12470 '-fsanitize=kernel-address' or '-fsanitize=address' The option 12471 cannot be combined with '-fsanitize=thread'. Note: By default the 12472 check is disabled at run time. To enable it, add 12473 'detect_invalid_pointer_pairs=2' to the environment variable 12474 'ASAN_OPTIONS'. Using 'detect_invalid_pointer_pairs=1' detects 12475 invalid operation only when both pointers are non-null. 12476 12477'-fsanitize=pointer-subtract' 12478 Instrument subtraction with pointer operands. The option must be 12479 combined with either '-fsanitize=kernel-address' or 12480 '-fsanitize=address' The option cannot be combined with 12481 '-fsanitize=thread'. Note: By default the check is disabled at run 12482 time. To enable it, add 'detect_invalid_pointer_pairs=2' to the 12483 environment variable 'ASAN_OPTIONS'. Using 12484 'detect_invalid_pointer_pairs=1' detects invalid operation only 12485 when both pointers are non-null. 12486 12487'-fsanitize=thread' 12488 Enable ThreadSanitizer, a fast data race detector. Memory access 12489 instructions are instrumented to detect data race bugs. See 12490 <https://github.com/google/sanitizers/wiki#threadsanitizer> for 12491 more details. The run-time behavior can be influenced using the 12492 'TSAN_OPTIONS' environment variable; see 12493 <https://github.com/google/sanitizers/wiki/ThreadSanitizerFlags> 12494 for a list of supported options. The option cannot be combined 12495 with '-fsanitize=address', '-fsanitize=leak'. 12496 12497 Note that sanitized atomic builtins cannot throw exceptions when 12498 operating on invalid memory addresses with non-call exceptions 12499 ('-fnon-call-exceptions'). 12500 12501'-fsanitize=leak' 12502 Enable LeakSanitizer, a memory leak detector. This option only 12503 matters for linking of executables and the executable is linked 12504 against a library that overrides 'malloc' and other allocator 12505 functions. See 12506 <https://github.com/google/sanitizers/wiki/AddressSanitizerLeakSanitizer> 12507 for more details. The run-time behavior can be influenced using 12508 the 'LSAN_OPTIONS' environment variable. The option cannot be 12509 combined with '-fsanitize=thread'. 12510 12511'-fsanitize=undefined' 12512 Enable UndefinedBehaviorSanitizer, a fast undefined behavior 12513 detector. Various computations are instrumented to detect 12514 undefined behavior at runtime. See 12515 <https://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html> for 12516 more details. The run-time behavior can be influenced using the 12517 'UBSAN_OPTIONS' environment variable. Current suboptions are: 12518 12519 '-fsanitize=shift' 12520 This option enables checking that the result of a shift 12521 operation is not undefined. Note that what exactly is 12522 considered undefined differs slightly between C and C++, as 12523 well as between ISO C90 and C99, etc. This option has two 12524 suboptions, '-fsanitize=shift-base' and 12525 '-fsanitize=shift-exponent'. 12526 12527 '-fsanitize=shift-exponent' 12528 This option enables checking that the second argument of a 12529 shift operation is not negative and is smaller than the 12530 precision of the promoted first argument. 12531 12532 '-fsanitize=shift-base' 12533 If the second argument of a shift operation is within range, 12534 check that the result of a shift operation is not undefined. 12535 Note that what exactly is considered undefined differs 12536 slightly between C and C++, as well as between ISO C90 and 12537 C99, etc. 12538 12539 '-fsanitize=integer-divide-by-zero' 12540 Detect integer division by zero as well as 'INT_MIN / -1' 12541 division. 12542 12543 '-fsanitize=unreachable' 12544 With this option, the compiler turns the 12545 '__builtin_unreachable' call into a diagnostics message call 12546 instead. When reaching the '__builtin_unreachable' call, the 12547 behavior is undefined. 12548 12549 '-fsanitize=vla-bound' 12550 This option instructs the compiler to check that the size of a 12551 variable length array is positive. 12552 12553 '-fsanitize=null' 12554 This option enables pointer checking. Particularly, the 12555 application built with this option turned on will issue an 12556 error message when it tries to dereference a NULL pointer, or 12557 if a reference (possibly an rvalue reference) is bound to a 12558 NULL pointer, or if a method is invoked on an object pointed 12559 by a NULL pointer. 12560 12561 '-fsanitize=return' 12562 This option enables return statement checking. Programs built 12563 with this option turned on will issue an error message when 12564 the end of a non-void function is reached without actually 12565 returning a value. This option works in C++ only. 12566 12567 '-fsanitize=signed-integer-overflow' 12568 This option enables signed integer overflow checking. We 12569 check that the result of '+', '*', and both unary and binary 12570 '-' does not overflow in the signed arithmetics. Note, 12571 integer promotion rules must be taken into account. That is, 12572 the following is not an overflow: 12573 signed char a = SCHAR_MAX; 12574 a++; 12575 12576 '-fsanitize=bounds' 12577 This option enables instrumentation of array bounds. Various 12578 out of bounds accesses are detected. Flexible array members, 12579 flexible array member-like arrays, and initializers of 12580 variables with static storage are not instrumented. 12581 12582 '-fsanitize=bounds-strict' 12583 This option enables strict instrumentation of array bounds. 12584 Most out of bounds accesses are detected, including flexible 12585 array members and flexible array member-like arrays. 12586 Initializers of variables with static storage are not 12587 instrumented. 12588 12589 '-fsanitize=alignment' 12590 12591 This option enables checking of alignment of pointers when 12592 they are dereferenced, or when a reference is bound to 12593 insufficiently aligned target, or when a method or constructor 12594 is invoked on insufficiently aligned object. 12595 12596 '-fsanitize=object-size' 12597 This option enables instrumentation of memory references using 12598 the '__builtin_object_size' function. Various out of bounds 12599 pointer accesses are detected. 12600 12601 '-fsanitize=float-divide-by-zero' 12602 Detect floating-point division by zero. Unlike other similar 12603 options, '-fsanitize=float-divide-by-zero' is not enabled by 12604 '-fsanitize=undefined', since floating-point division by zero 12605 can be a legitimate way of obtaining infinities and NaNs. 12606 12607 '-fsanitize=float-cast-overflow' 12608 This option enables floating-point type to integer conversion 12609 checking. We check that the result of the conversion does not 12610 overflow. Unlike other similar options, 12611 '-fsanitize=float-cast-overflow' is not enabled by 12612 '-fsanitize=undefined'. This option does not work well with 12613 'FE_INVALID' exceptions enabled. 12614 12615 '-fsanitize=nonnull-attribute' 12616 12617 This option enables instrumentation of calls, checking whether 12618 null values are not passed to arguments marked as requiring a 12619 non-null value by the 'nonnull' function attribute. 12620 12621 '-fsanitize=returns-nonnull-attribute' 12622 12623 This option enables instrumentation of return statements in 12624 functions marked with 'returns_nonnull' function attribute, to 12625 detect returning of null values from such functions. 12626 12627 '-fsanitize=bool' 12628 12629 This option enables instrumentation of loads from bool. If a 12630 value other than 0/1 is loaded, a run-time error is issued. 12631 12632 '-fsanitize=enum' 12633 12634 This option enables instrumentation of loads from an enum 12635 type. If a value outside the range of values for the enum 12636 type is loaded, a run-time error is issued. 12637 12638 '-fsanitize=vptr' 12639 12640 This option enables instrumentation of C++ member function 12641 calls, member accesses and some conversions between pointers 12642 to base and derived classes, to verify the referenced object 12643 has the correct dynamic type. 12644 12645 '-fsanitize=pointer-overflow' 12646 12647 This option enables instrumentation of pointer arithmetics. 12648 If the pointer arithmetics overflows, a run-time error is 12649 issued. 12650 12651 '-fsanitize=builtin' 12652 12653 This option enables instrumentation of arguments to selected 12654 builtin functions. If an invalid value is passed to such 12655 arguments, a run-time error is issued. E.g. passing 0 as the 12656 argument to '__builtin_ctz' or '__builtin_clz' invokes 12657 undefined behavior and is diagnosed by this option. 12658 12659 While '-ftrapv' causes traps for signed overflows to be emitted, 12660 '-fsanitize=undefined' gives a diagnostic message. This currently 12661 works only for the C family of languages. 12662 12663'-fno-sanitize=all' 12664 12665 This option disables all previously enabled sanitizers. 12666 '-fsanitize=all' is not allowed, as some sanitizers cannot be used 12667 together. 12668 12669'-fasan-shadow-offset=NUMBER' 12670 This option forces GCC to use custom shadow offset in 12671 AddressSanitizer checks. It is useful for experimenting with 12672 different shadow memory layouts in Kernel AddressSanitizer. 12673 12674'-fsanitize-sections=S1,S2,...' 12675 Sanitize global variables in selected user-defined sections. SI 12676 may contain wildcards. 12677 12678'-fsanitize-recover[=OPTS]' 12679 '-fsanitize-recover=' controls error recovery mode for sanitizers 12680 mentioned in comma-separated list of OPTS. Enabling this option 12681 for a sanitizer component causes it to attempt to continue running 12682 the program as if no error happened. This means multiple runtime 12683 errors can be reported in a single program run, and the exit code 12684 of the program may indicate success even when errors have been 12685 reported. The '-fno-sanitize-recover=' option can be used to alter 12686 this behavior: only the first detected error is reported and 12687 program then exits with a non-zero exit code. 12688 12689 Currently this feature only works for '-fsanitize=undefined' (and 12690 its suboptions except for '-fsanitize=unreachable' and 12691 '-fsanitize=return'), '-fsanitize=float-cast-overflow', 12692 '-fsanitize=float-divide-by-zero', '-fsanitize=bounds-strict', 12693 '-fsanitize=kernel-address' and '-fsanitize=address'. For these 12694 sanitizers error recovery is turned on by default, except 12695 '-fsanitize=address', for which this feature is experimental. 12696 '-fsanitize-recover=all' and '-fno-sanitize-recover=all' is also 12697 accepted, the former enables recovery for all sanitizers that 12698 support it, the latter disables recovery for all sanitizers that 12699 support it. 12700 12701 Even if a recovery mode is turned on the compiler side, it needs to 12702 be also enabled on the runtime library side, otherwise the failures 12703 are still fatal. The runtime library defaults to 'halt_on_error=0' 12704 for ThreadSanitizer and UndefinedBehaviorSanitizer, while default 12705 value for AddressSanitizer is 'halt_on_error=1'. This can be 12706 overridden through setting the 'halt_on_error' flag in the 12707 corresponding environment variable. 12708 12709 Syntax without an explicit OPTS parameter is deprecated. It is 12710 equivalent to specifying an OPTS list of: 12711 12712 undefined,float-cast-overflow,float-divide-by-zero,bounds-strict 12713 12714'-fsanitize-address-use-after-scope' 12715 Enable sanitization of local variables to detect use-after-scope 12716 bugs. The option sets '-fstack-reuse' to 'none'. 12717 12718'-fsanitize-undefined-trap-on-error' 12719 The '-fsanitize-undefined-trap-on-error' option instructs the 12720 compiler to report undefined behavior using '__builtin_trap' rather 12721 than a 'libubsan' library routine. The advantage of this is that 12722 the 'libubsan' library is not needed and is not linked in, so this 12723 is usable even in freestanding environments. 12724 12725'-fsanitize-coverage=trace-pc' 12726 Enable coverage-guided fuzzing code instrumentation. Inserts a 12727 call to '__sanitizer_cov_trace_pc' into every basic block. 12728 12729'-fsanitize-coverage=trace-cmp' 12730 Enable dataflow guided fuzzing code instrumentation. Inserts a 12731 call to '__sanitizer_cov_trace_cmp1', '__sanitizer_cov_trace_cmp2', 12732 '__sanitizer_cov_trace_cmp4' or '__sanitizer_cov_trace_cmp8' for 12733 integral comparison with both operands variable or 12734 '__sanitizer_cov_trace_const_cmp1', 12735 '__sanitizer_cov_trace_const_cmp2', 12736 '__sanitizer_cov_trace_const_cmp4' or 12737 '__sanitizer_cov_trace_const_cmp8' for integral comparison with one 12738 operand constant, '__sanitizer_cov_trace_cmpf' or 12739 '__sanitizer_cov_trace_cmpd' for float or double comparisons and 12740 '__sanitizer_cov_trace_switch' for switch statements. 12741 12742'-fcf-protection=[full|branch|return|none|check]' 12743 Enable code instrumentation of control-flow transfers to increase 12744 program security by checking that target addresses of control-flow 12745 transfer instructions (such as indirect function call, function 12746 return, indirect jump) are valid. This prevents diverting the flow 12747 of control to an unexpected target. This is intended to protect 12748 against such threats as Return-oriented Programming (ROP), and 12749 similarly call/jmp-oriented programming (COP/JOP). 12750 12751 The value 'branch' tells the compiler to implement checking of 12752 validity of control-flow transfer at the point of indirect branch 12753 instructions, i.e. call/jmp instructions. The value 'return' 12754 implements checking of validity at the point of returning from a 12755 function. The value 'full' is an alias for specifying both 12756 'branch' and 'return'. The value 'none' turns off instrumentation. 12757 12758 The value 'check' is used for the final link with link-time 12759 optimization (LTO). An error is issued if LTO object files are 12760 compiled with different '-fcf-protection' values. The value 12761 'check' is ignored at the compile time. 12762 12763 The macro '__CET__' is defined when '-fcf-protection' is used. The 12764 first bit of '__CET__' is set to 1 for the value 'branch' and the 12765 second bit of '__CET__' is set to 1 for the 'return'. 12766 12767 You can also use the 'nocf_check' attribute to identify which 12768 functions and calls should be skipped from instrumentation (*note 12769 Function Attributes::). 12770 12771 Currently the x86 GNU/Linux target provides an implementation based 12772 on Intel Control-flow Enforcement Technology (CET) which works for 12773 i686 processor or newer. 12774 12775'-fstack-protector' 12776 Emit extra code to check for buffer overflows, such as stack 12777 smashing attacks. This is done by adding a guard variable to 12778 functions with vulnerable objects. This includes functions that 12779 call 'alloca', and functions with buffers larger than or equal to 8 12780 bytes. The guards are initialized when a function is entered and 12781 then checked when the function exits. If a guard check fails, an 12782 error message is printed and the program exits. Only variables 12783 that are actually allocated on the stack are considered, optimized 12784 away variables or variables allocated in registers don't count. 12785 12786'-fstack-protector-all' 12787 Like '-fstack-protector' except that all functions are protected. 12788 12789'-fstack-protector-strong' 12790 Like '-fstack-protector' but includes additional functions to be 12791 protected -- those that have local array definitions, or have 12792 references to local frame addresses. Only variables that are 12793 actually allocated on the stack are considered, optimized away 12794 variables or variables allocated in registers don't count. 12795 12796'-fstack-protector-explicit' 12797 Like '-fstack-protector' but only protects those functions which 12798 have the 'stack_protect' attribute. 12799 12800'-fstack-check' 12801 Generate code to verify that you do not go beyond the boundary of 12802 the stack. You should specify this flag if you are running in an 12803 environment with multiple threads, but you only rarely need to 12804 specify it in a single-threaded environment since stack overflow is 12805 automatically detected on nearly all systems if there is only one 12806 stack. 12807 12808 Note that this switch does not actually cause checking to be done; 12809 the operating system or the language runtime must do that. The 12810 switch causes generation of code to ensure that they see the stack 12811 being extended. 12812 12813 You can additionally specify a string parameter: 'no' means no 12814 checking, 'generic' means force the use of old-style checking, 12815 'specific' means use the best checking method and is equivalent to 12816 bare '-fstack-check'. 12817 12818 Old-style checking is a generic mechanism that requires no specific 12819 target support in the compiler but comes with the following 12820 drawbacks: 12821 12822 1. Modified allocation strategy for large objects: they are 12823 always allocated dynamically if their size exceeds a fixed 12824 threshold. Note this may change the semantics of some code. 12825 12826 2. Fixed limit on the size of the static frame of functions: when 12827 it is topped by a particular function, stack checking is not 12828 reliable and a warning is issued by the compiler. 12829 12830 3. Inefficiency: because of both the modified allocation strategy 12831 and the generic implementation, code performance is hampered. 12832 12833 Note that old-style stack checking is also the fallback method for 12834 'specific' if no target support has been added in the compiler. 12835 12836 '-fstack-check=' is designed for Ada's needs to detect infinite 12837 recursion and stack overflows. 'specific' is an excellent choice 12838 when compiling Ada code. It is not generally sufficient to protect 12839 against stack-clash attacks. To protect against those you want 12840 '-fstack-clash-protection'. 12841 12842'-fstack-clash-protection' 12843 Generate code to prevent stack clash style attacks. When this 12844 option is enabled, the compiler will only allocate one page of 12845 stack space at a time and each page is accessed immediately after 12846 allocation. Thus, it prevents allocations from jumping over any 12847 stack guard page provided by the operating system. 12848 12849 Most targets do not fully support stack clash protection. However, 12850 on those targets '-fstack-clash-protection' will protect dynamic 12851 stack allocations. '-fstack-clash-protection' may also provide 12852 limited protection for static stack allocations if the target 12853 supports '-fstack-check=specific'. 12854 12855'-fstack-limit-register=REG' 12856'-fstack-limit-symbol=SYM' 12857'-fno-stack-limit' 12858 Generate code to ensure that the stack does not grow beyond a 12859 certain value, either the value of a register or the address of a 12860 symbol. If a larger stack is required, a signal is raised at run 12861 time. For most targets, the signal is raised before the stack 12862 overruns the boundary, so it is possible to catch the signal 12863 without taking special precautions. 12864 12865 For instance, if the stack starts at absolute address '0x80000000' 12866 and grows downwards, you can use the flags 12867 '-fstack-limit-symbol=__stack_limit' and 12868 '-Wl,--defsym,__stack_limit=0x7ffe0000' to enforce a stack limit of 12869 128KB. Note that this may only work with the GNU linker. 12870 12871 You can locally override stack limit checking by using the 12872 'no_stack_limit' function attribute (*note Function Attributes::). 12873 12874'-fsplit-stack' 12875 Generate code to automatically split the stack before it overflows. 12876 The resulting program has a discontiguous stack which can only 12877 overflow if the program is unable to allocate any more memory. 12878 This is most useful when running threaded programs, as it is no 12879 longer necessary to calculate a good stack size to use for each 12880 thread. This is currently only implemented for the x86 targets 12881 running GNU/Linux. 12882 12883 When code compiled with '-fsplit-stack' calls code compiled without 12884 '-fsplit-stack', there may not be much stack space available for 12885 the latter code to run. If compiling all code, including library 12886 code, with '-fsplit-stack' is not an option, then the linker can 12887 fix up these calls so that the code compiled without 12888 '-fsplit-stack' always has a large stack. Support for this is 12889 implemented in the gold linker in GNU binutils release 2.21 and 12890 later. 12891 12892'-fvtable-verify=[std|preinit|none]' 12893 This option is only available when compiling C++ code. It turns on 12894 (or off, if using '-fvtable-verify=none') the security feature that 12895 verifies at run time, for every virtual call, that the vtable 12896 pointer through which the call is made is valid for the type of the 12897 object, and has not been corrupted or overwritten. If an invalid 12898 vtable pointer is detected at run time, an error is reported and 12899 execution of the program is immediately halted. 12900 12901 This option causes run-time data structures to be built at program 12902 startup, which are used for verifying the vtable pointers. The 12903 options 'std' and 'preinit' control the timing of when these data 12904 structures are built. In both cases the data structures are built 12905 before execution reaches 'main'. Using '-fvtable-verify=std' 12906 causes the data structures to be built after shared libraries have 12907 been loaded and initialized. '-fvtable-verify=preinit' causes them 12908 to be built before shared libraries have been loaded and 12909 initialized. 12910 12911 If this option appears multiple times in the command line with 12912 different values specified, 'none' takes highest priority over both 12913 'std' and 'preinit'; 'preinit' takes priority over 'std'. 12914 12915'-fvtv-debug' 12916 When used in conjunction with '-fvtable-verify=std' or 12917 '-fvtable-verify=preinit', causes debug versions of the runtime 12918 functions for the vtable verification feature to be called. This 12919 flag also causes the compiler to log information about which vtable 12920 pointers it finds for each class. This information is written to a 12921 file named 'vtv_set_ptr_data.log' in the directory named by the 12922 environment variable 'VTV_LOGS_DIR' if that is defined or the 12923 current working directory otherwise. 12924 12925 Note: This feature _appends_ data to the log file. If you want a 12926 fresh log file, be sure to delete any existing one. 12927 12928'-fvtv-counts' 12929 This is a debugging flag. When used in conjunction with 12930 '-fvtable-verify=std' or '-fvtable-verify=preinit', this causes the 12931 compiler to keep track of the total number of virtual calls it 12932 encounters and the number of verifications it inserts. It also 12933 counts the number of calls to certain run-time library functions 12934 that it inserts and logs this information for each compilation 12935 unit. The compiler writes this information to a file named 12936 'vtv_count_data.log' in the directory named by the environment 12937 variable 'VTV_LOGS_DIR' if that is defined or the current working 12938 directory otherwise. It also counts the size of the vtable pointer 12939 sets for each class, and writes this information to 12940 'vtv_class_set_sizes.log' in the same directory. 12941 12942 Note: This feature _appends_ data to the log files. To get fresh 12943 log files, be sure to delete any existing ones. 12944 12945'-finstrument-functions' 12946 Generate instrumentation calls for entry and exit to functions. 12947 Just after function entry and just before function exit, the 12948 following profiling functions are called with the address of the 12949 current function and its call site. (On some platforms, 12950 '__builtin_return_address' does not work beyond the current 12951 function, so the call site information may not be available to the 12952 profiling functions otherwise.) 12953 12954 void __cyg_profile_func_enter (void *this_fn, 12955 void *call_site); 12956 void __cyg_profile_func_exit (void *this_fn, 12957 void *call_site); 12958 12959 The first argument is the address of the start of the current 12960 function, which may be looked up exactly in the symbol table. 12961 12962 This instrumentation is also done for functions expanded inline in 12963 other functions. The profiling calls indicate where, conceptually, 12964 the inline function is entered and exited. This means that 12965 addressable versions of such functions must be available. If all 12966 your uses of a function are expanded inline, this may mean an 12967 additional expansion of code size. If you use 'extern inline' in 12968 your C code, an addressable version of such functions must be 12969 provided. (This is normally the case anyway, but if you get lucky 12970 and the optimizer always expands the functions inline, you might 12971 have gotten away without providing static copies.) 12972 12973 A function may be given the attribute 'no_instrument_function', in 12974 which case this instrumentation is not done. This can be used, for 12975 example, for the profiling functions listed above, high-priority 12976 interrupt routines, and any functions from which the profiling 12977 functions cannot safely be called (perhaps signal handlers, if the 12978 profiling routines generate output or allocate memory). *Note 12979 Common Function Attributes::. 12980 12981'-finstrument-functions-exclude-file-list=FILE,FILE,...' 12982 12983 Set the list of functions that are excluded from instrumentation 12984 (see the description of '-finstrument-functions'). If the file 12985 that contains a function definition matches with one of FILE, then 12986 that function is not instrumented. The match is done on 12987 substrings: if the FILE parameter is a substring of the file name, 12988 it is considered to be a match. 12989 12990 For example: 12991 12992 -finstrument-functions-exclude-file-list=/bits/stl,include/sys 12993 12994 excludes any inline function defined in files whose pathnames 12995 contain '/bits/stl' or 'include/sys'. 12996 12997 If, for some reason, you want to include letter ',' in one of SYM, 12998 write '\,'. For example, 12999 '-finstrument-functions-exclude-file-list='\,\,tmp'' (note the 13000 single quote surrounding the option). 13001 13002'-finstrument-functions-exclude-function-list=SYM,SYM,...' 13003 13004 This is similar to '-finstrument-functions-exclude-file-list', but 13005 this option sets the list of function names to be excluded from 13006 instrumentation. The function name to be matched is its 13007 user-visible name, such as 'vector<int> blah(const vector<int> &)', 13008 not the internal mangled name (e.g., '_Z4blahRSt6vectorIiSaIiEE'). 13009 The match is done on substrings: if the SYM parameter is a 13010 substring of the function name, it is considered to be a match. 13011 For C99 and C++ extended identifiers, the function name must be 13012 given in UTF-8, not using universal character names. 13013 13014'-fpatchable-function-entry=N[,M]' 13015 Generate N NOPs right at the beginning of each function, with the 13016 function entry point before the Mth NOP. If M is omitted, it 13017 defaults to '0' so the function entry points to the address just at 13018 the first NOP. The NOP instructions reserve extra space which can 13019 be used to patch in any desired instrumentation at run time, 13020 provided that the code segment is writable. The amount of space is 13021 controllable indirectly via the number of NOPs; the NOP instruction 13022 used corresponds to the instruction emitted by the internal GCC 13023 back-end interface 'gen_nop'. This behavior is target-specific and 13024 may also depend on the architecture variant and/or other 13025 compilation options. 13026 13027 For run-time identification, the starting addresses of these areas, 13028 which correspond to their respective function entries minus M, are 13029 additionally collected in the '__patchable_function_entries' 13030 section of the resulting binary. 13031 13032 Note that the value of '__attribute__ ((patchable_function_entry 13033 (N,M)))' takes precedence over command-line option 13034 '-fpatchable-function-entry=N,M'. This can be used to increase the 13035 area size or to remove it completely on a single function. If 13036 'N=0', no pad location is recorded. 13037 13038 The NOP instructions are inserted at--and maybe before, depending 13039 on M--the function entry address, even before the prologue. 13040 13041 13042File: gcc.info, Node: Preprocessor Options, Next: Assembler Options, Prev: Instrumentation Options, Up: Invoking GCC 13043 130443.13 Options Controlling the Preprocessor 13045========================================= 13046 13047These options control the C preprocessor, which is run on each C source 13048file before actual compilation. 13049 13050 If you use the '-E' option, nothing is done except preprocessing. Some 13051of these options make sense only together with '-E' because they cause 13052the preprocessor output to be unsuitable for actual compilation. 13053 13054 In addition to the options listed here, there are a number of options 13055to control search paths for include files documented in *note Directory 13056Options::. Options to control preprocessor diagnostics are listed in 13057*note Warning Options::. 13058 13059'-D NAME' 13060 Predefine NAME as a macro, with definition '1'. 13061 13062'-D NAME=DEFINITION' 13063 The contents of DEFINITION are tokenized and processed as if they 13064 appeared during translation phase three in a '#define' directive. 13065 In particular, the definition is truncated by embedded newline 13066 characters. 13067 13068 If you are invoking the preprocessor from a shell or shell-like 13069 program you may need to use the shell's quoting syntax to protect 13070 characters such as spaces that have a meaning in the shell syntax. 13071 13072 If you wish to define a function-like macro on the command line, 13073 write its argument list with surrounding parentheses before the 13074 equals sign (if any). Parentheses are meaningful to most shells, 13075 so you should quote the option. With 'sh' and 'csh', 13076 '-D'NAME(ARGS...)=DEFINITION'' works. 13077 13078 '-D' and '-U' options are processed in the order they are given on 13079 the command line. All '-imacros FILE' and '-include FILE' options 13080 are processed after all '-D' and '-U' options. 13081 13082'-U NAME' 13083 Cancel any previous definition of NAME, either built in or provided 13084 with a '-D' option. 13085 13086'-include FILE' 13087 Process FILE as if '#include "file"' appeared as the first line of 13088 the primary source file. However, the first directory searched for 13089 FILE is the preprocessor's working directory _instead of_ the 13090 directory containing the main source file. If not found there, it 13091 is searched for in the remainder of the '#include "..."' search 13092 chain as normal. 13093 13094 If multiple '-include' options are given, the files are included in 13095 the order they appear on the command line. 13096 13097'-imacros FILE' 13098 Exactly like '-include', except that any output produced by 13099 scanning FILE is thrown away. Macros it defines remain defined. 13100 This allows you to acquire all the macros from a header without 13101 also processing its declarations. 13102 13103 All files specified by '-imacros' are processed before all files 13104 specified by '-include'. 13105 13106'-undef' 13107 Do not predefine any system-specific or GCC-specific macros. The 13108 standard predefined macros remain defined. 13109 13110'-pthread' 13111 Define additional macros required for using the POSIX threads 13112 library. You should use this option consistently for both 13113 compilation and linking. This option is supported on GNU/Linux 13114 targets, most other Unix derivatives, and also on x86 Cygwin and 13115 MinGW targets. 13116 13117'-M' 13118 Instead of outputting the result of preprocessing, output a rule 13119 suitable for 'make' describing the dependencies of the main source 13120 file. The preprocessor outputs one 'make' rule containing the 13121 object file name for that source file, a colon, and the names of 13122 all the included files, including those coming from '-include' or 13123 '-imacros' command-line options. 13124 13125 Unless specified explicitly (with '-MT' or '-MQ'), the object file 13126 name consists of the name of the source file with any suffix 13127 replaced with object file suffix and with any leading directory 13128 parts removed. If there are many included files then the rule is 13129 split into several lines using '\'-newline. The rule has no 13130 commands. 13131 13132 This option does not suppress the preprocessor's debug output, such 13133 as '-dM'. To avoid mixing such debug output with the dependency 13134 rules you should explicitly specify the dependency output file with 13135 '-MF', or use an environment variable like 'DEPENDENCIES_OUTPUT' 13136 (*note Environment Variables::). Debug output is still sent to the 13137 regular output stream as normal. 13138 13139 Passing '-M' to the driver implies '-E', and suppresses warnings 13140 with an implicit '-w'. 13141 13142'-MM' 13143 Like '-M' but do not mention header files that are found in system 13144 header directories, nor header files that are included, directly or 13145 indirectly, from such a header. 13146 13147 This implies that the choice of angle brackets or double quotes in 13148 an '#include' directive does not in itself determine whether that 13149 header appears in '-MM' dependency output. 13150 13151'-MF FILE' 13152 When used with '-M' or '-MM', specifies a file to write the 13153 dependencies to. If no '-MF' switch is given the preprocessor 13154 sends the rules to the same place it would send preprocessed 13155 output. 13156 13157 When used with the driver options '-MD' or '-MMD', '-MF' overrides 13158 the default dependency output file. 13159 13160 If FILE is '-', then the dependencies are written to 'stdout'. 13161 13162'-MG' 13163 In conjunction with an option such as '-M' requesting dependency 13164 generation, '-MG' assumes missing header files are generated files 13165 and adds them to the dependency list without raising an error. The 13166 dependency filename is taken directly from the '#include' directive 13167 without prepending any path. '-MG' also suppresses preprocessed 13168 output, as a missing header file renders this useless. 13169 13170 This feature is used in automatic updating of makefiles. 13171 13172'-MP' 13173 This option instructs CPP to add a phony target for each dependency 13174 other than the main file, causing each to depend on nothing. These 13175 dummy rules work around errors 'make' gives if you remove header 13176 files without updating the 'Makefile' to match. 13177 13178 This is typical output: 13179 13180 test.o: test.c test.h 13181 13182 test.h: 13183 13184'-MT TARGET' 13185 13186 Change the target of the rule emitted by dependency generation. By 13187 default CPP takes the name of the main input file, deletes any 13188 directory components and any file suffix such as '.c', and appends 13189 the platform's usual object suffix. The result is the target. 13190 13191 An '-MT' option sets the target to be exactly the string you 13192 specify. If you want multiple targets, you can specify them as a 13193 single argument to '-MT', or use multiple '-MT' options. 13194 13195 For example, '-MT '$(objpfx)foo.o'' might give 13196 13197 $(objpfx)foo.o: foo.c 13198 13199'-MQ TARGET' 13200 13201 Same as '-MT', but it quotes any characters which are special to 13202 Make. '-MQ '$(objpfx)foo.o'' gives 13203 13204 $$(objpfx)foo.o: foo.c 13205 13206 The default target is automatically quoted, as if it were given 13207 with '-MQ'. 13208 13209'-MD' 13210 '-MD' is equivalent to '-M -MF FILE', except that '-E' is not 13211 implied. The driver determines FILE based on whether an '-o' 13212 option is given. If it is, the driver uses its argument but with a 13213 suffix of '.d', otherwise it takes the name of the input file, 13214 removes any directory components and suffix, and applies a '.d' 13215 suffix. 13216 13217 If '-MD' is used in conjunction with '-E', any '-o' switch is 13218 understood to specify the dependency output file (*note -MF: 13219 dashMF.), but if used without '-E', each '-o' is understood to 13220 specify a target object file. 13221 13222 Since '-E' is not implied, '-MD' can be used to generate a 13223 dependency output file as a side effect of the compilation process. 13224 13225'-MMD' 13226 Like '-MD' except mention only user header files, not system header 13227 files. 13228 13229'-fpreprocessed' 13230 Indicate to the preprocessor that the input file has already been 13231 preprocessed. This suppresses things like macro expansion, 13232 trigraph conversion, escaped newline splicing, and processing of 13233 most directives. The preprocessor still recognizes and removes 13234 comments, so that you can pass a file preprocessed with '-C' to the 13235 compiler without problems. In this mode the integrated 13236 preprocessor is little more than a tokenizer for the front ends. 13237 13238 '-fpreprocessed' is implicit if the input file has one of the 13239 extensions '.i', '.ii' or '.mi'. These are the extensions that GCC 13240 uses for preprocessed files created by '-save-temps'. 13241 13242'-fdirectives-only' 13243 When preprocessing, handle directives, but do not expand macros. 13244 13245 The option's behavior depends on the '-E' and '-fpreprocessed' 13246 options. 13247 13248 With '-E', preprocessing is limited to the handling of directives 13249 such as '#define', '#ifdef', and '#error'. Other preprocessor 13250 operations, such as macro expansion and trigraph conversion are not 13251 performed. In addition, the '-dD' option is implicitly enabled. 13252 13253 With '-fpreprocessed', predefinition of command line and most 13254 builtin macros is disabled. Macros such as '__LINE__', which are 13255 contextually dependent, are handled normally. This enables 13256 compilation of files previously preprocessed with '-E 13257 -fdirectives-only'. 13258 13259 With both '-E' and '-fpreprocessed', the rules for '-fpreprocessed' 13260 take precedence. This enables full preprocessing of files 13261 previously preprocessed with '-E -fdirectives-only'. 13262 13263'-fdollars-in-identifiers' 13264 Accept '$' in identifiers. 13265 13266'-fextended-identifiers' 13267 Accept universal character names and extended characters in 13268 identifiers. This option is enabled by default for C99 (and later 13269 C standard versions) and C++. 13270 13271'-fno-canonical-system-headers' 13272 When preprocessing, do not shorten system header paths with 13273 canonicalization. 13274 13275'-fmax-include-depth=DEPTH' 13276 Set the maximum depth of the nested #include. The default is 200. 13277 13278'-ftabstop=WIDTH' 13279 Set the distance between tab stops. This helps the preprocessor 13280 report correct column numbers in warnings or errors, even if tabs 13281 appear on the line. If the value is less than 1 or greater than 13282 100, the option is ignored. The default is 8. 13283 13284'-ftrack-macro-expansion[=LEVEL]' 13285 Track locations of tokens across macro expansions. This allows the 13286 compiler to emit diagnostic about the current macro expansion stack 13287 when a compilation error occurs in a macro expansion. Using this 13288 option makes the preprocessor and the compiler consume more memory. 13289 The LEVEL parameter can be used to choose the level of precision of 13290 token location tracking thus decreasing the memory consumption if 13291 necessary. Value '0' of LEVEL de-activates this option. Value '1' 13292 tracks tokens locations in a degraded mode for the sake of minimal 13293 memory overhead. In this mode all tokens resulting from the 13294 expansion of an argument of a function-like macro have the same 13295 location. Value '2' tracks tokens locations completely. This 13296 value is the most memory hungry. When this option is given no 13297 argument, the default parameter value is '2'. 13298 13299 Note that '-ftrack-macro-expansion=2' is activated by default. 13300 13301'-fmacro-prefix-map=OLD=NEW' 13302 When preprocessing files residing in directory 'OLD', expand the 13303 '__FILE__' and '__BASE_FILE__' macros as if the files resided in 13304 directory 'NEW' instead. This can be used to change an absolute 13305 path to a relative path by using '.' for NEW which can result in 13306 more reproducible builds that are location independent. This 13307 option also affects '__builtin_FILE()' during compilation. See 13308 also '-ffile-prefix-map'. 13309 13310'-fexec-charset=CHARSET' 13311 Set the execution character set, used for string and character 13312 constants. The default is UTF-8. CHARSET can be any encoding 13313 supported by the system's 'iconv' library routine. 13314 13315'-fwide-exec-charset=CHARSET' 13316 Set the wide execution character set, used for wide string and 13317 character constants. The default is one of UTF-32BE, UTF-32LE, 13318 UTF-16BE, or UTF-16LE, whichever corresponds to the width of 13319 'wchar_t' and the big-endian or little-endian byte order being used 13320 for code generation. As with '-fexec-charset', CHARSET can be any 13321 encoding supported by the system's 'iconv' library routine; 13322 however, you will have problems with encodings that do not fit 13323 exactly in 'wchar_t'. 13324 13325'-finput-charset=CHARSET' 13326 Set the input character set, used for translation from the 13327 character set of the input file to the source character set used by 13328 GCC. If the locale does not specify, or GCC cannot get this 13329 information from the locale, the default is UTF-8. This can be 13330 overridden by either the locale or this command-line option. 13331 Currently the command-line option takes precedence if there's a 13332 conflict. CHARSET can be any encoding supported by the system's 13333 'iconv' library routine. 13334 13335'-fpch-deps' 13336 When using precompiled headers (*note Precompiled Headers::), this 13337 flag causes the dependency-output flags to also list the files from 13338 the precompiled header's dependencies. If not specified, only the 13339 precompiled header are listed and not the files that were used to 13340 create it, because those files are not consulted when a precompiled 13341 header is used. 13342 13343'-fpch-preprocess' 13344 This option allows use of a precompiled header (*note Precompiled 13345 Headers::) together with '-E'. It inserts a special '#pragma', 13346 '#pragma GCC pch_preprocess "FILENAME"' in the output to mark the 13347 place where the precompiled header was found, and its FILENAME. 13348 When '-fpreprocessed' is in use, GCC recognizes this '#pragma' and 13349 loads the PCH. 13350 13351 This option is off by default, because the resulting preprocessed 13352 output is only really suitable as input to GCC. It is switched on 13353 by '-save-temps'. 13354 13355 You should not write this '#pragma' in your own code, but it is 13356 safe to edit the filename if the PCH file is available in a 13357 different location. The filename may be absolute or it may be 13358 relative to GCC's current directory. 13359 13360'-fworking-directory' 13361 Enable generation of linemarkers in the preprocessor output that 13362 let the compiler know the current working directory at the time of 13363 preprocessing. When this option is enabled, the preprocessor 13364 emits, after the initial linemarker, a second linemarker with the 13365 current working directory followed by two slashes. GCC uses this 13366 directory, when it's present in the preprocessed input, as the 13367 directory emitted as the current working directory in some 13368 debugging information formats. This option is implicitly enabled 13369 if debugging information is enabled, but this can be inhibited with 13370 the negated form '-fno-working-directory'. If the '-P' flag is 13371 present in the command line, this option has no effect, since no 13372 '#line' directives are emitted whatsoever. 13373 13374'-A PREDICATE=ANSWER' 13375 Make an assertion with the predicate PREDICATE and answer ANSWER. 13376 This form is preferred to the older form '-A PREDICATE(ANSWER)', 13377 which is still supported, because it does not use shell special 13378 characters. 13379 13380'-A -PREDICATE=ANSWER' 13381 Cancel an assertion with the predicate PREDICATE and answer ANSWER. 13382 13383'-C' 13384 Do not discard comments. All comments are passed through to the 13385 output file, except for comments in processed directives, which are 13386 deleted along with the directive. 13387 13388 You should be prepared for side effects when using '-C'; it causes 13389 the preprocessor to treat comments as tokens in their own right. 13390 For example, comments appearing at the start of what would be a 13391 directive line have the effect of turning that line into an 13392 ordinary source line, since the first token on the line is no 13393 longer a '#'. 13394 13395'-CC' 13396 Do not discard comments, including during macro expansion. This is 13397 like '-C', except that comments contained within macros are also 13398 passed through to the output file where the macro is expanded. 13399 13400 In addition to the side effects of the '-C' option, the '-CC' 13401 option causes all C++-style comments inside a macro to be converted 13402 to C-style comments. This is to prevent later use of that macro 13403 from inadvertently commenting out the remainder of the source line. 13404 13405 The '-CC' option is generally used to support lint comments. 13406 13407'-P' 13408 Inhibit generation of linemarkers in the output from the 13409 preprocessor. This might be useful when running the preprocessor 13410 on something that is not C code, and will be sent to a program 13411 which might be confused by the linemarkers. 13412 13413'-traditional' 13414'-traditional-cpp' 13415 13416 Try to imitate the behavior of pre-standard C preprocessors, as 13417 opposed to ISO C preprocessors. See the GNU CPP manual for 13418 details. 13419 13420 Note that GCC does not otherwise attempt to emulate a pre-standard 13421 C compiler, and these options are only supported with the '-E' 13422 switch, or when invoking CPP explicitly. 13423 13424'-trigraphs' 13425 Support ISO C trigraphs. These are three-character sequences, all 13426 starting with '??', that are defined by ISO C to stand for single 13427 characters. For example, '??/' stands for '\', so ''??/n'' is a 13428 character constant for a newline. 13429 13430 The nine trigraphs and their replacements are 13431 13432 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??- 13433 Replacement: [ ] { } # \ ^ | ~ 13434 13435 By default, GCC ignores trigraphs, but in standard-conforming modes 13436 it converts them. See the '-std' and '-ansi' options. 13437 13438'-remap' 13439 Enable special code to work around file systems which only permit 13440 very short file names, such as MS-DOS. 13441 13442'-H' 13443 Print the name of each header file used, in addition to other 13444 normal activities. Each name is indented to show how deep in the 13445 '#include' stack it is. Precompiled header files are also printed, 13446 even if they are found to be invalid; an invalid precompiled header 13447 file is printed with '...x' and a valid one with '...!' . 13448 13449'-dLETTERS' 13450 Says to make debugging dumps during compilation as specified by 13451 LETTERS. The flags documented here are those relevant to the 13452 preprocessor. Other LETTERS are interpreted by the compiler 13453 proper, or reserved for future versions of GCC, and so are silently 13454 ignored. If you specify LETTERS whose behavior conflicts, the 13455 result is undefined. *Note Developer Options::, for more 13456 information. 13457 13458 '-dM' 13459 Instead of the normal output, generate a list of '#define' 13460 directives for all the macros defined during the execution of 13461 the preprocessor, including predefined macros. This gives you 13462 a way of finding out what is predefined in your version of the 13463 preprocessor. Assuming you have no file 'foo.h', the command 13464 13465 touch foo.h; cpp -dM foo.h 13466 13467 shows all the predefined macros. 13468 13469 If you use '-dM' without the '-E' option, '-dM' is interpreted 13470 as a synonym for '-fdump-rtl-mach'. *Note (gcc)Developer 13471 Options::. 13472 13473 '-dD' 13474 Like '-dM' except in two respects: it does _not_ include the 13475 predefined macros, and it outputs _both_ the '#define' 13476 directives and the result of preprocessing. Both kinds of 13477 output go to the standard output file. 13478 13479 '-dN' 13480 Like '-dD', but emit only the macro names, not their 13481 expansions. 13482 13483 '-dI' 13484 Output '#include' directives in addition to the result of 13485 preprocessing. 13486 13487 '-dU' 13488 Like '-dD' except that only macros that are expanded, or whose 13489 definedness is tested in preprocessor directives, are output; 13490 the output is delayed until the use or test of the macro; and 13491 '#undef' directives are also output for macros tested but 13492 undefined at the time. 13493 13494'-fdebug-cpp' 13495 This option is only useful for debugging GCC. When used from CPP or 13496 with '-E', it dumps debugging information about location maps. 13497 Every token in the output is preceded by the dump of the map its 13498 location belongs to. 13499 13500 When used from GCC without '-E', this option has no effect. 13501 13502'-Wp,OPTION' 13503 You can use '-Wp,OPTION' to bypass the compiler driver and pass 13504 OPTION directly through to the preprocessor. If OPTION contains 13505 commas, it is split into multiple options at the commas. However, 13506 many options are modified, translated or interpreted by the 13507 compiler driver before being passed to the preprocessor, and '-Wp' 13508 forcibly bypasses this phase. The preprocessor's direct interface 13509 is undocumented and subject to change, so whenever possible you 13510 should avoid using '-Wp' and let the driver handle the options 13511 instead. 13512 13513'-Xpreprocessor OPTION' 13514 Pass OPTION as an option to the preprocessor. You can use this to 13515 supply system-specific preprocessor options that GCC does not 13516 recognize. 13517 13518 If you want to pass an option that takes an argument, you must use 13519 '-Xpreprocessor' twice, once for the option and once for the 13520 argument. 13521 13522'-no-integrated-cpp' 13523 Perform preprocessing as a separate pass before compilation. By 13524 default, GCC performs preprocessing as an integrated part of input 13525 tokenization and parsing. If this option is provided, the 13526 appropriate language front end ('cc1', 'cc1plus', or 'cc1obj' for 13527 C, C++, and Objective-C, respectively) is instead invoked twice, 13528 once for preprocessing only and once for actual compilation of the 13529 preprocessed input. This option may be useful in conjunction with 13530 the '-B' or '-wrapper' options to specify an alternate preprocessor 13531 or perform additional processing of the program source between 13532 normal preprocessing and compilation. 13533 13534 13535File: gcc.info, Node: Assembler Options, Next: Link Options, Prev: Preprocessor Options, Up: Invoking GCC 13536 135373.14 Passing Options to the Assembler 13538===================================== 13539 13540You can pass options to the assembler. 13541 13542'-Wa,OPTION' 13543 Pass OPTION as an option to the assembler. If OPTION contains 13544 commas, it is split into multiple options at the commas. 13545 13546'-Xassembler OPTION' 13547 Pass OPTION as an option to the assembler. You can use this to 13548 supply system-specific assembler options that GCC does not 13549 recognize. 13550 13551 If you want to pass an option that takes an argument, you must use 13552 '-Xassembler' twice, once for the option and once for the argument. 13553 13554 13555File: gcc.info, Node: Link Options, Next: Directory Options, Prev: Assembler Options, Up: Invoking GCC 13556 135573.15 Options for Linking 13558======================== 13559 13560These options come into play when the compiler links object files into 13561an executable output file. They are meaningless if the compiler is not 13562doing a link step. 13563 13564'OBJECT-FILE-NAME' 13565 A file name that does not end in a special recognized suffix is 13566 considered to name an object file or library. (Object files are 13567 distinguished from libraries by the linker according to the file 13568 contents.) If linking is done, these object files are used as 13569 input to the linker. 13570 13571'-c' 13572'-S' 13573'-E' 13574 If any of these options is used, then the linker is not run, and 13575 object file names should not be used as arguments. *Note Overall 13576 Options::. 13577 13578'-flinker-output=TYPE' 13579 This option controls code generation of the link-time optimizer. 13580 By default the linker output is automatically determined by the 13581 linker plugin. For debugging the compiler and if incremental 13582 linking with a non-LTO object file is desired, it may be useful to 13583 control the type manually. 13584 13585 If TYPE is 'exec', code generation produces a static binary. In 13586 this case '-fpic' and '-fpie' are both disabled. 13587 13588 If TYPE is 'dyn', code generation produces a shared library. In 13589 this case '-fpic' or '-fPIC' is preserved, but not enabled 13590 automatically. This allows to build shared libraries without 13591 position-independent code on architectures where this is possible, 13592 i.e. on x86. 13593 13594 If TYPE is 'pie', code generation produces an '-fpie' executable. 13595 This results in similar optimizations as 'exec' except that '-fpie' 13596 is not disabled if specified at compilation time. 13597 13598 If TYPE is 'rel', the compiler assumes that incremental linking is 13599 done. The sections containing intermediate code for link-time 13600 optimization are merged, pre-optimized, and output to the resulting 13601 object file. In addition, if '-ffat-lto-objects' is specified, 13602 binary code is produced for future non-LTO linking. The object 13603 file produced by incremental linking is smaller than a static 13604 library produced from the same object files. At link time the 13605 result of incremental linking also loads faster than a static 13606 library assuming that the majority of objects in the library are 13607 used. 13608 13609 Finally 'nolto-rel' configures the compiler for incremental linking 13610 where code generation is forced, a final binary is produced, and 13611 the intermediate code for later link-time optimization is stripped. 13612 When multiple object files are linked together the resulting code 13613 is better optimized than with link-time optimizations disabled (for 13614 example, cross-module inlining happens), but most of benefits of 13615 whole program optimizations are lost. 13616 13617 During the incremental link (by '-r') the linker plugin defaults to 13618 'rel'. With current interfaces to GNU Binutils it is however not 13619 possible to incrementally link LTO objects and non-LTO objects into 13620 a single mixed object file. If any of object files in incremental 13621 link cannot be used for link-time optimization, the linker plugin 13622 issues a warning and uses 'nolto-rel'. To maintain whole program 13623 optimization, it is recommended to link such objects into static 13624 library instead. Alternatively it is possible to use H.J. Lu's 13625 binutils with support for mixed objects. 13626 13627'-fuse-ld=bfd' 13628 Use the 'bfd' linker instead of the default linker. 13629 13630'-fuse-ld=gold' 13631 Use the 'gold' linker instead of the default linker. 13632 13633'-fuse-ld=lld' 13634 Use the LLVM 'lld' linker instead of the default linker. 13635 13636'-lLIBRARY' 13637'-l LIBRARY' 13638 Search the library named LIBRARY when linking. (The second 13639 alternative with the library as a separate argument is only for 13640 POSIX compliance and is not recommended.) 13641 13642 The '-l' option is passed directly to the linker by GCC. Refer to 13643 your linker documentation for exact details. The general 13644 description below applies to the GNU linker. 13645 13646 The linker searches a standard list of directories for the library. 13647 The directories searched include several standard system 13648 directories plus any that you specify with '-L'. 13649 13650 Static libraries are archives of object files, and have file names 13651 like 'libLIBRARY.a'. Some targets also support shared libraries, 13652 which typically have names like 'libLIBRARY.so'. If both static 13653 and shared libraries are found, the linker gives preference to 13654 linking with the shared library unless the '-static' option is 13655 used. 13656 13657 It makes a difference where in the command you write this option; 13658 the linker searches and processes libraries and object files in the 13659 order they are specified. Thus, 'foo.o -lz bar.o' searches library 13660 'z' after file 'foo.o' but before 'bar.o'. If 'bar.o' refers to 13661 functions in 'z', those functions may not be loaded. 13662 13663'-lobjc' 13664 You need this special case of the '-l' option in order to link an 13665 Objective-C or Objective-C++ program. 13666 13667'-nostartfiles' 13668 Do not use the standard system startup files when linking. The 13669 standard system libraries are used normally, unless '-nostdlib', 13670 '-nolibc', or '-nodefaultlibs' is used. 13671 13672'-nodefaultlibs' 13673 Do not use the standard system libraries when linking. Only the 13674 libraries you specify are passed to the linker, and options 13675 specifying linkage of the system libraries, such as 13676 '-static-libgcc' or '-shared-libgcc', are ignored. The standard 13677 startup files are used normally, unless '-nostartfiles' is used. 13678 13679 The compiler may generate calls to 'memcmp', 'memset', 'memcpy' and 13680 'memmove'. These entries are usually resolved by entries in libc. 13681 These entry points should be supplied through some other mechanism 13682 when this option is specified. 13683 13684'-nolibc' 13685 Do not use the C library or system libraries tightly coupled with 13686 it when linking. Still link with the startup files, 'libgcc' or 13687 toolchain provided language support libraries such as 'libgnat', 13688 'libgfortran' or 'libstdc++' unless options preventing their 13689 inclusion are used as well. This typically removes '-lc' from the 13690 link command line, as well as system libraries that normally go 13691 with it and become meaningless when absence of a C library is 13692 assumed, for example '-lpthread' or '-lm' in some configurations. 13693 This is intended for bare-board targets when there is indeed no C 13694 library available. 13695 13696'-nostdlib' 13697 Do not use the standard system startup files or libraries when 13698 linking. No startup files and only the libraries you specify are 13699 passed to the linker, and options specifying linkage of the system 13700 libraries, such as '-static-libgcc' or '-shared-libgcc', are 13701 ignored. 13702 13703 The compiler may generate calls to 'memcmp', 'memset', 'memcpy' and 13704 'memmove'. These entries are usually resolved by entries in libc. 13705 These entry points should be supplied through some other mechanism 13706 when this option is specified. 13707 13708 One of the standard libraries bypassed by '-nostdlib' and 13709 '-nodefaultlibs' is 'libgcc.a', a library of internal subroutines 13710 which GCC uses to overcome shortcomings of particular machines, or 13711 special needs for some languages. (*Note Interfacing to GCC 13712 Output: (gccint)Interface, for more discussion of 'libgcc.a'.) In 13713 most cases, you need 'libgcc.a' even when you want to avoid other 13714 standard libraries. In other words, when you specify '-nostdlib' 13715 or '-nodefaultlibs' you should usually specify '-lgcc' as well. 13716 This ensures that you have no unresolved references to internal GCC 13717 library subroutines. (An example of such an internal subroutine is 13718 '__main', used to ensure C++ constructors are called; *note 13719 'collect2': (gccint)Collect2.) 13720 13721'-e ENTRY' 13722'--entry=ENTRY' 13723 13724 Specify that the program entry point is ENTRY. The argument is 13725 interpreted by the linker; the GNU linker accepts either a symbol 13726 name or an address. 13727 13728'-pie' 13729 Produce a dynamically linked position independent executable on 13730 targets that support it. For predictable results, you must also 13731 specify the same set of options used for compilation ('-fpie', 13732 '-fPIE', or model suboptions) when you specify this linker option. 13733 13734'-no-pie' 13735 Don't produce a dynamically linked position independent executable. 13736 13737'-static-pie' 13738 Produce a static position independent executable on targets that 13739 support it. A static position independent executable is similar to 13740 a static executable, but can be loaded at any address without a 13741 dynamic linker. For predictable results, you must also specify the 13742 same set of options used for compilation ('-fpie', '-fPIE', or 13743 model suboptions) when you specify this linker option. 13744 13745'-pthread' 13746 Link with the POSIX threads library. This option is supported on 13747 GNU/Linux targets, most other Unix derivatives, and also on x86 13748 Cygwin and MinGW targets. On some targets this option also sets 13749 flags for the preprocessor, so it should be used consistently for 13750 both compilation and linking. 13751 13752'-r' 13753 Produce a relocatable object as output. This is also known as 13754 partial linking. 13755 13756'-rdynamic' 13757 Pass the flag '-export-dynamic' to the ELF linker, on targets that 13758 support it. This instructs the linker to add all symbols, not only 13759 used ones, to the dynamic symbol table. This option is needed for 13760 some uses of 'dlopen' or to allow obtaining backtraces from within 13761 a program. 13762 13763'-s' 13764 Remove all symbol table and relocation information from the 13765 executable. 13766 13767'-static' 13768 On systems that support dynamic linking, this overrides '-pie' and 13769 prevents linking with the shared libraries. On other systems, this 13770 option has no effect. 13771 13772'-shared' 13773 Produce a shared object which can then be linked with other objects 13774 to form an executable. Not all systems support this option. For 13775 predictable results, you must also specify the same set of options 13776 used for compilation ('-fpic', '-fPIC', or model suboptions) when 13777 you specify this linker option.(1) 13778 13779'-shared-libgcc' 13780'-static-libgcc' 13781 On systems that provide 'libgcc' as a shared library, these options 13782 force the use of either the shared or static version, respectively. 13783 If no shared version of 'libgcc' was built when the compiler was 13784 configured, these options have no effect. 13785 13786 There are several situations in which an application should use the 13787 shared 'libgcc' instead of the static version. The most common of 13788 these is when the application wishes to throw and catch exceptions 13789 across different shared libraries. In that case, each of the 13790 libraries as well as the application itself should use the shared 13791 'libgcc'. 13792 13793 Therefore, the G++ driver automatically adds '-shared-libgcc' 13794 whenever you build a shared library or a main executable, because 13795 C++ programs typically use exceptions, so this is the right thing 13796 to do. 13797 13798 If, instead, you use the GCC driver to create shared libraries, you 13799 may find that they are not always linked with the shared 'libgcc'. 13800 If GCC finds, at its configuration time, that you have a non-GNU 13801 linker or a GNU linker that does not support option 13802 '--eh-frame-hdr', it links the shared version of 'libgcc' into 13803 shared libraries by default. Otherwise, it takes advantage of the 13804 linker and optimizes away the linking with the shared version of 13805 'libgcc', linking with the static version of libgcc by default. 13806 This allows exceptions to propagate through such shared libraries, 13807 without incurring relocation costs at library load time. 13808 13809 However, if a library or main executable is supposed to throw or 13810 catch exceptions, you must link it using the G++ driver, or using 13811 the option '-shared-libgcc', such that it is linked with the shared 13812 'libgcc'. 13813 13814'-static-libasan' 13815 When the '-fsanitize=address' option is used to link a program, the 13816 GCC driver automatically links against 'libasan'. If 'libasan' is 13817 available as a shared library, and the '-static' option is not 13818 used, then this links against the shared version of 'libasan'. The 13819 '-static-libasan' option directs the GCC driver to link 'libasan' 13820 statically, without necessarily linking other libraries statically. 13821 13822'-static-libtsan' 13823 When the '-fsanitize=thread' option is used to link a program, the 13824 GCC driver automatically links against 'libtsan'. If 'libtsan' is 13825 available as a shared library, and the '-static' option is not 13826 used, then this links against the shared version of 'libtsan'. The 13827 '-static-libtsan' option directs the GCC driver to link 'libtsan' 13828 statically, without necessarily linking other libraries statically. 13829 13830'-static-liblsan' 13831 When the '-fsanitize=leak' option is used to link a program, the 13832 GCC driver automatically links against 'liblsan'. If 'liblsan' is 13833 available as a shared library, and the '-static' option is not 13834 used, then this links against the shared version of 'liblsan'. The 13835 '-static-liblsan' option directs the GCC driver to link 'liblsan' 13836 statically, without necessarily linking other libraries statically. 13837 13838'-static-libubsan' 13839 When the '-fsanitize=undefined' option is used to link a program, 13840 the GCC driver automatically links against 'libubsan'. If 13841 'libubsan' is available as a shared library, and the '-static' 13842 option is not used, then this links against the shared version of 13843 'libubsan'. The '-static-libubsan' option directs the GCC driver 13844 to link 'libubsan' statically, without necessarily linking other 13845 libraries statically. 13846 13847'-static-libstdc++' 13848 When the 'g++' program is used to link a C++ program, it normally 13849 automatically links against 'libstdc++'. If 'libstdc++' is 13850 available as a shared library, and the '-static' option is not 13851 used, then this links against the shared version of 'libstdc++'. 13852 That is normally fine. However, it is sometimes useful to freeze 13853 the version of 'libstdc++' used by the program without going all 13854 the way to a fully static link. The '-static-libstdc++' option 13855 directs the 'g++' driver to link 'libstdc++' statically, without 13856 necessarily linking other libraries statically. 13857 13858'-symbolic' 13859 Bind references to global symbols when building a shared object. 13860 Warn about any unresolved references (unless overridden by the link 13861 editor option '-Xlinker -z -Xlinker defs'). Only a few systems 13862 support this option. 13863 13864'-T SCRIPT' 13865 Use SCRIPT as the linker script. This option is supported by most 13866 systems using the GNU linker. On some targets, such as bare-board 13867 targets without an operating system, the '-T' option may be 13868 required when linking to avoid references to undefined symbols. 13869 13870'-Xlinker OPTION' 13871 Pass OPTION as an option to the linker. You can use this to supply 13872 system-specific linker options that GCC does not recognize. 13873 13874 If you want to pass an option that takes a separate argument, you 13875 must use '-Xlinker' twice, once for the option and once for the 13876 argument. For example, to pass '-assert definitions', you must 13877 write '-Xlinker -assert -Xlinker definitions'. It does not work to 13878 write '-Xlinker "-assert definitions"', because this passes the 13879 entire string as a single argument, which is not what the linker 13880 expects. 13881 13882 When using the GNU linker, it is usually more convenient to pass 13883 arguments to linker options using the 'OPTION=VALUE' syntax than as 13884 separate arguments. For example, you can specify '-Xlinker 13885 -Map=output.map' rather than '-Xlinker -Map -Xlinker output.map'. 13886 Other linkers may not support this syntax for command-line options. 13887 13888'-Wl,OPTION' 13889 Pass OPTION as an option to the linker. If OPTION contains commas, 13890 it is split into multiple options at the commas. You can use this 13891 syntax to pass an argument to the option. For example, 13892 '-Wl,-Map,output.map' passes '-Map output.map' to the linker. When 13893 using the GNU linker, you can also get the same effect with 13894 '-Wl,-Map=output.map'. 13895 13896'-u SYMBOL' 13897 Pretend the symbol SYMBOL is undefined, to force linking of library 13898 modules to define it. You can use '-u' multiple times with 13899 different symbols to force loading of additional library modules. 13900 13901'-z KEYWORD' 13902 '-z' is passed directly on to the linker along with the keyword 13903 KEYWORD. See the section in the documentation of your linker for 13904 permitted values and their meanings. 13905 13906 ---------- Footnotes ---------- 13907 13908 (1) On some systems, 'gcc -shared' needs to build supplementary stub 13909code for constructors to work. On multi-libbed systems, 'gcc -shared' 13910must select the correct support libraries to link against. Failing to 13911supply the correct flags may lead to subtle defects. Supplying them in 13912cases where they are not necessary is innocuous. 13913 13914 13915File: gcc.info, Node: Directory Options, Next: Code Gen Options, Prev: Link Options, Up: Invoking GCC 13916 139173.16 Options for Directory Search 13918================================= 13919 13920These options specify directories to search for header files, for 13921libraries and for parts of the compiler: 13922 13923'-I DIR' 13924'-iquote DIR' 13925'-isystem DIR' 13926'-idirafter DIR' 13927 Add the directory DIR to the list of directories to be searched for 13928 header files during preprocessing. If DIR begins with '=' or 13929 '$SYSROOT', then the '=' or '$SYSROOT' is replaced by the sysroot 13930 prefix; see '--sysroot' and '-isysroot'. 13931 13932 Directories specified with '-iquote' apply only to the quote form 13933 of the directive, '#include "FILE"'. Directories specified with 13934 '-I', '-isystem', or '-idirafter' apply to lookup for both the 13935 '#include "FILE"' and '#include <FILE>' directives. 13936 13937 You can specify any number or combination of these options on the 13938 command line to search for header files in several directories. 13939 The lookup order is as follows: 13940 13941 1. For the quote form of the include directive, the directory of 13942 the current file is searched first. 13943 13944 2. For the quote form of the include directive, the directories 13945 specified by '-iquote' options are searched in left-to-right 13946 order, as they appear on the command line. 13947 13948 3. Directories specified with '-I' options are scanned in 13949 left-to-right order. 13950 13951 4. Directories specified with '-isystem' options are scanned in 13952 left-to-right order. 13953 13954 5. Standard system directories are scanned. 13955 13956 6. Directories specified with '-idirafter' options are scanned in 13957 left-to-right order. 13958 13959 You can use '-I' to override a system header file, substituting 13960 your own version, since these directories are searched before the 13961 standard system header file directories. However, you should not 13962 use this option to add directories that contain vendor-supplied 13963 system header files; use '-isystem' for that. 13964 13965 The '-isystem' and '-idirafter' options also mark the directory as 13966 a system directory, so that it gets the same special treatment that 13967 is applied to the standard system directories. 13968 13969 If a standard system include directory, or a directory specified 13970 with '-isystem', is also specified with '-I', the '-I' option is 13971 ignored. The directory is still searched but as a system directory 13972 at its normal position in the system include chain. This is to 13973 ensure that GCC's procedure to fix buggy system headers and the 13974 ordering for the '#include_next' directive are not inadvertently 13975 changed. If you really need to change the search order for system 13976 directories, use the '-nostdinc' and/or '-isystem' options. 13977 13978'-I-' 13979 Split the include path. This option has been deprecated. Please 13980 use '-iquote' instead for '-I' directories before the '-I-' and 13981 remove the '-I-' option. 13982 13983 Any directories specified with '-I' options before '-I-' are 13984 searched only for headers requested with '#include "FILE"'; they 13985 are not searched for '#include <FILE>'. If additional directories 13986 are specified with '-I' options after the '-I-', those directories 13987 are searched for all '#include' directives. 13988 13989 In addition, '-I-' inhibits the use of the directory of the current 13990 file directory as the first search directory for '#include "FILE"'. 13991 There is no way to override this effect of '-I-'. 13992 13993'-iprefix PREFIX' 13994 Specify PREFIX as the prefix for subsequent '-iwithprefix' options. 13995 If the prefix represents a directory, you should include the final 13996 '/'. 13997 13998'-iwithprefix DIR' 13999'-iwithprefixbefore DIR' 14000 Append DIR to the prefix specified previously with '-iprefix', and 14001 add the resulting directory to the include search path. 14002 '-iwithprefixbefore' puts it in the same place '-I' would; 14003 '-iwithprefix' puts it where '-idirafter' would. 14004 14005'-isysroot DIR' 14006 This option is like the '--sysroot' option, but applies only to 14007 header files (except for Darwin targets, where it applies to both 14008 header files and libraries). See the '--sysroot' option for more 14009 information. 14010 14011'-imultilib DIR' 14012 Use DIR as a subdirectory of the directory containing 14013 target-specific C++ headers. 14014 14015'-nostdinc' 14016 Do not search the standard system directories for header files. 14017 Only the directories explicitly specified with '-I', '-iquote', 14018 '-isystem', and/or '-idirafter' options (and the directory of the 14019 current file, if appropriate) are searched. 14020 14021'-nostdinc++' 14022 Do not search for header files in the C++-specific standard 14023 directories, but do still search the other standard directories. 14024 (This option is used when building the C++ library.) 14025 14026'-iplugindir=DIR' 14027 Set the directory to search for plugins that are passed by 14028 '-fplugin=NAME' instead of '-fplugin=PATH/NAME.so'. This option is 14029 not meant to be used by the user, but only passed by the driver. 14030 14031'-LDIR' 14032 Add directory DIR to the list of directories to be searched for 14033 '-l'. 14034 14035'-BPREFIX' 14036 This option specifies where to find the executables, libraries, 14037 include files, and data files of the compiler itself. 14038 14039 The compiler driver program runs one or more of the subprograms 14040 'cpp', 'cc1', 'as' and 'ld'. It tries PREFIX as a prefix for each 14041 program it tries to run, both with and without 'MACHINE/VERSION/' 14042 for the corresponding target machine and compiler version. 14043 14044 For each subprogram to be run, the compiler driver first tries the 14045 '-B' prefix, if any. If that name is not found, or if '-B' is not 14046 specified, the driver tries two standard prefixes, '/usr/lib/gcc/' 14047 and '/usr/local/lib/gcc/'. If neither of those results in a file 14048 name that is found, the unmodified program name is searched for 14049 using the directories specified in your 'PATH' environment 14050 variable. 14051 14052 The compiler checks to see if the path provided by '-B' refers to a 14053 directory, and if necessary it adds a directory separator character 14054 at the end of the path. 14055 14056 '-B' prefixes that effectively specify directory names also apply 14057 to libraries in the linker, because the compiler translates these 14058 options into '-L' options for the linker. They also apply to 14059 include files in the preprocessor, because the compiler translates 14060 these options into '-isystem' options for the preprocessor. In 14061 this case, the compiler appends 'include' to the prefix. 14062 14063 The runtime support file 'libgcc.a' can also be searched for using 14064 the '-B' prefix, if needed. If it is not found there, the two 14065 standard prefixes above are tried, and that is all. The file is 14066 left out of the link if it is not found by those means. 14067 14068 Another way to specify a prefix much like the '-B' prefix is to use 14069 the environment variable 'GCC_EXEC_PREFIX'. *Note Environment 14070 Variables::. 14071 14072 As a special kludge, if the path provided by '-B' is 14073 '[dir/]stageN/', where N is a number in the range 0 to 9, then it 14074 is replaced by '[dir/]include'. This is to help with 14075 boot-strapping the compiler. 14076 14077'-no-canonical-prefixes' 14078 Do not expand any symbolic links, resolve references to '/../' or 14079 '/./', or make the path absolute when generating a relative prefix. 14080 14081'--sysroot=DIR' 14082 Use DIR as the logical root directory for headers and libraries. 14083 For example, if the compiler normally searches for headers in 14084 '/usr/include' and libraries in '/usr/lib', it instead searches 14085 'DIR/usr/include' and 'DIR/usr/lib'. 14086 14087 If you use both this option and the '-isysroot' option, then the 14088 '--sysroot' option applies to libraries, but the '-isysroot' option 14089 applies to header files. 14090 14091 The GNU linker (beginning with version 2.16) has the necessary 14092 support for this option. If your linker does not support this 14093 option, the header file aspect of '--sysroot' still works, but the 14094 library aspect does not. 14095 14096'--no-sysroot-suffix' 14097 For some targets, a suffix is added to the root directory specified 14098 with '--sysroot', depending on the other options used, so that 14099 headers may for example be found in 'DIR/SUFFIX/usr/include' 14100 instead of 'DIR/usr/include'. This option disables the addition of 14101 such a suffix. 14102 14103 14104File: gcc.info, Node: Code Gen Options, Next: Developer Options, Prev: Directory Options, Up: Invoking GCC 14105 141063.17 Options for Code Generation Conventions 14107============================================ 14108 14109These machine-independent options control the interface conventions used 14110in code generation. 14111 14112 Most of them have both positive and negative forms; the negative form 14113of '-ffoo' is '-fno-foo'. In the table below, only one of the forms is 14114listed--the one that is not the default. You can figure out the other 14115form by either removing 'no-' or adding it. 14116 14117'-fstack-reuse=REUSE-LEVEL' 14118 This option controls stack space reuse for user declared local/auto 14119 variables and compiler generated temporaries. REUSE_LEVEL can be 14120 'all', 'named_vars', or 'none'. 'all' enables stack reuse for all 14121 local variables and temporaries, 'named_vars' enables the reuse 14122 only for user defined local variables with names, and 'none' 14123 disables stack reuse completely. The default value is 'all'. The 14124 option is needed when the program extends the lifetime of a scoped 14125 local variable or a compiler generated temporary beyond the end 14126 point defined by the language. When a lifetime of a variable ends, 14127 and if the variable lives in memory, the optimizing compiler has 14128 the freedom to reuse its stack space with other temporaries or 14129 scoped local variables whose live range does not overlap with it. 14130 Legacy code extending local lifetime is likely to break with the 14131 stack reuse optimization. 14132 14133 For example, 14134 14135 int *p; 14136 { 14137 int local1; 14138 14139 p = &local1; 14140 local1 = 10; 14141 .... 14142 } 14143 { 14144 int local2; 14145 local2 = 20; 14146 ... 14147 } 14148 14149 if (*p == 10) // out of scope use of local1 14150 { 14151 14152 } 14153 14154 Another example: 14155 14156 struct A 14157 { 14158 A(int k) : i(k), j(k) { } 14159 int i; 14160 int j; 14161 }; 14162 14163 A *ap; 14164 14165 void foo(const A& ar) 14166 { 14167 ap = &ar; 14168 } 14169 14170 void bar() 14171 { 14172 foo(A(10)); // temp object's lifetime ends when foo returns 14173 14174 { 14175 A a(20); 14176 .... 14177 } 14178 ap->i+= 10; // ap references out of scope temp whose space 14179 // is reused with a. What is the value of ap->i? 14180 } 14181 14182 14183 The lifetime of a compiler generated temporary is well defined by 14184 the C++ standard. When a lifetime of a temporary ends, and if the 14185 temporary lives in memory, the optimizing compiler has the freedom 14186 to reuse its stack space with other temporaries or scoped local 14187 variables whose live range does not overlap with it. However some 14188 of the legacy code relies on the behavior of older compilers in 14189 which temporaries' stack space is not reused, the aggressive stack 14190 reuse can lead to runtime errors. This option is used to control 14191 the temporary stack reuse optimization. 14192 14193'-ftrapv' 14194 This option generates traps for signed overflow on addition, 14195 subtraction, multiplication operations. The options '-ftrapv' and 14196 '-fwrapv' override each other, so using '-ftrapv' '-fwrapv' on the 14197 command-line results in '-fwrapv' being effective. Note that only 14198 active options override, so using '-ftrapv' '-fwrapv' '-fno-wrapv' 14199 on the command-line results in '-ftrapv' being effective. 14200 14201'-fwrapv' 14202 This option instructs the compiler to assume that signed arithmetic 14203 overflow of addition, subtraction and multiplication wraps around 14204 using twos-complement representation. This flag enables some 14205 optimizations and disables others. The options '-ftrapv' and 14206 '-fwrapv' override each other, so using '-ftrapv' '-fwrapv' on the 14207 command-line results in '-fwrapv' being effective. Note that only 14208 active options override, so using '-ftrapv' '-fwrapv' '-fno-wrapv' 14209 on the command-line results in '-ftrapv' being effective. 14210 14211'-fwrapv-pointer' 14212 This option instructs the compiler to assume that pointer 14213 arithmetic overflow on addition and subtraction wraps around using 14214 twos-complement representation. This flag disables some 14215 optimizations which assume pointer overflow is invalid. 14216 14217'-fstrict-overflow' 14218 This option implies '-fno-wrapv' '-fno-wrapv-pointer' and when 14219 negated implies '-fwrapv' '-fwrapv-pointer'. 14220 14221'-fexceptions' 14222 Enable exception handling. Generates extra code needed to 14223 propagate exceptions. For some targets, this implies GCC generates 14224 frame unwind information for all functions, which can produce 14225 significant data size overhead, although it does not affect 14226 execution. If you do not specify this option, GCC enables it by 14227 default for languages like C++ that normally require exception 14228 handling, and disables it for languages like C that do not normally 14229 require it. However, you may need to enable this option when 14230 compiling C code that needs to interoperate properly with exception 14231 handlers written in C++. You may also wish to disable this option 14232 if you are compiling older C++ programs that don't use exception 14233 handling. 14234 14235'-fnon-call-exceptions' 14236 Generate code that allows trapping instructions to throw 14237 exceptions. Note that this requires platform-specific runtime 14238 support that does not exist everywhere. Moreover, it only allows 14239 _trapping_ instructions to throw exceptions, i.e. memory references 14240 or floating-point instructions. It does not allow exceptions to be 14241 thrown from arbitrary signal handlers such as 'SIGALRM'. 14242 14243'-fdelete-dead-exceptions' 14244 Consider that instructions that may throw exceptions but don't 14245 otherwise contribute to the execution of the program can be 14246 optimized away. This option is enabled by default for the Ada 14247 front end, as permitted by the Ada language specification. 14248 Optimization passes that cause dead exceptions to be removed are 14249 enabled independently at different optimization levels. 14250 14251'-funwind-tables' 14252 Similar to '-fexceptions', except that it just generates any needed 14253 static data, but does not affect the generated code in any other 14254 way. You normally do not need to enable this option; instead, a 14255 language processor that needs this handling enables it on your 14256 behalf. 14257 14258'-fasynchronous-unwind-tables' 14259 Generate unwind table in DWARF format, if supported by target 14260 machine. The table is exact at each instruction boundary, so it 14261 can be used for stack unwinding from asynchronous events (such as 14262 debugger or garbage collector). 14263 14264'-fno-gnu-unique' 14265 On systems with recent GNU assembler and C library, the C++ 14266 compiler uses the 'STB_GNU_UNIQUE' binding to make sure that 14267 definitions of template static data members and static local 14268 variables in inline functions are unique even in the presence of 14269 'RTLD_LOCAL'; this is necessary to avoid problems with a library 14270 used by two different 'RTLD_LOCAL' plugins depending on a 14271 definition in one of them and therefore disagreeing with the other 14272 one about the binding of the symbol. But this causes 'dlclose' to 14273 be ignored for affected DSOs; if your program relies on 14274 reinitialization of a DSO via 'dlclose' and 'dlopen', you can use 14275 '-fno-gnu-unique'. 14276 14277'-fpcc-struct-return' 14278 Return "short" 'struct' and 'union' values in memory like longer 14279 ones, rather than in registers. This convention is less efficient, 14280 but it has the advantage of allowing intercallability between 14281 GCC-compiled files and files compiled with other compilers, 14282 particularly the Portable C Compiler (pcc). 14283 14284 The precise convention for returning structures in memory depends 14285 on the target configuration macros. 14286 14287 Short structures and unions are those whose size and alignment 14288 match that of some integer type. 14289 14290 *Warning:* code compiled with the '-fpcc-struct-return' switch is 14291 not binary compatible with code compiled with the 14292 '-freg-struct-return' switch. Use it to conform to a non-default 14293 application binary interface. 14294 14295'-freg-struct-return' 14296 Return 'struct' and 'union' values in registers when possible. 14297 This is more efficient for small structures than 14298 '-fpcc-struct-return'. 14299 14300 If you specify neither '-fpcc-struct-return' nor 14301 '-freg-struct-return', GCC defaults to whichever convention is 14302 standard for the target. If there is no standard convention, GCC 14303 defaults to '-fpcc-struct-return', except on targets where GCC is 14304 the principal compiler. In those cases, we can choose the 14305 standard, and we chose the more efficient register return 14306 alternative. 14307 14308 *Warning:* code compiled with the '-freg-struct-return' switch is 14309 not binary compatible with code compiled with the 14310 '-fpcc-struct-return' switch. Use it to conform to a non-default 14311 application binary interface. 14312 14313'-fshort-enums' 14314 Allocate to an 'enum' type only as many bytes as it needs for the 14315 declared range of possible values. Specifically, the 'enum' type 14316 is equivalent to the smallest integer type that has enough room. 14317 14318 *Warning:* the '-fshort-enums' switch causes GCC to generate code 14319 that is not binary compatible with code generated without that 14320 switch. Use it to conform to a non-default application binary 14321 interface. 14322 14323'-fshort-wchar' 14324 Override the underlying type for 'wchar_t' to be 'short unsigned 14325 int' instead of the default for the target. This option is useful 14326 for building programs to run under WINE. 14327 14328 *Warning:* the '-fshort-wchar' switch causes GCC to generate code 14329 that is not binary compatible with code generated without that 14330 switch. Use it to conform to a non-default application binary 14331 interface. 14332 14333'-fcommon' 14334 In C code, this option controls the placement of global variables 14335 defined without an initializer, known as "tentative definitions" in 14336 the C standard. Tentative definitions are distinct from 14337 declarations of a variable with the 'extern' keyword, which do not 14338 allocate storage. 14339 14340 The default is '-fno-common', which specifies that the compiler 14341 places uninitialized global variables in the BSS section of the 14342 object file. This inhibits the merging of tentative definitions by 14343 the linker so you get a multiple-definition error if the same 14344 variable is accidentally defined in more than one compilation unit. 14345 14346 The '-fcommon' places uninitialized global variables in a common 14347 block. This allows the linker to resolve all tentative definitions 14348 of the same variable in different compilation units to the same 14349 object, or to a non-tentative definition. This behavior is 14350 inconsistent with C++, and on many targets implies a speed and code 14351 size penalty on global variable references. It is mainly useful to 14352 enable legacy code to link without errors. 14353 14354'-fno-ident' 14355 Ignore the '#ident' directive. 14356 14357'-finhibit-size-directive' 14358 Don't output a '.size' assembler directive, or anything else that 14359 would cause trouble if the function is split in the middle, and the 14360 two halves are placed at locations far apart in memory. This 14361 option is used when compiling 'crtstuff.c'; you should not need to 14362 use it for anything else. 14363 14364'-fverbose-asm' 14365 Put extra commentary information in the generated assembly code to 14366 make it more readable. This option is generally only of use to 14367 those who actually need to read the generated assembly code 14368 (perhaps while debugging the compiler itself). 14369 14370 '-fno-verbose-asm', the default, causes the extra information to be 14371 omitted and is useful when comparing two assembler files. 14372 14373 The added comments include: 14374 14375 * information on the compiler version and command-line options, 14376 14377 * the source code lines associated with the assembly 14378 instructions, in the form FILENAME:LINENUMBER:CONTENT OF LINE, 14379 14380 * hints on which high-level expressions correspond to the 14381 various assembly instruction operands. 14382 14383 For example, given this C source file: 14384 14385 int test (int n) 14386 { 14387 int i; 14388 int total = 0; 14389 14390 for (i = 0; i < n; i++) 14391 total += i * i; 14392 14393 return total; 14394 } 14395 14396 compiling to (x86_64) assembly via '-S' and emitting the result 14397 direct to stdout via '-o' '-' 14398 14399 gcc -S test.c -fverbose-asm -Os -o - 14400 14401 gives output similar to this: 14402 14403 .file "test.c" 14404 # GNU C11 (GCC) version 7.0.0 20160809 (experimental) (x86_64-pc-linux-gnu) 14405 [...snip...] 14406 # options passed: 14407 [...snip...] 14408 14409 .text 14410 .globl test 14411 .type test, @function 14412 test: 14413 .LFB0: 14414 .cfi_startproc 14415 # test.c:4: int total = 0; 14416 xorl %eax, %eax # <retval> 14417 # test.c:6: for (i = 0; i < n; i++) 14418 xorl %edx, %edx # i 14419 .L2: 14420 # test.c:6: for (i = 0; i < n; i++) 14421 cmpl %edi, %edx # n, i 14422 jge .L5 #, 14423 # test.c:7: total += i * i; 14424 movl %edx, %ecx # i, tmp92 14425 imull %edx, %ecx # i, tmp92 14426 # test.c:6: for (i = 0; i < n; i++) 14427 incl %edx # i 14428 # test.c:7: total += i * i; 14429 addl %ecx, %eax # tmp92, <retval> 14430 jmp .L2 # 14431 .L5: 14432 # test.c:10: } 14433 ret 14434 .cfi_endproc 14435 .LFE0: 14436 .size test, .-test 14437 .ident "GCC: (GNU) 7.0.0 20160809 (experimental)" 14438 .section .note.GNU-stack,"",@progbits 14439 14440 The comments are intended for humans rather than machines and hence 14441 the precise format of the comments is subject to change. 14442 14443'-frecord-gcc-switches' 14444 This switch causes the command line used to invoke the compiler to 14445 be recorded into the object file that is being created. This 14446 switch is only implemented on some targets and the exact format of 14447 the recording is target and binary file format dependent, but it 14448 usually takes the form of a section containing ASCII text. This 14449 switch is related to the '-fverbose-asm' switch, but that switch 14450 only records information in the assembler output file as comments, 14451 so it never reaches the object file. See also 14452 '-grecord-gcc-switches' for another way of storing compiler options 14453 into the object file. 14454 14455'-fpic' 14456 Generate position-independent code (PIC) suitable for use in a 14457 shared library, if supported for the target machine. Such code 14458 accesses all constant addresses through a global offset table 14459 (GOT). The dynamic loader resolves the GOT entries when the 14460 program starts (the dynamic loader is not part of GCC; it is part 14461 of the operating system). If the GOT size for the linked 14462 executable exceeds a machine-specific maximum size, you get an 14463 error message from the linker indicating that '-fpic' does not 14464 work; in that case, recompile with '-fPIC' instead. (These 14465 maximums are 8k on the SPARC, 28k on AArch64 and 32k on the m68k 14466 and RS/6000. The x86 has no such limit.) 14467 14468 Position-independent code requires special support, and therefore 14469 works only on certain machines. For the x86, GCC supports PIC for 14470 System V but not for the Sun 386i. Code generated for the IBM 14471 RS/6000 is always position-independent. 14472 14473 When this flag is set, the macros '__pic__' and '__PIC__' are 14474 defined to 1. 14475 14476'-fPIC' 14477 If supported for the target machine, emit position-independent 14478 code, suitable for dynamic linking and avoiding any limit on the 14479 size of the global offset table. This option makes a difference on 14480 AArch64, m68k, PowerPC and SPARC. 14481 14482 Position-independent code requires special support, and therefore 14483 works only on certain machines. 14484 14485 When this flag is set, the macros '__pic__' and '__PIC__' are 14486 defined to 2. 14487 14488'-fpie' 14489'-fPIE' 14490 These options are similar to '-fpic' and '-fPIC', but the generated 14491 position-independent code can be only linked into executables. 14492 Usually these options are used to compile code that will be linked 14493 using the '-pie' GCC option. 14494 14495 '-fpie' and '-fPIE' both define the macros '__pie__' and '__PIE__'. 14496 The macros have the value 1 for '-fpie' and 2 for '-fPIE'. 14497 14498'-fno-plt' 14499 Do not use the PLT for external function calls in 14500 position-independent code. Instead, load the callee address at 14501 call sites from the GOT and branch to it. This leads to more 14502 efficient code by eliminating PLT stubs and exposing GOT loads to 14503 optimizations. On architectures such as 32-bit x86 where PLT stubs 14504 expect the GOT pointer in a specific register, this gives more 14505 register allocation freedom to the compiler. Lazy binding requires 14506 use of the PLT; with '-fno-plt' all external symbols are resolved 14507 at load time. 14508 14509 Alternatively, the function attribute 'noplt' can be used to avoid 14510 calls through the PLT for specific external functions. 14511 14512 In position-dependent code, a few targets also convert calls to 14513 functions that are marked to not use the PLT to use the GOT 14514 instead. 14515 14516'-fno-jump-tables' 14517 Do not use jump tables for switch statements even where it would be 14518 more efficient than other code generation strategies. This option 14519 is of use in conjunction with '-fpic' or '-fPIC' for building code 14520 that forms part of a dynamic linker and cannot reference the 14521 address of a jump table. On some targets, jump tables do not 14522 require a GOT and this option is not needed. 14523 14524'-ffixed-REG' 14525 Treat the register named REG as a fixed register; generated code 14526 should never refer to it (except perhaps as a stack pointer, frame 14527 pointer or in some other fixed role). 14528 14529 REG must be the name of a register. The register names accepted 14530 are machine-specific and are defined in the 'REGISTER_NAMES' macro 14531 in the machine description macro file. 14532 14533 This flag does not have a negative form, because it specifies a 14534 three-way choice. 14535 14536'-fcall-used-REG' 14537 Treat the register named REG as an allocable register that is 14538 clobbered by function calls. It may be allocated for temporaries 14539 or variables that do not live across a call. Functions compiled 14540 this way do not save and restore the register REG. 14541 14542 It is an error to use this flag with the frame pointer or stack 14543 pointer. Use of this flag for other registers that have fixed 14544 pervasive roles in the machine's execution model produces 14545 disastrous results. 14546 14547 This flag does not have a negative form, because it specifies a 14548 three-way choice. 14549 14550'-fcall-saved-REG' 14551 Treat the register named REG as an allocable register saved by 14552 functions. It may be allocated even for temporaries or variables 14553 that live across a call. Functions compiled this way save and 14554 restore the register REG if they use it. 14555 14556 It is an error to use this flag with the frame pointer or stack 14557 pointer. Use of this flag for other registers that have fixed 14558 pervasive roles in the machine's execution model produces 14559 disastrous results. 14560 14561 A different sort of disaster results from the use of this flag for 14562 a register in which function values may be returned. 14563 14564 This flag does not have a negative form, because it specifies a 14565 three-way choice. 14566 14567'-fpack-struct[=N]' 14568 Without a value specified, pack all structure members together 14569 without holes. When a value is specified (which must be a small 14570 power of two), pack structure members according to this value, 14571 representing the maximum alignment (that is, objects with default 14572 alignment requirements larger than this are output potentially 14573 unaligned at the next fitting location. 14574 14575 *Warning:* the '-fpack-struct' switch causes GCC to generate code 14576 that is not binary compatible with code generated without that 14577 switch. Additionally, it makes the code suboptimal. Use it to 14578 conform to a non-default application binary interface. 14579 14580'-fleading-underscore' 14581 This option and its counterpart, '-fno-leading-underscore', 14582 forcibly change the way C symbols are represented in the object 14583 file. One use is to help link with legacy assembly code. 14584 14585 *Warning:* the '-fleading-underscore' switch causes GCC to generate 14586 code that is not binary compatible with code generated without that 14587 switch. Use it to conform to a non-default application binary 14588 interface. Not all targets provide complete support for this 14589 switch. 14590 14591'-ftls-model=MODEL' 14592 Alter the thread-local storage model to be used (*note 14593 Thread-Local::). The MODEL argument should be one of 14594 'global-dynamic', 'local-dynamic', 'initial-exec' or 'local-exec'. 14595 Note that the choice is subject to optimization: the compiler may 14596 use a more efficient model for symbols not visible outside of the 14597 translation unit, or if '-fpic' is not given on the command line. 14598 14599 The default without '-fpic' is 'initial-exec'; with '-fpic' the 14600 default is 'global-dynamic'. 14601 14602'-ftrampolines' 14603 For targets that normally need trampolines for nested functions, 14604 always generate them instead of using descriptors. Otherwise, for 14605 targets that do not need them, like for example HP-PA or IA-64, do 14606 nothing. 14607 14608 A trampoline is a small piece of code that is created at run time 14609 on the stack when the address of a nested function is taken, and is 14610 used to call the nested function indirectly. Therefore, it 14611 requires the stack to be made executable in order for the program 14612 to work properly. 14613 14614 '-fno-trampolines' is enabled by default on a language by language 14615 basis to let the compiler avoid generating them, if it computes 14616 that this is safe, and replace them with descriptors. Descriptors 14617 are made up of data only, but the generated code must be prepared 14618 to deal with them. As of this writing, '-fno-trampolines' is 14619 enabled by default only for Ada. 14620 14621 Moreover, code compiled with '-ftrampolines' and code compiled with 14622 '-fno-trampolines' are not binary compatible if nested functions 14623 are present. This option must therefore be used on a program-wide 14624 basis and be manipulated with extreme care. 14625 14626'-fvisibility=[default|internal|hidden|protected]' 14627 Set the default ELF image symbol visibility to the specified 14628 option--all symbols are marked with this unless overridden within 14629 the code. Using this feature can very substantially improve 14630 linking and load times of shared object libraries, produce more 14631 optimized code, provide near-perfect API export and prevent symbol 14632 clashes. It is *strongly* recommended that you use this in any 14633 shared objects you distribute. 14634 14635 Despite the nomenclature, 'default' always means public; i.e., 14636 available to be linked against from outside the shared object. 14637 'protected' and 'internal' are pretty useless in real-world usage 14638 so the only other commonly used option is 'hidden'. The default if 14639 '-fvisibility' isn't specified is 'default', i.e., make every 14640 symbol public. 14641 14642 A good explanation of the benefits offered by ensuring ELF symbols 14643 have the correct visibility is given by "How To Write Shared 14644 Libraries" by Ulrich Drepper (which can be found at 14645 <https://www.akkadia.org/drepper/>)--however a superior solution 14646 made possible by this option to marking things hidden when the 14647 default is public is to make the default hidden and mark things 14648 public. This is the norm with DLLs on Windows and with 14649 '-fvisibility=hidden' and '__attribute__ ((visibility("default")))' 14650 instead of '__declspec(dllexport)' you get almost identical 14651 semantics with identical syntax. This is a great boon to those 14652 working with cross-platform projects. 14653 14654 For those adding visibility support to existing code, you may find 14655 '#pragma GCC visibility' of use. This works by you enclosing the 14656 declarations you wish to set visibility for with (for example) 14657 '#pragma GCC visibility push(hidden)' and '#pragma GCC visibility 14658 pop'. Bear in mind that symbol visibility should be viewed *as 14659 part of the API interface contract* and thus all new code should 14660 always specify visibility when it is not the default; i.e., 14661 declarations only for use within the local DSO should *always* be 14662 marked explicitly as hidden as so to avoid PLT indirection 14663 overheads--making this abundantly clear also aids readability and 14664 self-documentation of the code. Note that due to ISO C++ 14665 specification requirements, 'operator new' and 'operator delete' 14666 must always be of default visibility. 14667 14668 Be aware that headers from outside your project, in particular 14669 system headers and headers from any other library you use, may not 14670 be expecting to be compiled with visibility other than the default. 14671 You may need to explicitly say '#pragma GCC visibility 14672 push(default)' before including any such headers. 14673 14674 'extern' declarations are not affected by '-fvisibility', so a lot 14675 of code can be recompiled with '-fvisibility=hidden' with no 14676 modifications. However, this means that calls to 'extern' 14677 functions with no explicit visibility use the PLT, so it is more 14678 effective to use '__attribute ((visibility))' and/or '#pragma GCC 14679 visibility' to tell the compiler which 'extern' declarations should 14680 be treated as hidden. 14681 14682 Note that '-fvisibility' does affect C++ vague linkage entities. 14683 This means that, for instance, an exception class that is be thrown 14684 between DSOs must be explicitly marked with default visibility so 14685 that the 'type_info' nodes are unified between the DSOs. 14686 14687 An overview of these techniques, their benefits and how to use them 14688 is at <http://gcc.gnu.org/wiki/Visibility>. 14689 14690'-fstrict-volatile-bitfields' 14691 This option should be used if accesses to volatile bit-fields (or 14692 other structure fields, although the compiler usually honors those 14693 types anyway) should use a single access of the width of the 14694 field's type, aligned to a natural alignment if possible. For 14695 example, targets with memory-mapped peripheral registers might 14696 require all such accesses to be 16 bits wide; with this flag you 14697 can declare all peripheral bit-fields as 'unsigned short' (assuming 14698 short is 16 bits on these targets) to force GCC to use 16-bit 14699 accesses instead of, perhaps, a more efficient 32-bit access. 14700 14701 If this option is disabled, the compiler uses the most efficient 14702 instruction. In the previous example, that might be a 32-bit load 14703 instruction, even though that accesses bytes that do not contain 14704 any portion of the bit-field, or memory-mapped registers unrelated 14705 to the one being updated. 14706 14707 In some cases, such as when the 'packed' attribute is applied to a 14708 structure field, it may not be possible to access the field with a 14709 single read or write that is correctly aligned for the target 14710 machine. In this case GCC falls back to generating multiple 14711 accesses rather than code that will fault or truncate the result at 14712 run time. 14713 14714 Note: Due to restrictions of the C/C++11 memory model, write 14715 accesses are not allowed to touch non bit-field members. It is 14716 therefore recommended to define all bits of the field's type as 14717 bit-field members. 14718 14719 The default value of this option is determined by the application 14720 binary interface for the target processor. 14721 14722'-fsync-libcalls' 14723 This option controls whether any out-of-line instance of the 14724 '__sync' family of functions may be used to implement the C++11 14725 '__atomic' family of functions. 14726 14727 The default value of this option is enabled, thus the only useful 14728 form of the option is '-fno-sync-libcalls'. This option is used in 14729 the implementation of the 'libatomic' runtime library. 14730 14731 14732File: gcc.info, Node: Developer Options, Next: Submodel Options, Prev: Code Gen Options, Up: Invoking GCC 14733 147343.18 GCC Developer Options 14735========================== 14736 14737This section describes command-line options that are primarily of 14738interest to GCC developers, including options to support compiler 14739testing and investigation of compiler bugs and compile-time performance 14740problems. This includes options that produce debug dumps at various 14741points in the compilation; that print statistics such as memory use and 14742execution time; and that print information about GCC's configuration, 14743such as where it searches for libraries. You should rarely need to use 14744any of these options for ordinary compilation and linking tasks. 14745 14746 Many developer options that cause GCC to dump output to a file take an 14747optional '=FILENAME' suffix. You can specify 'stdout' or '-' to dump to 14748standard output, and 'stderr' for standard error. 14749 14750 If '=FILENAME' is omitted, a default dump file name is constructed by 14751concatenating the base dump file name, a pass number, phase letter, and 14752pass name. The base dump file name is the name of output file produced 14753by the compiler if explicitly specified and not an executable; otherwise 14754it is the source file name. The pass number is determined by the order 14755passes are registered with the compiler's pass manager. This is 14756generally the same as the order of execution, but passes registered by 14757plugins, target-specific passes, or passes that are otherwise registered 14758late are numbered higher than the pass named 'final', even if they are 14759executed earlier. The phase letter is one of 'i' (inter-procedural 14760analysis), 'l' (language-specific), 'r' (RTL), or 't' (tree). The files 14761are created in the directory of the output file. 14762 14763'-fcallgraph-info' 14764'-fcallgraph-info=MARKERS' 14765 Makes the compiler output callgraph information for the program, on 14766 a per-object-file basis. The information is generated in the 14767 common VCG format. It can be decorated with additional, per-node 14768 and/or per-edge information, if a list of comma-separated markers 14769 is additionally specified. When the 'su' marker is specified, the 14770 callgraph is decorated with stack usage information; it is 14771 equivalent to '-fstack-usage'. When the 'da' marker is specified, 14772 the callgraph is decorated with information about dynamically 14773 allocated objects. 14774 14775 When compiling with '-flto', no callgraph information is output 14776 along with the object file. At LTO link time, '-fcallgraph-info' 14777 may generate multiple callgraph information files next to 14778 intermediate LTO output files. 14779 14780'-dLETTERS' 14781'-fdump-rtl-PASS' 14782'-fdump-rtl-PASS=FILENAME' 14783 Says to make debugging dumps during compilation at times specified 14784 by LETTERS. This is used for debugging the RTL-based passes of the 14785 compiler. 14786 14787 Some '-dLETTERS' switches have different meaning when '-E' is used 14788 for preprocessing. *Note Preprocessor Options::, for information 14789 about preprocessor-specific dump options. 14790 14791 Debug dumps can be enabled with a '-fdump-rtl' switch or some '-d' 14792 option LETTERS. Here are the possible letters for use in PASS and 14793 LETTERS, and their meanings: 14794 14795 '-fdump-rtl-alignments' 14796 Dump after branch alignments have been computed. 14797 14798 '-fdump-rtl-asmcons' 14799 Dump after fixing rtl statements that have unsatisfied in/out 14800 constraints. 14801 14802 '-fdump-rtl-auto_inc_dec' 14803 Dump after auto-inc-dec discovery. This pass is only run on 14804 architectures that have auto inc or auto dec instructions. 14805 14806 '-fdump-rtl-barriers' 14807 Dump after cleaning up the barrier instructions. 14808 14809 '-fdump-rtl-bbpart' 14810 Dump after partitioning hot and cold basic blocks. 14811 14812 '-fdump-rtl-bbro' 14813 Dump after block reordering. 14814 14815 '-fdump-rtl-btl1' 14816 '-fdump-rtl-btl2' 14817 '-fdump-rtl-btl1' and '-fdump-rtl-btl2' enable dumping after 14818 the two branch target load optimization passes. 14819 14820 '-fdump-rtl-bypass' 14821 Dump after jump bypassing and control flow optimizations. 14822 14823 '-fdump-rtl-combine' 14824 Dump after the RTL instruction combination pass. 14825 14826 '-fdump-rtl-compgotos' 14827 Dump after duplicating the computed gotos. 14828 14829 '-fdump-rtl-ce1' 14830 '-fdump-rtl-ce2' 14831 '-fdump-rtl-ce3' 14832 '-fdump-rtl-ce1', '-fdump-rtl-ce2', and '-fdump-rtl-ce3' 14833 enable dumping after the three if conversion passes. 14834 14835 '-fdump-rtl-cprop_hardreg' 14836 Dump after hard register copy propagation. 14837 14838 '-fdump-rtl-csa' 14839 Dump after combining stack adjustments. 14840 14841 '-fdump-rtl-cse1' 14842 '-fdump-rtl-cse2' 14843 '-fdump-rtl-cse1' and '-fdump-rtl-cse2' enable dumping after 14844 the two common subexpression elimination passes. 14845 14846 '-fdump-rtl-dce' 14847 Dump after the standalone dead code elimination passes. 14848 14849 '-fdump-rtl-dbr' 14850 Dump after delayed branch scheduling. 14851 14852 '-fdump-rtl-dce1' 14853 '-fdump-rtl-dce2' 14854 '-fdump-rtl-dce1' and '-fdump-rtl-dce2' enable dumping after 14855 the two dead store elimination passes. 14856 14857 '-fdump-rtl-eh' 14858 Dump after finalization of EH handling code. 14859 14860 '-fdump-rtl-eh_ranges' 14861 Dump after conversion of EH handling range regions. 14862 14863 '-fdump-rtl-expand' 14864 Dump after RTL generation. 14865 14866 '-fdump-rtl-fwprop1' 14867 '-fdump-rtl-fwprop2' 14868 '-fdump-rtl-fwprop1' and '-fdump-rtl-fwprop2' enable dumping 14869 after the two forward propagation passes. 14870 14871 '-fdump-rtl-gcse1' 14872 '-fdump-rtl-gcse2' 14873 '-fdump-rtl-gcse1' and '-fdump-rtl-gcse2' enable dumping after 14874 global common subexpression elimination. 14875 14876 '-fdump-rtl-init-regs' 14877 Dump after the initialization of the registers. 14878 14879 '-fdump-rtl-initvals' 14880 Dump after the computation of the initial value sets. 14881 14882 '-fdump-rtl-into_cfglayout' 14883 Dump after converting to cfglayout mode. 14884 14885 '-fdump-rtl-ira' 14886 Dump after iterated register allocation. 14887 14888 '-fdump-rtl-jump' 14889 Dump after the second jump optimization. 14890 14891 '-fdump-rtl-loop2' 14892 '-fdump-rtl-loop2' enables dumping after the rtl loop 14893 optimization passes. 14894 14895 '-fdump-rtl-mach' 14896 Dump after performing the machine dependent reorganization 14897 pass, if that pass exists. 14898 14899 '-fdump-rtl-mode_sw' 14900 Dump after removing redundant mode switches. 14901 14902 '-fdump-rtl-rnreg' 14903 Dump after register renumbering. 14904 14905 '-fdump-rtl-outof_cfglayout' 14906 Dump after converting from cfglayout mode. 14907 14908 '-fdump-rtl-peephole2' 14909 Dump after the peephole pass. 14910 14911 '-fdump-rtl-postreload' 14912 Dump after post-reload optimizations. 14913 14914 '-fdump-rtl-pro_and_epilogue' 14915 Dump after generating the function prologues and epilogues. 14916 14917 '-fdump-rtl-sched1' 14918 '-fdump-rtl-sched2' 14919 '-fdump-rtl-sched1' and '-fdump-rtl-sched2' enable dumping 14920 after the basic block scheduling passes. 14921 14922 '-fdump-rtl-ree' 14923 Dump after sign/zero extension elimination. 14924 14925 '-fdump-rtl-seqabstr' 14926 Dump after common sequence discovery. 14927 14928 '-fdump-rtl-shorten' 14929 Dump after shortening branches. 14930 14931 '-fdump-rtl-sibling' 14932 Dump after sibling call optimizations. 14933 14934 '-fdump-rtl-split1' 14935 '-fdump-rtl-split2' 14936 '-fdump-rtl-split3' 14937 '-fdump-rtl-split4' 14938 '-fdump-rtl-split5' 14939 These options enable dumping after five rounds of instruction 14940 splitting. 14941 14942 '-fdump-rtl-sms' 14943 Dump after modulo scheduling. This pass is only run on some 14944 architectures. 14945 14946 '-fdump-rtl-stack' 14947 Dump after conversion from GCC's "flat register file" 14948 registers to the x87's stack-like registers. This pass is 14949 only run on x86 variants. 14950 14951 '-fdump-rtl-subreg1' 14952 '-fdump-rtl-subreg2' 14953 '-fdump-rtl-subreg1' and '-fdump-rtl-subreg2' enable dumping 14954 after the two subreg expansion passes. 14955 14956 '-fdump-rtl-unshare' 14957 Dump after all rtl has been unshared. 14958 14959 '-fdump-rtl-vartrack' 14960 Dump after variable tracking. 14961 14962 '-fdump-rtl-vregs' 14963 Dump after converting virtual registers to hard registers. 14964 14965 '-fdump-rtl-web' 14966 Dump after live range splitting. 14967 14968 '-fdump-rtl-regclass' 14969 '-fdump-rtl-subregs_of_mode_init' 14970 '-fdump-rtl-subregs_of_mode_finish' 14971 '-fdump-rtl-dfinit' 14972 '-fdump-rtl-dfinish' 14973 These dumps are defined but always produce empty files. 14974 14975 '-da' 14976 '-fdump-rtl-all' 14977 Produce all the dumps listed above. 14978 14979 '-dA' 14980 Annotate the assembler output with miscellaneous debugging 14981 information. 14982 14983 '-dD' 14984 Dump all macro definitions, at the end of preprocessing, in 14985 addition to normal output. 14986 14987 '-dH' 14988 Produce a core dump whenever an error occurs. 14989 14990 '-dp' 14991 Annotate the assembler output with a comment indicating which 14992 pattern and alternative is used. The length and cost of each 14993 instruction are also printed. 14994 14995 '-dP' 14996 Dump the RTL in the assembler output as a comment before each 14997 instruction. Also turns on '-dp' annotation. 14998 14999 '-dx' 15000 Just generate RTL for a function instead of compiling it. 15001 Usually used with '-fdump-rtl-expand'. 15002 15003'-fdump-debug' 15004 Dump debugging information generated during the debug generation 15005 phase. 15006 15007'-fdump-earlydebug' 15008 Dump debugging information generated during the early debug 15009 generation phase. 15010 15011'-fdump-noaddr' 15012 When doing debugging dumps, suppress address output. This makes it 15013 more feasible to use diff on debugging dumps for compiler 15014 invocations with different compiler binaries and/or different text 15015 / bss / data / heap / stack / dso start locations. 15016 15017'-freport-bug' 15018 Collect and dump debug information into a temporary file if an 15019 internal compiler error (ICE) occurs. 15020 15021'-fdump-unnumbered' 15022 When doing debugging dumps, suppress instruction numbers and 15023 address output. This makes it more feasible to use diff on 15024 debugging dumps for compiler invocations with different options, in 15025 particular with and without '-g'. 15026 15027'-fdump-unnumbered-links' 15028 When doing debugging dumps (see '-d' option above), suppress 15029 instruction numbers for the links to the previous and next 15030 instructions in a sequence. 15031 15032'-fdump-ipa-SWITCH' 15033'-fdump-ipa-SWITCH-OPTIONS' 15034 Control the dumping at various stages of inter-procedural analysis 15035 language tree to a file. The file name is generated by appending a 15036 switch specific suffix to the source file name, and the file is 15037 created in the same directory as the output file. The following 15038 dumps are possible: 15039 15040 'all' 15041 Enables all inter-procedural analysis dumps. 15042 15043 'cgraph' 15044 Dumps information about call-graph optimization, unused 15045 function removal, and inlining decisions. 15046 15047 'inline' 15048 Dump after function inlining. 15049 15050 Additionally, the options '-optimized', '-missed', '-note', and 15051 '-all' can be provided, with the same meaning as for '-fopt-info', 15052 defaulting to '-optimized'. 15053 15054 For example, '-fdump-ipa-inline-optimized-missed' will emit 15055 information on callsites that were inlined, along with callsites 15056 that were not inlined. 15057 15058 By default, the dump will contain messages about successful 15059 optimizations (equivalent to '-optimized') together with low-level 15060 details about the analysis. 15061 15062'-fdump-lang-all' 15063'-fdump-lang-SWITCH' 15064'-fdump-lang-SWITCH-OPTIONS' 15065'-fdump-lang-SWITCH-OPTIONS=FILENAME' 15066 Control the dumping of language-specific information. The OPTIONS 15067 and FILENAME portions behave as described in the '-fdump-tree' 15068 option. The following SWITCH values are accepted: 15069 15070 'all' 15071 15072 Enable all language-specific dumps. 15073 15074 'class' 15075 Dump class hierarchy information. Virtual table information 15076 is emitted unless ''slim'' is specified. This option is 15077 applicable to C++ only. 15078 15079 'raw' 15080 Dump the raw internal tree data. This option is applicable to 15081 C++ only. 15082 15083'-fdump-passes' 15084 Print on 'stderr' the list of optimization passes that are turned 15085 on and off by the current command-line options. 15086 15087'-fdump-statistics-OPTION' 15088 Enable and control dumping of pass statistics in a separate file. 15089 The file name is generated by appending a suffix ending in 15090 '.statistics' to the source file name, and the file is created in 15091 the same directory as the output file. If the '-OPTION' form is 15092 used, '-stats' causes counters to be summed over the whole 15093 compilation unit while '-details' dumps every event as the passes 15094 generate them. The default with no option is to sum counters for 15095 each function compiled. 15096 15097'-fdump-tree-all' 15098'-fdump-tree-SWITCH' 15099'-fdump-tree-SWITCH-OPTIONS' 15100'-fdump-tree-SWITCH-OPTIONS=FILENAME' 15101 Control the dumping at various stages of processing the 15102 intermediate language tree to a file. If the '-OPTIONS' form is 15103 used, OPTIONS is a list of '-' separated options which control the 15104 details of the dump. Not all options are applicable to all dumps; 15105 those that are not meaningful are ignored. The following options 15106 are available 15107 15108 'address' 15109 Print the address of each node. Usually this is not 15110 meaningful as it changes according to the environment and 15111 source file. Its primary use is for tying up a dump file with 15112 a debug environment. 15113 'asmname' 15114 If 'DECL_ASSEMBLER_NAME' has been set for a given decl, use 15115 that in the dump instead of 'DECL_NAME'. Its primary use is 15116 ease of use working backward from mangled names in the 15117 assembly file. 15118 'slim' 15119 When dumping front-end intermediate representations, inhibit 15120 dumping of members of a scope or body of a function merely 15121 because that scope has been reached. Only dump such items 15122 when they are directly reachable by some other path. 15123 15124 When dumping pretty-printed trees, this option inhibits 15125 dumping the bodies of control structures. 15126 15127 When dumping RTL, print the RTL in slim (condensed) form 15128 instead of the default LISP-like representation. 15129 'raw' 15130 Print a raw representation of the tree. By default, trees are 15131 pretty-printed into a C-like representation. 15132 'details' 15133 Enable more detailed dumps (not honored by every dump option). 15134 Also include information from the optimization passes. 15135 'stats' 15136 Enable dumping various statistics about the pass (not honored 15137 by every dump option). 15138 'blocks' 15139 Enable showing basic block boundaries (disabled in raw dumps). 15140 'graph' 15141 For each of the other indicated dump files 15142 ('-fdump-rtl-PASS'), dump a representation of the control flow 15143 graph suitable for viewing with GraphViz to 15144 'FILE.PASSID.PASS.dot'. Each function in the file is 15145 pretty-printed as a subgraph, so that GraphViz can render them 15146 all in a single plot. 15147 15148 This option currently only works for RTL dumps, and the RTL is 15149 always dumped in slim form. 15150 'vops' 15151 Enable showing virtual operands for every statement. 15152 'lineno' 15153 Enable showing line numbers for statements. 15154 'uid' 15155 Enable showing the unique ID ('DECL_UID') for each variable. 15156 'verbose' 15157 Enable showing the tree dump for each statement. 15158 'eh' 15159 Enable showing the EH region number holding each statement. 15160 'scev' 15161 Enable showing scalar evolution analysis details. 15162 'optimized' 15163 Enable showing optimization information (only available in 15164 certain passes). 15165 'missed' 15166 Enable showing missed optimization information (only available 15167 in certain passes). 15168 'note' 15169 Enable other detailed optimization information (only available 15170 in certain passes). 15171 'all' 15172 Turn on all options, except 'raw', 'slim', 'verbose' and 15173 'lineno'. 15174 'optall' 15175 Turn on all optimization options, i.e., 'optimized', 'missed', 15176 and 'note'. 15177 15178 To determine what tree dumps are available or find the dump for a 15179 pass of interest follow the steps below. 15180 15181 1. Invoke GCC with '-fdump-passes' and in the 'stderr' output 15182 look for a code that corresponds to the pass you are 15183 interested in. For example, the codes 'tree-evrp', 15184 'tree-vrp1', and 'tree-vrp2' correspond to the three Value 15185 Range Propagation passes. The number at the end distinguishes 15186 distinct invocations of the same pass. 15187 2. To enable the creation of the dump file, append the pass code 15188 to the '-fdump-' option prefix and invoke GCC with it. For 15189 example, to enable the dump from the Early Value Range 15190 Propagation pass, invoke GCC with the '-fdump-tree-evrp' 15191 option. Optionally, you may specify the name of the dump 15192 file. If you don't specify one, GCC creates as described 15193 below. 15194 3. Find the pass dump in a file whose name is composed of three 15195 components separated by a period: the name of the source file 15196 GCC was invoked to compile, a numeric suffix indicating the 15197 pass number followed by the letter 't' for tree passes (and 15198 the letter 'r' for RTL passes), and finally the pass code. 15199 For example, the Early VRP pass dump might be in a file named 15200 'myfile.c.038t.evrp' in the current working directory. Note 15201 that the numeric codes are not stable and may change from one 15202 version of GCC to another. 15203 15204'-fopt-info' 15205'-fopt-info-OPTIONS' 15206'-fopt-info-OPTIONS=FILENAME' 15207 Controls optimization dumps from various optimization passes. If 15208 the '-OPTIONS' form is used, OPTIONS is a list of '-' separated 15209 option keywords to select the dump details and optimizations. 15210 15211 The OPTIONS can be divided into three groups: 15212 1. options describing what kinds of messages should be emitted, 15213 2. options describing the verbosity of the dump, and 15214 3. options describing which optimizations should be included. 15215 The options from each group can be freely mixed as they are 15216 non-overlapping. However, in case of any conflicts, the later 15217 options override the earlier options on the command line. 15218 15219 The following options control which kinds of messages should be 15220 emitted: 15221 15222 'optimized' 15223 Print information when an optimization is successfully 15224 applied. It is up to a pass to decide which information is 15225 relevant. For example, the vectorizer passes print the source 15226 location of loops which are successfully vectorized. 15227 'missed' 15228 Print information about missed optimizations. Individual 15229 passes control which information to include in the output. 15230 'note' 15231 Print verbose information about optimizations, such as certain 15232 transformations, more detailed messages about decisions etc. 15233 'all' 15234 Print detailed optimization information. This includes 15235 'optimized', 'missed', and 'note'. 15236 15237 The following option controls the dump verbosity: 15238 15239 'internals' 15240 By default, only "high-level" messages are emitted. This 15241 option enables additional, more detailed, messages, which are 15242 likely to only be of interest to GCC developers. 15243 15244 One or more of the following option keywords can be used to 15245 describe a group of optimizations: 15246 15247 'ipa' 15248 Enable dumps from all interprocedural optimizations. 15249 'loop' 15250 Enable dumps from all loop optimizations. 15251 'inline' 15252 Enable dumps from all inlining optimizations. 15253 'omp' 15254 Enable dumps from all OMP (Offloading and Multi Processing) 15255 optimizations. 15256 'vec' 15257 Enable dumps from all vectorization optimizations. 15258 'optall' 15259 Enable dumps from all optimizations. This is a superset of 15260 the optimization groups listed above. 15261 15262 If OPTIONS is omitted, it defaults to 'optimized-optall', which 15263 means to dump messages about successful optimizations from all the 15264 passes, omitting messages that are treated as "internals". 15265 15266 If the FILENAME is provided, then the dumps from all the applicable 15267 optimizations are concatenated into the FILENAME. Otherwise the 15268 dump is output onto 'stderr'. Though multiple '-fopt-info' options 15269 are accepted, only one of them can include a FILENAME. If other 15270 filenames are provided then all but the first such option are 15271 ignored. 15272 15273 Note that the output FILENAME is overwritten in case of multiple 15274 translation units. If a combined output from multiple translation 15275 units is desired, 'stderr' should be used instead. 15276 15277 In the following example, the optimization info is output to 15278 'stderr': 15279 15280 gcc -O3 -fopt-info 15281 15282 This example: 15283 gcc -O3 -fopt-info-missed=missed.all 15284 15285 outputs missed optimization report from all the passes into 15286 'missed.all', and this one: 15287 15288 gcc -O2 -ftree-vectorize -fopt-info-vec-missed 15289 15290 prints information about missed optimization opportunities from 15291 vectorization passes on 'stderr'. Note that 15292 '-fopt-info-vec-missed' is equivalent to '-fopt-info-missed-vec'. 15293 The order of the optimization group names and message types listed 15294 after '-fopt-info' does not matter. 15295 15296 As another example, 15297 gcc -O3 -fopt-info-inline-optimized-missed=inline.txt 15298 15299 outputs information about missed optimizations as well as optimized 15300 locations from all the inlining passes into 'inline.txt'. 15301 15302 Finally, consider: 15303 15304 gcc -fopt-info-vec-missed=vec.miss -fopt-info-loop-optimized=loop.opt 15305 15306 Here the two output filenames 'vec.miss' and 'loop.opt' are in 15307 conflict since only one output file is allowed. In this case, only 15308 the first option takes effect and the subsequent options are 15309 ignored. Thus only 'vec.miss' is produced which contains dumps 15310 from the vectorizer about missed opportunities. 15311 15312'-fsave-optimization-record' 15313 Write a SRCFILE.opt-record.json.gz file detailing what 15314 optimizations were performed, for those optimizations that support 15315 '-fopt-info'. 15316 15317 This option is experimental and the format of the data within the 15318 compressed JSON file is subject to change. 15319 15320 It is roughly equivalent to a machine-readable version of 15321 '-fopt-info-all', as a collection of messages with source file, 15322 line number and column number, with the following additional data 15323 for each message: 15324 15325 * the execution count of the code being optimized, along with 15326 metadata about whether this was from actual profile data, or 15327 just an estimate, allowing consumers to prioritize messages by 15328 code hotness, 15329 15330 * the function name of the code being optimized, where 15331 applicable, 15332 15333 * the "inlining chain" for the code being optimized, so that 15334 when a function is inlined into several different places 15335 (which might themselves be inlined), the reader can 15336 distinguish between the copies, 15337 15338 * objects identifying those parts of the message that refer to 15339 expressions, statements or symbol-table nodes, which of these 15340 categories they are, and, when available, their source code 15341 location, 15342 15343 * the GCC pass that emitted the message, and 15344 15345 * the location in GCC's own code from which the message was 15346 emitted 15347 15348 Additionally, some messages are logically nested within other 15349 messages, reflecting implementation details of the optimization 15350 passes. 15351 15352'-fsched-verbose=N' 15353 On targets that use instruction scheduling, this option controls 15354 the amount of debugging output the scheduler prints to the dump 15355 files. 15356 15357 For N greater than zero, '-fsched-verbose' outputs the same 15358 information as '-fdump-rtl-sched1' and '-fdump-rtl-sched2'. For N 15359 greater than one, it also output basic block probabilities, 15360 detailed ready list information and unit/insn info. For N greater 15361 than two, it includes RTL at abort point, control-flow and regions 15362 info. And for N over four, '-fsched-verbose' also includes 15363 dependence info. 15364 15365'-fenable-KIND-PASS' 15366'-fdisable-KIND-PASS=RANGE-LIST' 15367 15368 This is a set of options that are used to explicitly disable/enable 15369 optimization passes. These options are intended for use for 15370 debugging GCC. Compiler users should use regular options for 15371 enabling/disabling passes instead. 15372 15373 '-fdisable-ipa-PASS' 15374 Disable IPA pass PASS. PASS is the pass name. If the same 15375 pass is statically invoked in the compiler multiple times, the 15376 pass name should be appended with a sequential number starting 15377 from 1. 15378 15379 '-fdisable-rtl-PASS' 15380 '-fdisable-rtl-PASS=RANGE-LIST' 15381 Disable RTL pass PASS. PASS is the pass name. If the same 15382 pass is statically invoked in the compiler multiple times, the 15383 pass name should be appended with a sequential number starting 15384 from 1. RANGE-LIST is a comma-separated list of function 15385 ranges or assembler names. Each range is a number pair 15386 separated by a colon. The range is inclusive in both ends. 15387 If the range is trivial, the number pair can be simplified as 15388 a single number. If the function's call graph node's UID 15389 falls within one of the specified ranges, the PASS is disabled 15390 for that function. The UID is shown in the function header of 15391 a dump file, and the pass names can be dumped by using option 15392 '-fdump-passes'. 15393 15394 '-fdisable-tree-PASS' 15395 '-fdisable-tree-PASS=RANGE-LIST' 15396 Disable tree pass PASS. See '-fdisable-rtl' for the 15397 description of option arguments. 15398 15399 '-fenable-ipa-PASS' 15400 Enable IPA pass PASS. PASS is the pass name. If the same 15401 pass is statically invoked in the compiler multiple times, the 15402 pass name should be appended with a sequential number starting 15403 from 1. 15404 15405 '-fenable-rtl-PASS' 15406 '-fenable-rtl-PASS=RANGE-LIST' 15407 Enable RTL pass PASS. See '-fdisable-rtl' for option argument 15408 description and examples. 15409 15410 '-fenable-tree-PASS' 15411 '-fenable-tree-PASS=RANGE-LIST' 15412 Enable tree pass PASS. See '-fdisable-rtl' for the 15413 description of option arguments. 15414 15415 Here are some examples showing uses of these options. 15416 15417 15418 # disable ccp1 for all functions 15419 -fdisable-tree-ccp1 15420 # disable complete unroll for function whose cgraph node uid is 1 15421 -fenable-tree-cunroll=1 15422 # disable gcse2 for functions at the following ranges [1,1], 15423 # [300,400], and [400,1000] 15424 # disable gcse2 for functions foo and foo2 15425 -fdisable-rtl-gcse2=foo,foo2 15426 # disable early inlining 15427 -fdisable-tree-einline 15428 # disable ipa inlining 15429 -fdisable-ipa-inline 15430 # enable tree full unroll 15431 -fenable-tree-unroll 15432 15433 15434'-fchecking' 15435'-fchecking=N' 15436 Enable internal consistency checking. The default depends on the 15437 compiler configuration. '-fchecking=2' enables further internal 15438 consistency checking that might affect code generation. 15439 15440'-frandom-seed=STRING' 15441 This option provides a seed that GCC uses in place of random 15442 numbers in generating certain symbol names that have to be 15443 different in every compiled file. It is also used to place unique 15444 stamps in coverage data files and the object files that produce 15445 them. You can use the '-frandom-seed' option to produce 15446 reproducibly identical object files. 15447 15448 The STRING can either be a number (decimal, octal or hex) or an 15449 arbitrary string (in which case it's converted to a number by 15450 computing CRC32). 15451 15452 The STRING should be different for every file you compile. 15453 15454'-save-temps' 15455'-save-temps=cwd' 15456 Store the usual "temporary" intermediate files permanently; place 15457 them in the current directory and name them based on the source 15458 file. Thus, compiling 'foo.c' with '-c -save-temps' produces files 15459 'foo.i' and 'foo.s', as well as 'foo.o'. This creates a 15460 preprocessed 'foo.i' output file even though the compiler now 15461 normally uses an integrated preprocessor. 15462 15463 When used in combination with the '-x' command-line option, 15464 '-save-temps' is sensible enough to avoid over writing an input 15465 source file with the same extension as an intermediate file. The 15466 corresponding intermediate file may be obtained by renaming the 15467 source file before using '-save-temps'. 15468 15469 If you invoke GCC in parallel, compiling several different source 15470 files that share a common base name in different subdirectories or 15471 the same source file compiled for multiple output destinations, it 15472 is likely that the different parallel compilers will interfere with 15473 each other, and overwrite the temporary files. For instance: 15474 15475 gcc -save-temps -o outdir1/foo.o indir1/foo.c& 15476 gcc -save-temps -o outdir2/foo.o indir2/foo.c& 15477 15478 may result in 'foo.i' and 'foo.o' being written to simultaneously 15479 by both compilers. 15480 15481'-save-temps=obj' 15482 Store the usual "temporary" intermediate files permanently. If the 15483 '-o' option is used, the temporary files are based on the object 15484 file. If the '-o' option is not used, the '-save-temps=obj' switch 15485 behaves like '-save-temps'. 15486 15487 For example: 15488 15489 gcc -save-temps=obj -c foo.c 15490 gcc -save-temps=obj -c bar.c -o dir/xbar.o 15491 gcc -save-temps=obj foobar.c -o dir2/yfoobar 15492 15493 creates 'foo.i', 'foo.s', 'dir/xbar.i', 'dir/xbar.s', 15494 'dir2/yfoobar.i', 'dir2/yfoobar.s', and 'dir2/yfoobar.o'. 15495 15496'-time[=FILE]' 15497 Report the CPU time taken by each subprocess in the compilation 15498 sequence. For C source files, this is the compiler proper and 15499 assembler (plus the linker if linking is done). 15500 15501 Without the specification of an output file, the output looks like 15502 this: 15503 15504 # cc1 0.12 0.01 15505 # as 0.00 0.01 15506 15507 The first number on each line is the "user time", that is time 15508 spent executing the program itself. The second number is "system 15509 time", time spent executing operating system routines on behalf of 15510 the program. Both numbers are in seconds. 15511 15512 With the specification of an output file, the output is appended to 15513 the named file, and it looks like this: 15514 15515 0.12 0.01 cc1 OPTIONS 15516 0.00 0.01 as OPTIONS 15517 15518 The "user time" and the "system time" are moved before the program 15519 name, and the options passed to the program are displayed, so that 15520 one can later tell what file was being compiled, and with which 15521 options. 15522 15523'-fdump-final-insns[=FILE]' 15524 Dump the final internal representation (RTL) to FILE. If the 15525 optional argument is omitted (or if FILE is '.'), the name of the 15526 dump file is determined by appending '.gkd' to the compilation 15527 output file name. 15528 15529'-fcompare-debug[=OPTS]' 15530 If no error occurs during compilation, run the compiler a second 15531 time, adding OPTS and '-fcompare-debug-second' to the arguments 15532 passed to the second compilation. Dump the final internal 15533 representation in both compilations, and print an error if they 15534 differ. 15535 15536 If the equal sign is omitted, the default '-gtoggle' is used. 15537 15538 The environment variable 'GCC_COMPARE_DEBUG', if defined, non-empty 15539 and nonzero, implicitly enables '-fcompare-debug'. If 15540 'GCC_COMPARE_DEBUG' is defined to a string starting with a dash, 15541 then it is used for OPTS, otherwise the default '-gtoggle' is used. 15542 15543 '-fcompare-debug=', with the equal sign but without OPTS, is 15544 equivalent to '-fno-compare-debug', which disables the dumping of 15545 the final representation and the second compilation, preventing 15546 even 'GCC_COMPARE_DEBUG' from taking effect. 15547 15548 To verify full coverage during '-fcompare-debug' testing, set 15549 'GCC_COMPARE_DEBUG' to say '-fcompare-debug-not-overridden', which 15550 GCC rejects as an invalid option in any actual compilation (rather 15551 than preprocessing, assembly or linking). To get just a warning, 15552 setting 'GCC_COMPARE_DEBUG' to '-w%n-fcompare-debug not overridden' 15553 will do. 15554 15555'-fcompare-debug-second' 15556 This option is implicitly passed to the compiler for the second 15557 compilation requested by '-fcompare-debug', along with options to 15558 silence warnings, and omitting other options that would cause the 15559 compiler to produce output to files or to standard output as a side 15560 effect. Dump files and preserved temporary files are renamed so as 15561 to contain the '.gk' additional extension during the second 15562 compilation, to avoid overwriting those generated by the first. 15563 15564 When this option is passed to the compiler driver, it causes the 15565 _first_ compilation to be skipped, which makes it useful for little 15566 other than debugging the compiler proper. 15567 15568'-gtoggle' 15569 Turn off generation of debug info, if leaving out this option 15570 generates it, or turn it on at level 2 otherwise. The position of 15571 this argument in the command line does not matter; it takes effect 15572 after all other options are processed, and it does so only once, no 15573 matter how many times it is given. This is mainly intended to be 15574 used with '-fcompare-debug'. 15575 15576'-fvar-tracking-assignments-toggle' 15577 Toggle '-fvar-tracking-assignments', in the same way that 15578 '-gtoggle' toggles '-g'. 15579 15580'-Q' 15581 Makes the compiler print out each function name as it is compiled, 15582 and print some statistics about each pass when it finishes. 15583 15584'-ftime-report' 15585 Makes the compiler print some statistics about the time consumed by 15586 each pass when it finishes. 15587 15588'-ftime-report-details' 15589 Record the time consumed by infrastructure parts separately for 15590 each pass. 15591 15592'-fira-verbose=N' 15593 Control the verbosity of the dump file for the integrated register 15594 allocator. The default value is 5. If the value N is greater or 15595 equal to 10, the dump output is sent to stderr using the same 15596 format as N minus 10. 15597 15598'-flto-report' 15599 Prints a report with internal details on the workings of the 15600 link-time optimizer. The contents of this report vary from version 15601 to version. It is meant to be useful to GCC developers when 15602 processing object files in LTO mode (via '-flto'). 15603 15604 Disabled by default. 15605 15606'-flto-report-wpa' 15607 Like '-flto-report', but only print for the WPA phase of link-time 15608 optimization. 15609 15610'-fmem-report' 15611 Makes the compiler print some statistics about permanent memory 15612 allocation when it finishes. 15613 15614'-fmem-report-wpa' 15615 Makes the compiler print some statistics about permanent memory 15616 allocation for the WPA phase only. 15617 15618'-fpre-ipa-mem-report' 15619'-fpost-ipa-mem-report' 15620 Makes the compiler print some statistics about permanent memory 15621 allocation before or after interprocedural optimization. 15622 15623'-fprofile-report' 15624 Makes the compiler print some statistics about consistency of the 15625 (estimated) profile and effect of individual passes. 15626 15627'-fstack-usage' 15628 Makes the compiler output stack usage information for the program, 15629 on a per-function basis. The filename for the dump is made by 15630 appending '.su' to the AUXNAME. AUXNAME is generated from the name 15631 of the output file, if explicitly specified and it is not an 15632 executable, otherwise it is the basename of the source file. An 15633 entry is made up of three fields: 15634 15635 * The name of the function. 15636 * A number of bytes. 15637 * One or more qualifiers: 'static', 'dynamic', 'bounded'. 15638 15639 The qualifier 'static' means that the function manipulates the 15640 stack statically: a fixed number of bytes are allocated for the 15641 frame on function entry and released on function exit; no stack 15642 adjustments are otherwise made in the function. The second field 15643 is this fixed number of bytes. 15644 15645 The qualifier 'dynamic' means that the function manipulates the 15646 stack dynamically: in addition to the static allocation described 15647 above, stack adjustments are made in the body of the function, for 15648 example to push/pop arguments around function calls. If the 15649 qualifier 'bounded' is also present, the amount of these 15650 adjustments is bounded at compile time and the second field is an 15651 upper bound of the total amount of stack used by the function. If 15652 it is not present, the amount of these adjustments is not bounded 15653 at compile time and the second field only represents the bounded 15654 part. 15655 15656'-fstats' 15657 Emit statistics about front-end processing at the end of the 15658 compilation. This option is supported only by the C++ front end, 15659 and the information is generally only useful to the G++ development 15660 team. 15661 15662'-fdbg-cnt-list' 15663 Print the name and the counter upper bound for all debug counters. 15664 15665'-fdbg-cnt=COUNTER-VALUE-LIST' 15666 Set the internal debug counter lower and upper bound. 15667 COUNTER-VALUE-LIST is a comma-separated list of 15668 NAME:LOWER_BOUND1-UPPER_BOUND1 [:LOWER_BOUND2-UPPER_BOUND2...] 15669 tuples which sets the name of the counter and list of closed 15670 intervals. The LOWER_BOUND is optional and is zero initialized if 15671 not set. For example, with '-fdbg-cnt=dce:2-4:10-11,tail_call:10', 15672 'dbg_cnt(dce)' returns true only for second, third, fourth, tenth 15673 and eleventh invocation. For 'dbg_cnt(tail_call)' true is returned 15674 for first 10 invocations. 15675 15676'-print-file-name=LIBRARY' 15677 Print the full absolute name of the library file LIBRARY that would 15678 be used when linking--and don't do anything else. With this 15679 option, GCC does not compile or link anything; it just prints the 15680 file name. 15681 15682'-print-multi-directory' 15683 Print the directory name corresponding to the multilib selected by 15684 any other switches present in the command line. This directory is 15685 supposed to exist in 'GCC_EXEC_PREFIX'. 15686 15687'-print-multi-lib' 15688 Print the mapping from multilib directory names to compiler 15689 switches that enable them. The directory name is separated from 15690 the switches by ';', and each switch starts with an '@' instead of 15691 the '-', without spaces between multiple switches. This is 15692 supposed to ease shell processing. 15693 15694'-print-multi-os-directory' 15695 Print the path to OS libraries for the selected multilib, relative 15696 to some 'lib' subdirectory. If OS libraries are present in the 15697 'lib' subdirectory and no multilibs are used, this is usually just 15698 '.', if OS libraries are present in 'libSUFFIX' sibling directories 15699 this prints e.g. '../lib64', '../lib' or '../lib32', or if OS 15700 libraries are present in 'lib/SUBDIR' subdirectories it prints e.g. 15701 'amd64', 'sparcv9' or 'ev6'. 15702 15703'-print-multiarch' 15704 Print the path to OS libraries for the selected multiarch, relative 15705 to some 'lib' subdirectory. 15706 15707'-print-prog-name=PROGRAM' 15708 Like '-print-file-name', but searches for a program such as 'cpp'. 15709 15710'-print-libgcc-file-name' 15711 Same as '-print-file-name=libgcc.a'. 15712 15713 This is useful when you use '-nostdlib' or '-nodefaultlibs' but you 15714 do want to link with 'libgcc.a'. You can do: 15715 15716 gcc -nostdlib FILES... `gcc -print-libgcc-file-name` 15717 15718'-print-search-dirs' 15719 Print the name of the configured installation directory and a list 15720 of program and library directories 'gcc' searches--and don't do 15721 anything else. 15722 15723 This is useful when 'gcc' prints the error message 'installation 15724 problem, cannot exec cpp0: No such file or directory'. To resolve 15725 this you either need to put 'cpp0' and the other compiler 15726 components where 'gcc' expects to find them, or you can set the 15727 environment variable 'GCC_EXEC_PREFIX' to the directory where you 15728 installed them. Don't forget the trailing '/'. *Note Environment 15729 Variables::. 15730 15731'-print-sysroot' 15732 Print the target sysroot directory that is used during compilation. 15733 This is the target sysroot specified either at configure time or 15734 using the '--sysroot' option, possibly with an extra suffix that 15735 depends on compilation options. If no target sysroot is specified, 15736 the option prints nothing. 15737 15738'-print-sysroot-headers-suffix' 15739 Print the suffix added to the target sysroot when searching for 15740 headers, or give an error if the compiler is not configured with 15741 such a suffix--and don't do anything else. 15742 15743'-dumpmachine' 15744 Print the compiler's target machine (for example, 15745 'i686-pc-linux-gnu')--and don't do anything else. 15746 15747'-dumpversion' 15748 Print the compiler version (for example, '3.0', '6.3.0' or 15749 '7')--and don't do anything else. This is the compiler version 15750 used in filesystem paths and specs. Depending on how the compiler 15751 has been configured it can be just a single number (major version), 15752 two numbers separated by a dot (major and minor version) or three 15753 numbers separated by dots (major, minor and patchlevel version). 15754 15755'-dumpfullversion' 15756 Print the full compiler version--and don't do anything else. The 15757 output is always three numbers separated by dots, major, minor and 15758 patchlevel version. 15759 15760'-dumpspecs' 15761 Print the compiler's built-in specs--and don't do anything else. 15762 (This is used when GCC itself is being built.) *Note Spec Files::. 15763 15764 15765File: gcc.info, Node: Submodel Options, Next: Spec Files, Prev: Developer Options, Up: Invoking GCC 15766 157673.19 Machine-Dependent Options 15768============================== 15769 15770Each target machine supported by GCC can have its own options--for 15771example, to allow you to compile for a particular processor variant or 15772ABI, or to control optimizations specific to that machine. By 15773convention, the names of machine-specific options start with '-m'. 15774 15775 Some configurations of the compiler also support additional 15776target-specific options, usually for compatibility with other compilers 15777on the same platform. 15778 15779* Menu: 15780 15781* AArch64 Options:: 15782* Adapteva Epiphany Options:: 15783* AMD GCN Options:: 15784* ARC Options:: 15785* ARM Options:: 15786* AVR Options:: 15787* Blackfin Options:: 15788* C6X Options:: 15789* CRIS Options:: 15790* CR16 Options:: 15791* C-SKY Options:: 15792* Darwin Options:: 15793* DEC Alpha Options:: 15794* eBPF Options:: 15795* FR30 Options:: 15796* FT32 Options:: 15797* FRV Options:: 15798* GNU/Linux Options:: 15799* H8/300 Options:: 15800* HPPA Options:: 15801* IA-64 Options:: 15802* LM32 Options:: 15803* M32C Options:: 15804* M32R/D Options:: 15805* M680x0 Options:: 15806* MCore Options:: 15807* MeP Options:: 15808* MicroBlaze Options:: 15809* MIPS Options:: 15810* MMIX Options:: 15811* MN10300 Options:: 15812* Moxie Options:: 15813* MSP430 Options:: 15814* NDS32 Options:: 15815* Nios II Options:: 15816* Nvidia PTX Options:: 15817* OpenRISC Options:: 15818* PDP-11 Options:: 15819* picoChip Options:: 15820* PowerPC Options:: 15821* PRU Options:: 15822* RISC-V Options:: 15823* RL78 Options:: 15824* RS/6000 and PowerPC Options:: 15825* RX Options:: 15826* S/390 and zSeries Options:: 15827* Score Options:: 15828* SH Options:: 15829* Solaris 2 Options:: 15830* SPARC Options:: 15831* System V Options:: 15832* TILE-Gx Options:: 15833* TILEPro Options:: 15834* V850 Options:: 15835* VAX Options:: 15836* Visium Options:: 15837* VMS Options:: 15838* VxWorks Options:: 15839* x86 Options:: 15840* x86 Windows Options:: 15841* Xstormy16 Options:: 15842* Xtensa Options:: 15843* zSeries Options:: 15844 15845 15846File: gcc.info, Node: AArch64 Options, Next: Adapteva Epiphany Options, Up: Submodel Options 15847 158483.19.1 AArch64 Options 15849---------------------- 15850 15851These options are defined for AArch64 implementations: 15852 15853'-mabi=NAME' 15854 Generate code for the specified data model. Permissible values are 15855 'ilp32' for SysV-like data model where int, long int and pointers 15856 are 32 bits, and 'lp64' for SysV-like data model where int is 32 15857 bits, but long int and pointers are 64 bits. 15858 15859 The default depends on the specific target configuration. Note 15860 that the LP64 and ILP32 ABIs are not link-compatible; you must 15861 compile your entire program with the same ABI, and link with a 15862 compatible set of libraries. 15863 15864'-mbig-endian' 15865 Generate big-endian code. This is the default when GCC is 15866 configured for an 'aarch64_be-*-*' target. 15867 15868'-mgeneral-regs-only' 15869 Generate code which uses only the general-purpose registers. This 15870 will prevent the compiler from using floating-point and Advanced 15871 SIMD registers but will not impose any restrictions on the 15872 assembler. 15873 15874'-mlittle-endian' 15875 Generate little-endian code. This is the default when GCC is 15876 configured for an 'aarch64-*-*' but not an 'aarch64_be-*-*' target. 15877 15878'-mcmodel=tiny' 15879 Generate code for the tiny code model. The program and its 15880 statically defined symbols must be within 1MB of each other. 15881 Programs can be statically or dynamically linked. 15882 15883'-mcmodel=small' 15884 Generate code for the small code model. The program and its 15885 statically defined symbols must be within 4GB of each other. 15886 Programs can be statically or dynamically linked. This is the 15887 default code model. 15888 15889'-mcmodel=large' 15890 Generate code for the large code model. This makes no assumptions 15891 about addresses and sizes of sections. Programs can be statically 15892 linked only. The '-mcmodel=large' option is incompatible with 15893 '-mabi=ilp32', '-fpic' and '-fPIC'. 15894 15895'-mstrict-align' 15896'-mno-strict-align' 15897 Avoid or allow generating memory accesses that may not be aligned 15898 on a natural object boundary as described in the architecture 15899 specification. 15900 15901'-momit-leaf-frame-pointer' 15902'-mno-omit-leaf-frame-pointer' 15903 Omit or keep the frame pointer in leaf functions. The former 15904 behavior is the default. 15905 15906'-mstack-protector-guard=GUARD' 15907'-mstack-protector-guard-reg=REG' 15908'-mstack-protector-guard-offset=OFFSET' 15909 Generate stack protection code using canary at GUARD. Supported 15910 locations are 'global' for a global canary or 'sysreg' for a canary 15911 in an appropriate system register. 15912 15913 With the latter choice the options 15914 '-mstack-protector-guard-reg=REG' and 15915 '-mstack-protector-guard-offset=OFFSET' furthermore specify which 15916 system register to use as base register for reading the canary, and 15917 from what offset from that base register. There is no default 15918 register or offset as this is entirely for use within the Linux 15919 kernel. 15920 15921'-mstack-protector-guard=GUARD' 15922'-mstack-protector-guard-reg=REG' 15923'-mstack-protector-guard-offset=OFFSET' 15924 Generate stack protection code using canary at GUARD. Supported 15925 locations are 'global' for a global canary or 'sysreg' for a canary 15926 in an appropriate system register. 15927 15928 With the latter choice the options 15929 '-mstack-protector-guard-reg=REG' and 15930 '-mstack-protector-guard-offset=OFFSET' furthermore specify which 15931 system register to use as base register for reading the canary, and 15932 from what offset from that base register. There is no default 15933 register or offset as this is entirely for use within the Linux 15934 kernel. 15935 15936'-mtls-dialect=desc' 15937 Use TLS descriptors as the thread-local storage mechanism for 15938 dynamic accesses of TLS variables. This is the default. 15939 15940'-mtls-dialect=traditional' 15941 Use traditional TLS as the thread-local storage mechanism for 15942 dynamic accesses of TLS variables. 15943 15944'-mtls-size=SIZE' 15945 Specify bit size of immediate TLS offsets. Valid values are 12, 15946 24, 32, 48. This option requires binutils 2.26 or newer. 15947 15948'-mfix-cortex-a53-835769' 15949'-mno-fix-cortex-a53-835769' 15950 Enable or disable the workaround for the ARM Cortex-A53 erratum 15951 number 835769. This involves inserting a NOP instruction between 15952 memory instructions and 64-bit integer multiply-accumulate 15953 instructions. 15954 15955'-mfix-cortex-a53-843419' 15956'-mno-fix-cortex-a53-843419' 15957 Enable or disable the workaround for the ARM Cortex-A53 erratum 15958 number 843419. This erratum workaround is made at link time and 15959 this will only pass the corresponding flag to the linker. 15960 15961'-mlow-precision-recip-sqrt' 15962'-mno-low-precision-recip-sqrt' 15963 Enable or disable the reciprocal square root approximation. This 15964 option only has an effect if '-ffast-math' or 15965 '-funsafe-math-optimizations' is used as well. Enabling this 15966 reduces precision of reciprocal square root results to about 16 15967 bits for single precision and to 32 bits for double precision. 15968 15969'-mlow-precision-sqrt' 15970'-mno-low-precision-sqrt' 15971 Enable or disable the square root approximation. This option only 15972 has an effect if '-ffast-math' or '-funsafe-math-optimizations' is 15973 used as well. Enabling this reduces precision of square root 15974 results to about 16 bits for single precision and to 32 bits for 15975 double precision. If enabled, it implies 15976 '-mlow-precision-recip-sqrt'. 15977 15978'-mlow-precision-div' 15979'-mno-low-precision-div' 15980 Enable or disable the division approximation. This option only has 15981 an effect if '-ffast-math' or '-funsafe-math-optimizations' is used 15982 as well. Enabling this reduces precision of division results to 15983 about 16 bits for single precision and to 32 bits for double 15984 precision. 15985 15986'-mtrack-speculation' 15987'-mno-track-speculation' 15988 Enable or disable generation of additional code to track 15989 speculative execution through conditional branches. The tracking 15990 state can then be used by the compiler when expanding calls to 15991 '__builtin_speculation_safe_copy' to permit a more efficient code 15992 sequence to be generated. 15993 15994'-moutline-atomics' 15995'-mno-outline-atomics' 15996 Enable or disable calls to out-of-line helpers to implement atomic 15997 operations. These helpers will, at runtime, determine if the LSE 15998 instructions from ARMv8.1-A can be used; if not, they will use the 15999 load/store-exclusive instructions that are present in the base 16000 ARMv8.0 ISA. 16001 16002 This option is only applicable when compiling for the base ARMv8.0 16003 instruction set. If using a later revision, e.g. 16004 '-march=armv8.1-a' or '-march=armv8-a+lse', the ARMv8.1-Atomics 16005 instructions will be used directly. The same applies when using 16006 '-mcpu=' when the selected cpu supports the 'lse' feature. This 16007 option is on by default. 16008 16009'-march=NAME' 16010 Specify the name of the target architecture and, optionally, one or 16011 more feature modifiers. This option has the form 16012 '-march=ARCH{+[no]FEATURE}*'. 16013 16014 The table below summarizes the permissible values for ARCH and the 16015 features that they enable by default: 16016 16017 ARCH value Architecture Includes by default 16018 -------------------------------------------------------------------------- 16019 'armv8-a' Armv8-A '+fp', '+simd' 16020 'armv8.1-a' Armv8.1-A 'armv8-a', '+crc', '+lse', '+rdma' 16021 'armv8.2-a' Armv8.2-A 'armv8.1-a' 16022 'armv8.3-a' Armv8.3-A 'armv8.2-a' 16023 'armv8.4-a' Armv8.4-A 'armv8.3-a', '+fp16fml', '+dotprod' 16024 'armv8.5-a' Armv8.5-A 'armv8.4-a', '+sb', '+ssbs', '+predres' 16025 'armv8.6-a' Armv8.6-A 'armv8.5-a', '+bf16', '+i8mm' 16026 16027 The value 'native' is available on native AArch64 GNU/Linux and 16028 causes the compiler to pick the architecture of the host system. 16029 This option has no effect if the compiler is unable to recognize 16030 the architecture of the host system, 16031 16032 The permissible values for FEATURE are listed in the sub-section on 16033 *note '-march' and '-mcpu' Feature Modifiers: 16034 aarch64-feature-modifiers. Where conflicting feature modifiers are 16035 specified, the right-most feature is used. 16036 16037 GCC uses NAME to determine what kind of instructions it can emit 16038 when generating assembly code. If '-march' is specified without 16039 either of '-mtune' or '-mcpu' also being specified, the code is 16040 tuned to perform well across a range of target processors 16041 implementing the target architecture. 16042 16043'-mtune=NAME' 16044 Specify the name of the target processor for which GCC should tune 16045 the performance of the code. Permissible values for this option 16046 are: 'generic', 'cortex-a35', 'cortex-a53', 'cortex-a55', 16047 'cortex-a57', 'cortex-a72', 'cortex-a73', 'cortex-a75', 16048 'cortex-a76', 'cortex-a76ae', 'cortex-a77', 'cortex-a65', 16049 'cortex-a65ae', 'cortex-a34', 'ares', 'exynos-m1', 'emag', 16050 'falkor', 'neoverse-e1', 'neoverse-n1', 'neoverse-n2', 16051 'neoverse-v1', 'neoverse-v2', 'neoverse-512tvb', 'qdf24xx', 16052 'saphira', 'phecda', 'xgene1', 'vulcan', 'octeontx', 'octeontx81', 16053 'octeontx83', 'octeontx2', 'octeontx2t98', 'octeontx2t96' 16054 'octeontx2t93', 'octeontx2f95', 'octeontx2f95n', 'octeontx2f95mm', 16055 'a64fx', 'thunderx', 'thunderxt88', 'thunderxt88p1', 'thunderxt81', 16056 'tsv110', 'thunderxt83', 'thunderx2t99', 'thunderx3t110', 'zeus', 16057 'cortex-a57.cortex-a53', 'cortex-a72.cortex-a53', 16058 'cortex-a73.cortex-a35', 'cortex-a73.cortex-a53', 16059 'cortex-a75.cortex-a55', 'cortex-a76.cortex-a55', 'ampere1', 16060 'ampere1a', 'native'. 16061 16062 The values 'cortex-a57.cortex-a53', 'cortex-a72.cortex-a53', 16063 'cortex-a73.cortex-a35', 'cortex-a73.cortex-a53', 16064 'cortex-a75.cortex-a55', 'cortex-a76.cortex-a55' specify that GCC 16065 should tune for a big.LITTLE system. 16066 16067 The value 'neoverse-512tvb' specifies that GCC should tune for 16068 Neoverse cores that (a) implement SVE and (b) have a total vector 16069 bandwidth of 512 bits per cycle. In other words, the option tells 16070 GCC to tune for Neoverse cores that can execute 4 128-bit Advanced 16071 SIMD arithmetic instructions a cycle and that can execute an 16072 equivalent number of SVE arithmetic instructions per cycle (2 for 16073 256-bit SVE, 4 for 128-bit SVE). This is more general than tuning 16074 for a specific core like Neoverse V1 but is more specific than the 16075 default tuning described below. 16076 16077 Additionally on native AArch64 GNU/Linux systems the value 'native' 16078 tunes performance to the host system. This option has no effect if 16079 the compiler is unable to recognize the processor of the host 16080 system. 16081 16082 Where none of '-mtune=', '-mcpu=' or '-march=' are specified, the 16083 code is tuned to perform well across a range of target processors. 16084 16085 This option cannot be suffixed by feature modifiers. 16086 16087'-mcpu=NAME' 16088 Specify the name of the target processor, optionally suffixed by 16089 one or more feature modifiers. This option has the form 16090 '-mcpu=CPU{+[no]FEATURE}*', where the permissible values for CPU 16091 are the same as those available for '-mtune'. The permissible 16092 values for FEATURE are documented in the sub-section on *note 16093 '-march' and '-mcpu' Feature Modifiers: aarch64-feature-modifiers. 16094 Where conflicting feature modifiers are specified, the right-most 16095 feature is used. 16096 16097 GCC uses NAME to determine what kind of instructions it can emit 16098 when generating assembly code (as if by '-march') and to determine 16099 the target processor for which to tune for performance (as if by 16100 '-mtune'). Where this option is used in conjunction with '-march' 16101 or '-mtune', those options take precedence over the appropriate 16102 part of this option. 16103 16104 '-mcpu=neoverse-512tvb' is special in that it does not refer to a 16105 specific core, but instead refers to all Neoverse cores that (a) 16106 implement SVE and (b) have a total vector bandwidth of 512 bits a 16107 cycle. Unless overridden by '-march', '-mcpu=neoverse-512tvb' 16108 generates code that can run on a Neoverse V1 core, since Neoverse 16109 V1 is the first Neoverse core with these properties. Unless 16110 overridden by '-mtune', '-mcpu=neoverse-512tvb' tunes code in the 16111 same way as for '-mtune=neoverse-512tvb'. 16112 16113'-moverride=STRING' 16114 Override tuning decisions made by the back-end in response to a 16115 '-mtune=' switch. The syntax, semantics, and accepted values for 16116 STRING in this option are not guaranteed to be consistent across 16117 releases. 16118 16119 This option is only intended to be useful when developing GCC. 16120 16121'-mverbose-cost-dump' 16122 Enable verbose cost model dumping in the debug dump files. This 16123 option is provided for use in debugging the compiler. 16124 16125'-mpc-relative-literal-loads' 16126'-mno-pc-relative-literal-loads' 16127 Enable or disable PC-relative literal loads. With this option 16128 literal pools are accessed using a single instruction and emitted 16129 after each function. This limits the maximum size of functions to 16130 1MB. This is enabled by default for '-mcmodel=tiny'. 16131 16132'-msign-return-address=SCOPE' 16133 Select the function scope on which return address signing will be 16134 applied. Permissible values are 'none', which disables return 16135 address signing, 'non-leaf', which enables pointer signing for 16136 functions which are not leaf functions, and 'all', which enables 16137 pointer signing for all functions. The default value is 'none'. 16138 This option has been deprecated by -mbranch-protection. 16139 16140'-mbranch-protection=NONE|STANDARD|PAC-RET[+LEAF+B-KEY]|BTI' 16141 Select the branch protection features to use. 'none' is the 16142 default and turns off all types of branch protection. 'standard' 16143 turns on all types of branch protection features. If a feature has 16144 additional tuning options, then 'standard' sets it to its standard 16145 level. 'pac-ret[+LEAF]' turns on return address signing to its 16146 standard level: signing functions that save the return address to 16147 memory (non-leaf functions will practically always do this) using 16148 the a-key. The optional argument 'leaf' can be used to extend the 16149 signing to include leaf functions. The optional argument 'b-key' 16150 can be used to sign the functions with the B-key instead of the 16151 A-key. 'bti' turns on branch target identification mechanism. 16152 16153'-mharden-sls=OPTS' 16154 Enable compiler hardening against straight line speculation (SLS). 16155 OPTS is a comma-separated list of the following options: 16156 'retbr' 16157 'blr' 16158 In addition, '-mharden-sls=all' enables all SLS hardening while 16159 '-mharden-sls=none' disables all SLS hardening. 16160 16161'-msve-vector-bits=BITS' 16162 Specify the number of bits in an SVE vector register. This option 16163 only has an effect when SVE is enabled. 16164 16165 GCC supports two forms of SVE code generation: "vector-length 16166 agnostic" output that works with any size of vector register and 16167 "vector-length specific" output that allows GCC to make assumptions 16168 about the vector length when it is useful for optimization reasons. 16169 The possible values of 'bits' are: 'scalable', '128', '256', '512', 16170 '1024' and '2048'. Specifying 'scalable' selects vector-length 16171 agnostic output. At present '-msve-vector-bits=128' also generates 16172 vector-length agnostic output for big-endian targets. All other 16173 values generate vector-length specific code. The behavior of these 16174 values may change in future releases and no value except 'scalable' 16175 should be relied on for producing code that is portable across 16176 different hardware SVE vector lengths. 16177 16178 The default is '-msve-vector-bits=scalable', which produces 16179 vector-length agnostic code. 16180 161813.19.1.1 '-march' and '-mcpu' Feature Modifiers 16182............................................... 16183 16184Feature modifiers used with '-march' and '-mcpu' can be any of the 16185following and their inverses 'noFEATURE': 16186 16187'crc' 16188 Enable CRC extension. This is on by default for 16189 '-march=armv8.1-a'. 16190'crypto' 16191 Enable Crypto extension. This also enables Advanced SIMD and 16192 floating-point instructions. 16193'fp' 16194 Enable floating-point instructions. This is on by default for all 16195 possible values for options '-march' and '-mcpu'. 16196'simd' 16197 Enable Advanced SIMD instructions. This also enables 16198 floating-point instructions. This is on by default for all 16199 possible values for options '-march' and '-mcpu'. 16200'sve' 16201 Enable Scalable Vector Extension instructions. This also enables 16202 Advanced SIMD and floating-point instructions. 16203'lse' 16204 Enable Large System Extension instructions. This is on by default 16205 for '-march=armv8.1-a'. 16206'rdma' 16207 Enable Round Double Multiply Accumulate instructions. This is on 16208 by default for '-march=armv8.1-a'. 16209'fp16' 16210 Enable FP16 extension. This also enables floating-point 16211 instructions. 16212'fp16fml' 16213 Enable FP16 fmla extension. This also enables FP16 extensions and 16214 floating-point instructions. This option is enabled by default for 16215 '-march=armv8.4-a'. Use of this option with architectures prior to 16216 Armv8.2-A is not supported. 16217 16218'rcpc' 16219 Enable the RcPc extension. This does not change code generation 16220 from GCC, but is passed on to the assembler, enabling inline asm 16221 statements to use instructions from the RcPc extension. 16222'dotprod' 16223 Enable the Dot Product extension. This also enables Advanced SIMD 16224 instructions. 16225'aes' 16226 Enable the Armv8-a aes and pmull crypto extension. This also 16227 enables Advanced SIMD instructions. 16228'sha2' 16229 Enable the Armv8-a sha2 crypto extension. This also enables 16230 Advanced SIMD instructions. 16231'sha3' 16232 Enable the sha512 and sha3 crypto extension. This also enables 16233 Advanced SIMD instructions. Use of this option with architectures 16234 prior to Armv8.2-A is not supported. 16235'sm4' 16236 Enable the sm3 and sm4 crypto extension. This also enables 16237 Advanced SIMD instructions. Use of this option with architectures 16238 prior to Armv8.2-A is not supported. 16239'profile' 16240 Enable the Statistical Profiling extension. This option is only to 16241 enable the extension at the assembler level and does not affect 16242 code generation. 16243'rng' 16244 Enable the Armv8.5-a Random Number instructions. This option is 16245 only to enable the extension at the assembler level and does not 16246 affect code generation. 16247'memtag' 16248 Enable the Armv8.5-a Memory Tagging Extensions. Use of this option 16249 with architectures prior to Armv8.5-A is not supported. 16250'sb' 16251 Enable the Armv8-a Speculation Barrier instruction. This option is 16252 only to enable the extension at the assembler level and does not 16253 affect code generation. This option is enabled by default for 16254 '-march=armv8.5-a'. 16255'ssbs' 16256 Enable the Armv8-a Speculative Store Bypass Safe instruction. This 16257 option is only to enable the extension at the assembler level and 16258 does not affect code generation. This option is enabled by default 16259 for '-march=armv8.5-a'. 16260'predres' 16261 Enable the Armv8-a Execution and Data Prediction Restriction 16262 instructions. This option is only to enable the extension at the 16263 assembler level and does not affect code generation. This option 16264 is enabled by default for '-march=armv8.5-a'. 16265'sve2' 16266 Enable the Armv8-a Scalable Vector Extension 2. This also enables 16267 SVE instructions. 16268'sve2-bitperm' 16269 Enable SVE2 bitperm instructions. This also enables SVE2 16270 instructions. 16271'sve2-sm4' 16272 Enable SVE2 sm4 instructions. This also enables SVE2 instructions. 16273'sve2-aes' 16274 Enable SVE2 aes instructions. This also enables SVE2 instructions. 16275'sve2-sha3' 16276 Enable SVE2 sha3 instructions. This also enables SVE2 16277 instructions. 16278'tme' 16279 Enable the Transactional Memory Extension. 16280'i8mm' 16281 Enable 8-bit Integer Matrix Multiply instructions. This also 16282 enables Advanced SIMD and floating-point instructions. This option 16283 is enabled by default for '-march=armv8.6-a'. Use of this option 16284 with architectures prior to Armv8.2-A is not supported. 16285'f32mm' 16286 Enable 32-bit Floating point Matrix Multiply instructions. This 16287 also enables SVE instructions. Use of this option with 16288 architectures prior to Armv8.2-A is not supported. 16289'f64mm' 16290 Enable 64-bit Floating point Matrix Multiply instructions. This 16291 also enables SVE instructions. Use of this option with 16292 architectures prior to Armv8.2-A is not supported. 16293'bf16' 16294 Enable brain half-precision floating-point instructions. This also 16295 enables Advanced SIMD and floating-point instructions. This option 16296 is enabled by default for '-march=armv8.6-a'. Use of this option 16297 with architectures prior to Armv8.2-A is not supported. 16298 16299 Feature 'crypto' implies 'aes', 'sha2', and 'simd', which implies 'fp'. 16300Conversely, 'nofp' implies 'nosimd', which implies 'nocrypto', 'noaes' 16301and 'nosha2'. 16302 16303 16304File: gcc.info, Node: Adapteva Epiphany Options, Next: AMD GCN Options, Prev: AArch64 Options, Up: Submodel Options 16305 163063.19.2 Adapteva Epiphany Options 16307-------------------------------- 16308 16309These '-m' options are defined for Adapteva Epiphany: 16310 16311'-mhalf-reg-file' 16312 Don't allocate any register in the range 'r32'...'r63'. That 16313 allows code to run on hardware variants that lack these registers. 16314 16315'-mprefer-short-insn-regs' 16316 Preferentially allocate registers that allow short instruction 16317 generation. This can result in increased instruction count, so 16318 this may either reduce or increase overall code size. 16319 16320'-mbranch-cost=NUM' 16321 Set the cost of branches to roughly NUM "simple" instructions. 16322 This cost is only a heuristic and is not guaranteed to produce 16323 consistent results across releases. 16324 16325'-mcmove' 16326 Enable the generation of conditional moves. 16327 16328'-mnops=NUM' 16329 Emit NUM NOPs before every other generated instruction. 16330 16331'-mno-soft-cmpsf' 16332 For single-precision floating-point comparisons, emit an 'fsub' 16333 instruction and test the flags. This is faster than a software 16334 comparison, but can get incorrect results in the presence of NaNs, 16335 or when two different small numbers are compared such that their 16336 difference is calculated as zero. The default is '-msoft-cmpsf', 16337 which uses slower, but IEEE-compliant, software comparisons. 16338 16339'-mstack-offset=NUM' 16340 Set the offset between the top of the stack and the stack pointer. 16341 E.g., a value of 8 means that the eight bytes in the range 16342 'sp+0...sp+7' can be used by leaf functions without stack 16343 allocation. Values other than '8' or '16' are untested and 16344 unlikely to work. Note also that this option changes the ABI; 16345 compiling a program with a different stack offset than the 16346 libraries have been compiled with generally does not work. This 16347 option can be useful if you want to evaluate if a different stack 16348 offset would give you better code, but to actually use a different 16349 stack offset to build working programs, it is recommended to 16350 configure the toolchain with the appropriate 16351 '--with-stack-offset=NUM' option. 16352 16353'-mno-round-nearest' 16354 Make the scheduler assume that the rounding mode has been set to 16355 truncating. The default is '-mround-nearest'. 16356 16357'-mlong-calls' 16358 If not otherwise specified by an attribute, assume all calls might 16359 be beyond the offset range of the 'b' / 'bl' instructions, and 16360 therefore load the function address into a register before 16361 performing a (otherwise direct) call. This is the default. 16362 16363'-mshort-calls' 16364 If not otherwise specified by an attribute, assume all direct calls 16365 are in the range of the 'b' / 'bl' instructions, so use these 16366 instructions for direct calls. The default is '-mlong-calls'. 16367 16368'-msmall16' 16369 Assume addresses can be loaded as 16-bit unsigned values. This 16370 does not apply to function addresses for which '-mlong-calls' 16371 semantics are in effect. 16372 16373'-mfp-mode=MODE' 16374 Set the prevailing mode of the floating-point unit. This 16375 determines the floating-point mode that is provided and expected at 16376 function call and return time. Making this mode match the mode you 16377 predominantly need at function start can make your programs smaller 16378 and faster by avoiding unnecessary mode switches. 16379 16380 MODE can be set to one the following values: 16381 16382 'caller' 16383 Any mode at function entry is valid, and retained or restored 16384 when the function returns, and when it calls other functions. 16385 This mode is useful for compiling libraries or other 16386 compilation units you might want to incorporate into different 16387 programs with different prevailing FPU modes, and the 16388 convenience of being able to use a single object file 16389 outweighs the size and speed overhead for any extra mode 16390 switching that might be needed, compared with what would be 16391 needed with a more specific choice of prevailing FPU mode. 16392 16393 'truncate' 16394 This is the mode used for floating-point calculations with 16395 truncating (i.e. round towards zero) rounding mode. That 16396 includes conversion from floating point to integer. 16397 16398 'round-nearest' 16399 This is the mode used for floating-point calculations with 16400 round-to-nearest-or-even rounding mode. 16401 16402 'int' 16403 This is the mode used to perform integer calculations in the 16404 FPU, e.g. integer multiply, or integer 16405 multiply-and-accumulate. 16406 16407 The default is '-mfp-mode=caller' 16408 16409'-mno-split-lohi' 16410'-mno-postinc' 16411'-mno-postmodify' 16412 Code generation tweaks that disable, respectively, splitting of 16413 32-bit loads, generation of post-increment addresses, and 16414 generation of post-modify addresses. The defaults are 16415 'msplit-lohi', '-mpost-inc', and '-mpost-modify'. 16416 16417'-mnovect-double' 16418 Change the preferred SIMD mode to SImode. The default is 16419 '-mvect-double', which uses DImode as preferred SIMD mode. 16420 16421'-max-vect-align=NUM' 16422 The maximum alignment for SIMD vector mode types. NUM may be 4 or 16423 8. The default is 8. Note that this is an ABI change, even though 16424 many library function interfaces are unaffected if they don't use 16425 SIMD vector modes in places that affect size and/or alignment of 16426 relevant types. 16427 16428'-msplit-vecmove-early' 16429 Split vector moves into single word moves before reload. In theory 16430 this can give better register allocation, but so far the reverse 16431 seems to be generally the case. 16432 16433'-m1reg-REG' 16434 Specify a register to hold the constant -1, which makes loading 16435 small negative constants and certain bitmasks faster. Allowable 16436 values for REG are 'r43' and 'r63', which specify use of that 16437 register as a fixed register, and 'none', which means that no 16438 register is used for this purpose. The default is '-m1reg-none'. 16439 16440 16441File: gcc.info, Node: AMD GCN Options, Next: ARC Options, Prev: Adapteva Epiphany Options, Up: Submodel Options 16442 164433.19.3 AMD GCN Options 16444---------------------- 16445 16446These options are defined specifically for the AMD GCN port. 16447 16448'-march=GPU' 16449'-mtune=GPU' 16450 Set architecture type or tuning for GPU. Supported values for GPU 16451 are 16452 16453 'fiji' 16454 Compile for GCN3 Fiji devices (gfx803). 16455 16456 'gfx900' 16457 Compile for GCN5 Vega 10 devices (gfx900). 16458 16459 'gfx906' 16460 Compile for GCN5 Vega 20 devices (gfx906). 16461 16462'-mstack-size=BYTES' 16463 Specify how many BYTES of stack space will be requested for each 16464 GPU thread (wave-front). Beware that there may be many threads and 16465 limited memory available. The size of the stack allocation may 16466 also have an impact on run-time performance. The default is 32KB 16467 when using OpenACC or OpenMP, and 1MB otherwise. 16468 16469 16470File: gcc.info, Node: ARC Options, Next: ARM Options, Prev: AMD GCN Options, Up: Submodel Options 16471 164723.19.4 ARC Options 16473------------------ 16474 16475The following options control the architecture variant for which code is 16476being compiled: 16477 16478'-mbarrel-shifter' 16479 Generate instructions supported by barrel shifter. This is the 16480 default unless '-mcpu=ARC601' or '-mcpu=ARCEM' is in effect. 16481 16482'-mjli-always' 16483 Force to call a function using jli_s instruction. This option is 16484 valid only for ARCv2 architecture. 16485 16486'-mcpu=CPU' 16487 Set architecture type, register usage, and instruction scheduling 16488 parameters for CPU. There are also shortcut alias options 16489 available for backward compatibility and convenience. Supported 16490 values for CPU are 16491 16492 'arc600' 16493 Compile for ARC600. Aliases: '-mA6', '-mARC600'. 16494 16495 'arc601' 16496 Compile for ARC601. Alias: '-mARC601'. 16497 16498 'arc700' 16499 Compile for ARC700. Aliases: '-mA7', '-mARC700'. This is the 16500 default when configured with '--with-cpu=arc700'. 16501 16502 'arcem' 16503 Compile for ARC EM. 16504 16505 'archs' 16506 Compile for ARC HS. 16507 16508 'em' 16509 Compile for ARC EM CPU with no hardware extensions. 16510 16511 'em4' 16512 Compile for ARC EM4 CPU. 16513 16514 'em4_dmips' 16515 Compile for ARC EM4 DMIPS CPU. 16516 16517 'em4_fpus' 16518 Compile for ARC EM4 DMIPS CPU with the single-precision 16519 floating-point extension. 16520 16521 'em4_fpuda' 16522 Compile for ARC EM4 DMIPS CPU with single-precision 16523 floating-point and double assist instructions. 16524 16525 'hs' 16526 Compile for ARC HS CPU with no hardware extensions except the 16527 atomic instructions. 16528 16529 'hs34' 16530 Compile for ARC HS34 CPU. 16531 16532 'hs38' 16533 Compile for ARC HS38 CPU. 16534 16535 'hs38_linux' 16536 Compile for ARC HS38 CPU with all hardware extensions on. 16537 16538 'arc600_norm' 16539 Compile for ARC 600 CPU with 'norm' instructions enabled. 16540 16541 'arc600_mul32x16' 16542 Compile for ARC 600 CPU with 'norm' and 32x16-bit multiply 16543 instructions enabled. 16544 16545 'arc600_mul64' 16546 Compile for ARC 600 CPU with 'norm' and 'mul64'-family 16547 instructions enabled. 16548 16549 'arc601_norm' 16550 Compile for ARC 601 CPU with 'norm' instructions enabled. 16551 16552 'arc601_mul32x16' 16553 Compile for ARC 601 CPU with 'norm' and 32x16-bit multiply 16554 instructions enabled. 16555 16556 'arc601_mul64' 16557 Compile for ARC 601 CPU with 'norm' and 'mul64'-family 16558 instructions enabled. 16559 16560 'nps400' 16561 Compile for ARC 700 on NPS400 chip. 16562 16563 'em_mini' 16564 Compile for ARC EM minimalist configuration featuring reduced 16565 register set. 16566 16567'-mdpfp' 16568'-mdpfp-compact' 16569 Generate double-precision FPX instructions, tuned for the compact 16570 implementation. 16571 16572'-mdpfp-fast' 16573 Generate double-precision FPX instructions, tuned for the fast 16574 implementation. 16575 16576'-mno-dpfp-lrsr' 16577 Disable 'lr' and 'sr' instructions from using FPX extension aux 16578 registers. 16579 16580'-mea' 16581 Generate extended arithmetic instructions. Currently only 'divaw', 16582 'adds', 'subs', and 'sat16' are supported. Only valid for 16583 '-mcpu=ARC700'. 16584 16585'-mno-mpy' 16586 Do not generate 'mpy'-family instructions for ARC700. This option 16587 is deprecated. 16588 16589'-mmul32x16' 16590 Generate 32x16-bit multiply and multiply-accumulate instructions. 16591 16592'-mmul64' 16593 Generate 'mul64' and 'mulu64' instructions. Only valid for 16594 '-mcpu=ARC600'. 16595 16596'-mnorm' 16597 Generate 'norm' instructions. This is the default if 16598 '-mcpu=ARC700' is in effect. 16599 16600'-mspfp' 16601'-mspfp-compact' 16602 Generate single-precision FPX instructions, tuned for the compact 16603 implementation. 16604 16605'-mspfp-fast' 16606 Generate single-precision FPX instructions, tuned for the fast 16607 implementation. 16608 16609'-msimd' 16610 Enable generation of ARC SIMD instructions via target-specific 16611 builtins. Only valid for '-mcpu=ARC700'. 16612 16613'-msoft-float' 16614 This option ignored; it is provided for compatibility purposes 16615 only. Software floating-point code is emitted by default, and this 16616 default can overridden by FPX options; '-mspfp', '-mspfp-compact', 16617 or '-mspfp-fast' for single precision, and '-mdpfp', 16618 '-mdpfp-compact', or '-mdpfp-fast' for double precision. 16619 16620'-mswap' 16621 Generate 'swap' instructions. 16622 16623'-matomic' 16624 This enables use of the locked load/store conditional extension to 16625 implement atomic memory built-in functions. Not available for ARC 16626 6xx or ARC EM cores. 16627 16628'-mdiv-rem' 16629 Enable 'div' and 'rem' instructions for ARCv2 cores. 16630 16631'-mcode-density' 16632 Enable code density instructions for ARC EM. This option is on by 16633 default for ARC HS. 16634 16635'-mll64' 16636 Enable double load/store operations for ARC HS cores. 16637 16638'-mtp-regno=REGNO' 16639 Specify thread pointer register number. 16640 16641'-mmpy-option=MULTO' 16642 Compile ARCv2 code with a multiplier design option. You can 16643 specify the option using either a string or numeric value for 16644 MULTO. 'wlh1' is the default value. The recognized values are: 16645 16646 '0' 16647 'none' 16648 No multiplier available. 16649 16650 '1' 16651 'w' 16652 16x16 multiplier, fully pipelined. The following instructions 16653 are enabled: 'mpyw' and 'mpyuw'. 16654 16655 '2' 16656 'wlh1' 16657 32x32 multiplier, fully pipelined (1 stage). The following 16658 instructions are additionally enabled: 'mpy', 'mpyu', 'mpym', 16659 'mpymu', and 'mpy_s'. 16660 16661 '3' 16662 'wlh2' 16663 32x32 multiplier, fully pipelined (2 stages). The following 16664 instructions are additionally enabled: 'mpy', 'mpyu', 'mpym', 16665 'mpymu', and 'mpy_s'. 16666 16667 '4' 16668 'wlh3' 16669 Two 16x16 multipliers, blocking, sequential. The following 16670 instructions are additionally enabled: 'mpy', 'mpyu', 'mpym', 16671 'mpymu', and 'mpy_s'. 16672 16673 '5' 16674 'wlh4' 16675 One 16x16 multiplier, blocking, sequential. The following 16676 instructions are additionally enabled: 'mpy', 'mpyu', 'mpym', 16677 'mpymu', and 'mpy_s'. 16678 16679 '6' 16680 'wlh5' 16681 One 32x4 multiplier, blocking, sequential. The following 16682 instructions are additionally enabled: 'mpy', 'mpyu', 'mpym', 16683 'mpymu', and 'mpy_s'. 16684 16685 '7' 16686 'plus_dmpy' 16687 ARC HS SIMD support. 16688 16689 '8' 16690 'plus_macd' 16691 ARC HS SIMD support. 16692 16693 '9' 16694 'plus_qmacw' 16695 ARC HS SIMD support. 16696 16697 This option is only available for ARCv2 cores. 16698 16699'-mfpu=FPU' 16700 Enables support for specific floating-point hardware extensions for 16701 ARCv2 cores. Supported values for FPU are: 16702 16703 'fpus' 16704 Enables support for single-precision floating-point hardware 16705 extensions. 16706 16707 'fpud' 16708 Enables support for double-precision floating-point hardware 16709 extensions. The single-precision floating-point extension is 16710 also enabled. Not available for ARC EM. 16711 16712 'fpuda' 16713 Enables support for double-precision floating-point hardware 16714 extensions using double-precision assist instructions. The 16715 single-precision floating-point extension is also enabled. 16716 This option is only available for ARC EM. 16717 16718 'fpuda_div' 16719 Enables support for double-precision floating-point hardware 16720 extensions using double-precision assist instructions. The 16721 single-precision floating-point, square-root, and divide 16722 extensions are also enabled. This option is only available 16723 for ARC EM. 16724 16725 'fpuda_fma' 16726 Enables support for double-precision floating-point hardware 16727 extensions using double-precision assist instructions. The 16728 single-precision floating-point and fused multiply and add 16729 hardware extensions are also enabled. This option is only 16730 available for ARC EM. 16731 16732 'fpuda_all' 16733 Enables support for double-precision floating-point hardware 16734 extensions using double-precision assist instructions. All 16735 single-precision floating-point hardware extensions are also 16736 enabled. This option is only available for ARC EM. 16737 16738 'fpus_div' 16739 Enables support for single-precision floating-point, 16740 square-root and divide hardware extensions. 16741 16742 'fpud_div' 16743 Enables support for double-precision floating-point, 16744 square-root and divide hardware extensions. This option 16745 includes option 'fpus_div'. Not available for ARC EM. 16746 16747 'fpus_fma' 16748 Enables support for single-precision floating-point and fused 16749 multiply and add hardware extensions. 16750 16751 'fpud_fma' 16752 Enables support for double-precision floating-point and fused 16753 multiply and add hardware extensions. This option includes 16754 option 'fpus_fma'. Not available for ARC EM. 16755 16756 'fpus_all' 16757 Enables support for all single-precision floating-point 16758 hardware extensions. 16759 16760 'fpud_all' 16761 Enables support for all single- and double-precision 16762 floating-point hardware extensions. Not available for ARC EM. 16763 16764'-mirq-ctrl-saved=REGISTER-RANGE, BLINK, LP_COUNT' 16765 Specifies general-purposes registers that the processor 16766 automatically saves/restores on interrupt entry and exit. 16767 REGISTER-RANGE is specified as two registers separated by a dash. 16768 The register range always starts with 'r0', the upper limit is 'fp' 16769 register. BLINK and LP_COUNT are optional. This option is only 16770 valid for ARC EM and ARC HS cores. 16771 16772'-mrgf-banked-regs=NUMBER' 16773 Specifies the number of registers replicated in second register 16774 bank on entry to fast interrupt. Fast interrupts are interrupts 16775 with the highest priority level P0. These interrupts save only PC 16776 and STATUS32 registers to avoid memory transactions during 16777 interrupt entry and exit sequences. Use this option when you are 16778 using fast interrupts in an ARC V2 family processor. Permitted 16779 values are 4, 8, 16, and 32. 16780 16781'-mlpc-width=WIDTH' 16782 Specify the width of the 'lp_count' register. Valid values for 16783 WIDTH are 8, 16, 20, 24, 28 and 32 bits. The default width is 16784 fixed to 32 bits. If the width is less than 32, the compiler does 16785 not attempt to transform loops in your program to use the 16786 zero-delay loop mechanism unless it is known that the 'lp_count' 16787 register can hold the required loop-counter value. Depending on 16788 the width specified, the compiler and run-time library might 16789 continue to use the loop mechanism for various needs. This option 16790 defines macro '__ARC_LPC_WIDTH__' with the value of WIDTH. 16791 16792'-mrf16' 16793 This option instructs the compiler to generate code for a 16-entry 16794 register file. This option defines the '__ARC_RF16__' preprocessor 16795 macro. 16796 16797'-mbranch-index' 16798 Enable use of 'bi' or 'bih' instructions to implement jump tables. 16799 16800 The following options are passed through to the assembler, and also 16801define preprocessor macro symbols. 16802 16803'-mdsp-packa' 16804 Passed down to the assembler to enable the DSP Pack A extensions. 16805 Also sets the preprocessor symbol '__Xdsp_packa'. This option is 16806 deprecated. 16807 16808'-mdvbf' 16809 Passed down to the assembler to enable the dual Viterbi butterfly 16810 extension. Also sets the preprocessor symbol '__Xdvbf'. This 16811 option is deprecated. 16812 16813'-mlock' 16814 Passed down to the assembler to enable the locked load/store 16815 conditional extension. Also sets the preprocessor symbol 16816 '__Xlock'. 16817 16818'-mmac-d16' 16819 Passed down to the assembler. Also sets the preprocessor symbol 16820 '__Xxmac_d16'. This option is deprecated. 16821 16822'-mmac-24' 16823 Passed down to the assembler. Also sets the preprocessor symbol 16824 '__Xxmac_24'. This option is deprecated. 16825 16826'-mrtsc' 16827 Passed down to the assembler to enable the 64-bit time-stamp 16828 counter extension instruction. Also sets the preprocessor symbol 16829 '__Xrtsc'. This option is deprecated. 16830 16831'-mswape' 16832 Passed down to the assembler to enable the swap byte ordering 16833 extension instruction. Also sets the preprocessor symbol 16834 '__Xswape'. 16835 16836'-mtelephony' 16837 Passed down to the assembler to enable dual- and single-operand 16838 instructions for telephony. Also sets the preprocessor symbol 16839 '__Xtelephony'. This option is deprecated. 16840 16841'-mxy' 16842 Passed down to the assembler to enable the XY memory extension. 16843 Also sets the preprocessor symbol '__Xxy'. 16844 16845 The following options control how the assembly code is annotated: 16846 16847'-misize' 16848 Annotate assembler instructions with estimated addresses. 16849 16850'-mannotate-align' 16851 Explain what alignment considerations lead to the decision to make 16852 an instruction short or long. 16853 16854 The following options are passed through to the linker: 16855 16856'-marclinux' 16857 Passed through to the linker, to specify use of the 'arclinux' 16858 emulation. This option is enabled by default in tool chains built 16859 for 'arc-linux-uclibc' and 'arceb-linux-uclibc' targets when 16860 profiling is not requested. 16861 16862'-marclinux_prof' 16863 Passed through to the linker, to specify use of the 'arclinux_prof' 16864 emulation. This option is enabled by default in tool chains built 16865 for 'arc-linux-uclibc' and 'arceb-linux-uclibc' targets when 16866 profiling is requested. 16867 16868 The following options control the semantics of generated code: 16869 16870'-mlong-calls' 16871 Generate calls as register indirect calls, thus providing access to 16872 the full 32-bit address range. 16873 16874'-mmedium-calls' 16875 Don't use less than 25-bit addressing range for calls, which is the 16876 offset available for an unconditional branch-and-link instruction. 16877 Conditional execution of function calls is suppressed, to allow use 16878 of the 25-bit range, rather than the 21-bit range with conditional 16879 branch-and-link. This is the default for tool chains built for 16880 'arc-linux-uclibc' and 'arceb-linux-uclibc' targets. 16881 16882'-G NUM' 16883 Put definitions of externally-visible data in a small data section 16884 if that data is no bigger than NUM bytes. The default value of NUM 16885 is 4 for any ARC configuration, or 8 when we have double load/store 16886 operations. 16887 16888'-mno-sdata' 16889 Do not generate sdata references. This is the default for tool 16890 chains built for 'arc-linux-uclibc' and 'arceb-linux-uclibc' 16891 targets. 16892 16893'-mvolatile-cache' 16894 Use ordinarily cached memory accesses for volatile references. 16895 This is the default. 16896 16897'-mno-volatile-cache' 16898 Enable cache bypass for volatile references. 16899 16900 The following options fine tune code generation: 16901'-malign-call' 16902 Do alignment optimizations for call instructions. 16903 16904'-mauto-modify-reg' 16905 Enable the use of pre/post modify with register displacement. 16906 16907'-mbbit-peephole' 16908 Enable bbit peephole2. 16909 16910'-mno-brcc' 16911 This option disables a target-specific pass in 'arc_reorg' to 16912 generate compare-and-branch ('brCC') instructions. It has no 16913 effect on generation of these instructions driven by the combiner 16914 pass. 16915 16916'-mcase-vector-pcrel' 16917 Use PC-relative switch case tables to enable case table shortening. 16918 This is the default for '-Os'. 16919 16920'-mcompact-casesi' 16921 Enable compact 'casesi' pattern. This is the default for '-Os', 16922 and only available for ARCv1 cores. This option is deprecated. 16923 16924'-mno-cond-exec' 16925 Disable the ARCompact-specific pass to generate conditional 16926 execution instructions. 16927 16928 Due to delay slot scheduling and interactions between operand 16929 numbers, literal sizes, instruction lengths, and the support for 16930 conditional execution, the target-independent pass to generate 16931 conditional execution is often lacking, so the ARC port has kept a 16932 special pass around that tries to find more conditional execution 16933 generation opportunities after register allocation, branch 16934 shortening, and delay slot scheduling have been done. This pass 16935 generally, but not always, improves performance and code size, at 16936 the cost of extra compilation time, which is why there is an option 16937 to switch it off. If you have a problem with call instructions 16938 exceeding their allowable offset range because they are 16939 conditionalized, you should consider using '-mmedium-calls' 16940 instead. 16941 16942'-mearly-cbranchsi' 16943 Enable pre-reload use of the 'cbranchsi' pattern. 16944 16945'-mexpand-adddi' 16946 Expand 'adddi3' and 'subdi3' at RTL generation time into 'add.f', 16947 'adc' etc. This option is deprecated. 16948 16949'-mindexed-loads' 16950 Enable the use of indexed loads. This can be problematic because 16951 some optimizers then assume that indexed stores exist, which is not 16952 the case. 16953 16954'-mlra' 16955 Enable Local Register Allocation. This is still experimental for 16956 ARC, so by default the compiler uses standard reload (i.e. 16957 '-mno-lra'). 16958 16959'-mlra-priority-none' 16960 Don't indicate any priority for target registers. 16961 16962'-mlra-priority-compact' 16963 Indicate target register priority for r0..r3 / r12..r15. 16964 16965'-mlra-priority-noncompact' 16966 Reduce target register priority for r0..r3 / r12..r15. 16967 16968'-mmillicode' 16969 When optimizing for size (using '-Os'), prologues and epilogues 16970 that have to save or restore a large number of registers are often 16971 shortened by using call to a special function in libgcc; this is 16972 referred to as a _millicode_ call. As these calls can pose 16973 performance issues, and/or cause linking issues when linking in a 16974 nonstandard way, this option is provided to turn on or off 16975 millicode call generation. 16976 16977'-mcode-density-frame' 16978 This option enable the compiler to emit 'enter' and 'leave' 16979 instructions. These instructions are only valid for CPUs with 16980 code-density feature. 16981 16982'-mmixed-code' 16983 Tweak register allocation to help 16-bit instruction generation. 16984 This generally has the effect of decreasing the average instruction 16985 size while increasing the instruction count. 16986 16987'-mq-class' 16988 Ths option is deprecated. Enable 'q' instruction alternatives. 16989 This is the default for '-Os'. 16990 16991'-mRcq' 16992 Enable 'Rcq' constraint handling. Most short code generation 16993 depends on this. This is the default. 16994 16995'-mRcw' 16996 Enable 'Rcw' constraint handling. Most ccfsm condexec mostly 16997 depends on this. This is the default. 16998 16999'-msize-level=LEVEL' 17000 Fine-tune size optimization with regards to instruction lengths and 17001 alignment. The recognized values for LEVEL are: 17002 '0' 17003 No size optimization. This level is deprecated and treated 17004 like '1'. 17005 17006 '1' 17007 Short instructions are used opportunistically. 17008 17009 '2' 17010 In addition, alignment of loops and of code after barriers are 17011 dropped. 17012 17013 '3' 17014 In addition, optional data alignment is dropped, and the 17015 option 'Os' is enabled. 17016 17017 This defaults to '3' when '-Os' is in effect. Otherwise, the 17018 behavior when this is not set is equivalent to level '1'. 17019 17020'-mtune=CPU' 17021 Set instruction scheduling parameters for CPU, overriding any 17022 implied by '-mcpu='. 17023 17024 Supported values for CPU are 17025 17026 'ARC600' 17027 Tune for ARC600 CPU. 17028 17029 'ARC601' 17030 Tune for ARC601 CPU. 17031 17032 'ARC700' 17033 Tune for ARC700 CPU with standard multiplier block. 17034 17035 'ARC700-xmac' 17036 Tune for ARC700 CPU with XMAC block. 17037 17038 'ARC725D' 17039 Tune for ARC725D CPU. 17040 17041 'ARC750D' 17042 Tune for ARC750D CPU. 17043 17044'-mmultcost=NUM' 17045 Cost to assume for a multiply instruction, with '4' being equal to 17046 a normal instruction. 17047 17048'-munalign-prob-threshold=PROBABILITY' 17049 Set probability threshold for unaligning branches. When tuning for 17050 'ARC700' and optimizing for speed, branches without filled delay 17051 slot are preferably emitted unaligned and long, unless profiling 17052 indicates that the probability for the branch to be taken is below 17053 PROBABILITY. *Note Cross-profiling::. The default is 17054 (REG_BR_PROB_BASE/2), i.e. 5000. 17055 17056 The following options are maintained for backward compatibility, but 17057are now deprecated and will be removed in a future release: 17058 17059'-margonaut' 17060 Obsolete FPX. 17061 17062'-mbig-endian' 17063'-EB' 17064 Compile code for big-endian targets. Use of these options is now 17065 deprecated. Big-endian code is supported by configuring GCC to 17066 build 'arceb-elf32' and 'arceb-linux-uclibc' targets, for which big 17067 endian is the default. 17068 17069'-mlittle-endian' 17070'-EL' 17071 Compile code for little-endian targets. Use of these options is 17072 now deprecated. Little-endian code is supported by configuring GCC 17073 to build 'arc-elf32' and 'arc-linux-uclibc' targets, for which 17074 little endian is the default. 17075 17076'-mbarrel_shifter' 17077 Replaced by '-mbarrel-shifter'. 17078 17079'-mdpfp_compact' 17080 Replaced by '-mdpfp-compact'. 17081 17082'-mdpfp_fast' 17083 Replaced by '-mdpfp-fast'. 17084 17085'-mdsp_packa' 17086 Replaced by '-mdsp-packa'. 17087 17088'-mEA' 17089 Replaced by '-mea'. 17090 17091'-mmac_24' 17092 Replaced by '-mmac-24'. 17093 17094'-mmac_d16' 17095 Replaced by '-mmac-d16'. 17096 17097'-mspfp_compact' 17098 Replaced by '-mspfp-compact'. 17099 17100'-mspfp_fast' 17101 Replaced by '-mspfp-fast'. 17102 17103'-mtune=CPU' 17104 Values 'arc600', 'arc601', 'arc700' and 'arc700-xmac' for CPU are 17105 replaced by 'ARC600', 'ARC601', 'ARC700' and 'ARC700-xmac' 17106 respectively. 17107 17108'-multcost=NUM' 17109 Replaced by '-mmultcost'. 17110 17111 17112File: gcc.info, Node: ARM Options, Next: AVR Options, Prev: ARC Options, Up: Submodel Options 17113 171143.19.5 ARM Options 17115------------------ 17116 17117These '-m' options are defined for the ARM port: 17118 17119'-mabi=NAME' 17120 Generate code for the specified ABI. Permissible values are: 17121 'apcs-gnu', 'atpcs', 'aapcs', 'aapcs-linux' and 'iwmmxt'. 17122 17123'-mapcs-frame' 17124 Generate a stack frame that is compliant with the ARM Procedure 17125 Call Standard for all functions, even if this is not strictly 17126 necessary for correct execution of the code. Specifying 17127 '-fomit-frame-pointer' with this option causes the stack frames not 17128 to be generated for leaf functions. The default is 17129 '-mno-apcs-frame'. This option is deprecated. 17130 17131'-mapcs' 17132 This is a synonym for '-mapcs-frame' and is deprecated. 17133 17134'-mthumb-interwork' 17135 Generate code that supports calling between the ARM and Thumb 17136 instruction sets. Without this option, on pre-v5 architectures, 17137 the two instruction sets cannot be reliably used inside one 17138 program. The default is '-mno-thumb-interwork', since slightly 17139 larger code is generated when '-mthumb-interwork' is specified. In 17140 AAPCS configurations this option is meaningless. 17141 17142'-mno-sched-prolog' 17143 Prevent the reordering of instructions in the function prologue, or 17144 the merging of those instruction with the instructions in the 17145 function's body. This means that all functions start with a 17146 recognizable set of instructions (or in fact one of a choice from a 17147 small set of different function prologues), and this information 17148 can be used to locate the start of functions inside an executable 17149 piece of code. The default is '-msched-prolog'. 17150 17151'-mfloat-abi=NAME' 17152 Specifies which floating-point ABI to use. Permissible values are: 17153 'soft', 'softfp' and 'hard'. 17154 17155 Specifying 'soft' causes GCC to generate output containing library 17156 calls for floating-point operations. 'softfp' allows the 17157 generation of code using hardware floating-point instructions, but 17158 still uses the soft-float calling conventions. 'hard' allows 17159 generation of floating-point instructions and uses FPU-specific 17160 calling conventions. 17161 17162 The default depends on the specific target configuration. Note 17163 that the hard-float and soft-float ABIs are not link-compatible; 17164 you must compile your entire program with the same ABI, and link 17165 with a compatible set of libraries. 17166 17167'-mgeneral-regs-only' 17168 Generate code which uses only the general-purpose registers. This 17169 will prevent the compiler from using floating-point and Advanced 17170 SIMD registers but will not impose any restrictions on the 17171 assembler. 17172 17173'-mlittle-endian' 17174 Generate code for a processor running in little-endian mode. This 17175 is the default for all standard configurations. 17176 17177'-mbig-endian' 17178 Generate code for a processor running in big-endian mode; the 17179 default is to compile code for a little-endian processor. 17180 17181'-mbe8' 17182'-mbe32' 17183 When linking a big-endian image select between BE8 and BE32 17184 formats. The option has no effect for little-endian images and is 17185 ignored. The default is dependent on the selected target 17186 architecture. For ARMv6 and later architectures the default is 17187 BE8, for older architectures the default is BE32. BE32 format has 17188 been deprecated by ARM. 17189 17190'-march=NAME[+extension...]' 17191 This specifies the name of the target ARM architecture. GCC uses 17192 this name to determine what kind of instructions it can emit when 17193 generating assembly code. This option can be used in conjunction 17194 with or instead of the '-mcpu=' option. 17195 17196 Permissible names are: 'armv4t', 'armv5t', 'armv5te', 'armv6', 17197 'armv6j', 'armv6k', 'armv6kz', 'armv6t2', 'armv6z', 'armv6zk', 17198 'armv7', 'armv7-a', 'armv7ve', 'armv8-a', 'armv8.1-a', 'armv8.2-a', 17199 'armv8.3-a', 'armv8.4-a', 'armv8.5-a', 'armv8.6-a', 'armv7-r', 17200 'armv8-r', 'armv6-m', 'armv6s-m', 'armv7-m', 'armv7e-m', 17201 'armv8-m.base', 'armv8-m.main', 'armv8.1-m.main', 'iwmmxt' and 17202 'iwmmxt2'. 17203 17204 Additionally, the following architectures, which lack support for 17205 the Thumb execution state, are recognized but support is 17206 deprecated: 'armv4'. 17207 17208 Many of the architectures support extensions. These can be added 17209 by appending '+EXTENSION' to the architecture name. Extension 17210 options are processed in order and capabilities accumulate. An 17211 extension will also enable any necessary base extensions upon which 17212 it depends. For example, the '+crypto' extension will always 17213 enable the '+simd' extension. The exception to the additive 17214 construction is for extensions that are prefixed with '+no...': 17215 these extensions disable the specified option and any other 17216 extensions that may depend on the presence of that extension. 17217 17218 For example, '-march=armv7-a+simd+nofp+vfpv4' is equivalent to 17219 writing '-march=armv7-a+vfpv4' since the '+simd' option is entirely 17220 disabled by the '+nofp' option that follows it. 17221 17222 Most extension names are generically named, but have an effect that 17223 is dependent upon the architecture to which it is applied. For 17224 example, the '+simd' option can be applied to both 'armv7-a' and 17225 'armv8-a' architectures, but will enable the original ARMv7-A 17226 Advanced SIMD (Neon) extensions for 'armv7-a' and the ARMv8-A 17227 variant for 'armv8-a'. 17228 17229 The table below lists the supported extensions for each 17230 architecture. Architectures not mentioned do not support any 17231 extensions. 17232 17233 'armv5te' 17234 'armv6' 17235 'armv6j' 17236 'armv6k' 17237 'armv6kz' 17238 'armv6t2' 17239 'armv6z' 17240 'armv6zk' 17241 '+fp' 17242 The VFPv2 floating-point instructions. The extension 17243 '+vfpv2' can be used as an alias for this extension. 17244 17245 '+nofp' 17246 Disable the floating-point instructions. 17247 17248 'armv7' 17249 The common subset of the ARMv7-A, ARMv7-R and ARMv7-M 17250 architectures. 17251 '+fp' 17252 The VFPv3 floating-point instructions, with 16 17253 double-precision registers. The extension '+vfpv3-d16' 17254 can be used as an alias for this extension. Note that 17255 floating-point is not supported by the base ARMv7-M 17256 architecture, but is compatible with both the ARMv7-A and 17257 ARMv7-R architectures. 17258 17259 '+nofp' 17260 Disable the floating-point instructions. 17261 17262 'armv7-a' 17263 '+mp' 17264 The multiprocessing extension. 17265 17266 '+sec' 17267 The security extension. 17268 17269 '+fp' 17270 The VFPv3 floating-point instructions, with 16 17271 double-precision registers. The extension '+vfpv3-d16' 17272 can be used as an alias for this extension. 17273 17274 '+simd' 17275 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point 17276 instructions. The extensions '+neon' and '+neon-vfpv3' 17277 can be used as aliases for this extension. 17278 17279 '+vfpv3' 17280 The VFPv3 floating-point instructions, with 32 17281 double-precision registers. 17282 17283 '+vfpv3-d16-fp16' 17284 The VFPv3 floating-point instructions, with 16 17285 double-precision registers and the half-precision 17286 floating-point conversion operations. 17287 17288 '+vfpv3-fp16' 17289 The VFPv3 floating-point instructions, with 32 17290 double-precision registers and the half-precision 17291 floating-point conversion operations. 17292 17293 '+vfpv4-d16' 17294 The VFPv4 floating-point instructions, with 16 17295 double-precision registers. 17296 17297 '+vfpv4' 17298 The VFPv4 floating-point instructions, with 32 17299 double-precision registers. 17300 17301 '+neon-fp16' 17302 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point 17303 instructions, with the half-precision floating-point 17304 conversion operations. 17305 17306 '+neon-vfpv4' 17307 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point 17308 instructions. 17309 17310 '+nosimd' 17311 Disable the Advanced SIMD instructions (does not disable 17312 floating point). 17313 17314 '+nofp' 17315 Disable the floating-point and Advanced SIMD 17316 instructions. 17317 17318 'armv7ve' 17319 The extended version of the ARMv7-A architecture with support 17320 for virtualization. 17321 '+fp' 17322 The VFPv4 floating-point instructions, with 16 17323 double-precision registers. The extension '+vfpv4-d16' 17324 can be used as an alias for this extension. 17325 17326 '+simd' 17327 The Advanced SIMD (Neon) v2 and the VFPv4 floating-point 17328 instructions. The extension '+neon-vfpv4' can be used as 17329 an alias for this extension. 17330 17331 '+vfpv3-d16' 17332 The VFPv3 floating-point instructions, with 16 17333 double-precision registers. 17334 17335 '+vfpv3' 17336 The VFPv3 floating-point instructions, with 32 17337 double-precision registers. 17338 17339 '+vfpv3-d16-fp16' 17340 The VFPv3 floating-point instructions, with 16 17341 double-precision registers and the half-precision 17342 floating-point conversion operations. 17343 17344 '+vfpv3-fp16' 17345 The VFPv3 floating-point instructions, with 32 17346 double-precision registers and the half-precision 17347 floating-point conversion operations. 17348 17349 '+vfpv4-d16' 17350 The VFPv4 floating-point instructions, with 16 17351 double-precision registers. 17352 17353 '+vfpv4' 17354 The VFPv4 floating-point instructions, with 32 17355 double-precision registers. 17356 17357 '+neon' 17358 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point 17359 instructions. The extension '+neon-vfpv3' can be used as 17360 an alias for this extension. 17361 17362 '+neon-fp16' 17363 The Advanced SIMD (Neon) v1 and the VFPv3 floating-point 17364 instructions, with the half-precision floating-point 17365 conversion operations. 17366 17367 '+nosimd' 17368 Disable the Advanced SIMD instructions (does not disable 17369 floating point). 17370 17371 '+nofp' 17372 Disable the floating-point and Advanced SIMD 17373 instructions. 17374 17375 'armv8-a' 17376 '+crc' 17377 The Cyclic Redundancy Check (CRC) instructions. 17378 '+simd' 17379 The ARMv8-A Advanced SIMD and floating-point 17380 instructions. 17381 '+crypto' 17382 The cryptographic instructions. 17383 '+nocrypto' 17384 Disable the cryptographic instructions. 17385 '+nofp' 17386 Disable the floating-point, Advanced SIMD and 17387 cryptographic instructions. 17388 '+sb' 17389 Speculation Barrier Instruction. 17390 '+predres' 17391 Execution and Data Prediction Restriction Instructions. 17392 17393 'armv8.1-a' 17394 '+simd' 17395 The ARMv8.1-A Advanced SIMD and floating-point 17396 instructions. 17397 17398 '+crypto' 17399 The cryptographic instructions. This also enables the 17400 Advanced SIMD and floating-point instructions. 17401 17402 '+nocrypto' 17403 Disable the cryptographic instructions. 17404 17405 '+nofp' 17406 Disable the floating-point, Advanced SIMD and 17407 cryptographic instructions. 17408 17409 '+sb' 17410 Speculation Barrier Instruction. 17411 17412 '+predres' 17413 Execution and Data Prediction Restriction Instructions. 17414 17415 'armv8.2-a' 17416 'armv8.3-a' 17417 '+fp16' 17418 The half-precision floating-point data processing 17419 instructions. This also enables the Advanced SIMD and 17420 floating-point instructions. 17421 17422 '+fp16fml' 17423 The half-precision floating-point fmla extension. This 17424 also enables the half-precision floating-point extension 17425 and Advanced SIMD and floating-point instructions. 17426 17427 '+simd' 17428 The ARMv8.1-A Advanced SIMD and floating-point 17429 instructions. 17430 17431 '+crypto' 17432 The cryptographic instructions. This also enables the 17433 Advanced SIMD and floating-point instructions. 17434 17435 '+dotprod' 17436 Enable the Dot Product extension. This also enables 17437 Advanced SIMD instructions. 17438 17439 '+nocrypto' 17440 Disable the cryptographic extension. 17441 17442 '+nofp' 17443 Disable the floating-point, Advanced SIMD and 17444 cryptographic instructions. 17445 17446 '+sb' 17447 Speculation Barrier Instruction. 17448 17449 '+predres' 17450 Execution and Data Prediction Restriction Instructions. 17451 17452 '+i8mm' 17453 8-bit Integer Matrix Multiply instructions. This also 17454 enables Advanced SIMD and floating-point instructions. 17455 17456 '+bf16' 17457 Brain half-precision floating-point instructions. This 17458 also enables Advanced SIMD and floating-point 17459 instructions. 17460 17461 'armv8.4-a' 17462 '+fp16' 17463 The half-precision floating-point data processing 17464 instructions. This also enables the Advanced SIMD and 17465 floating-point instructions as well as the Dot Product 17466 extension and the half-precision floating-point fmla 17467 extension. 17468 17469 '+simd' 17470 The ARMv8.3-A Advanced SIMD and floating-point 17471 instructions as well as the Dot Product extension. 17472 17473 '+crypto' 17474 The cryptographic instructions. This also enables the 17475 Advanced SIMD and floating-point instructions as well as 17476 the Dot Product extension. 17477 17478 '+nocrypto' 17479 Disable the cryptographic extension. 17480 17481 '+nofp' 17482 Disable the floating-point, Advanced SIMD and 17483 cryptographic instructions. 17484 17485 '+sb' 17486 Speculation Barrier Instruction. 17487 17488 '+predres' 17489 Execution and Data Prediction Restriction Instructions. 17490 17491 '+i8mm' 17492 8-bit Integer Matrix Multiply instructions. This also 17493 enables Advanced SIMD and floating-point instructions. 17494 17495 '+bf16' 17496 Brain half-precision floating-point instructions. This 17497 also enables Advanced SIMD and floating-point 17498 instructions. 17499 17500 'armv8.5-a' 17501 '+fp16' 17502 The half-precision floating-point data processing 17503 instructions. This also enables the Advanced SIMD and 17504 floating-point instructions as well as the Dot Product 17505 extension and the half-precision floating-point fmla 17506 extension. 17507 17508 '+simd' 17509 The ARMv8.3-A Advanced SIMD and floating-point 17510 instructions as well as the Dot Product extension. 17511 17512 '+crypto' 17513 The cryptographic instructions. This also enables the 17514 Advanced SIMD and floating-point instructions as well as 17515 the Dot Product extension. 17516 17517 '+nocrypto' 17518 Disable the cryptographic extension. 17519 17520 '+nofp' 17521 Disable the floating-point, Advanced SIMD and 17522 cryptographic instructions. 17523 17524 '+i8mm' 17525 8-bit Integer Matrix Multiply instructions. This also 17526 enables Advanced SIMD and floating-point instructions. 17527 17528 '+bf16' 17529 Brain half-precision floating-point instructions. This 17530 also enables Advanced SIMD and floating-point 17531 instructions. 17532 17533 'armv8.6-a' 17534 '+fp16' 17535 The half-precision floating-point data processing 17536 instructions. This also enables the Advanced SIMD and 17537 floating-point instructions as well as the Dot Product 17538 extension and the half-precision floating-point fmla 17539 extension. 17540 17541 '+simd' 17542 The ARMv8.3-A Advanced SIMD and floating-point 17543 instructions as well as the Dot Product extension. 17544 17545 '+crypto' 17546 The cryptographic instructions. This also enables the 17547 Advanced SIMD and floating-point instructions as well as 17548 the Dot Product extension. 17549 17550 '+nocrypto' 17551 Disable the cryptographic extension. 17552 17553 '+nofp' 17554 Disable the floating-point, Advanced SIMD and 17555 cryptographic instructions. 17556 17557 '+i8mm' 17558 8-bit Integer Matrix Multiply instructions. This also 17559 enables Advanced SIMD and floating-point instructions. 17560 17561 '+bf16' 17562 Brain half-precision floating-point instructions. This 17563 also enables Advanced SIMD and floating-point 17564 instructions. 17565 17566 'armv7-r' 17567 '+fp.sp' 17568 The single-precision VFPv3 floating-point instructions. 17569 The extension '+vfpv3xd' can be used as an alias for this 17570 extension. 17571 17572 '+fp' 17573 The VFPv3 floating-point instructions with 16 17574 double-precision registers. The extension +vfpv3-d16 can 17575 be used as an alias for this extension. 17576 17577 '+vfpv3xd-d16-fp16' 17578 The single-precision VFPv3 floating-point instructions 17579 with 16 double-precision registers and the half-precision 17580 floating-point conversion operations. 17581 17582 '+vfpv3-d16-fp16' 17583 The VFPv3 floating-point instructions with 16 17584 double-precision registers and the half-precision 17585 floating-point conversion operations. 17586 17587 '+nofp' 17588 Disable the floating-point extension. 17589 17590 '+idiv' 17591 The ARM-state integer division instructions. 17592 17593 '+noidiv' 17594 Disable the ARM-state integer division extension. 17595 17596 'armv7e-m' 17597 '+fp' 17598 The single-precision VFPv4 floating-point instructions. 17599 17600 '+fpv5' 17601 The single-precision FPv5 floating-point instructions. 17602 17603 '+fp.dp' 17604 The single- and double-precision FPv5 floating-point 17605 instructions. 17606 17607 '+nofp' 17608 Disable the floating-point extensions. 17609 17610 'armv8.1-m.main' 17611 17612 '+dsp' 17613 The DSP instructions. 17614 17615 '+mve' 17616 The M-Profile Vector Extension (MVE) integer 17617 instructions. 17618 17619 '+mve.fp' 17620 The M-Profile Vector Extension (MVE) integer and single 17621 precision floating-point instructions. 17622 17623 '+fp' 17624 The single-precision floating-point instructions. 17625 17626 '+fp.dp' 17627 The single- and double-precision floating-point 17628 instructions. 17629 17630 '+nofp' 17631 Disable the floating-point extension. 17632 17633 '+cdecp0, +cdecp1, ... , +cdecp7' 17634 Enable the Custom Datapath Extension (CDE) on selected 17635 coprocessors according to the numbers given in the 17636 options in the range 0 to 7. 17637 17638 'armv8-m.main' 17639 '+dsp' 17640 The DSP instructions. 17641 17642 '+nodsp' 17643 Disable the DSP extension. 17644 17645 '+fp' 17646 The single-precision floating-point instructions. 17647 17648 '+fp.dp' 17649 The single- and double-precision floating-point 17650 instructions. 17651 17652 '+nofp' 17653 Disable the floating-point extension. 17654 17655 '+cdecp0, +cdecp1, ... , +cdecp7' 17656 Enable the Custom Datapath Extension (CDE) on selected 17657 coprocessors according to the numbers given in the 17658 options in the range 0 to 7. 17659 17660 'armv8-r' 17661 '+crc' 17662 The Cyclic Redundancy Check (CRC) instructions. 17663 '+fp.sp' 17664 The single-precision FPv5 floating-point instructions. 17665 '+simd' 17666 The ARMv8-A Advanced SIMD and floating-point 17667 instructions. 17668 '+crypto' 17669 The cryptographic instructions. 17670 '+nocrypto' 17671 Disable the cryptographic instructions. 17672 '+nofp' 17673 Disable the floating-point, Advanced SIMD and 17674 cryptographic instructions. 17675 17676 '-march=native' causes the compiler to auto-detect the architecture 17677 of the build computer. At present, this feature is only supported 17678 on GNU/Linux, and not all architectures are recognized. If the 17679 auto-detect is unsuccessful the option has no effect. 17680 17681'-mtune=NAME' 17682 This option specifies the name of the target ARM processor for 17683 which GCC should tune the performance of the code. For some ARM 17684 implementations better performance can be obtained by using this 17685 option. Permissible names are: 'arm7tdmi', 'arm7tdmi-s', 17686 'arm710t', 'arm720t', 'arm740t', 'strongarm', 'strongarm110', 17687 'strongarm1100', 0'strongarm1110', 'arm8', 'arm810', 'arm9', 17688 'arm9e', 'arm920', 'arm920t', 'arm922t', 'arm946e-s', 'arm966e-s', 17689 'arm968e-s', 'arm926ej-s', 'arm940t', 'arm9tdmi', 'arm10tdmi', 17690 'arm1020t', 'arm1026ej-s', 'arm10e', 'arm1020e', 'arm1022e', 17691 'arm1136j-s', 'arm1136jf-s', 'mpcore', 'mpcorenovfp', 17692 'arm1156t2-s', 'arm1156t2f-s', 'arm1176jz-s', 'arm1176jzf-s', 17693 'generic-armv7-a', 'cortex-a5', 'cortex-a7', 'cortex-a8', 17694 'cortex-a9', 'cortex-a12', 'cortex-a15', 'cortex-a17', 17695 'cortex-a32', 'cortex-a35', 'cortex-a53', 'cortex-a55', 17696 'cortex-a57', 'cortex-a72', 'cortex-a73', 'cortex-a75', 17697 'cortex-a76', 'cortex-a76ae', 'cortex-a77', 'ares', 'cortex-r4', 17698 'cortex-r4f', 'cortex-r5', 'cortex-r7', 'cortex-r8', 'cortex-r52', 17699 'cortex-m0', 'cortex-m0plus', 'cortex-m1', 'cortex-m3', 17700 'cortex-m4', 'cortex-m7', 'cortex-m23', 'cortex-m33', 17701 'cortex-m35p', 'cortex-m55', 'cortex-m1.small-multiply', 17702 'cortex-m0.small-multiply', 'cortex-m0plus.small-multiply', 17703 'exynos-m1', 'marvell-pj4', 'neoverse-n1', 'neoverse-n2', 17704 'neoverse-v1', 'xscale', 'iwmmxt', 'iwmmxt2', 'ep9312', 'fa526', 17705 'fa626', 'fa606te', 'fa626te', 'fmp626', 'fa726te', 'xgene1'. 17706 17707 Additionally, this option can specify that GCC should tune the 17708 performance of the code for a big.LITTLE system. Permissible names 17709 are: 'cortex-a15.cortex-a7', 'cortex-a17.cortex-a7', 17710 'cortex-a57.cortex-a53', 'cortex-a72.cortex-a53', 17711 'cortex-a72.cortex-a35', 'cortex-a73.cortex-a53', 17712 'cortex-a75.cortex-a55', 'cortex-a76.cortex-a55'. 17713 17714 '-mtune=generic-ARCH' specifies that GCC should tune the 17715 performance for a blend of processors within architecture ARCH. 17716 The aim is to generate code that run well on the current most 17717 popular processors, balancing between optimizations that benefit 17718 some CPUs in the range, and avoiding performance pitfalls of other 17719 CPUs. The effects of this option may change in future GCC versions 17720 as CPU models come and go. 17721 17722 '-mtune' permits the same extension options as '-mcpu', but the 17723 extension options do not affect the tuning of the generated code. 17724 17725 '-mtune=native' causes the compiler to auto-detect the CPU of the 17726 build computer. At present, this feature is only supported on 17727 GNU/Linux, and not all architectures are recognized. If the 17728 auto-detect is unsuccessful the option has no effect. 17729 17730'-mcpu=NAME[+extension...]' 17731 This specifies the name of the target ARM processor. GCC uses this 17732 name to derive the name of the target ARM architecture (as if 17733 specified by '-march') and the ARM processor type for which to tune 17734 for performance (as if specified by '-mtune'). Where this option 17735 is used in conjunction with '-march' or '-mtune', those options 17736 take precedence over the appropriate part of this option. 17737 17738 Many of the supported CPUs implement optional architectural 17739 extensions. Where this is so the architectural extensions are 17740 normally enabled by default. If implementations that lack the 17741 extension exist, then the extension syntax can be used to disable 17742 those extensions that have been omitted. For floating-point and 17743 Advanced SIMD (Neon) instructions, the settings of the options 17744 '-mfloat-abi' and '-mfpu' must also be considered: floating-point 17745 and Advanced SIMD instructions will only be used if '-mfloat-abi' 17746 is not set to 'soft'; and any setting of '-mfpu' other than 'auto' 17747 will override the available floating-point and SIMD extension 17748 instructions. 17749 17750 For example, 'cortex-a9' can be found in three major 17751 configurations: integer only, with just a floating-point unit or 17752 with floating-point and Advanced SIMD. The default is to enable all 17753 the instructions, but the extensions '+nosimd' and '+nofp' can be 17754 used to disable just the SIMD or both the SIMD and floating-point 17755 instructions respectively. 17756 17757 Permissible names for this option are the same as those for 17758 '-mtune'. 17759 17760 The following extension options are common to the listed CPUs: 17761 17762 '+nodsp' 17763 Disable the DSP instructions on 'cortex-m33', 'cortex-m35p'. 17764 17765 '+nofp' 17766 Disables the floating-point instructions on 'arm9e', 17767 'arm946e-s', 'arm966e-s', 'arm968e-s', 'arm10e', 'arm1020e', 17768 'arm1022e', 'arm926ej-s', 'arm1026ej-s', 'cortex-r5', 17769 'cortex-r7', 'cortex-r8', 'cortex-m4', 'cortex-m7', 17770 'cortex-m33' and 'cortex-m35p'. Disables the floating-point 17771 and SIMD instructions on 'generic-armv7-a', 'cortex-a5', 17772 'cortex-a7', 'cortex-a8', 'cortex-a9', 'cortex-a12', 17773 'cortex-a15', 'cortex-a17', 'cortex-a15.cortex-a7', 17774 'cortex-a17.cortex-a7', 'cortex-a32', 'cortex-a35', 17775 'cortex-a53' and 'cortex-a55'. 17776 17777 '+nofp.dp' 17778 Disables the double-precision component of the floating-point 17779 instructions on 'cortex-r5', 'cortex-r7', 'cortex-r8', 17780 'cortex-r52' and 'cortex-m7'. 17781 17782 '+nosimd' 17783 Disables the SIMD (but not floating-point) instructions on 17784 'generic-armv7-a', 'cortex-a5', 'cortex-a7' and 'cortex-a9'. 17785 17786 '+crypto' 17787 Enables the cryptographic instructions on 'cortex-a32', 17788 'cortex-a35', 'cortex-a53', 'cortex-a55', 'cortex-a57', 17789 'cortex-a72', 'cortex-a73', 'cortex-a75', 'exynos-m1', 17790 'xgene1', 'cortex-a57.cortex-a53', 'cortex-a72.cortex-a53', 17791 'cortex-a73.cortex-a35', 'cortex-a73.cortex-a53' and 17792 'cortex-a75.cortex-a55'. 17793 17794 Additionally the 'generic-armv7-a' pseudo target defaults to VFPv3 17795 with 16 double-precision registers. It supports the following 17796 extension options: 'mp', 'sec', 'vfpv3-d16', 'vfpv3', 17797 'vfpv3-d16-fp16', 'vfpv3-fp16', 'vfpv4-d16', 'vfpv4', 'neon', 17798 'neon-vfpv3', 'neon-fp16', 'neon-vfpv4'. The meanings are the same 17799 as for the extensions to '-march=armv7-a'. 17800 17801 '-mcpu=generic-ARCH' is also permissible, and is equivalent to 17802 '-march=ARCH -mtune=generic-ARCH'. See '-mtune' for more 17803 information. 17804 17805 '-mcpu=native' causes the compiler to auto-detect the CPU of the 17806 build computer. At present, this feature is only supported on 17807 GNU/Linux, and not all architectures are recognized. If the 17808 auto-detect is unsuccessful the option has no effect. 17809 17810'-mfpu=NAME' 17811 This specifies what floating-point hardware (or hardware emulation) 17812 is available on the target. Permissible names are: 'auto', 17813 'vfpv2', 'vfpv3', 'vfpv3-fp16', 'vfpv3-d16', 'vfpv3-d16-fp16', 17814 'vfpv3xd', 'vfpv3xd-fp16', 'neon-vfpv3', 'neon-fp16', 'vfpv4', 17815 'vfpv4-d16', 'fpv4-sp-d16', 'neon-vfpv4', 'fpv5-d16', 17816 'fpv5-sp-d16', 'fp-armv8', 'neon-fp-armv8' and 17817 'crypto-neon-fp-armv8'. Note that 'neon' is an alias for 17818 'neon-vfpv3' and 'vfp' is an alias for 'vfpv2'. 17819 17820 The setting 'auto' is the default and is special. It causes the 17821 compiler to select the floating-point and Advanced SIMD 17822 instructions based on the settings of '-mcpu' and '-march'. 17823 17824 If the selected floating-point hardware includes the NEON extension 17825 (e.g. '-mfpu=neon'), note that floating-point operations are not 17826 generated by GCC's auto-vectorization pass unless 17827 '-funsafe-math-optimizations' is also specified. This is because 17828 NEON hardware does not fully implement the IEEE 754 standard for 17829 floating-point arithmetic (in particular denormal values are 17830 treated as zero), so the use of NEON instructions may lead to a 17831 loss of precision. 17832 17833 You can also set the fpu name at function level by using the 17834 'target("fpu=")' function attributes (*note ARM Function 17835 Attributes::) or pragmas (*note Function Specific Option 17836 Pragmas::). 17837 17838'-mfp16-format=NAME' 17839 Specify the format of the '__fp16' half-precision floating-point 17840 type. Permissible names are 'none', 'ieee', and 'alternative'; the 17841 default is 'none', in which case the '__fp16' type is not defined. 17842 *Note Half-Precision::, for more information. 17843 17844'-mstructure-size-boundary=N' 17845 The sizes of all structures and unions are rounded up to a multiple 17846 of the number of bits set by this option. Permissible values are 17847 8, 32 and 64. The default value varies for different toolchains. 17848 For the COFF targeted toolchain the default value is 8. A value of 17849 64 is only allowed if the underlying ABI supports it. 17850 17851 Specifying a larger number can produce faster, more efficient code, 17852 but can also increase the size of the program. Different values 17853 are potentially incompatible. Code compiled with one value cannot 17854 necessarily expect to work with code or libraries compiled with 17855 another value, if they exchange information using structures or 17856 unions. 17857 17858 This option is deprecated. 17859 17860'-mabort-on-noreturn' 17861 Generate a call to the function 'abort' at the end of a 'noreturn' 17862 function. It is executed if the function tries to return. 17863 17864'-mlong-calls' 17865'-mno-long-calls' 17866 Tells the compiler to perform function calls by first loading the 17867 address of the function into a register and then performing a 17868 subroutine call on this register. This switch is needed if the 17869 target function lies outside of the 64-megabyte addressing range of 17870 the offset-based version of subroutine call instruction. 17871 17872 Even if this switch is enabled, not all function calls are turned 17873 into long calls. The heuristic is that static functions, functions 17874 that have the 'short_call' attribute, functions that are inside the 17875 scope of a '#pragma no_long_calls' directive, and functions whose 17876 definitions have already been compiled within the current 17877 compilation unit are not turned into long calls. The exceptions to 17878 this rule are that weak function definitions, functions with the 17879 'long_call' attribute or the 'section' attribute, and functions 17880 that are within the scope of a '#pragma long_calls' directive are 17881 always turned into long calls. 17882 17883 This feature is not enabled by default. Specifying 17884 '-mno-long-calls' restores the default behavior, as does placing 17885 the function calls within the scope of a '#pragma long_calls_off' 17886 directive. Note these switches have no effect on how the compiler 17887 generates code to handle function calls via function pointers. 17888 17889'-msingle-pic-base' 17890 Treat the register used for PIC addressing as read-only, rather 17891 than loading it in the prologue for each function. The runtime 17892 system is responsible for initializing this register with an 17893 appropriate value before execution begins. 17894 17895'-mpic-register=REG' 17896 Specify the register to be used for PIC addressing. For standard 17897 PIC base case, the default is any suitable register determined by 17898 compiler. For single PIC base case, the default is 'R9' if target 17899 is EABI based or stack-checking is enabled, otherwise the default 17900 is 'R10'. 17901 17902'-mpic-data-is-text-relative' 17903 Assume that the displacement between the text and data segments is 17904 fixed at static link time. This permits using PC-relative 17905 addressing operations to access data known to be in the data 17906 segment. For non-VxWorks RTP targets, this option is enabled by 17907 default. When disabled on such targets, it will enable 17908 '-msingle-pic-base' by default. 17909 17910'-mpoke-function-name' 17911 Write the name of each function into the text section, directly 17912 preceding the function prologue. The generated code is similar to 17913 this: 17914 17915 t0 17916 .ascii "arm_poke_function_name", 0 17917 .align 17918 t1 17919 .word 0xff000000 + (t1 - t0) 17920 arm_poke_function_name 17921 mov ip, sp 17922 stmfd sp!, {fp, ip, lr, pc} 17923 sub fp, ip, #4 17924 17925 When performing a stack backtrace, code can inspect the value of 17926 'pc' stored at 'fp + 0'. If the trace function then looks at 17927 location 'pc - 12' and the top 8 bits are set, then we know that 17928 there is a function name embedded immediately preceding this 17929 location and has length '((pc[-3]) & 0xff000000)'. 17930 17931'-mthumb' 17932'-marm' 17933 17934 Select between generating code that executes in ARM and Thumb 17935 states. The default for most configurations is to generate code 17936 that executes in ARM state, but the default can be changed by 17937 configuring GCC with the '--with-mode='STATE configure option. 17938 17939 You can also override the ARM and Thumb mode for each function by 17940 using the 'target("thumb")' and 'target("arm")' function attributes 17941 (*note ARM Function Attributes::) or pragmas (*note Function 17942 Specific Option Pragmas::). 17943 17944'-mflip-thumb' 17945 Switch ARM/Thumb modes on alternating functions. This option is 17946 provided for regression testing of mixed Thumb/ARM code generation, 17947 and is not intended for ordinary use in compiling code. 17948 17949'-mtpcs-frame' 17950 Generate a stack frame that is compliant with the Thumb Procedure 17951 Call Standard for all non-leaf functions. (A leaf function is one 17952 that does not call any other functions.) The default is 17953 '-mno-tpcs-frame'. 17954 17955'-mtpcs-leaf-frame' 17956 Generate a stack frame that is compliant with the Thumb Procedure 17957 Call Standard for all leaf functions. (A leaf function is one that 17958 does not call any other functions.) The default is 17959 '-mno-apcs-leaf-frame'. 17960 17961'-mcallee-super-interworking' 17962 Gives all externally visible functions in the file being compiled 17963 an ARM instruction set header which switches to Thumb mode before 17964 executing the rest of the function. This allows these functions to 17965 be called from non-interworking code. This option is not valid in 17966 AAPCS configurations because interworking is enabled by default. 17967 17968'-mcaller-super-interworking' 17969 Allows calls via function pointers (including virtual functions) to 17970 execute correctly regardless of whether the target code has been 17971 compiled for interworking or not. There is a small overhead in the 17972 cost of executing a function pointer if this option is enabled. 17973 This option is not valid in AAPCS configurations because 17974 interworking is enabled by default. 17975 17976'-mtp=NAME' 17977 Specify the access model for the thread local storage pointer. The 17978 valid models are 'soft', which generates calls to 17979 '__aeabi_read_tp', 'cp15', which fetches the thread pointer from 17980 'cp15' directly (supported in the arm6k architecture), and 'auto', 17981 which uses the best available method for the selected processor. 17982 The default setting is 'auto'. 17983 17984'-mtls-dialect=DIALECT' 17985 Specify the dialect to use for accessing thread local storage. Two 17986 DIALECTs are supported--'gnu' and 'gnu2'. The 'gnu' dialect 17987 selects the original GNU scheme for supporting local and global 17988 dynamic TLS models. The 'gnu2' dialect selects the GNU descriptor 17989 scheme, which provides better performance for shared libraries. 17990 The GNU descriptor scheme is compatible with the original scheme, 17991 but does require new assembler, linker and library support. 17992 Initial and local exec TLS models are unaffected by this option and 17993 always use the original scheme. 17994 17995'-mword-relocations' 17996 Only generate absolute relocations on word-sized values (i.e. 17997 R_ARM_ABS32). This is enabled by default on targets (uClinux, 17998 SymbianOS) where the runtime loader imposes this restriction, and 17999 when '-fpic' or '-fPIC' is specified. This option conflicts with 18000 '-mslow-flash-data'. 18001 18002'-mfix-cortex-m3-ldrd' 18003 Some Cortex-M3 cores can cause data corruption when 'ldrd' 18004 instructions with overlapping destination and base registers are 18005 used. This option avoids generating these instructions. This 18006 option is enabled by default when '-mcpu=cortex-m3' is specified. 18007 18008'-munaligned-access' 18009'-mno-unaligned-access' 18010 Enables (or disables) reading and writing of 16- and 32- bit values 18011 from addresses that are not 16- or 32- bit aligned. By default 18012 unaligned access is disabled for all pre-ARMv6, all ARMv6-M and for 18013 ARMv8-M Baseline architectures, and enabled for all other 18014 architectures. If unaligned access is not enabled then words in 18015 packed data structures are accessed a byte at a time. 18016 18017 The ARM attribute 'Tag_CPU_unaligned_access' is set in the 18018 generated object file to either true or false, depending upon the 18019 setting of this option. If unaligned access is enabled then the 18020 preprocessor symbol '__ARM_FEATURE_UNALIGNED' is also defined. 18021 18022'-mneon-for-64bits' 18023 This option is deprecated and has no effect. 18024 18025'-mslow-flash-data' 18026 Assume loading data from flash is slower than fetching instruction. 18027 Therefore literal load is minimized for better performance. This 18028 option is only supported when compiling for ARMv7 M-profile and off 18029 by default. It conflicts with '-mword-relocations'. 18030 18031'-masm-syntax-unified' 18032 Assume inline assembler is using unified asm syntax. The default 18033 is currently off which implies divided syntax. This option has no 18034 impact on Thumb2. However, this may change in future releases of 18035 GCC. Divided syntax should be considered deprecated. 18036 18037'-mrestrict-it' 18038 Restricts generation of IT blocks to conform to the rules of 18039 ARMv8-A. IT blocks can only contain a single 16-bit instruction 18040 from a select set of instructions. This option is on by default 18041 for ARMv8-A Thumb mode. 18042 18043'-mprint-tune-info' 18044 Print CPU tuning information as comment in assembler file. This is 18045 an option used only for regression testing of the compiler and not 18046 intended for ordinary use in compiling code. This option is 18047 disabled by default. 18048 18049'-mverbose-cost-dump' 18050 Enable verbose cost model dumping in the debug dump files. This 18051 option is provided for use in debugging the compiler. 18052 18053'-mpure-code' 18054 Do not allow constant data to be placed in code sections. 18055 Additionally, when compiling for ELF object format give all text 18056 sections the ELF processor-specific section attribute 18057 'SHF_ARM_PURECODE'. This option is only available when generating 18058 non-pic code for M-profile targets. 18059 18060'-mcmse' 18061 Generate secure code as per the "ARMv8-M Security Extensions: 18062 Requirements on Development Tools Engineering Specification", which 18063 can be found on 18064 <https://developer.arm.com/documentation/ecm0359818/latest/>. 18065 18066'-mfix-cmse-cve-2021-35465' 18067 Mitigate against a potential security issue with the 'VLLDM' 18068 instruction in some M-profile devices when using CMSE 18069 (CVE-2021-365465). This option is enabled by default when the 18070 option '-mcpu=' is used with 'cortex-m33', 'cortex-m35p' or 18071 'cortex-m55'. The option '-mno-fix-cmse-cve-2021-35465' can be 18072 used to disable the mitigation. 18073 18074'-mfdpic' 18075'-mno-fdpic' 18076 Select the FDPIC ABI, which uses 64-bit function descriptors to 18077 represent pointers to functions. When the compiler is configured 18078 for 'arm-*-uclinuxfdpiceabi' targets, this option is on by default 18079 and implies '-fPIE' if none of the PIC/PIE-related options is 18080 provided. On other targets, it only enables the FDPIC-specific 18081 code generation features, and the user should explicitly provide 18082 the PIC/PIE-related options as needed. 18083 18084 Note that static linking is not supported because it would still 18085 involve the dynamic linker when the program self-relocates. If 18086 such behavior is acceptable, use -static and -Wl,-dynamic-linker 18087 options. 18088 18089 The opposite '-mno-fdpic' option is useful (and required) to build 18090 the Linux kernel using the same ('arm-*-uclinuxfdpiceabi') 18091 toolchain as the one used to build the userland programs. 18092 18093 18094File: gcc.info, Node: AVR Options, Next: Blackfin Options, Prev: ARM Options, Up: Submodel Options 18095 180963.19.6 AVR Options 18097------------------ 18098 18099These options are defined for AVR implementations: 18100 18101'-mmcu=MCU' 18102 Specify Atmel AVR instruction set architectures (ISA) or MCU type. 18103 18104 The default for this option is 'avr2'. 18105 18106 GCC supports the following AVR devices and ISAs: 18107 18108 'avr2' 18109 "Classic" devices with up to 8 KiB of program memory. 18110 MCU = 'attiny22', 'attiny26', 'at90s2313', 'at90s2323', 18111 'at90s2333', 'at90s2343', 'at90s4414', 'at90s4433', 18112 'at90s4434', 'at90c8534', 'at90s8515', 'at90s8535'. 18113 18114 'avr25' 18115 "Classic" devices with up to 8 KiB of program memory and with 18116 the 'MOVW' instruction. 18117 MCU = 'attiny13', 'attiny13a', 'attiny24', 'attiny24a', 18118 'attiny25', 'attiny261', 'attiny261a', 'attiny2313', 18119 'attiny2313a', 'attiny43u', 'attiny44', 'attiny44a', 18120 'attiny45', 'attiny48', 'attiny441', 'attiny461', 18121 'attiny461a', 'attiny4313', 'attiny84', 'attiny84a', 18122 'attiny85', 'attiny87', 'attiny88', 'attiny828', 'attiny841', 18123 'attiny861', 'attiny861a', 'ata5272', 'ata6616c', 'at86rf401'. 18124 18125 'avr3' 18126 "Classic" devices with 16 KiB up to 64 KiB of program memory. 18127 MCU = 'at76c711', 'at43usb355'. 18128 18129 'avr31' 18130 "Classic" devices with 128 KiB of program memory. 18131 MCU = 'atmega103', 'at43usb320'. 18132 18133 'avr35' 18134 "Classic" devices with 16 KiB up to 64 KiB of program memory 18135 and with the 'MOVW' instruction. 18136 MCU = 'attiny167', 'attiny1634', 'atmega8u2', 'atmega16u2', 18137 'atmega32u2', 'ata5505', 'ata6617c', 'ata664251', 'at90usb82', 18138 'at90usb162'. 18139 18140 'avr4' 18141 "Enhanced" devices with up to 8 KiB of program memory. 18142 MCU = 'atmega48', 'atmega48a', 'atmega48p', 'atmega48pa', 18143 'atmega48pb', 'atmega8', 'atmega8a', 'atmega8hva', 'atmega88', 18144 'atmega88a', 'atmega88p', 'atmega88pa', 'atmega88pb', 18145 'atmega8515', 'atmega8535', 'ata6285', 'ata6286', 'ata6289', 18146 'ata6612c', 'at90pwm1', 'at90pwm2', 'at90pwm2b', 'at90pwm3', 18147 'at90pwm3b', 'at90pwm81'. 18148 18149 'avr5' 18150 "Enhanced" devices with 16 KiB up to 64 KiB of program memory. 18151 18152 MCU = 'atmega16', 'atmega16a', 'atmega16hva', 'atmega16hva2', 18153 'atmega16hvb', 'atmega16hvbrevb', 'atmega16m1', 'atmega16u4', 18154 'atmega161', 'atmega162', 'atmega163', 'atmega164a', 18155 'atmega164p', 'atmega164pa', 'atmega165', 'atmega165a', 18156 'atmega165p', 'atmega165pa', 'atmega168', 'atmega168a', 18157 'atmega168p', 'atmega168pa', 'atmega168pb', 'atmega169', 18158 'atmega169a', 'atmega169p', 'atmega169pa', 'atmega32', 18159 'atmega32a', 'atmega32c1', 'atmega32hvb', 'atmega32hvbrevb', 18160 'atmega32m1', 'atmega32u4', 'atmega32u6', 'atmega323', 18161 'atmega324a', 'atmega324p', 'atmega324pa', 'atmega325', 18162 'atmega325a', 'atmega325p', 'atmega325pa', 'atmega328', 18163 'atmega328p', 'atmega328pb', 'atmega329', 'atmega329a', 18164 'atmega329p', 'atmega329pa', 'atmega3250', 'atmega3250a', 18165 'atmega3250p', 'atmega3250pa', 'atmega3290', 'atmega3290a', 18166 'atmega3290p', 'atmega3290pa', 'atmega406', 'atmega64', 18167 'atmega64a', 'atmega64c1', 'atmega64hve', 'atmega64hve2', 18168 'atmega64m1', 'atmega64rfr2', 'atmega640', 'atmega644', 18169 'atmega644a', 'atmega644p', 'atmega644pa', 'atmega644rfr2', 18170 'atmega645', 'atmega645a', 'atmega645p', 'atmega649', 18171 'atmega649a', 'atmega649p', 'atmega6450', 'atmega6450a', 18172 'atmega6450p', 'atmega6490', 'atmega6490a', 'atmega6490p', 18173 'ata5795', 'ata5790', 'ata5790n', 'ata5791', 'ata6613c', 18174 'ata6614q', 'ata5782', 'ata5831', 'ata8210', 'ata8510', 18175 'ata5702m322', 'at90pwm161', 'at90pwm216', 'at90pwm316', 18176 'at90can32', 'at90can64', 'at90scr100', 'at90usb646', 18177 'at90usb647', 'at94k', 'm3000'. 18178 18179 'avr51' 18180 "Enhanced" devices with 128 KiB of program memory. 18181 MCU = 'atmega128', 'atmega128a', 'atmega128rfa1', 18182 'atmega128rfr2', 'atmega1280', 'atmega1281', 'atmega1284', 18183 'atmega1284p', 'atmega1284rfr2', 'at90can128', 'at90usb1286', 18184 'at90usb1287'. 18185 18186 'avr6' 18187 "Enhanced" devices with 3-byte PC, i.e. with more than 128 KiB 18188 of program memory. 18189 MCU = 'atmega256rfr2', 'atmega2560', 'atmega2561', 18190 'atmega2564rfr2'. 18191 18192 'avrxmega2' 18193 "XMEGA" devices with more than 8 KiB and up to 64 KiB of 18194 program memory. 18195 MCU = 'atxmega8e5', 'atxmega16a4', 'atxmega16a4u', 18196 'atxmega16c4', 'atxmega16d4', 'atxmega16e5', 'atxmega32a4', 18197 'atxmega32a4u', 'atxmega32c3', 'atxmega32c4', 'atxmega32d3', 18198 'atxmega32d4', 'atxmega32e5'. 18199 18200 'avrxmega3' 18201 "XMEGA" devices with up to 64 KiB of combined program memory 18202 and RAM, and with program memory visible in the RAM address 18203 space. 18204 MCU = 'attiny202', 'attiny204', 'attiny212', 'attiny214', 18205 'attiny402', 'attiny404', 'attiny406', 'attiny412', 18206 'attiny414', 'attiny416', 'attiny417', 'attiny804', 18207 'attiny806', 'attiny807', 'attiny814', 'attiny816', 18208 'attiny817', 'attiny1604', 'attiny1606', 'attiny1607', 18209 'attiny1614', 'attiny1616', 'attiny1617', 'attiny3214', 18210 'attiny3216', 'attiny3217', 'atmega808', 'atmega809', 18211 'atmega1608', 'atmega1609', 'atmega3208', 'atmega3209', 18212 'atmega4808', 'atmega4809'. 18213 18214 'avrxmega4' 18215 "XMEGA" devices with more than 64 KiB and up to 128 KiB of 18216 program memory. 18217 MCU = 'atxmega64a3', 'atxmega64a3u', 'atxmega64a4u', 18218 'atxmega64b1', 'atxmega64b3', 'atxmega64c3', 'atxmega64d3', 18219 'atxmega64d4'. 18220 18221 'avrxmega5' 18222 "XMEGA" devices with more than 64 KiB and up to 128 KiB of 18223 program memory and more than 64 KiB of RAM. 18224 MCU = 'atxmega64a1', 'atxmega64a1u'. 18225 18226 'avrxmega6' 18227 "XMEGA" devices with more than 128 KiB of program memory. 18228 MCU = 'atxmega128a3', 'atxmega128a3u', 'atxmega128b1', 18229 'atxmega128b3', 'atxmega128c3', 'atxmega128d3', 18230 'atxmega128d4', 'atxmega192a3', 'atxmega192a3u', 18231 'atxmega192c3', 'atxmega192d3', 'atxmega256a3', 18232 'atxmega256a3b', 'atxmega256a3bu', 'atxmega256a3u', 18233 'atxmega256c3', 'atxmega256d3', 'atxmega384c3', 18234 'atxmega384d3'. 18235 18236 'avrxmega7' 18237 "XMEGA" devices with more than 128 KiB of program memory and 18238 more than 64 KiB of RAM. 18239 MCU = 'atxmega128a1', 'atxmega128a1u', 'atxmega128a4u'. 18240 18241 'avrtiny' 18242 "TINY" Tiny core devices with 512 B up to 4 KiB of program 18243 memory. 18244 MCU = 'attiny4', 'attiny5', 'attiny9', 'attiny10', 'attiny20', 18245 'attiny40'. 18246 18247 'avr1' 18248 This ISA is implemented by the minimal AVR core and supported 18249 for assembler only. 18250 MCU = 'attiny11', 'attiny12', 'attiny15', 'attiny28', 18251 'at90s1200'. 18252 18253'-mabsdata' 18254 18255 Assume that all data in static storage can be accessed by LDS / STS 18256 instructions. This option has only an effect on reduced Tiny 18257 devices like ATtiny40. See also the 'absdata' *note variable 18258 attribute: AVR Variable Attributes. 18259 18260'-maccumulate-args' 18261 Accumulate outgoing function arguments and acquire/release the 18262 needed stack space for outgoing function arguments once in function 18263 prologue/epilogue. Without this option, outgoing arguments are 18264 pushed before calling a function and popped afterwards. 18265 18266 Popping the arguments after the function call can be expensive on 18267 AVR so that accumulating the stack space might lead to smaller 18268 executables because arguments need not be removed from the stack 18269 after such a function call. 18270 18271 This option can lead to reduced code size for functions that 18272 perform several calls to functions that get their arguments on the 18273 stack like calls to printf-like functions. 18274 18275'-mbranch-cost=COST' 18276 Set the branch costs for conditional branch instructions to COST. 18277 Reasonable values for COST are small, non-negative integers. The 18278 default branch cost is 0. 18279 18280'-mcall-prologues' 18281 Functions prologues/epilogues are expanded as calls to appropriate 18282 subroutines. Code size is smaller. 18283 18284'-mdouble=BITS' 18285'-mlong-double=BITS' 18286 Set the size (in bits) of the 'double' or 'long double' type, 18287 respectively. Possible values for BITS are 32 and 64. Whether or 18288 not a specific value for BITS is allowed depends on the 18289 '--with-double=' and '--with-long-double=' 18290 configure options (https://gcc.gnu.org/install/configure.html#avr), 18291 and the same applies for the default values of the options. 18292 18293'-mgas-isr-prologues' 18294 Interrupt service routines (ISRs) may use the '__gcc_isr' pseudo 18295 instruction supported by GNU Binutils. If this option is on, the 18296 feature can still be disabled for individual ISRs by means of the 18297 *note 'no_gccisr': AVR Function Attributes. function attribute. 18298 This feature is activated per default if optimization is on (but 18299 not with '-Og', *note Optimize Options::), and if GNU Binutils 18300 support PR21683 (https://sourceware.org/PR21683). 18301 18302'-mint8' 18303 Assume 'int' to be 8-bit integer. This affects the sizes of all 18304 types: a 'char' is 1 byte, an 'int' is 1 byte, a 'long' is 2 bytes, 18305 and 'long long' is 4 bytes. Please note that this option does not 18306 conform to the C standards, but it results in smaller code size. 18307 18308'-mmain-is-OS_task' 18309 Do not save registers in 'main'. The effect is the same like 18310 attaching attribute *note 'OS_task': AVR Function Attributes. to 18311 'main'. It is activated per default if optimization is on. 18312 18313'-mn-flash=NUM' 18314 Assume that the flash memory has a size of NUM times 64 KiB. 18315 18316'-mno-interrupts' 18317 Generated code is not compatible with hardware interrupts. Code 18318 size is smaller. 18319 18320'-mrelax' 18321 Try to replace 'CALL' resp. 'JMP' instruction by the shorter 18322 'RCALL' resp. 'RJMP' instruction if applicable. Setting '-mrelax' 18323 just adds the '--mlink-relax' option to the assembler's command 18324 line and the '--relax' option to the linker's command line. 18325 18326 Jump relaxing is performed by the linker because jump offsets are 18327 not known before code is located. Therefore, the assembler code 18328 generated by the compiler is the same, but the instructions in the 18329 executable may differ from instructions in the assembler code. 18330 18331 Relaxing must be turned on if linker stubs are needed, see the 18332 section on 'EIND' and linker stubs below. 18333 18334'-mrmw' 18335 Assume that the device supports the Read-Modify-Write instructions 18336 'XCH', 'LAC', 'LAS' and 'LAT'. 18337 18338'-mshort-calls' 18339 18340 Assume that 'RJMP' and 'RCALL' can target the whole program memory. 18341 18342 This option is used internally for multilib selection. It is not 18343 an optimization option, and you don't need to set it by hand. 18344 18345'-msp8' 18346 Treat the stack pointer register as an 8-bit register, i.e. assume 18347 the high byte of the stack pointer is zero. In general, you don't 18348 need to set this option by hand. 18349 18350 This option is used internally by the compiler to select and build 18351 multilibs for architectures 'avr2' and 'avr25'. These 18352 architectures mix devices with and without 'SPH'. For any setting 18353 other than '-mmcu=avr2' or '-mmcu=avr25' the compiler driver adds 18354 or removes this option from the compiler proper's command line, 18355 because the compiler then knows if the device or architecture has 18356 an 8-bit stack pointer and thus no 'SPH' register or not. 18357 18358'-mstrict-X' 18359 Use address register 'X' in a way proposed by the hardware. This 18360 means that 'X' is only used in indirect, post-increment or 18361 pre-decrement addressing. 18362 18363 Without this option, the 'X' register may be used in the same way 18364 as 'Y' or 'Z' which then is emulated by additional instructions. 18365 For example, loading a value with 'X+const' addressing with a small 18366 non-negative 'const < 64' to a register RN is performed as 18367 18368 adiw r26, const ; X += const 18369 ld RN, X ; RN = *X 18370 sbiw r26, const ; X -= const 18371 18372'-mtiny-stack' 18373 Only change the lower 8 bits of the stack pointer. 18374 18375'-mfract-convert-truncate' 18376 Allow to use truncation instead of rounding towards zero for 18377 fractional fixed-point types. 18378 18379'-nodevicelib' 18380 Don't link against AVR-LibC's device specific library 'lib<mcu>.a'. 18381 18382'-nodevicespecs' 18383 Don't add '-specs=device-specs/specs-MCU' to the compiler driver's 18384 command line. The user takes responsibility for supplying the 18385 sub-processes like compiler proper, assembler and linker with 18386 appropriate command line options. This means that the user has to 18387 supply her private device specs file by means of 18388 '-specs=PATH-TO-SPECS-FILE'. There is no more need for option 18389 '-mmcu=MCU'. 18390 18391 This option can also serve as a replacement for the older way of 18392 specifying custom device-specs files that needed '-B SOME-PATH' to 18393 point to a directory which contains a folder named 'device-specs' 18394 which contains a specs file named 'specs-MCU', where MCU was 18395 specified by '-mmcu=MCU'. 18396 18397'-Waddr-space-convert' 18398 Warn about conversions between address spaces in the case where the 18399 resulting address space is not contained in the incoming address 18400 space. 18401 18402'-Wmisspelled-isr' 18403 Warn if the ISR is misspelled, i.e. without __vector prefix. 18404 Enabled by default. 18405 184063.19.6.1 'EIND' and Devices with More Than 128 Ki Bytes of Flash 18407................................................................ 18408 18409Pointers in the implementation are 16 bits wide. The address of a 18410function or label is represented as word address so that indirect jumps 18411and calls can target any code address in the range of 64 Ki words. 18412 18413 In order to facilitate indirect jump on devices with more than 128 Ki 18414bytes of program memory space, there is a special function register 18415called 'EIND' that serves as most significant part of the target address 18416when 'EICALL' or 'EIJMP' instructions are used. 18417 18418 Indirect jumps and calls on these devices are handled as follows by the 18419compiler and are subject to some limitations: 18420 18421 * The compiler never sets 'EIND'. 18422 18423 * The compiler uses 'EIND' implicitly in 'EICALL'/'EIJMP' 18424 instructions or might read 'EIND' directly in order to emulate an 18425 indirect call/jump by means of a 'RET' instruction. 18426 18427 * The compiler assumes that 'EIND' never changes during the startup 18428 code or during the application. In particular, 'EIND' is not 18429 saved/restored in function or interrupt service routine 18430 prologue/epilogue. 18431 18432 * For indirect calls to functions and computed goto, the linker 18433 generates _stubs_. Stubs are jump pads sometimes also called 18434 _trampolines_. Thus, the indirect call/jump jumps to such a stub. 18435 The stub contains a direct jump to the desired address. 18436 18437 * Linker relaxation must be turned on so that the linker generates 18438 the stubs correctly in all situations. See the compiler option 18439 '-mrelax' and the linker option '--relax'. There are corner cases 18440 where the linker is supposed to generate stubs but aborts without 18441 relaxation and without a helpful error message. 18442 18443 * The default linker script is arranged for code with 'EIND = 0'. If 18444 code is supposed to work for a setup with 'EIND != 0', a custom 18445 linker script has to be used in order to place the sections whose 18446 name start with '.trampolines' into the segment where 'EIND' points 18447 to. 18448 18449 * The startup code from libgcc never sets 'EIND'. Notice that 18450 startup code is a blend of code from libgcc and AVR-LibC. For the 18451 impact of AVR-LibC on 'EIND', see the 18452 AVR-LibC user manual (http://nongnu.org/avr-libc/user-manual/). 18453 18454 * It is legitimate for user-specific startup code to set up 'EIND' 18455 early, for example by means of initialization code located in 18456 section '.init3'. Such code runs prior to general startup code 18457 that initializes RAM and calls constructors, but after the bit of 18458 startup code from AVR-LibC that sets 'EIND' to the segment where 18459 the vector table is located. 18460 #include <avr/io.h> 18461 18462 static void 18463 __attribute__((section(".init3"),naked,used,no_instrument_function)) 18464 init3_set_eind (void) 18465 { 18466 __asm volatile ("ldi r24,pm_hh8(__trampolines_start)\n\t" 18467 "out %i0,r24" :: "n" (&EIND) : "r24","memory"); 18468 } 18469 18470 The '__trampolines_start' symbol is defined in the linker script. 18471 18472 * Stubs are generated automatically by the linker if the following 18473 two conditions are met: 18474 18475 - The address of a label is taken by means of the 'gs' modifier 18476 (short for _generate stubs_) like so: 18477 LDI r24, lo8(gs(FUNC)) 18478 LDI r25, hi8(gs(FUNC)) 18479 - The final location of that label is in a code segment 18480 _outside_ the segment where the stubs are located. 18481 18482 * The compiler emits such 'gs' modifiers for code labels in the 18483 following situations: 18484 - Taking address of a function or code label. 18485 - Computed goto. 18486 - If prologue-save function is used, see '-mcall-prologues' 18487 command-line option. 18488 - Switch/case dispatch tables. If you do not want such dispatch 18489 tables you can specify the '-fno-jump-tables' command-line 18490 option. 18491 - C and C++ constructors/destructors called during 18492 startup/shutdown. 18493 - If the tools hit a 'gs()' modifier explained above. 18494 18495 * Jumping to non-symbolic addresses like so is _not_ supported: 18496 18497 int main (void) 18498 { 18499 /* Call function at word address 0x2 */ 18500 return ((int(*)(void)) 0x2)(); 18501 } 18502 18503 Instead, a stub has to be set up, i.e. the function has to be 18504 called through a symbol ('func_4' in the example): 18505 18506 int main (void) 18507 { 18508 extern int func_4 (void); 18509 18510 /* Call function at byte address 0x4 */ 18511 return func_4(); 18512 } 18513 18514 and the application be linked with '-Wl,--defsym,func_4=0x4'. 18515 Alternatively, 'func_4' can be defined in the linker script. 18516 185173.19.6.2 Handling of the 'RAMPD', 'RAMPX', 'RAMPY' and 'RAMPZ' Special Function Registers 18518......................................................................................... 18519 18520Some AVR devices support memories larger than the 64 KiB range that can 18521be accessed with 16-bit pointers. To access memory locations outside 18522this 64 KiB range, the content of a 'RAMP' register is used as high part 18523of the address: The 'X', 'Y', 'Z' address register is concatenated with 18524the 'RAMPX', 'RAMPY', 'RAMPZ' special function register, respectively, 18525to get a wide address. Similarly, 'RAMPD' is used together with direct 18526addressing. 18527 18528 * The startup code initializes the 'RAMP' special function registers 18529 with zero. 18530 18531 * If a *note named address space: AVR Named Address Spaces. other 18532 than generic or '__flash' is used, then 'RAMPZ' is set as needed 18533 before the operation. 18534 18535 * If the device supports RAM larger than 64 KiB and the compiler 18536 needs to change 'RAMPZ' to accomplish an operation, 'RAMPZ' is 18537 reset to zero after the operation. 18538 18539 * If the device comes with a specific 'RAMP' register, the ISR 18540 prologue/epilogue saves/restores that SFR and initializes it with 18541 zero in case the ISR code might (implicitly) use it. 18542 18543 * RAM larger than 64 KiB is not supported by GCC for AVR targets. If 18544 you use inline assembler to read from locations outside the 16-bit 18545 address range and change one of the 'RAMP' registers, you must 18546 reset it to zero after the access. 18547 185483.19.6.3 AVR Built-in Macros 18549............................ 18550 18551GCC defines several built-in macros so that the user code can test for 18552the presence or absence of features. Almost any of the following 18553built-in macros are deduced from device capabilities and thus triggered 18554by the '-mmcu=' command-line option. 18555 18556 For even more AVR-specific built-in macros see *note AVR Named Address 18557Spaces:: and *note AVR Built-in Functions::. 18558 18559'__AVR_ARCH__' 18560 Build-in macro that resolves to a decimal number that identifies 18561 the architecture and depends on the '-mmcu=MCU' option. Possible 18562 values are: 18563 18564 '2', '25', '3', '31', '35', '4', '5', '51', '6' 18565 18566 for MCU='avr2', 'avr25', 'avr3', 'avr31', 'avr35', 'avr4', 'avr5', 18567 'avr51', 'avr6', 18568 18569 respectively and 18570 18571 '100', '102', '103', '104', '105', '106', '107' 18572 18573 for MCU='avrtiny', 'avrxmega2', 'avrxmega3', 'avrxmega4', 18574 'avrxmega5', 'avrxmega6', 'avrxmega7', respectively. If MCU 18575 specifies a device, this built-in macro is set accordingly. For 18576 example, with '-mmcu=atmega8' the macro is defined to '4'. 18577 18578'__AVR_DEVICE__' 18579 Setting '-mmcu=DEVICE' defines this built-in macro which reflects 18580 the device's name. For example, '-mmcu=atmega8' defines the 18581 built-in macro '__AVR_ATmega8__', '-mmcu=attiny261a' defines 18582 '__AVR_ATtiny261A__', etc. 18583 18584 The built-in macros' names follow the scheme '__AVR_DEVICE__' where 18585 DEVICE is the device name as from the AVR user manual. The 18586 difference between DEVICE in the built-in macro and DEVICE in 18587 '-mmcu=DEVICE' is that the latter is always lowercase. 18588 18589 If DEVICE is not a device but only a core architecture like 18590 'avr51', this macro is not defined. 18591 18592'__AVR_DEVICE_NAME__' 18593 Setting '-mmcu=DEVICE' defines this built-in macro to the device's 18594 name. For example, with '-mmcu=atmega8' the macro is defined to 18595 'atmega8'. 18596 18597 If DEVICE is not a device but only a core architecture like 18598 'avr51', this macro is not defined. 18599 18600'__AVR_XMEGA__' 18601 The device / architecture belongs to the XMEGA family of devices. 18602 18603'__AVR_HAVE_ELPM__' 18604 The device has the 'ELPM' instruction. 18605 18606'__AVR_HAVE_ELPMX__' 18607 The device has the 'ELPM RN,Z' and 'ELPM RN,Z+' instructions. 18608 18609'__AVR_HAVE_MOVW__' 18610 The device has the 'MOVW' instruction to perform 16-bit 18611 register-register moves. 18612 18613'__AVR_HAVE_LPMX__' 18614 The device has the 'LPM RN,Z' and 'LPM RN,Z+' instructions. 18615 18616'__AVR_HAVE_MUL__' 18617 The device has a hardware multiplier. 18618 18619'__AVR_HAVE_JMP_CALL__' 18620 The device has the 'JMP' and 'CALL' instructions. This is the case 18621 for devices with more than 8 KiB of program memory. 18622 18623'__AVR_HAVE_EIJMP_EICALL__' 18624'__AVR_3_BYTE_PC__' 18625 The device has the 'EIJMP' and 'EICALL' instructions. This is the 18626 case for devices with more than 128 KiB of program memory. This 18627 also means that the program counter (PC) is 3 bytes wide. 18628 18629'__AVR_2_BYTE_PC__' 18630 The program counter (PC) is 2 bytes wide. This is the case for 18631 devices with up to 128 KiB of program memory. 18632 18633'__AVR_HAVE_8BIT_SP__' 18634'__AVR_HAVE_16BIT_SP__' 18635 The stack pointer (SP) register is treated as 8-bit respectively 18636 16-bit register by the compiler. The definition of these macros is 18637 affected by '-mtiny-stack'. 18638 18639'__AVR_HAVE_SPH__' 18640'__AVR_SP8__' 18641 The device has the SPH (high part of stack pointer) special 18642 function register or has an 8-bit stack pointer, respectively. The 18643 definition of these macros is affected by '-mmcu=' and in the cases 18644 of '-mmcu=avr2' and '-mmcu=avr25' also by '-msp8'. 18645 18646'__AVR_HAVE_RAMPD__' 18647'__AVR_HAVE_RAMPX__' 18648'__AVR_HAVE_RAMPY__' 18649'__AVR_HAVE_RAMPZ__' 18650 The device has the 'RAMPD', 'RAMPX', 'RAMPY', 'RAMPZ' special 18651 function register, respectively. 18652 18653'__NO_INTERRUPTS__' 18654 This macro reflects the '-mno-interrupts' command-line option. 18655 18656'__AVR_ERRATA_SKIP__' 18657'__AVR_ERRATA_SKIP_JMP_CALL__' 18658 Some AVR devices (AT90S8515, ATmega103) must not skip 32-bit 18659 instructions because of a hardware erratum. Skip instructions are 18660 'SBRS', 'SBRC', 'SBIS', 'SBIC' and 'CPSE'. The second macro is 18661 only defined if '__AVR_HAVE_JMP_CALL__' is also set. 18662 18663'__AVR_ISA_RMW__' 18664 The device has Read-Modify-Write instructions (XCH, LAC, LAS and 18665 LAT). 18666 18667'__AVR_SFR_OFFSET__=OFFSET' 18668 Instructions that can address I/O special function registers 18669 directly like 'IN', 'OUT', 'SBI', etc. may use a different address 18670 as if addressed by an instruction to access RAM like 'LD' or 'STS'. 18671 This offset depends on the device architecture and has to be 18672 subtracted from the RAM address in order to get the respective 18673 I/O address. 18674 18675'__AVR_SHORT_CALLS__' 18676 The '-mshort-calls' command line option is set. 18677 18678'__AVR_PM_BASE_ADDRESS__=ADDR' 18679 Some devices support reading from flash memory by means of 'LD*' 18680 instructions. The flash memory is seen in the data address space 18681 at an offset of '__AVR_PM_BASE_ADDRESS__'. If this macro is not 18682 defined, this feature is not available. If defined, the address 18683 space is linear and there is no need to put '.rodata' into RAM. 18684 This is handled by the default linker description file, and is 18685 currently available for 'avrtiny' and 'avrxmega3'. Even more 18686 convenient, there is no need to use address spaces like '__flash' 18687 or features like attribute 'progmem' and 'pgm_read_*'. 18688 18689'__WITH_AVRLIBC__' 18690 The compiler is configured to be used together with AVR-Libc. See 18691 the '--with-avrlibc' configure option. 18692 18693'__HAVE_DOUBLE_MULTILIB__' 18694 Defined if '-mdouble=' acts as a multilib option. 18695 18696'__HAVE_DOUBLE32__' 18697'__HAVE_DOUBLE64__' 18698 Defined if the compiler supports 32-bit double resp. 64-bit 18699 double. The actual layout is specified by option '-mdouble='. 18700 18701'__DEFAULT_DOUBLE__' 18702 The size in bits of 'double' if '-mdouble=' is not set. To test 18703 the layout of 'double' in a program, use the built-in macro 18704 '__SIZEOF_DOUBLE__'. 18705 18706'__HAVE_LONG_DOUBLE32__' 18707'__HAVE_LONG_DOUBLE64__' 18708'__HAVE_LONG_DOUBLE_MULTILIB__' 18709'__DEFAULT_LONG_DOUBLE__' 18710 Same as above, but for 'long double' instead of 'double'. 18711 18712'__WITH_DOUBLE_COMPARISON__' 18713 Reflects the '--with-double-comparison={tristate|bool|libf7}' 18714 configure option (https://gcc.gnu.org/install/configure.html#avr) 18715 and is defined to '2' or '3'. 18716 18717'__WITH_LIBF7_LIBGCC__' 18718'__WITH_LIBF7_MATH__' 18719'__WITH_LIBF7_MATH_SYMBOLS__' 18720 Reflects the '--with-libf7={libgcc|math|math-symbols}' 18721 configure option (https://gcc.gnu.org/install/configure.html#avr). 18722 18723 18724File: gcc.info, Node: Blackfin Options, Next: C6X Options, Prev: AVR Options, Up: Submodel Options 18725 187263.19.7 Blackfin Options 18727----------------------- 18728 18729'-mcpu=CPU[-SIREVISION]' 18730 Specifies the name of the target Blackfin processor. Currently, 18731 CPU can be one of 'bf512', 'bf514', 'bf516', 'bf518', 'bf522', 18732 'bf523', 'bf524', 'bf525', 'bf526', 'bf527', 'bf531', 'bf532', 18733 'bf533', 'bf534', 'bf536', 'bf537', 'bf538', 'bf539', 'bf542', 18734 'bf544', 'bf547', 'bf548', 'bf549', 'bf542m', 'bf544m', 'bf547m', 18735 'bf548m', 'bf549m', 'bf561', 'bf592'. 18736 18737 The optional SIREVISION specifies the silicon revision of the 18738 target Blackfin processor. Any workarounds available for the 18739 targeted silicon revision are enabled. If SIREVISION is 'none', no 18740 workarounds are enabled. If SIREVISION is 'any', all workarounds 18741 for the targeted processor are enabled. The '__SILICON_REVISION__' 18742 macro is defined to two hexadecimal digits representing the major 18743 and minor numbers in the silicon revision. If SIREVISION is 18744 'none', the '__SILICON_REVISION__' is not defined. If SIREVISION 18745 is 'any', the '__SILICON_REVISION__' is defined to be '0xffff'. If 18746 this optional SIREVISION is not used, GCC assumes the latest known 18747 silicon revision of the targeted Blackfin processor. 18748 18749 GCC defines a preprocessor macro for the specified CPU. For the 18750 'bfin-elf' toolchain, this option causes the hardware BSP provided 18751 by libgloss to be linked in if '-msim' is not given. 18752 18753 Without this option, 'bf532' is used as the processor by default. 18754 18755 Note that support for 'bf561' is incomplete. For 'bf561', only the 18756 preprocessor macro is defined. 18757 18758'-msim' 18759 Specifies that the program will be run on the simulator. This 18760 causes the simulator BSP provided by libgloss to be linked in. 18761 This option has effect only for 'bfin-elf' toolchain. Certain 18762 other options, such as '-mid-shared-library' and '-mfdpic', imply 18763 '-msim'. 18764 18765'-momit-leaf-frame-pointer' 18766 Don't keep the frame pointer in a register for leaf functions. 18767 This avoids the instructions to save, set up and restore frame 18768 pointers and makes an extra register available in leaf functions. 18769 18770'-mspecld-anomaly' 18771 When enabled, the compiler ensures that the generated code does not 18772 contain speculative loads after jump instructions. If this option 18773 is used, '__WORKAROUND_SPECULATIVE_LOADS' is defined. 18774 18775'-mno-specld-anomaly' 18776 Don't generate extra code to prevent speculative loads from 18777 occurring. 18778 18779'-mcsync-anomaly' 18780 When enabled, the compiler ensures that the generated code does not 18781 contain CSYNC or SSYNC instructions too soon after conditional 18782 branches. If this option is used, '__WORKAROUND_SPECULATIVE_SYNCS' 18783 is defined. 18784 18785'-mno-csync-anomaly' 18786 Don't generate extra code to prevent CSYNC or SSYNC instructions 18787 from occurring too soon after a conditional branch. 18788 18789'-mlow64k' 18790 When enabled, the compiler is free to take advantage of the 18791 knowledge that the entire program fits into the low 64k of memory. 18792 18793'-mno-low64k' 18794 Assume that the program is arbitrarily large. This is the default. 18795 18796'-mstack-check-l1' 18797 Do stack checking using information placed into L1 scratchpad 18798 memory by the uClinux kernel. 18799 18800'-mid-shared-library' 18801 Generate code that supports shared libraries via the library ID 18802 method. This allows for execute in place and shared libraries in 18803 an environment without virtual memory management. This option 18804 implies '-fPIC'. With a 'bfin-elf' target, this option implies 18805 '-msim'. 18806 18807'-mno-id-shared-library' 18808 Generate code that doesn't assume ID-based shared libraries are 18809 being used. This is the default. 18810 18811'-mleaf-id-shared-library' 18812 Generate code that supports shared libraries via the library ID 18813 method, but assumes that this library or executable won't link 18814 against any other ID shared libraries. That allows the compiler to 18815 use faster code for jumps and calls. 18816 18817'-mno-leaf-id-shared-library' 18818 Do not assume that the code being compiled won't link against any 18819 ID shared libraries. Slower code is generated for jump and call 18820 insns. 18821 18822'-mshared-library-id=n' 18823 Specifies the identification number of the ID-based shared library 18824 being compiled. Specifying a value of 0 generates more compact 18825 code; specifying other values forces the allocation of that number 18826 to the current library but is no more space- or time-efficient than 18827 omitting this option. 18828 18829'-msep-data' 18830 Generate code that allows the data segment to be located in a 18831 different area of memory from the text segment. This allows for 18832 execute in place in an environment without virtual memory 18833 management by eliminating relocations against the text section. 18834 18835'-mno-sep-data' 18836 Generate code that assumes that the data segment follows the text 18837 segment. This is the default. 18838 18839'-mlong-calls' 18840'-mno-long-calls' 18841 Tells the compiler to perform function calls by first loading the 18842 address of the function into a register and then performing a 18843 subroutine call on this register. This switch is needed if the 18844 target function lies outside of the 24-bit addressing range of the 18845 offset-based version of subroutine call instruction. 18846 18847 This feature is not enabled by default. Specifying 18848 '-mno-long-calls' restores the default behavior. Note these 18849 switches have no effect on how the compiler generates code to 18850 handle function calls via function pointers. 18851 18852'-mfast-fp' 18853 Link with the fast floating-point library. This library relaxes 18854 some of the IEEE floating-point standard's rules for checking 18855 inputs against Not-a-Number (NAN), in the interest of performance. 18856 18857'-minline-plt' 18858 Enable inlining of PLT entries in function calls to functions that 18859 are not known to bind locally. It has no effect without '-mfdpic'. 18860 18861'-mmulticore' 18862 Build a standalone application for multicore Blackfin processors. 18863 This option causes proper start files and link scripts supporting 18864 multicore to be used, and defines the macro '__BFIN_MULTICORE'. It 18865 can only be used with '-mcpu=bf561[-SIREVISION]'. 18866 18867 This option can be used with '-mcorea' or '-mcoreb', which selects 18868 the one-application-per-core programming model. Without '-mcorea' 18869 or '-mcoreb', the single-application/dual-core programming model is 18870 used. In this model, the main function of Core B should be named 18871 as 'coreb_main'. 18872 18873 If this option is not used, the single-core application programming 18874 model is used. 18875 18876'-mcorea' 18877 Build a standalone application for Core A of BF561 when using the 18878 one-application-per-core programming model. Proper start files and 18879 link scripts are used to support Core A, and the macro 18880 '__BFIN_COREA' is defined. This option can only be used in 18881 conjunction with '-mmulticore'. 18882 18883'-mcoreb' 18884 Build a standalone application for Core B of BF561 when using the 18885 one-application-per-core programming model. Proper start files and 18886 link scripts are used to support Core B, and the macro 18887 '__BFIN_COREB' is defined. When this option is used, 'coreb_main' 18888 should be used instead of 'main'. This option can only be used in 18889 conjunction with '-mmulticore'. 18890 18891'-msdram' 18892 Build a standalone application for SDRAM. Proper start files and 18893 link scripts are used to put the application into SDRAM, and the 18894 macro '__BFIN_SDRAM' is defined. The loader should initialize 18895 SDRAM before loading the application. 18896 18897'-micplb' 18898 Assume that ICPLBs are enabled at run time. This has an effect on 18899 certain anomaly workarounds. For Linux targets, the default is to 18900 assume ICPLBs are enabled; for standalone applications the default 18901 is off. 18902 18903 18904File: gcc.info, Node: C6X Options, Next: CRIS Options, Prev: Blackfin Options, Up: Submodel Options 18905 189063.19.8 C6X Options 18907------------------ 18908 18909'-march=NAME' 18910 This specifies the name of the target architecture. GCC uses this 18911 name to determine what kind of instructions it can emit when 18912 generating assembly code. Permissible names are: 'c62x', 'c64x', 18913 'c64x+', 'c67x', 'c67x+', 'c674x'. 18914 18915'-mbig-endian' 18916 Generate code for a big-endian target. 18917 18918'-mlittle-endian' 18919 Generate code for a little-endian target. This is the default. 18920 18921'-msim' 18922 Choose startup files and linker script suitable for the simulator. 18923 18924'-msdata=default' 18925 Put small global and static data in the '.neardata' section, which 18926 is pointed to by register 'B14'. Put small uninitialized global 18927 and static data in the '.bss' section, which is adjacent to the 18928 '.neardata' section. Put small read-only data into the '.rodata' 18929 section. The corresponding sections used for large pieces of data 18930 are '.fardata', '.far' and '.const'. 18931 18932'-msdata=all' 18933 Put all data, not just small objects, into the sections reserved 18934 for small data, and use addressing relative to the 'B14' register 18935 to access them. 18936 18937'-msdata=none' 18938 Make no use of the sections reserved for small data, and use 18939 absolute addresses to access all data. Put all initialized global 18940 and static data in the '.fardata' section, and all uninitialized 18941 data in the '.far' section. Put all constant data into the 18942 '.const' section. 18943 18944 18945File: gcc.info, Node: CRIS Options, Next: CR16 Options, Prev: C6X Options, Up: Submodel Options 18946 189473.19.9 CRIS Options 18948------------------- 18949 18950These options are defined specifically for the CRIS ports. 18951 18952'-march=ARCHITECTURE-TYPE' 18953'-mcpu=ARCHITECTURE-TYPE' 18954 Generate code for the specified architecture. The choices for 18955 ARCHITECTURE-TYPE are 'v3', 'v8' and 'v10' for respectively 18956 ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is 'v0' except for 18957 cris-axis-linux-gnu, where the default is 'v10'. 18958 18959'-mtune=ARCHITECTURE-TYPE' 18960 Tune to ARCHITECTURE-TYPE everything applicable about the generated 18961 code, except for the ABI and the set of available instructions. 18962 The choices for ARCHITECTURE-TYPE are the same as for 18963 '-march=ARCHITECTURE-TYPE'. 18964 18965'-mmax-stack-frame=N' 18966 Warn when the stack frame of a function exceeds N bytes. 18967 18968'-metrax4' 18969'-metrax100' 18970 The options '-metrax4' and '-metrax100' are synonyms for 18971 '-march=v3' and '-march=v8' respectively. 18972 18973'-mmul-bug-workaround' 18974'-mno-mul-bug-workaround' 18975 Work around a bug in the 'muls' and 'mulu' instructions for CPU 18976 models where it applies. This option is active by default. 18977 18978'-mpdebug' 18979 Enable CRIS-specific verbose debug-related information in the 18980 assembly code. This option also has the effect of turning off the 18981 '#NO_APP' formatted-code indicator to the assembler at the 18982 beginning of the assembly file. 18983 18984'-mcc-init' 18985 Do not use condition-code results from previous instruction; always 18986 emit compare and test instructions before use of condition codes. 18987 18988'-mno-side-effects' 18989 Do not emit instructions with side effects in addressing modes 18990 other than post-increment. 18991 18992'-mstack-align' 18993'-mno-stack-align' 18994'-mdata-align' 18995'-mno-data-align' 18996'-mconst-align' 18997'-mno-const-align' 18998 These options ('no-' options) arrange (eliminate arrangements) for 18999 the stack frame, individual data and constants to be aligned for 19000 the maximum single data access size for the chosen CPU model. The 19001 default is to arrange for 32-bit alignment. ABI details such as 19002 structure layout are not affected by these options. 19003 19004'-m32-bit' 19005'-m16-bit' 19006'-m8-bit' 19007 Similar to the stack- data- and const-align options above, these 19008 options arrange for stack frame, writable data and constants to all 19009 be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit 19010 alignment. 19011 19012'-mno-prologue-epilogue' 19013'-mprologue-epilogue' 19014 With '-mno-prologue-epilogue', the normal function prologue and 19015 epilogue which set up the stack frame are omitted and no return 19016 instructions or return sequences are generated in the code. Use 19017 this option only together with visual inspection of the compiled 19018 code: no warnings or errors are generated when call-saved registers 19019 must be saved, or storage for local variables needs to be 19020 allocated. 19021 19022'-mno-gotplt' 19023'-mgotplt' 19024 With '-fpic' and '-fPIC', don't generate (do generate) instruction 19025 sequences that load addresses for functions from the PLT part of 19026 the GOT rather than (traditional on other architectures) calls to 19027 the PLT. The default is '-mgotplt'. 19028 19029'-melf' 19030 Legacy no-op option only recognized with the cris-axis-elf and 19031 cris-axis-linux-gnu targets. 19032 19033'-mlinux' 19034 Legacy no-op option only recognized with the cris-axis-linux-gnu 19035 target. 19036 19037'-sim' 19038 This option, recognized for the cris-axis-elf, arranges to link 19039 with input-output functions from a simulator library. Code, 19040 initialized data and zero-initialized data are allocated 19041 consecutively. 19042 19043'-sim2' 19044 Like '-sim', but pass linker options to locate initialized data at 19045 0x40000000 and zero-initialized data at 0x80000000. 19046 19047 19048File: gcc.info, Node: CR16 Options, Next: C-SKY Options, Prev: CRIS Options, Up: Submodel Options 19049 190503.19.10 CR16 Options 19051-------------------- 19052 19053These options are defined specifically for the CR16 ports. 19054 19055'-mmac' 19056 Enable the use of multiply-accumulate instructions. Disabled by 19057 default. 19058 19059'-mcr16cplus' 19060'-mcr16c' 19061 Generate code for CR16C or CR16C+ architecture. CR16C+ 19062 architecture is default. 19063 19064'-msim' 19065 Links the library libsim.a which is in compatible with simulator. 19066 Applicable to ELF compiler only. 19067 19068'-mint32' 19069 Choose integer type as 32-bit wide. 19070 19071'-mbit-ops' 19072 Generates 'sbit'/'cbit' instructions for bit manipulations. 19073 19074'-mdata-model=MODEL' 19075 Choose a data model. The choices for MODEL are 'near', 'far' or 19076 'medium'. 'medium' is default. However, 'far' is not valid with 19077 '-mcr16c', as the CR16C architecture does not support the far data 19078 model. 19079 19080 19081File: gcc.info, Node: C-SKY Options, Next: Darwin Options, Prev: CR16 Options, Up: Submodel Options 19082 190833.19.11 C-SKY Options 19084--------------------- 19085 19086GCC supports these options when compiling for C-SKY V2 processors. 19087 19088'-march=ARCH' 19089 Specify the C-SKY target architecture. Valid values for ARCH are: 19090 'ck801', 'ck802', 'ck803', 'ck807', and 'ck810'. The default is 19091 'ck810'. 19092 19093'-mcpu=CPU' 19094 Specify the C-SKY target processor. Valid values for CPU are: 19095 'ck801', 'ck801t', 'ck802', 'ck802t', 'ck802j', 'ck803', 'ck803h', 19096 'ck803t', 'ck803ht', 'ck803f', 'ck803fh', 'ck803e', 'ck803eh', 19097 'ck803et', 'ck803eht', 'ck803ef', 'ck803efh', 'ck803ft', 19098 'ck803eft', 'ck803efht', 'ck803r1', 'ck803hr1', 'ck803tr1', 19099 'ck803htr1', 'ck803fr1', 'ck803fhr1', 'ck803er1', 'ck803ehr1', 19100 'ck803etr1', 'ck803ehtr1', 'ck803efr1', 'ck803efhr1', 'ck803ftr1', 19101 'ck803eftr1', 'ck803efhtr1', 'ck803s', 'ck803st', 'ck803se', 19102 'ck803sf', 'ck803sef', 'ck803seft', 'ck807e', 'ck807ef', 'ck807', 19103 'ck807f', 'ck810e', 'ck810et', 'ck810ef', 'ck810eft', 'ck810', 19104 'ck810v', 'ck810f', 'ck810t', 'ck810fv', 'ck810tv', 'ck810ft', and 19105 'ck810ftv'. 19106 19107'-mbig-endian' 19108'-EB' 19109'-mlittle-endian' 19110'-EL' 19111 19112 Select big- or little-endian code. The default is little-endian. 19113 19114'-mhard-float' 19115'-msoft-float' 19116 19117 Select hardware or software floating-point implementations. The 19118 default is soft float. 19119 19120'-mdouble-float' 19121'-mno-double-float' 19122 When '-mhard-float' is in effect, enable generation of 19123 double-precision float instructions. This is the default except 19124 when compiling for CK803. 19125 19126'-mfdivdu' 19127'-mno-fdivdu' 19128 When '-mhard-float' is in effect, enable generation of 'frecipd', 19129 'fsqrtd', and 'fdivd' instructions. This is the default except 19130 when compiling for CK803. 19131 19132'-mfpu=FPU' 19133 Select the floating-point processor. This option can only be used 19134 with '-mhard-float'. Values for FPU are 'fpv2_sf' (equivalent to 19135 '-mno-double-float -mno-fdivdu'), 'fpv2' ('-mdouble-float 19136 -mno-divdu'), and 'fpv2_divd' ('-mdouble-float -mdivdu'). 19137 19138'-melrw' 19139'-mno-elrw' 19140 Enable the extended 'lrw' instruction. This option defaults to on 19141 for CK801 and off otherwise. 19142 19143'-mistack' 19144'-mno-istack' 19145 Enable interrupt stack instructions; the default is off. 19146 19147 The '-mistack' option is required to handle the 'interrupt' and 19148 'isr' function attributes (*note C-SKY Function Attributes::). 19149 19150'-mmp' 19151 Enable multiprocessor instructions; the default is off. 19152 19153'-mcp' 19154 Enable coprocessor instructions; the default is off. 19155 19156'-mcache' 19157 Enable coprocessor instructions; the default is off. 19158 19159'-msecurity' 19160 Enable C-SKY security instructions; the default is off. 19161 19162'-mtrust' 19163 Enable C-SKY trust instructions; the default is off. 19164 19165'-mdsp' 19166'-medsp' 19167'-mvdsp' 19168 Enable C-SKY DSP, Enhanced DSP, or Vector DSP instructions, 19169 respectively. All of these options default to off. 19170 19171'-mdiv' 19172'-mno-div' 19173 Generate divide instructions. Default is off. 19174 19175'-msmart' 19176'-mno-smart' 19177 Generate code for Smart Mode, using only registers numbered 0-7 to 19178 allow use of 16-bit instructions. This option is ignored for CK801 19179 where this is the required behavior, and it defaults to on for 19180 CK802. For other targets, the default is off. 19181 19182'-mhigh-registers' 19183'-mno-high-registers' 19184 Generate code using the high registers numbered 16-31. This option 19185 is not supported on CK801, CK802, or CK803, and is enabled by 19186 default for other processors. 19187 19188'-manchor' 19189'-mno-anchor' 19190 Generate code using global anchor symbol addresses. 19191 19192'-mpushpop' 19193'-mno-pushpop' 19194 Generate code using 'push' and 'pop' instructions. This option 19195 defaults to on. 19196 19197'-mmultiple-stld' 19198'-mstm' 19199'-mno-multiple-stld' 19200'-mno-stm' 19201 Generate code using 'stm' and 'ldm' instructions. This option 19202 isn't supported on CK801 but is enabled by default on other 19203 processors. 19204 19205'-mconstpool' 19206'-mno-constpool' 19207 Create constant pools in the compiler instead of deferring it to 19208 the assembler. This option is the default and required for correct 19209 code generation on CK801 and CK802, and is optional on other 19210 processors. 19211 19212'-mstack-size' 19213'-mno-stack-size' 19214 Emit '.stack_size' directives for each function in the assembly 19215 output. This option defaults to off. 19216 19217'-mccrt' 19218'-mno-ccrt' 19219 Generate code for the C-SKY compiler runtime instead of libgcc. 19220 This option defaults to off. 19221 19222'-mbranch-cost=N' 19223 Set the branch costs to roughly 'n' instructions. The default is 19224 1. 19225 19226'-msched-prolog' 19227'-mno-sched-prolog' 19228 Permit scheduling of function prologue and epilogue sequences. 19229 Using this option can result in code that is not compliant with the 19230 C-SKY V2 ABI prologue requirements and that cannot be debugged or 19231 backtraced. It is disabled by default. 19232 19233 19234File: gcc.info, Node: Darwin Options, Next: DEC Alpha Options, Prev: C-SKY Options, Up: Submodel Options 19235 192363.19.12 Darwin Options 19237---------------------- 19238 19239These options are defined for all architectures running the Darwin 19240operating system. 19241 19242 FSF GCC on Darwin does not create "fat" object files; it creates an 19243object file for the single architecture that GCC was built to target. 19244Apple's GCC on Darwin does create "fat" files if multiple '-arch' 19245options are used; it does so by running the compiler or linker multiple 19246times and joining the results together with 'lipo'. 19247 19248 The subtype of the file created (like 'ppc7400' or 'ppc970' or 'i686') 19249is determined by the flags that specify the ISA that GCC is targeting, 19250like '-mcpu' or '-march'. The '-force_cpusubtype_ALL' option can be 19251used to override this. 19252 19253 The Darwin tools vary in their behavior when presented with an ISA 19254mismatch. The assembler, 'as', only permits instructions to be used 19255that are valid for the subtype of the file it is generating, so you 19256cannot put 64-bit instructions in a 'ppc750' object file. The linker 19257for shared libraries, '/usr/bin/libtool', fails and prints an error if 19258asked to create a shared library with a less restrictive subtype than 19259its input files (for instance, trying to put a 'ppc970' object file in a 19260'ppc7400' library). The linker for executables, 'ld', quietly gives the 19261executable the most restrictive subtype of any of its input files. 19262 19263'-FDIR' 19264 Add the framework directory DIR to the head of the list of 19265 directories to be searched for header files. These directories are 19266 interleaved with those specified by '-I' options and are scanned in 19267 a left-to-right order. 19268 19269 A framework directory is a directory with frameworks in it. A 19270 framework is a directory with a 'Headers' and/or 'PrivateHeaders' 19271 directory contained directly in it that ends in '.framework'. The 19272 name of a framework is the name of this directory excluding the 19273 '.framework'. Headers associated with the framework are found in 19274 one of those two directories, with 'Headers' being searched first. 19275 A subframework is a framework directory that is in a framework's 19276 'Frameworks' directory. Includes of subframework headers can only 19277 appear in a header of a framework that contains the subframework, 19278 or in a sibling subframework header. Two subframeworks are 19279 siblings if they occur in the same framework. A subframework 19280 should not have the same name as a framework; a warning is issued 19281 if this is violated. Currently a subframework cannot have 19282 subframeworks; in the future, the mechanism may be extended to 19283 support this. The standard frameworks can be found in 19284 '/System/Library/Frameworks' and '/Library/Frameworks'. An example 19285 include looks like '#include <Framework/header.h>', where 19286 'Framework' denotes the name of the framework and 'header.h' is 19287 found in the 'PrivateHeaders' or 'Headers' directory. 19288 19289'-iframeworkDIR' 19290 Like '-F' except the directory is a treated as a system directory. 19291 The main difference between this '-iframework' and '-F' is that 19292 with '-iframework' the compiler does not warn about constructs 19293 contained within header files found via DIR. This option is valid 19294 only for the C family of languages. 19295 19296'-gused' 19297 Emit debugging information for symbols that are used. For stabs 19298 debugging format, this enables '-feliminate-unused-debug-symbols'. 19299 This is by default ON. 19300 19301'-gfull' 19302 Emit debugging information for all symbols and types. 19303 19304'-mmacosx-version-min=VERSION' 19305 The earliest version of MacOS X that this executable will run on is 19306 VERSION. Typical values of VERSION include '10.1', '10.2', and 19307 '10.3.9'. 19308 19309 If the compiler was built to use the system's headers by default, 19310 then the default for this option is the system version on which the 19311 compiler is running, otherwise the default is to make choices that 19312 are compatible with as many systems and code bases as possible. 19313 19314'-mkernel' 19315 Enable kernel development mode. The '-mkernel' option sets 19316 '-static', '-fno-common', '-fno-use-cxa-atexit', '-fno-exceptions', 19317 '-fno-non-call-exceptions', '-fapple-kext', '-fno-weak' and 19318 '-fno-rtti' where applicable. This mode also sets '-mno-altivec', 19319 '-msoft-float', '-fno-builtin' and '-mlong-branch' for PowerPC 19320 targets. 19321 19322'-mone-byte-bool' 19323 Override the defaults for 'bool' so that 'sizeof(bool)==1'. By 19324 default 'sizeof(bool)' is '4' when compiling for Darwin/PowerPC and 19325 '1' when compiling for Darwin/x86, so this option has no effect on 19326 x86. 19327 19328 *Warning:* The '-mone-byte-bool' switch causes GCC to generate code 19329 that is not binary compatible with code generated without that 19330 switch. Using this switch may require recompiling all other 19331 modules in a program, including system libraries. Use this switch 19332 to conform to a non-default data model. 19333 19334'-mfix-and-continue' 19335'-ffix-and-continue' 19336'-findirect-data' 19337 Generate code suitable for fast turnaround development, such as to 19338 allow GDB to dynamically load '.o' files into already-running 19339 programs. '-findirect-data' and '-ffix-and-continue' are provided 19340 for backwards compatibility. 19341 19342'-all_load' 19343 Loads all members of static archive libraries. See man ld(1) for 19344 more information. 19345 19346'-arch_errors_fatal' 19347 Cause the errors having to do with files that have the wrong 19348 architecture to be fatal. 19349 19350'-bind_at_load' 19351 Causes the output file to be marked such that the dynamic linker 19352 will bind all undefined references when the file is loaded or 19353 launched. 19354 19355'-bundle' 19356 Produce a Mach-o bundle format file. See man ld(1) for more 19357 information. 19358 19359'-bundle_loader EXECUTABLE' 19360 This option specifies the EXECUTABLE that will load the build 19361 output file being linked. See man ld(1) for more information. 19362 19363'-dynamiclib' 19364 When passed this option, GCC produces a dynamic library instead of 19365 an executable when linking, using the Darwin 'libtool' command. 19366 19367'-force_cpusubtype_ALL' 19368 This causes GCC's output file to have the 'ALL' subtype, instead of 19369 one controlled by the '-mcpu' or '-march' option. 19370 19371'-allowable_client CLIENT_NAME' 19372'-client_name' 19373'-compatibility_version' 19374'-current_version' 19375'-dead_strip' 19376'-dependency-file' 19377'-dylib_file' 19378'-dylinker_install_name' 19379'-dynamic' 19380'-exported_symbols_list' 19381'-filelist' 19382'-flat_namespace' 19383'-force_flat_namespace' 19384'-headerpad_max_install_names' 19385'-image_base' 19386'-init' 19387'-install_name' 19388'-keep_private_externs' 19389'-multi_module' 19390'-multiply_defined' 19391'-multiply_defined_unused' 19392'-noall_load' 19393'-no_dead_strip_inits_and_terms' 19394'-nofixprebinding' 19395'-nomultidefs' 19396'-noprebind' 19397'-noseglinkedit' 19398'-pagezero_size' 19399'-prebind' 19400'-prebind_all_twolevel_modules' 19401'-private_bundle' 19402'-read_only_relocs' 19403'-sectalign' 19404'-sectobjectsymbols' 19405'-whyload' 19406'-seg1addr' 19407'-sectcreate' 19408'-sectobjectsymbols' 19409'-sectorder' 19410'-segaddr' 19411'-segs_read_only_addr' 19412'-segs_read_write_addr' 19413'-seg_addr_table' 19414'-seg_addr_table_filename' 19415'-seglinkedit' 19416'-segprot' 19417'-segs_read_only_addr' 19418'-segs_read_write_addr' 19419'-single_module' 19420'-static' 19421'-sub_library' 19422'-sub_umbrella' 19423'-twolevel_namespace' 19424'-umbrella' 19425'-undefined' 19426'-unexported_symbols_list' 19427'-weak_reference_mismatches' 19428'-whatsloaded' 19429 These options are passed to the Darwin linker. The Darwin linker 19430 man page describes them in detail. 19431 19432 19433File: gcc.info, Node: DEC Alpha Options, Next: eBPF Options, Prev: Darwin Options, Up: Submodel Options 19434 194353.19.13 DEC Alpha Options 19436------------------------- 19437 19438These '-m' options are defined for the DEC Alpha implementations: 19439 19440'-mno-soft-float' 19441'-msoft-float' 19442 Use (do not use) the hardware floating-point instructions for 19443 floating-point operations. When '-msoft-float' is specified, 19444 functions in 'libgcc.a' are used to perform floating-point 19445 operations. Unless they are replaced by routines that emulate the 19446 floating-point operations, or compiled in such a way as to call 19447 such emulations routines, these routines issue floating-point 19448 operations. If you are compiling for an Alpha without 19449 floating-point operations, you must ensure that the library is 19450 built so as not to call them. 19451 19452 Note that Alpha implementations without floating-point operations 19453 are required to have floating-point registers. 19454 19455'-mfp-reg' 19456'-mno-fp-regs' 19457 Generate code that uses (does not use) the floating-point register 19458 set. '-mno-fp-regs' implies '-msoft-float'. If the floating-point 19459 register set is not used, floating-point operands are passed in 19460 integer registers as if they were integers and floating-point 19461 results are passed in '$0' instead of '$f0'. This is a 19462 non-standard calling sequence, so any function with a 19463 floating-point argument or return value called by code compiled 19464 with '-mno-fp-regs' must also be compiled with that option. 19465 19466 A typical use of this option is building a kernel that does not 19467 use, and hence need not save and restore, any floating-point 19468 registers. 19469 19470'-mieee' 19471 The Alpha architecture implements floating-point hardware optimized 19472 for maximum performance. It is mostly compliant with the IEEE 19473 floating-point standard. However, for full compliance, software 19474 assistance is required. This option generates code fully 19475 IEEE-compliant code _except_ that the INEXACT-FLAG is not 19476 maintained (see below). If this option is turned on, the 19477 preprocessor macro '_IEEE_FP' is defined during compilation. The 19478 resulting code is less efficient but is able to correctly support 19479 denormalized numbers and exceptional IEEE values such as 19480 not-a-number and plus/minus infinity. Other Alpha compilers call 19481 this option '-ieee_with_no_inexact'. 19482 19483'-mieee-with-inexact' 19484 This is like '-mieee' except the generated code also maintains the 19485 IEEE INEXACT-FLAG. Turning on this option causes the generated 19486 code to implement fully-compliant IEEE math. In addition to 19487 '_IEEE_FP', '_IEEE_FP_EXACT' is defined as a preprocessor macro. 19488 On some Alpha implementations the resulting code may execute 19489 significantly slower than the code generated by default. Since 19490 there is very little code that depends on the INEXACT-FLAG, you 19491 should normally not specify this option. Other Alpha compilers 19492 call this option '-ieee_with_inexact'. 19493 19494'-mfp-trap-mode=TRAP-MODE' 19495 This option controls what floating-point related traps are enabled. 19496 Other Alpha compilers call this option '-fptm TRAP-MODE'. The trap 19497 mode can be set to one of four values: 19498 19499 'n' 19500 This is the default (normal) setting. The only traps that are 19501 enabled are the ones that cannot be disabled in software 19502 (e.g., division by zero trap). 19503 19504 'u' 19505 In addition to the traps enabled by 'n', underflow traps are 19506 enabled as well. 19507 19508 'su' 19509 Like 'u', but the instructions are marked to be safe for 19510 software completion (see Alpha architecture manual for 19511 details). 19512 19513 'sui' 19514 Like 'su', but inexact traps are enabled as well. 19515 19516'-mfp-rounding-mode=ROUNDING-MODE' 19517 Selects the IEEE rounding mode. Other Alpha compilers call this 19518 option '-fprm ROUNDING-MODE'. The ROUNDING-MODE can be one of: 19519 19520 'n' 19521 Normal IEEE rounding mode. Floating-point numbers are rounded 19522 towards the nearest machine number or towards the even machine 19523 number in case of a tie. 19524 19525 'm' 19526 Round towards minus infinity. 19527 19528 'c' 19529 Chopped rounding mode. Floating-point numbers are rounded 19530 towards zero. 19531 19532 'd' 19533 Dynamic rounding mode. A field in the floating-point control 19534 register (FPCR, see Alpha architecture reference manual) 19535 controls the rounding mode in effect. The C library 19536 initializes this register for rounding towards plus infinity. 19537 Thus, unless your program modifies the FPCR, 'd' corresponds 19538 to round towards plus infinity. 19539 19540'-mtrap-precision=TRAP-PRECISION' 19541 In the Alpha architecture, floating-point traps are imprecise. 19542 This means without software assistance it is impossible to recover 19543 from a floating trap and program execution normally needs to be 19544 terminated. GCC can generate code that can assist operating system 19545 trap handlers in determining the exact location that caused a 19546 floating-point trap. Depending on the requirements of an 19547 application, different levels of precisions can be selected: 19548 19549 'p' 19550 Program precision. This option is the default and means a 19551 trap handler can only identify which program caused a 19552 floating-point exception. 19553 19554 'f' 19555 Function precision. The trap handler can determine the 19556 function that caused a floating-point exception. 19557 19558 'i' 19559 Instruction precision. The trap handler can determine the 19560 exact instruction that caused a floating-point exception. 19561 19562 Other Alpha compilers provide the equivalent options called 19563 '-scope_safe' and '-resumption_safe'. 19564 19565'-mieee-conformant' 19566 This option marks the generated code as IEEE conformant. You must 19567 not use this option unless you also specify '-mtrap-precision=i' 19568 and either '-mfp-trap-mode=su' or '-mfp-trap-mode=sui'. Its only 19569 effect is to emit the line '.eflag 48' in the function prologue of 19570 the generated assembly file. 19571 19572'-mbuild-constants' 19573 Normally GCC examines a 32- or 64-bit integer constant to see if it 19574 can construct it from smaller constants in two or three 19575 instructions. If it cannot, it outputs the constant as a literal 19576 and generates code to load it from the data segment at run time. 19577 19578 Use this option to require GCC to construct _all_ integer constants 19579 using code, even if it takes more instructions (the maximum is 19580 six). 19581 19582 You typically use this option to build a shared library dynamic 19583 loader. Itself a shared library, it must relocate itself in memory 19584 before it can find the variables and constants in its own data 19585 segment. 19586 19587'-mbwx' 19588'-mno-bwx' 19589'-mcix' 19590'-mno-cix' 19591'-mfix' 19592'-mno-fix' 19593'-mmax' 19594'-mno-max' 19595 Indicate whether GCC should generate code to use the optional BWX, 19596 CIX, FIX and MAX instruction sets. The default is to use the 19597 instruction sets supported by the CPU type specified via '-mcpu=' 19598 option or that of the CPU on which GCC was built if none is 19599 specified. 19600 19601'-mfloat-vax' 19602'-mfloat-ieee' 19603 Generate code that uses (does not use) VAX F and G floating-point 19604 arithmetic instead of IEEE single and double precision. 19605 19606'-mexplicit-relocs' 19607'-mno-explicit-relocs' 19608 Older Alpha assemblers provided no way to generate symbol 19609 relocations except via assembler macros. Use of these macros does 19610 not allow optimal instruction scheduling. GNU binutils as of 19611 version 2.12 supports a new syntax that allows the compiler to 19612 explicitly mark which relocations should apply to which 19613 instructions. This option is mostly useful for debugging, as GCC 19614 detects the capabilities of the assembler when it is built and sets 19615 the default accordingly. 19616 19617'-msmall-data' 19618'-mlarge-data' 19619 When '-mexplicit-relocs' is in effect, static data is accessed via 19620 "gp-relative" relocations. When '-msmall-data' is used, objects 8 19621 bytes long or smaller are placed in a "small data area" (the 19622 '.sdata' and '.sbss' sections) and are accessed via 16-bit 19623 relocations off of the '$gp' register. This limits the size of the 19624 small data area to 64KB, but allows the variables to be directly 19625 accessed via a single instruction. 19626 19627 The default is '-mlarge-data'. With this option the data area is 19628 limited to just below 2GB. Programs that require more than 2GB of 19629 data must use 'malloc' or 'mmap' to allocate the data in the heap 19630 instead of in the program's data segment. 19631 19632 When generating code for shared libraries, '-fpic' implies 19633 '-msmall-data' and '-fPIC' implies '-mlarge-data'. 19634 19635'-msmall-text' 19636'-mlarge-text' 19637 When '-msmall-text' is used, the compiler assumes that the code of 19638 the entire program (or shared library) fits in 4MB, and is thus 19639 reachable with a branch instruction. When '-msmall-data' is used, 19640 the compiler can assume that all local symbols share the same '$gp' 19641 value, and thus reduce the number of instructions required for a 19642 function call from 4 to 1. 19643 19644 The default is '-mlarge-text'. 19645 19646'-mcpu=CPU_TYPE' 19647 Set the instruction set and instruction scheduling parameters for 19648 machine type CPU_TYPE. You can specify either the 'EV' style name 19649 or the corresponding chip number. GCC supports scheduling 19650 parameters for the EV4, EV5 and EV6 family of processors and 19651 chooses the default values for the instruction set from the 19652 processor you specify. If you do not specify a processor type, GCC 19653 defaults to the processor on which the compiler was built. 19654 19655 Supported values for CPU_TYPE are 19656 19657 'ev4' 19658 'ev45' 19659 '21064' 19660 Schedules as an EV4 and has no instruction set extensions. 19661 19662 'ev5' 19663 '21164' 19664 Schedules as an EV5 and has no instruction set extensions. 19665 19666 'ev56' 19667 '21164a' 19668 Schedules as an EV5 and supports the BWX extension. 19669 19670 'pca56' 19671 '21164pc' 19672 '21164PC' 19673 Schedules as an EV5 and supports the BWX and MAX extensions. 19674 19675 'ev6' 19676 '21264' 19677 Schedules as an EV6 and supports the BWX, FIX, and MAX 19678 extensions. 19679 19680 'ev67' 19681 '21264a' 19682 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX 19683 extensions. 19684 19685 Native toolchains also support the value 'native', which selects 19686 the best architecture option for the host processor. 19687 '-mcpu=native' has no effect if GCC does not recognize the 19688 processor. 19689 19690'-mtune=CPU_TYPE' 19691 Set only the instruction scheduling parameters for machine type 19692 CPU_TYPE. The instruction set is not changed. 19693 19694 Native toolchains also support the value 'native', which selects 19695 the best architecture option for the host processor. 19696 '-mtune=native' has no effect if GCC does not recognize the 19697 processor. 19698 19699'-mmemory-latency=TIME' 19700 Sets the latency the scheduler should assume for typical memory 19701 references as seen by the application. This number is highly 19702 dependent on the memory access patterns used by the application and 19703 the size of the external cache on the machine. 19704 19705 Valid options for TIME are 19706 19707 'NUMBER' 19708 A decimal number representing clock cycles. 19709 19710 'L1' 19711 'L2' 19712 'L3' 19713 'main' 19714 The compiler contains estimates of the number of clock cycles 19715 for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches 19716 (also called Dcache, Scache, and Bcache), as well as to main 19717 memory. Note that L3 is only valid for EV5. 19718 19719 19720File: gcc.info, Node: eBPF Options, Next: FR30 Options, Prev: DEC Alpha Options, Up: Submodel Options 19721 197223.19.14 eBPF Options 19723-------------------- 19724 19725'-mframe-limit=BYTES' 19726 This specifies the hard limit for frame sizes, in bytes. 19727 Currently, the value that can be specified should be less than or 19728 equal to '32767'. Defaults to whatever limit is imposed by the 19729 version of the Linux kernel targeted. 19730 19731'-mkernel=VERSION' 19732 This specifies the minimum version of the kernel that will run the 19733 compiled program. GCC uses this version to determine which 19734 instructions to use, what kernel helpers to allow, etc. Currently, 19735 VERSION can be one of '4.0', '4.1', '4.2', '4.3', '4.4', '4.5', 19736 '4.6', '4.7', '4.8', '4.9', '4.10', '4.11', '4.12', '4.13', '4.14', 19737 '4.15', '4.16', '4.17', '4.18', '4.19', '4.20', '5.0', '5.1', 19738 '5.2', 'latest' and 'native'. 19739 19740'-mbig-endian' 19741 Generate code for a big-endian target. 19742 19743'-mlittle-endian' 19744 Generate code for a little-endian target. This is the default. 19745 19746'-mxbpf' 19747 Generate code for an expanded version of BPF, which relaxes some of 19748 the restrictions imposed by the BPF architecture: 19749 - Save and restore callee-saved registers at function entry and 19750 exit, respectively. 19751 19752 19753File: gcc.info, Node: FR30 Options, Next: FT32 Options, Prev: eBPF Options, Up: Submodel Options 19754 197553.19.15 FR30 Options 19756-------------------- 19757 19758These options are defined specifically for the FR30 port. 19759 19760'-msmall-model' 19761 Use the small address space model. This can produce smaller code, 19762 but it does assume that all symbolic values and addresses fit into 19763 a 20-bit range. 19764 19765'-mno-lsim' 19766 Assume that runtime support has been provided and so there is no 19767 need to include the simulator library ('libsim.a') on the linker 19768 command line. 19769 19770 19771File: gcc.info, Node: FT32 Options, Next: FRV Options, Prev: FR30 Options, Up: Submodel Options 19772 197733.19.16 FT32 Options 19774-------------------- 19775 19776These options are defined specifically for the FT32 port. 19777 19778'-msim' 19779 Specifies that the program will be run on the simulator. This 19780 causes an alternate runtime startup and library to be linked. You 19781 must not use this option when generating programs that will run on 19782 real hardware; you must provide your own runtime library for 19783 whatever I/O functions are needed. 19784 19785'-mlra' 19786 Enable Local Register Allocation. This is still experimental for 19787 FT32, so by default the compiler uses standard reload. 19788 19789'-mnodiv' 19790 Do not use div and mod instructions. 19791 19792'-mft32b' 19793 Enable use of the extended instructions of the FT32B processor. 19794 19795'-mcompress' 19796 Compress all code using the Ft32B code compression scheme. 19797 19798'-mnopm' 19799 Do not generate code that reads program memory. 19800 19801 19802File: gcc.info, Node: FRV Options, Next: GNU/Linux Options, Prev: FT32 Options, Up: Submodel Options 19803 198043.19.17 FRV Options 19805------------------- 19806 19807'-mgpr-32' 19808 19809 Only use the first 32 general-purpose registers. 19810 19811'-mgpr-64' 19812 19813 Use all 64 general-purpose registers. 19814 19815'-mfpr-32' 19816 19817 Use only the first 32 floating-point registers. 19818 19819'-mfpr-64' 19820 19821 Use all 64 floating-point registers. 19822 19823'-mhard-float' 19824 19825 Use hardware instructions for floating-point operations. 19826 19827'-msoft-float' 19828 19829 Use library routines for floating-point operations. 19830 19831'-malloc-cc' 19832 19833 Dynamically allocate condition code registers. 19834 19835'-mfixed-cc' 19836 19837 Do not try to dynamically allocate condition code registers, only 19838 use 'icc0' and 'fcc0'. 19839 19840'-mdword' 19841 19842 Change ABI to use double word insns. 19843 19844'-mno-dword' 19845 19846 Do not use double word instructions. 19847 19848'-mdouble' 19849 19850 Use floating-point double instructions. 19851 19852'-mno-double' 19853 19854 Do not use floating-point double instructions. 19855 19856'-mmedia' 19857 19858 Use media instructions. 19859 19860'-mno-media' 19861 19862 Do not use media instructions. 19863 19864'-mmuladd' 19865 19866 Use multiply and add/subtract instructions. 19867 19868'-mno-muladd' 19869 19870 Do not use multiply and add/subtract instructions. 19871 19872'-mfdpic' 19873 19874 Select the FDPIC ABI, which uses function descriptors to represent 19875 pointers to functions. Without any PIC/PIE-related options, it 19876 implies '-fPIE'. With '-fpic' or '-fpie', it assumes GOT entries 19877 and small data are within a 12-bit range from the GOT base address; 19878 with '-fPIC' or '-fPIE', GOT offsets are computed with 32 bits. 19879 With a 'bfin-elf' target, this option implies '-msim'. 19880 19881'-minline-plt' 19882 19883 Enable inlining of PLT entries in function calls to functions that 19884 are not known to bind locally. It has no effect without '-mfdpic'. 19885 It's enabled by default if optimizing for speed and compiling for 19886 shared libraries (i.e., '-fPIC' or '-fpic'), or when an 19887 optimization option such as '-O3' or above is present in the 19888 command line. 19889 19890'-mTLS' 19891 19892 Assume a large TLS segment when generating thread-local code. 19893 19894'-mtls' 19895 19896 Do not assume a large TLS segment when generating thread-local 19897 code. 19898 19899'-mgprel-ro' 19900 19901 Enable the use of 'GPREL' relocations in the FDPIC ABI for data 19902 that is known to be in read-only sections. It's enabled by 19903 default, except for '-fpic' or '-fpie': even though it may help 19904 make the global offset table smaller, it trades 1 instruction for 19905 4. With '-fPIC' or '-fPIE', it trades 3 instructions for 4, one of 19906 which may be shared by multiple symbols, and it avoids the need for 19907 a GOT entry for the referenced symbol, so it's more likely to be a 19908 win. If it is not, '-mno-gprel-ro' can be used to disable it. 19909 19910'-multilib-library-pic' 19911 19912 Link with the (library, not FD) pic libraries. It's implied by 19913 '-mlibrary-pic', as well as by '-fPIC' and '-fpic' without 19914 '-mfdpic'. You should never have to use it explicitly. 19915 19916'-mlinked-fp' 19917 19918 Follow the EABI requirement of always creating a frame pointer 19919 whenever a stack frame is allocated. This option is enabled by 19920 default and can be disabled with '-mno-linked-fp'. 19921 19922'-mlong-calls' 19923 19924 Use indirect addressing to call functions outside the current 19925 compilation unit. This allows the functions to be placed anywhere 19926 within the 32-bit address space. 19927 19928'-malign-labels' 19929 19930 Try to align labels to an 8-byte boundary by inserting NOPs into 19931 the previous packet. This option only has an effect when VLIW 19932 packing is enabled. It doesn't create new packets; it merely adds 19933 NOPs to existing ones. 19934 19935'-mlibrary-pic' 19936 19937 Generate position-independent EABI code. 19938 19939'-macc-4' 19940 19941 Use only the first four media accumulator registers. 19942 19943'-macc-8' 19944 19945 Use all eight media accumulator registers. 19946 19947'-mpack' 19948 19949 Pack VLIW instructions. 19950 19951'-mno-pack' 19952 19953 Do not pack VLIW instructions. 19954 19955'-mno-eflags' 19956 19957 Do not mark ABI switches in e_flags. 19958 19959'-mcond-move' 19960 19961 Enable the use of conditional-move instructions (default). 19962 19963 This switch is mainly for debugging the compiler and will likely be 19964 removed in a future version. 19965 19966'-mno-cond-move' 19967 19968 Disable the use of conditional-move instructions. 19969 19970 This switch is mainly for debugging the compiler and will likely be 19971 removed in a future version. 19972 19973'-mscc' 19974 19975 Enable the use of conditional set instructions (default). 19976 19977 This switch is mainly for debugging the compiler and will likely be 19978 removed in a future version. 19979 19980'-mno-scc' 19981 19982 Disable the use of conditional set instructions. 19983 19984 This switch is mainly for debugging the compiler and will likely be 19985 removed in a future version. 19986 19987'-mcond-exec' 19988 19989 Enable the use of conditional execution (default). 19990 19991 This switch is mainly for debugging the compiler and will likely be 19992 removed in a future version. 19993 19994'-mno-cond-exec' 19995 19996 Disable the use of conditional execution. 19997 19998 This switch is mainly for debugging the compiler and will likely be 19999 removed in a future version. 20000 20001'-mvliw-branch' 20002 20003 Run a pass to pack branches into VLIW instructions (default). 20004 20005 This switch is mainly for debugging the compiler and will likely be 20006 removed in a future version. 20007 20008'-mno-vliw-branch' 20009 20010 Do not run a pass to pack branches into VLIW instructions. 20011 20012 This switch is mainly for debugging the compiler and will likely be 20013 removed in a future version. 20014 20015'-mmulti-cond-exec' 20016 20017 Enable optimization of '&&' and '||' in conditional execution 20018 (default). 20019 20020 This switch is mainly for debugging the compiler and will likely be 20021 removed in a future version. 20022 20023'-mno-multi-cond-exec' 20024 20025 Disable optimization of '&&' and '||' in conditional execution. 20026 20027 This switch is mainly for debugging the compiler and will likely be 20028 removed in a future version. 20029 20030'-mnested-cond-exec' 20031 20032 Enable nested conditional execution optimizations (default). 20033 20034 This switch is mainly for debugging the compiler and will likely be 20035 removed in a future version. 20036 20037'-mno-nested-cond-exec' 20038 20039 Disable nested conditional execution optimizations. 20040 20041 This switch is mainly for debugging the compiler and will likely be 20042 removed in a future version. 20043 20044'-moptimize-membar' 20045 20046 This switch removes redundant 'membar' instructions from the 20047 compiler-generated code. It is enabled by default. 20048 20049'-mno-optimize-membar' 20050 20051 This switch disables the automatic removal of redundant 'membar' 20052 instructions from the generated code. 20053 20054'-mtomcat-stats' 20055 20056 Cause gas to print out tomcat statistics. 20057 20058'-mcpu=CPU' 20059 20060 Select the processor type for which to generate code. Possible 20061 values are 'frv', 'fr550', 'tomcat', 'fr500', 'fr450', 'fr405', 20062 'fr400', 'fr300' and 'simple'. 20063 20064 20065File: gcc.info, Node: GNU/Linux Options, Next: H8/300 Options, Prev: FRV Options, Up: Submodel Options 20066 200673.19.18 GNU/Linux Options 20068------------------------- 20069 20070These '-m' options are defined for GNU/Linux targets: 20071 20072'-mglibc' 20073 Use the GNU C library. This is the default except on 20074 '*-*-linux-*uclibc*', '*-*-linux-*musl*' and '*-*-linux-*android*' 20075 targets. 20076 20077'-muclibc' 20078 Use uClibc C library. This is the default on '*-*-linux-*uclibc*' 20079 targets. 20080 20081'-mmusl' 20082 Use the musl C library. This is the default on '*-*-linux-*musl*' 20083 targets. 20084 20085'-mbionic' 20086 Use Bionic C library. This is the default on '*-*-linux-*android*' 20087 targets. 20088 20089'-mandroid' 20090 Compile code compatible with Android platform. This is the default 20091 on '*-*-linux-*android*' targets. 20092 20093 When compiling, this option enables '-mbionic', '-fPIC', 20094 '-fno-exceptions' and '-fno-rtti' by default. When linking, this 20095 option makes the GCC driver pass Android-specific options to the 20096 linker. Finally, this option causes the preprocessor macro 20097 '__ANDROID__' to be defined. 20098 20099'-tno-android-cc' 20100 Disable compilation effects of '-mandroid', i.e., do not enable 20101 '-mbionic', '-fPIC', '-fno-exceptions' and '-fno-rtti' by default. 20102 20103'-tno-android-ld' 20104 Disable linking effects of '-mandroid', i.e., pass standard Linux 20105 linking options to the linker. 20106 20107 20108File: gcc.info, Node: H8/300 Options, Next: HPPA Options, Prev: GNU/Linux Options, Up: Submodel Options 20109 201103.19.19 H8/300 Options 20111---------------------- 20112 20113These '-m' options are defined for the H8/300 implementations: 20114 20115'-mrelax' 20116 Shorten some address references at link time, when possible; uses 20117 the linker option '-relax'. *Note 'ld' and the H8/300: (ld)H8/300, 20118 for a fuller description. 20119 20120'-mh' 20121 Generate code for the H8/300H. 20122 20123'-ms' 20124 Generate code for the H8S. 20125 20126'-mn' 20127 Generate code for the H8S and H8/300H in the normal mode. This 20128 switch must be used either with '-mh' or '-ms'. 20129 20130'-ms2600' 20131 Generate code for the H8S/2600. This switch must be used with 20132 '-ms'. 20133 20134'-mexr' 20135 Extended registers are stored on stack before execution of function 20136 with monitor attribute. Default option is '-mexr'. This option is 20137 valid only for H8S targets. 20138 20139'-mno-exr' 20140 Extended registers are not stored on stack before execution of 20141 function with monitor attribute. Default option is '-mno-exr'. 20142 This option is valid only for H8S targets. 20143 20144'-mint32' 20145 Make 'int' data 32 bits by default. 20146 20147'-malign-300' 20148 On the H8/300H and H8S, use the same alignment rules as for the 20149 H8/300. The default for the H8/300H and H8S is to align longs and 20150 floats on 4-byte boundaries. '-malign-300' causes them to be 20151 aligned on 2-byte boundaries. This option has no effect on the 20152 H8/300. 20153 20154 20155File: gcc.info, Node: HPPA Options, Next: IA-64 Options, Prev: H8/300 Options, Up: Submodel Options 20156 201573.19.20 HPPA Options 20158-------------------- 20159 20160These '-m' options are defined for the HPPA family of computers: 20161 20162'-march=ARCHITECTURE-TYPE' 20163 Generate code for the specified architecture. The choices for 20164 ARCHITECTURE-TYPE are '1.0' for PA 1.0, '1.1' for PA 1.1, and '2.0' 20165 for PA 2.0 processors. Refer to '/usr/lib/sched.models' on an 20166 HP-UX system to determine the proper architecture option for your 20167 machine. Code compiled for lower numbered architectures runs on 20168 higher numbered architectures, but not the other way around. 20169 20170'-mpa-risc-1-0' 20171'-mpa-risc-1-1' 20172'-mpa-risc-2-0' 20173 Synonyms for '-march=1.0', '-march=1.1', and '-march=2.0' 20174 respectively. 20175 20176'-mcaller-copies' 20177 The caller copies function arguments passed by hidden reference. 20178 This option should be used with care as it is not compatible with 20179 the default 32-bit runtime. However, only aggregates larger than 20180 eight bytes are passed by hidden reference and the option provides 20181 better compatibility with OpenMP. 20182 20183'-mjump-in-delay' 20184 This option is ignored and provided for compatibility purposes 20185 only. 20186 20187'-mdisable-fpregs' 20188 Prevent floating-point registers from being used in any manner. 20189 This is necessary for compiling kernels that perform lazy context 20190 switching of floating-point registers. If you use this option and 20191 attempt to perform floating-point operations, the compiler aborts. 20192 20193'-mdisable-indexing' 20194 Prevent the compiler from using indexing address modes. This 20195 avoids some rather obscure problems when compiling MIG generated 20196 code under MACH. 20197 20198'-mno-space-regs' 20199 Generate code that assumes the target has no space registers. This 20200 allows GCC to generate faster indirect calls and use unscaled index 20201 address modes. 20202 20203 Such code is suitable for level 0 PA systems and kernels. 20204 20205'-mfast-indirect-calls' 20206 Generate code that assumes calls never cross space boundaries. 20207 This allows GCC to emit code that performs faster indirect calls. 20208 20209 This option does not work in the presence of shared libraries or 20210 nested functions. 20211 20212'-mfixed-range=REGISTER-RANGE' 20213 Generate code treating the given register range as fixed registers. 20214 A fixed register is one that the register allocator cannot use. 20215 This is useful when compiling kernel code. A register range is 20216 specified as two registers separated by a dash. Multiple register 20217 ranges can be specified separated by a comma. 20218 20219'-mlong-load-store' 20220 Generate 3-instruction load and store sequences as sometimes 20221 required by the HP-UX 10 linker. This is equivalent to the '+k' 20222 option to the HP compilers. 20223 20224'-mportable-runtime' 20225 Use the portable calling conventions proposed by HP for ELF 20226 systems. 20227 20228'-mgas' 20229 Enable the use of assembler directives only GAS understands. 20230 20231'-mschedule=CPU-TYPE' 20232 Schedule code according to the constraints for the machine type 20233 CPU-TYPE. The choices for CPU-TYPE are '700' '7100', '7100LC', 20234 '7200', '7300' and '8000'. Refer to '/usr/lib/sched.models' on an 20235 HP-UX system to determine the proper scheduling option for your 20236 machine. The default scheduling is '8000'. 20237 20238'-mlinker-opt' 20239 Enable the optimization pass in the HP-UX linker. Note this makes 20240 symbolic debugging impossible. It also triggers a bug in the HP-UX 20241 8 and HP-UX 9 linkers in which they give bogus error messages when 20242 linking some programs. 20243 20244'-msoft-float' 20245 Generate output containing library calls for floating point. 20246 *Warning:* the requisite libraries are not available for all HPPA 20247 targets. Normally the facilities of the machine's usual C compiler 20248 are used, but this cannot be done directly in cross-compilation. 20249 You must make your own arrangements to provide suitable library 20250 functions for cross-compilation. 20251 20252 '-msoft-float' changes the calling convention in the output file; 20253 therefore, it is only useful if you compile _all_ of a program with 20254 this option. In particular, you need to compile 'libgcc.a', the 20255 library that comes with GCC, with '-msoft-float' in order for this 20256 to work. 20257 20258'-msio' 20259 Generate the predefine, '_SIO', for server IO. The default is 20260 '-mwsio'. This generates the predefines, '__hp9000s700', 20261 '__hp9000s700__' and '_WSIO', for workstation IO. These options 20262 are available under HP-UX and HI-UX. 20263 20264'-mgnu-ld' 20265 Use options specific to GNU 'ld'. This passes '-shared' to 'ld' 20266 when building a shared library. It is the default when GCC is 20267 configured, explicitly or implicitly, with the GNU linker. This 20268 option does not affect which 'ld' is called; it only changes what 20269 parameters are passed to that 'ld'. The 'ld' that is called is 20270 determined by the '--with-ld' configure option, GCC's program 20271 search path, and finally by the user's 'PATH'. The linker used by 20272 GCC can be printed using 'which `gcc -print-prog-name=ld`'. This 20273 option is only available on the 64-bit HP-UX GCC, i.e. configured 20274 with 'hppa*64*-*-hpux*'. 20275 20276'-mhp-ld' 20277 Use options specific to HP 'ld'. This passes '-b' to 'ld' when 20278 building a shared library and passes '+Accept TypeMismatch' to 'ld' 20279 on all links. It is the default when GCC is configured, explicitly 20280 or implicitly, with the HP linker. This option does not affect 20281 which 'ld' is called; it only changes what parameters are passed to 20282 that 'ld'. The 'ld' that is called is determined by the 20283 '--with-ld' configure option, GCC's program search path, and 20284 finally by the user's 'PATH'. The linker used by GCC can be 20285 printed using 'which `gcc -print-prog-name=ld`'. This option is 20286 only available on the 64-bit HP-UX GCC, i.e. configured with 20287 'hppa*64*-*-hpux*'. 20288 20289'-mlong-calls' 20290 Generate code that uses long call sequences. This ensures that a 20291 call is always able to reach linker generated stubs. The default 20292 is to generate long calls only when the distance from the call site 20293 to the beginning of the function or translation unit, as the case 20294 may be, exceeds a predefined limit set by the branch type being 20295 used. The limits for normal calls are 7,600,000 and 240,000 bytes, 20296 respectively for the PA 2.0 and PA 1.X architectures. Sibcalls are 20297 always limited at 240,000 bytes. 20298 20299 Distances are measured from the beginning of functions when using 20300 the '-ffunction-sections' option, or when using the '-mgas' and 20301 '-mno-portable-runtime' options together under HP-UX with the SOM 20302 linker. 20303 20304 It is normally not desirable to use this option as it degrades 20305 performance. However, it may be useful in large applications, 20306 particularly when partial linking is used to build the application. 20307 20308 The types of long calls used depends on the capabilities of the 20309 assembler and linker, and the type of code being generated. The 20310 impact on systems that support long absolute calls, and long pic 20311 symbol-difference or pc-relative calls should be relatively small. 20312 However, an indirect call is used on 32-bit ELF systems in pic code 20313 and it is quite long. 20314 20315'-munix=UNIX-STD' 20316 Generate compiler predefines and select a startfile for the 20317 specified UNIX standard. The choices for UNIX-STD are '93', '95' 20318 and '98'. '93' is supported on all HP-UX versions. '95' is 20319 available on HP-UX 10.10 and later. '98' is available on HP-UX 20320 11.11 and later. The default values are '93' for HP-UX 10.00, '95' 20321 for HP-UX 10.10 though to 11.00, and '98' for HP-UX 11.11 and 20322 later. 20323 20324 '-munix=93' provides the same predefines as GCC 3.3 and 3.4. 20325 '-munix=95' provides additional predefines for 'XOPEN_UNIX' and 20326 '_XOPEN_SOURCE_EXTENDED', and the startfile 'unix95.o'. 20327 '-munix=98' provides additional predefines for '_XOPEN_UNIX', 20328 '_XOPEN_SOURCE_EXTENDED', '_INCLUDE__STDC_A1_SOURCE' and 20329 '_INCLUDE_XOPEN_SOURCE_500', and the startfile 'unix98.o'. 20330 20331 It is _important_ to note that this option changes the interfaces 20332 for various library routines. It also affects the operational 20333 behavior of the C library. Thus, _extreme_ care is needed in using 20334 this option. 20335 20336 Library code that is intended to operate with more than one UNIX 20337 standard must test, set and restore the variable 20338 '__xpg4_extended_mask' as appropriate. Most GNU software doesn't 20339 provide this capability. 20340 20341'-nolibdld' 20342 Suppress the generation of link options to search libdld.sl when 20343 the '-static' option is specified on HP-UX 10 and later. 20344 20345'-static' 20346 The HP-UX implementation of setlocale in libc has a dependency on 20347 libdld.sl. There isn't an archive version of libdld.sl. Thus, 20348 when the '-static' option is specified, special link options are 20349 needed to resolve this dependency. 20350 20351 On HP-UX 10 and later, the GCC driver adds the necessary options to 20352 link with libdld.sl when the '-static' option is specified. This 20353 causes the resulting binary to be dynamic. On the 64-bit port, the 20354 linkers generate dynamic binaries by default in any case. The 20355 '-nolibdld' option can be used to prevent the GCC driver from 20356 adding these link options. 20357 20358'-threads' 20359 Add support for multithreading with the "dce thread" library under 20360 HP-UX. This option sets flags for both the preprocessor and 20361 linker. 20362 20363 20364File: gcc.info, Node: IA-64 Options, Next: LM32 Options, Prev: HPPA Options, Up: Submodel Options 20365 203663.19.21 IA-64 Options 20367--------------------- 20368 20369These are the '-m' options defined for the Intel IA-64 architecture. 20370 20371'-mbig-endian' 20372 Generate code for a big-endian target. This is the default for 20373 HP-UX. 20374 20375'-mlittle-endian' 20376 Generate code for a little-endian target. This is the default for 20377 AIX5 and GNU/Linux. 20378 20379'-mgnu-as' 20380'-mno-gnu-as' 20381 Generate (or don't) code for the GNU assembler. This is the 20382 default. 20383 20384'-mgnu-ld' 20385'-mno-gnu-ld' 20386 Generate (or don't) code for the GNU linker. This is the default. 20387 20388'-mno-pic' 20389 Generate code that does not use a global pointer register. The 20390 result is not position independent code, and violates the IA-64 20391 ABI. 20392 20393'-mvolatile-asm-stop' 20394'-mno-volatile-asm-stop' 20395 Generate (or don't) a stop bit immediately before and after 20396 volatile asm statements. 20397 20398'-mregister-names' 20399'-mno-register-names' 20400 Generate (or don't) 'in', 'loc', and 'out' register names for the 20401 stacked registers. This may make assembler output more readable. 20402 20403'-mno-sdata' 20404'-msdata' 20405 Disable (or enable) optimizations that use the small data section. 20406 This may be useful for working around optimizer bugs. 20407 20408'-mconstant-gp' 20409 Generate code that uses a single constant global pointer value. 20410 This is useful when compiling kernel code. 20411 20412'-mauto-pic' 20413 Generate code that is self-relocatable. This implies 20414 '-mconstant-gp'. This is useful when compiling firmware code. 20415 20416'-minline-float-divide-min-latency' 20417 Generate code for inline divides of floating-point values using the 20418 minimum latency algorithm. 20419 20420'-minline-float-divide-max-throughput' 20421 Generate code for inline divides of floating-point values using the 20422 maximum throughput algorithm. 20423 20424'-mno-inline-float-divide' 20425 Do not generate inline code for divides of floating-point values. 20426 20427'-minline-int-divide-min-latency' 20428 Generate code for inline divides of integer values using the 20429 minimum latency algorithm. 20430 20431'-minline-int-divide-max-throughput' 20432 Generate code for inline divides of integer values using the 20433 maximum throughput algorithm. 20434 20435'-mno-inline-int-divide' 20436 Do not generate inline code for divides of integer values. 20437 20438'-minline-sqrt-min-latency' 20439 Generate code for inline square roots using the minimum latency 20440 algorithm. 20441 20442'-minline-sqrt-max-throughput' 20443 Generate code for inline square roots using the maximum throughput 20444 algorithm. 20445 20446'-mno-inline-sqrt' 20447 Do not generate inline code for 'sqrt'. 20448 20449'-mfused-madd' 20450'-mno-fused-madd' 20451 Do (don't) generate code that uses the fused multiply/add or 20452 multiply/subtract instructions. The default is to use these 20453 instructions. 20454 20455'-mno-dwarf2-asm' 20456'-mdwarf2-asm' 20457 Don't (or do) generate assembler code for the DWARF line number 20458 debugging info. This may be useful when not using the GNU 20459 assembler. 20460 20461'-mearly-stop-bits' 20462'-mno-early-stop-bits' 20463 Allow stop bits to be placed earlier than immediately preceding the 20464 instruction that triggered the stop bit. This can improve 20465 instruction scheduling, but does not always do so. 20466 20467'-mfixed-range=REGISTER-RANGE' 20468 Generate code treating the given register range as fixed registers. 20469 A fixed register is one that the register allocator cannot use. 20470 This is useful when compiling kernel code. A register range is 20471 specified as two registers separated by a dash. Multiple register 20472 ranges can be specified separated by a comma. 20473 20474'-mtls-size=TLS-SIZE' 20475 Specify bit size of immediate TLS offsets. Valid values are 14, 20476 22, and 64. 20477 20478'-mtune=CPU-TYPE' 20479 Tune the instruction scheduling for a particular CPU, Valid values 20480 are 'itanium', 'itanium1', 'merced', 'itanium2', and 'mckinley'. 20481 20482'-milp32' 20483'-mlp64' 20484 Generate code for a 32-bit or 64-bit environment. The 32-bit 20485 environment sets int, long and pointer to 32 bits. The 64-bit 20486 environment sets int to 32 bits and long and pointer to 64 bits. 20487 These are HP-UX specific flags. 20488 20489'-mno-sched-br-data-spec' 20490'-msched-br-data-spec' 20491 (Dis/En)able data speculative scheduling before reload. This 20492 results in generation of 'ld.a' instructions and the corresponding 20493 check instructions ('ld.c' / 'chk.a'). The default setting is 20494 disabled. 20495 20496'-msched-ar-data-spec' 20497'-mno-sched-ar-data-spec' 20498 (En/Dis)able data speculative scheduling after reload. This 20499 results in generation of 'ld.a' instructions and the corresponding 20500 check instructions ('ld.c' / 'chk.a'). The default setting is 20501 enabled. 20502 20503'-mno-sched-control-spec' 20504'-msched-control-spec' 20505 (Dis/En)able control speculative scheduling. This feature is 20506 available only during region scheduling (i.e. before reload). This 20507 results in generation of the 'ld.s' instructions and the 20508 corresponding check instructions 'chk.s'. The default setting is 20509 disabled. 20510 20511'-msched-br-in-data-spec' 20512'-mno-sched-br-in-data-spec' 20513 (En/Dis)able speculative scheduling of the instructions that are 20514 dependent on the data speculative loads before reload. This is 20515 effective only with '-msched-br-data-spec' enabled. The default 20516 setting is enabled. 20517 20518'-msched-ar-in-data-spec' 20519'-mno-sched-ar-in-data-spec' 20520 (En/Dis)able speculative scheduling of the instructions that are 20521 dependent on the data speculative loads after reload. This is 20522 effective only with '-msched-ar-data-spec' enabled. The default 20523 setting is enabled. 20524 20525'-msched-in-control-spec' 20526'-mno-sched-in-control-spec' 20527 (En/Dis)able speculative scheduling of the instructions that are 20528 dependent on the control speculative loads. This is effective only 20529 with '-msched-control-spec' enabled. The default setting is 20530 enabled. 20531 20532'-mno-sched-prefer-non-data-spec-insns' 20533'-msched-prefer-non-data-spec-insns' 20534 If enabled, data-speculative instructions are chosen for schedule 20535 only if there are no other choices at the moment. This makes the 20536 use of the data speculation much more conservative. The default 20537 setting is disabled. 20538 20539'-mno-sched-prefer-non-control-spec-insns' 20540'-msched-prefer-non-control-spec-insns' 20541 If enabled, control-speculative instructions are chosen for 20542 schedule only if there are no other choices at the moment. This 20543 makes the use of the control speculation much more conservative. 20544 The default setting is disabled. 20545 20546'-mno-sched-count-spec-in-critical-path' 20547'-msched-count-spec-in-critical-path' 20548 If enabled, speculative dependencies are considered during 20549 computation of the instructions priorities. This makes the use of 20550 the speculation a bit more conservative. The default setting is 20551 disabled. 20552 20553'-msched-spec-ldc' 20554 Use a simple data speculation check. This option is on by default. 20555 20556'-msched-control-spec-ldc' 20557 Use a simple check for control speculation. This option is on by 20558 default. 20559 20560'-msched-stop-bits-after-every-cycle' 20561 Place a stop bit after every cycle when scheduling. This option is 20562 on by default. 20563 20564'-msched-fp-mem-deps-zero-cost' 20565 Assume that floating-point stores and loads are not likely to cause 20566 a conflict when placed into the same instruction group. This 20567 option is disabled by default. 20568 20569'-msel-sched-dont-check-control-spec' 20570 Generate checks for control speculation in selective scheduling. 20571 This flag is disabled by default. 20572 20573'-msched-max-memory-insns=MAX-INSNS' 20574 Limit on the number of memory insns per instruction group, giving 20575 lower priority to subsequent memory insns attempting to schedule in 20576 the same instruction group. Frequently useful to prevent cache 20577 bank conflicts. The default value is 1. 20578 20579'-msched-max-memory-insns-hard-limit' 20580 Makes the limit specified by 'msched-max-memory-insns' a hard 20581 limit, disallowing more than that number in an instruction group. 20582 Otherwise, the limit is "soft", meaning that non-memory operations 20583 are preferred when the limit is reached, but memory operations may 20584 still be scheduled. 20585 20586 20587File: gcc.info, Node: LM32 Options, Next: M32C Options, Prev: IA-64 Options, Up: Submodel Options 20588 205893.19.22 LM32 Options 20590-------------------- 20591 20592These '-m' options are defined for the LatticeMico32 architecture: 20593 20594'-mbarrel-shift-enabled' 20595 Enable barrel-shift instructions. 20596 20597'-mdivide-enabled' 20598 Enable divide and modulus instructions. 20599 20600'-mmultiply-enabled' 20601 Enable multiply instructions. 20602 20603'-msign-extend-enabled' 20604 Enable sign extend instructions. 20605 20606'-muser-enabled' 20607 Enable user-defined instructions. 20608 20609 20610File: gcc.info, Node: M32C Options, Next: M32R/D Options, Prev: LM32 Options, Up: Submodel Options 20611 206123.19.23 M32C Options 20613-------------------- 20614 20615'-mcpu=NAME' 20616 Select the CPU for which code is generated. NAME may be one of 20617 'r8c' for the R8C/Tiny series, 'm16c' for the M16C (up to /60) 20618 series, 'm32cm' for the M16C/80 series, or 'm32c' for the M32C/80 20619 series. 20620 20621'-msim' 20622 Specifies that the program will be run on the simulator. This 20623 causes an alternate runtime library to be linked in which supports, 20624 for example, file I/O. You must not use this option when 20625 generating programs that will run on real hardware; you must 20626 provide your own runtime library for whatever I/O functions are 20627 needed. 20628 20629'-memregs=NUMBER' 20630 Specifies the number of memory-based pseudo-registers GCC uses 20631 during code generation. These pseudo-registers are used like real 20632 registers, so there is a tradeoff between GCC's ability to fit the 20633 code into available registers, and the performance penalty of using 20634 memory instead of registers. Note that all modules in a program 20635 must be compiled with the same value for this option. Because of 20636 that, you must not use this option with GCC's default runtime 20637 libraries. 20638 20639 20640File: gcc.info, Node: M32R/D Options, Next: M680x0 Options, Prev: M32C Options, Up: Submodel Options 20641 206423.19.24 M32R/D Options 20643---------------------- 20644 20645These '-m' options are defined for Renesas M32R/D architectures: 20646 20647'-m32r2' 20648 Generate code for the M32R/2. 20649 20650'-m32rx' 20651 Generate code for the M32R/X. 20652 20653'-m32r' 20654 Generate code for the M32R. This is the default. 20655 20656'-mmodel=small' 20657 Assume all objects live in the lower 16MB of memory (so that their 20658 addresses can be loaded with the 'ld24' instruction), and assume 20659 all subroutines are reachable with the 'bl' instruction. This is 20660 the default. 20661 20662 The addressability of a particular object can be set with the 20663 'model' attribute. 20664 20665'-mmodel=medium' 20666 Assume objects may be anywhere in the 32-bit address space (the 20667 compiler generates 'seth/add3' instructions to load their 20668 addresses), and assume all subroutines are reachable with the 'bl' 20669 instruction. 20670 20671'-mmodel=large' 20672 Assume objects may be anywhere in the 32-bit address space (the 20673 compiler generates 'seth/add3' instructions to load their 20674 addresses), and assume subroutines may not be reachable with the 20675 'bl' instruction (the compiler generates the much slower 20676 'seth/add3/jl' instruction sequence). 20677 20678'-msdata=none' 20679 Disable use of the small data area. Variables are put into one of 20680 '.data', '.bss', or '.rodata' (unless the 'section' attribute has 20681 been specified). This is the default. 20682 20683 The small data area consists of sections '.sdata' and '.sbss'. 20684 Objects may be explicitly put in the small data area with the 20685 'section' attribute using one of these sections. 20686 20687'-msdata=sdata' 20688 Put small global and static data in the small data area, but do not 20689 generate special code to reference them. 20690 20691'-msdata=use' 20692 Put small global and static data in the small data area, and 20693 generate special instructions to reference them. 20694 20695'-G NUM' 20696 Put global and static objects less than or equal to NUM bytes into 20697 the small data or BSS sections instead of the normal data or BSS 20698 sections. The default value of NUM is 8. The '-msdata' option 20699 must be set to one of 'sdata' or 'use' for this option to have any 20700 effect. 20701 20702 All modules should be compiled with the same '-G NUM' value. 20703 Compiling with different values of NUM may or may not work; if it 20704 doesn't the linker gives an error message--incorrect code is not 20705 generated. 20706 20707'-mdebug' 20708 Makes the M32R-specific code in the compiler display some 20709 statistics that might help in debugging programs. 20710 20711'-malign-loops' 20712 Align all loops to a 32-byte boundary. 20713 20714'-mno-align-loops' 20715 Do not enforce a 32-byte alignment for loops. This is the default. 20716 20717'-missue-rate=NUMBER' 20718 Issue NUMBER instructions per cycle. NUMBER can only be 1 or 2. 20719 20720'-mbranch-cost=NUMBER' 20721 NUMBER can only be 1 or 2. If it is 1 then branches are preferred 20722 over conditional code, if it is 2, then the opposite applies. 20723 20724'-mflush-trap=NUMBER' 20725 Specifies the trap number to use to flush the cache. The default 20726 is 12. Valid numbers are between 0 and 15 inclusive. 20727 20728'-mno-flush-trap' 20729 Specifies that the cache cannot be flushed by using a trap. 20730 20731'-mflush-func=NAME' 20732 Specifies the name of the operating system function to call to 20733 flush the cache. The default is '_flush_cache', but a function 20734 call is only used if a trap is not available. 20735 20736'-mno-flush-func' 20737 Indicates that there is no OS function for flushing the cache. 20738 20739 20740File: gcc.info, Node: M680x0 Options, Next: MCore Options, Prev: M32R/D Options, Up: Submodel Options 20741 207423.19.25 M680x0 Options 20743---------------------- 20744 20745These are the '-m' options defined for M680x0 and ColdFire processors. 20746The default settings depend on which architecture was selected when the 20747compiler was configured; the defaults for the most common choices are 20748given below. 20749 20750'-march=ARCH' 20751 Generate code for a specific M680x0 or ColdFire instruction set 20752 architecture. Permissible values of ARCH for M680x0 architectures 20753 are: '68000', '68010', '68020', '68030', '68040', '68060' and 20754 'cpu32'. ColdFire architectures are selected according to 20755 Freescale's ISA classification and the permissible values are: 20756 'isaa', 'isaaplus', 'isab' and 'isac'. 20757 20758 GCC defines a macro '__mcfARCH__' whenever it is generating code 20759 for a ColdFire target. The ARCH in this macro is one of the 20760 '-march' arguments given above. 20761 20762 When used together, '-march' and '-mtune' select code that runs on 20763 a family of similar processors but that is optimized for a 20764 particular microarchitecture. 20765 20766'-mcpu=CPU' 20767 Generate code for a specific M680x0 or ColdFire processor. The 20768 M680x0 CPUs are: '68000', '68010', '68020', '68030', '68040', 20769 '68060', '68302', '68332' and 'cpu32'. The ColdFire CPUs are given 20770 by the table below, which also classifies the CPUs into families: 20771 20772 *Family* *'-mcpu' arguments* 20773 '51' '51' '51ac' '51ag' '51cn' '51em' '51je' '51jf' '51jg' 20774 '51jm' '51mm' '51qe' '51qm' 20775 '5206' '5202' '5204' '5206' 20776 '5206e' '5206e' 20777 '5208' '5207' '5208' 20778 '5211a' '5210a' '5211a' 20779 '5213' '5211' '5212' '5213' 20780 '5216' '5214' '5216' 20781 '52235' '52230' '52231' '52232' '52233' '52234' '52235' 20782 '5225' '5224' '5225' 20783 '52259' '52252' '52254' '52255' '52256' '52258' '52259' 20784 '5235' '5232' '5233' '5234' '5235' '523x' 20785 '5249' '5249' 20786 '5250' '5250' 20787 '5271' '5270' '5271' 20788 '5272' '5272' 20789 '5275' '5274' '5275' 20790 '5282' '5280' '5281' '5282' '528x' 20791 '53017' '53011' '53012' '53013' '53014' '53015' '53016' '53017' 20792 '5307' '5307' 20793 '5329' '5327' '5328' '5329' '532x' 20794 '5373' '5372' '5373' '537x' 20795 '5407' '5407' 20796 '5475' '5470' '5471' '5472' '5473' '5474' '5475' '547x' '5480' 20797 '5481' '5482' '5483' '5484' '5485' 20798 20799 '-mcpu=CPU' overrides '-march=ARCH' if ARCH is compatible with CPU. 20800 Other combinations of '-mcpu' and '-march' are rejected. 20801 20802 GCC defines the macro '__mcf_cpu_CPU' when ColdFire target CPU is 20803 selected. It also defines '__mcf_family_FAMILY', where the value 20804 of FAMILY is given by the table above. 20805 20806'-mtune=TUNE' 20807 Tune the code for a particular microarchitecture within the 20808 constraints set by '-march' and '-mcpu'. The M680x0 20809 microarchitectures are: '68000', '68010', '68020', '68030', 20810 '68040', '68060' and 'cpu32'. The ColdFire microarchitectures are: 20811 'cfv1', 'cfv2', 'cfv3', 'cfv4' and 'cfv4e'. 20812 20813 You can also use '-mtune=68020-40' for code that needs to run 20814 relatively well on 68020, 68030 and 68040 targets. 20815 '-mtune=68020-60' is similar but includes 68060 targets as well. 20816 These two options select the same tuning decisions as '-m68020-40' 20817 and '-m68020-60' respectively. 20818 20819 GCC defines the macros '__mcARCH' and '__mcARCH__' when tuning for 20820 680x0 architecture ARCH. It also defines 'mcARCH' unless either 20821 '-ansi' or a non-GNU '-std' option is used. If GCC is tuning for a 20822 range of architectures, as selected by '-mtune=68020-40' or 20823 '-mtune=68020-60', it defines the macros for every architecture in 20824 the range. 20825 20826 GCC also defines the macro '__mUARCH__' when tuning for ColdFire 20827 microarchitecture UARCH, where UARCH is one of the arguments given 20828 above. 20829 20830'-m68000' 20831'-mc68000' 20832 Generate output for a 68000. This is the default when the compiler 20833 is configured for 68000-based systems. It is equivalent to 20834 '-march=68000'. 20835 20836 Use this option for microcontrollers with a 68000 or EC000 core, 20837 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356. 20838 20839'-m68010' 20840 Generate output for a 68010. This is the default when the compiler 20841 is configured for 68010-based systems. It is equivalent to 20842 '-march=68010'. 20843 20844'-m68020' 20845'-mc68020' 20846 Generate output for a 68020. This is the default when the compiler 20847 is configured for 68020-based systems. It is equivalent to 20848 '-march=68020'. 20849 20850'-m68030' 20851 Generate output for a 68030. This is the default when the compiler 20852 is configured for 68030-based systems. It is equivalent to 20853 '-march=68030'. 20854 20855'-m68040' 20856 Generate output for a 68040. This is the default when the compiler 20857 is configured for 68040-based systems. It is equivalent to 20858 '-march=68040'. 20859 20860 This option inhibits the use of 68881/68882 instructions that have 20861 to be emulated by software on the 68040. Use this option if your 20862 68040 does not have code to emulate those instructions. 20863 20864'-m68060' 20865 Generate output for a 68060. This is the default when the compiler 20866 is configured for 68060-based systems. It is equivalent to 20867 '-march=68060'. 20868 20869 This option inhibits the use of 68020 and 68881/68882 instructions 20870 that have to be emulated by software on the 68060. Use this option 20871 if your 68060 does not have code to emulate those instructions. 20872 20873'-mcpu32' 20874 Generate output for a CPU32. This is the default when the compiler 20875 is configured for CPU32-based systems. It is equivalent to 20876 '-march=cpu32'. 20877 20878 Use this option for microcontrollers with a CPU32 or CPU32+ core, 20879 including the 68330, 68331, 68332, 68333, 68334, 68336, 68340, 20880 68341, 68349 and 68360. 20881 20882'-m5200' 20883 Generate output for a 520X ColdFire CPU. This is the default when 20884 the compiler is configured for 520X-based systems. It is 20885 equivalent to '-mcpu=5206', and is now deprecated in favor of that 20886 option. 20887 20888 Use this option for microcontroller with a 5200 core, including the 20889 MCF5202, MCF5203, MCF5204 and MCF5206. 20890 20891'-m5206e' 20892 Generate output for a 5206e ColdFire CPU. The option is now 20893 deprecated in favor of the equivalent '-mcpu=5206e'. 20894 20895'-m528x' 20896 Generate output for a member of the ColdFire 528X family. The 20897 option is now deprecated in favor of the equivalent '-mcpu=528x'. 20898 20899'-m5307' 20900 Generate output for a ColdFire 5307 CPU. The option is now 20901 deprecated in favor of the equivalent '-mcpu=5307'. 20902 20903'-m5407' 20904 Generate output for a ColdFire 5407 CPU. The option is now 20905 deprecated in favor of the equivalent '-mcpu=5407'. 20906 20907'-mcfv4e' 20908 Generate output for a ColdFire V4e family CPU (e.g. 547x/548x). 20909 This includes use of hardware floating-point instructions. The 20910 option is equivalent to '-mcpu=547x', and is now deprecated in 20911 favor of that option. 20912 20913'-m68020-40' 20914 Generate output for a 68040, without using any of the new 20915 instructions. This results in code that can run relatively 20916 efficiently on either a 68020/68881 or a 68030 or a 68040. The 20917 generated code does use the 68881 instructions that are emulated on 20918 the 68040. 20919 20920 The option is equivalent to '-march=68020' '-mtune=68020-40'. 20921 20922'-m68020-60' 20923 Generate output for a 68060, without using any of the new 20924 instructions. This results in code that can run relatively 20925 efficiently on either a 68020/68881 or a 68030 or a 68040. The 20926 generated code does use the 68881 instructions that are emulated on 20927 the 68060. 20928 20929 The option is equivalent to '-march=68020' '-mtune=68020-60'. 20930 20931'-mhard-float' 20932'-m68881' 20933 Generate floating-point instructions. This is the default for 20934 68020 and above, and for ColdFire devices that have an FPU. It 20935 defines the macro '__HAVE_68881__' on M680x0 targets and 20936 '__mcffpu__' on ColdFire targets. 20937 20938'-msoft-float' 20939 Do not generate floating-point instructions; use library calls 20940 instead. This is the default for 68000, 68010, and 68832 targets. 20941 It is also the default for ColdFire devices that have no FPU. 20942 20943'-mdiv' 20944'-mno-div' 20945 Generate (do not generate) ColdFire hardware divide and remainder 20946 instructions. If '-march' is used without '-mcpu', the default is 20947 "on" for ColdFire architectures and "off" for M680x0 architectures. 20948 Otherwise, the default is taken from the target CPU (either the 20949 default CPU, or the one specified by '-mcpu'). For example, the 20950 default is "off" for '-mcpu=5206' and "on" for '-mcpu=5206e'. 20951 20952 GCC defines the macro '__mcfhwdiv__' when this option is enabled. 20953 20954'-mshort' 20955 Consider type 'int' to be 16 bits wide, like 'short int'. 20956 Additionally, parameters passed on the stack are also aligned to a 20957 16-bit boundary even on targets whose API mandates promotion to 20958 32-bit. 20959 20960'-mno-short' 20961 Do not consider type 'int' to be 16 bits wide. This is the 20962 default. 20963 20964'-mnobitfield' 20965'-mno-bitfield' 20966 Do not use the bit-field instructions. The '-m68000', '-mcpu32' 20967 and '-m5200' options imply '-mnobitfield'. 20968 20969'-mbitfield' 20970 Do use the bit-field instructions. The '-m68020' option implies 20971 '-mbitfield'. This is the default if you use a configuration 20972 designed for a 68020. 20973 20974'-mrtd' 20975 Use a different function-calling convention, in which functions 20976 that take a fixed number of arguments return with the 'rtd' 20977 instruction, which pops their arguments while returning. This 20978 saves one instruction in the caller since there is no need to pop 20979 the arguments there. 20980 20981 This calling convention is incompatible with the one normally used 20982 on Unix, so you cannot use it if you need to call libraries 20983 compiled with the Unix compiler. 20984 20985 Also, you must provide function prototypes for all functions that 20986 take variable numbers of arguments (including 'printf'); otherwise 20987 incorrect code is generated for calls to those functions. 20988 20989 In addition, seriously incorrect code results if you call a 20990 function with too many arguments. (Normally, extra arguments are 20991 harmlessly ignored.) 20992 20993 The 'rtd' instruction is supported by the 68010, 68020, 68030, 20994 68040, 68060 and CPU32 processors, but not by the 68000 or 5200. 20995 20996 The default is '-mno-rtd'. 20997 20998'-malign-int' 20999'-mno-align-int' 21000 Control whether GCC aligns 'int', 'long', 'long long', 'float', 21001 'double', and 'long double' variables on a 32-bit boundary 21002 ('-malign-int') or a 16-bit boundary ('-mno-align-int'). Aligning 21003 variables on 32-bit boundaries produces code that runs somewhat 21004 faster on processors with 32-bit busses at the expense of more 21005 memory. 21006 21007 *Warning:* if you use the '-malign-int' switch, GCC aligns 21008 structures containing the above types differently than most 21009 published application binary interface specifications for the m68k. 21010 21011 Use the pc-relative addressing mode of the 68000 directly, instead 21012 of using a global offset table. At present, this option implies 21013 '-fpic', allowing at most a 16-bit offset for pc-relative 21014 addressing. '-fPIC' is not presently supported with '-mpcrel', 21015 though this could be supported for 68020 and higher processors. 21016 21017'-mno-strict-align' 21018'-mstrict-align' 21019 Do not (do) assume that unaligned memory references are handled by 21020 the system. 21021 21022'-msep-data' 21023 Generate code that allows the data segment to be located in a 21024 different area of memory from the text segment. This allows for 21025 execute-in-place in an environment without virtual memory 21026 management. This option implies '-fPIC'. 21027 21028'-mno-sep-data' 21029 Generate code that assumes that the data segment follows the text 21030 segment. This is the default. 21031 21032'-mid-shared-library' 21033 Generate code that supports shared libraries via the library ID 21034 method. This allows for execute-in-place and shared libraries in 21035 an environment without virtual memory management. This option 21036 implies '-fPIC'. 21037 21038'-mno-id-shared-library' 21039 Generate code that doesn't assume ID-based shared libraries are 21040 being used. This is the default. 21041 21042'-mshared-library-id=n' 21043 Specifies the identification number of the ID-based shared library 21044 being compiled. Specifying a value of 0 generates more compact 21045 code; specifying other values forces the allocation of that number 21046 to the current library, but is no more space- or time-efficient 21047 than omitting this option. 21048 21049'-mxgot' 21050'-mno-xgot' 21051 When generating position-independent code for ColdFire, generate 21052 code that works if the GOT has more than 8192 entries. This code 21053 is larger and slower than code generated without this option. On 21054 M680x0 processors, this option is not needed; '-fPIC' suffices. 21055 21056 GCC normally uses a single instruction to load values from the GOT. 21057 While this is relatively efficient, it only works if the GOT is 21058 smaller than about 64k. Anything larger causes the linker to 21059 report an error such as: 21060 21061 relocation truncated to fit: R_68K_GOT16O foobar 21062 21063 If this happens, you should recompile your code with '-mxgot'. It 21064 should then work with very large GOTs. However, code generated 21065 with '-mxgot' is less efficient, since it takes 4 instructions to 21066 fetch the value of a global symbol. 21067 21068 Note that some linkers, including newer versions of the GNU linker, 21069 can create multiple GOTs and sort GOT entries. If you have such a 21070 linker, you should only need to use '-mxgot' when compiling a 21071 single object file that accesses more than 8192 GOT entries. Very 21072 few do. 21073 21074 These options have no effect unless GCC is generating 21075 position-independent code. 21076 21077'-mlong-jump-table-offsets' 21078 Use 32-bit offsets in 'switch' tables. The default is to use 21079 16-bit offsets. 21080 21081 21082File: gcc.info, Node: MCore Options, Next: MeP Options, Prev: M680x0 Options, Up: Submodel Options 21083 210843.19.26 MCore Options 21085--------------------- 21086 21087These are the '-m' options defined for the Motorola M*Core processors. 21088 21089'-mhardlit' 21090'-mno-hardlit' 21091 Inline constants into the code stream if it can be done in two 21092 instructions or less. 21093 21094'-mdiv' 21095'-mno-div' 21096 Use the divide instruction. (Enabled by default). 21097 21098'-mrelax-immediate' 21099'-mno-relax-immediate' 21100 Allow arbitrary-sized immediates in bit operations. 21101 21102'-mwide-bitfields' 21103'-mno-wide-bitfields' 21104 Always treat bit-fields as 'int'-sized. 21105 21106'-m4byte-functions' 21107'-mno-4byte-functions' 21108 Force all functions to be aligned to a 4-byte boundary. 21109 21110'-mcallgraph-data' 21111'-mno-callgraph-data' 21112 Emit callgraph information. 21113 21114'-mslow-bytes' 21115'-mno-slow-bytes' 21116 Prefer word access when reading byte quantities. 21117 21118'-mlittle-endian' 21119'-mbig-endian' 21120 Generate code for a little-endian target. 21121 21122'-m210' 21123'-m340' 21124 Generate code for the 210 processor. 21125 21126'-mno-lsim' 21127 Assume that runtime support has been provided and so omit the 21128 simulator library ('libsim.a)' from the linker command line. 21129 21130'-mstack-increment=SIZE' 21131 Set the maximum amount for a single stack increment operation. 21132 Large values can increase the speed of programs that contain 21133 functions that need a large amount of stack space, but they can 21134 also trigger a segmentation fault if the stack is extended too 21135 much. The default value is 0x1000. 21136 21137 21138File: gcc.info, Node: MeP Options, Next: MicroBlaze Options, Prev: MCore Options, Up: Submodel Options 21139 211403.19.27 MeP Options 21141------------------- 21142 21143'-mabsdiff' 21144 Enables the 'abs' instruction, which is the absolute difference 21145 between two registers. 21146 21147'-mall-opts' 21148 Enables all the optional instructions--average, multiply, divide, 21149 bit operations, leading zero, absolute difference, min/max, clip, 21150 and saturation. 21151 21152'-maverage' 21153 Enables the 'ave' instruction, which computes the average of two 21154 registers. 21155 21156'-mbased=N' 21157 Variables of size N bytes or smaller are placed in the '.based' 21158 section by default. Based variables use the '$tp' register as a 21159 base register, and there is a 128-byte limit to the '.based' 21160 section. 21161 21162'-mbitops' 21163 Enables the bit operation instructions--bit test ('btstm'), set 21164 ('bsetm'), clear ('bclrm'), invert ('bnotm'), and test-and-set 21165 ('tas'). 21166 21167'-mc=NAME' 21168 Selects which section constant data is placed in. NAME may be 21169 'tiny', 'near', or 'far'. 21170 21171'-mclip' 21172 Enables the 'clip' instruction. Note that '-mclip' is not useful 21173 unless you also provide '-mminmax'. 21174 21175'-mconfig=NAME' 21176 Selects one of the built-in core configurations. Each MeP chip has 21177 one or more modules in it; each module has a core CPU and a variety 21178 of coprocessors, optional instructions, and peripherals. The 21179 'MeP-Integrator' tool, not part of GCC, provides these 21180 configurations through this option; using this option is the same 21181 as using all the corresponding command-line options. The default 21182 configuration is 'default'. 21183 21184'-mcop' 21185 Enables the coprocessor instructions. By default, this is a 32-bit 21186 coprocessor. Note that the coprocessor is normally enabled via the 21187 '-mconfig=' option. 21188 21189'-mcop32' 21190 Enables the 32-bit coprocessor's instructions. 21191 21192'-mcop64' 21193 Enables the 64-bit coprocessor's instructions. 21194 21195'-mivc2' 21196 Enables IVC2 scheduling. IVC2 is a 64-bit VLIW coprocessor. 21197 21198'-mdc' 21199 Causes constant variables to be placed in the '.near' section. 21200 21201'-mdiv' 21202 Enables the 'div' and 'divu' instructions. 21203 21204'-meb' 21205 Generate big-endian code. 21206 21207'-mel' 21208 Generate little-endian code. 21209 21210'-mio-volatile' 21211 Tells the compiler that any variable marked with the 'io' attribute 21212 is to be considered volatile. 21213 21214'-ml' 21215 Causes variables to be assigned to the '.far' section by default. 21216 21217'-mleadz' 21218 Enables the 'leadz' (leading zero) instruction. 21219 21220'-mm' 21221 Causes variables to be assigned to the '.near' section by default. 21222 21223'-mminmax' 21224 Enables the 'min' and 'max' instructions. 21225 21226'-mmult' 21227 Enables the multiplication and multiply-accumulate instructions. 21228 21229'-mno-opts' 21230 Disables all the optional instructions enabled by '-mall-opts'. 21231 21232'-mrepeat' 21233 Enables the 'repeat' and 'erepeat' instructions, used for 21234 low-overhead looping. 21235 21236'-ms' 21237 Causes all variables to default to the '.tiny' section. Note that 21238 there is a 65536-byte limit to this section. Accesses to these 21239 variables use the '%gp' base register. 21240 21241'-msatur' 21242 Enables the saturation instructions. Note that the compiler does 21243 not currently generate these itself, but this option is included 21244 for compatibility with other tools, like 'as'. 21245 21246'-msdram' 21247 Link the SDRAM-based runtime instead of the default ROM-based 21248 runtime. 21249 21250'-msim' 21251 Link the simulator run-time libraries. 21252 21253'-msimnovec' 21254 Link the simulator runtime libraries, excluding built-in support 21255 for reset and exception vectors and tables. 21256 21257'-mtf' 21258 Causes all functions to default to the '.far' section. Without 21259 this option, functions default to the '.near' section. 21260 21261'-mtiny=N' 21262 Variables that are N bytes or smaller are allocated to the '.tiny' 21263 section. These variables use the '$gp' base register. The default 21264 for this option is 4, but note that there's a 65536-byte limit to 21265 the '.tiny' section. 21266 21267 21268File: gcc.info, Node: MicroBlaze Options, Next: MIPS Options, Prev: MeP Options, Up: Submodel Options 21269 212703.19.28 MicroBlaze Options 21271-------------------------- 21272 21273'-msoft-float' 21274 Use software emulation for floating point (default). 21275 21276'-mhard-float' 21277 Use hardware floating-point instructions. 21278 21279'-mmemcpy' 21280 Do not optimize block moves, use 'memcpy'. 21281 21282'-mno-clearbss' 21283 This option is deprecated. Use '-fno-zero-initialized-in-bss' 21284 instead. 21285 21286'-mcpu=CPU-TYPE' 21287 Use features of, and schedule code for, the given CPU. Supported 21288 values are in the format 'vX.YY.Z', where X is a major version, YY 21289 is the minor version, and Z is compatibility code. Example values 21290 are 'v3.00.a', 'v4.00.b', 'v5.00.a', 'v5.00.b', 'v6.00.a'. 21291 21292'-mxl-soft-mul' 21293 Use software multiply emulation (default). 21294 21295'-mxl-soft-div' 21296 Use software emulation for divides (default). 21297 21298'-mxl-barrel-shift' 21299 Use the hardware barrel shifter. 21300 21301'-mxl-pattern-compare' 21302 Use pattern compare instructions. 21303 21304'-msmall-divides' 21305 Use table lookup optimization for small signed integer divisions. 21306 21307'-mxl-stack-check' 21308 This option is deprecated. Use '-fstack-check' instead. 21309 21310'-mxl-gp-opt' 21311 Use GP-relative '.sdata'/'.sbss' sections. 21312 21313'-mxl-multiply-high' 21314 Use multiply high instructions for high part of 32x32 multiply. 21315 21316'-mxl-float-convert' 21317 Use hardware floating-point conversion instructions. 21318 21319'-mxl-float-sqrt' 21320 Use hardware floating-point square root instruction. 21321 21322'-mbig-endian' 21323 Generate code for a big-endian target. 21324 21325'-mlittle-endian' 21326 Generate code for a little-endian target. 21327 21328'-mxl-reorder' 21329 Use reorder instructions (swap and byte reversed load/store). 21330 21331'-mxl-mode-APP-MODEL' 21332 Select application model APP-MODEL. Valid models are 21333 'executable' 21334 normal executable (default), uses startup code 'crt0.o'. 21335 21336 '-mpic-data-is-text-relative' 21337 Assume that the displacement between the text and data 21338 segments is fixed at static link time. This allows data to be 21339 referenced by offset from start of text address instead of GOT 21340 since PC-relative addressing is not supported. 21341 21342 'xmdstub' 21343 for use with Xilinx Microprocessor Debugger (XMD) based 21344 software intrusive debug agent called xmdstub. This uses 21345 startup file 'crt1.o' and sets the start address of the 21346 program to 0x800. 21347 21348 'bootstrap' 21349 for applications that are loaded using a bootloader. This 21350 model uses startup file 'crt2.o' which does not contain a 21351 processor reset vector handler. This is suitable for 21352 transferring control on a processor reset to the bootloader 21353 rather than the application. 21354 21355 'novectors' 21356 for applications that do not require any of the MicroBlaze 21357 vectors. This option may be useful for applications running 21358 within a monitoring application. This model uses 'crt3.o' as 21359 a startup file. 21360 21361 Option '-xl-mode-APP-MODEL' is a deprecated alias for 21362 '-mxl-mode-APP-MODEL'. 21363 21364 21365File: gcc.info, Node: MIPS Options, Next: MMIX Options, Prev: MicroBlaze Options, Up: Submodel Options 21366 213673.19.29 MIPS Options 21368-------------------- 21369 21370'-EB' 21371 Generate big-endian code. 21372 21373'-EL' 21374 Generate little-endian code. This is the default for 'mips*el-*-*' 21375 configurations. 21376 21377'-march=ARCH' 21378 Generate code that runs on ARCH, which can be the name of a generic 21379 MIPS ISA, or the name of a particular processor. The ISA names 21380 are: 'mips1', 'mips2', 'mips3', 'mips4', 'mips32', 'mips32r2', 21381 'mips32r3', 'mips32r5', 'mips32r6', 'mips64', 'mips64r2', 21382 'mips64r3', 'mips64r5' and 'mips64r6'. The processor names are: 21383 '4kc', '4km', '4kp', '4ksc', '4kec', '4kem', '4kep', '4ksd', '5kc', 21384 '5kf', '20kc', '24kc', '24kf2_1', '24kf1_1', '24kec', '24kef2_1', 21385 '24kef1_1', '34kc', '34kf2_1', '34kf1_1', '34kn', '74kc', 21386 '74kf2_1', '74kf1_1', '74kf3_2', '1004kc', '1004kf2_1', 21387 '1004kf1_1', 'i6400', 'i6500', 'interaptiv', 'loongson2e', 21388 'loongson2f', 'loongson3a', 'gs464', 'gs464e', 'gs264e', 'm4k', 21389 'm14k', 'm14kc', 'm14ke', 'm14kec', 'm5100', 'm5101', 'octeon', 21390 'octeon+', 'octeon2', 'octeon3', 'orion', 'p5600', 'p6600', 21391 'r2000', 'r3000', 'r3900', 'r4000', 'r4400', 'r4600', 'r4650', 21392 'r4700', 'r5900', 'r6000', 'r8000', 'rm7000', 'rm9000', 'r10000', 21393 'r12000', 'r14000', 'r16000', 'sb1', 'sr71000', 'vr4100', 'vr4111', 21394 'vr4120', 'vr4130', 'vr4300', 'vr5000', 'vr5400', 'vr5500', 'xlr' 21395 and 'xlp'. The special value 'from-abi' selects the most 21396 compatible architecture for the selected ABI (that is, 'mips1' for 21397 32-bit ABIs and 'mips3' for 64-bit ABIs). 21398 21399 The native Linux/GNU toolchain also supports the value 'native', 21400 which selects the best architecture option for the host processor. 21401 '-march=native' has no effect if GCC does not recognize the 21402 processor. 21403 21404 In processor names, a final '000' can be abbreviated as 'k' (for 21405 example, '-march=r2k'). Prefixes are optional, and 'vr' may be 21406 written 'r'. 21407 21408 Names of the form 'Nf2_1' refer to processors with FPUs clocked at 21409 half the rate of the core, names of the form 'Nf1_1' refer to 21410 processors with FPUs clocked at the same rate as the core, and 21411 names of the form 'Nf3_2' refer to processors with FPUs clocked a 21412 ratio of 3:2 with respect to the core. For compatibility reasons, 21413 'Nf' is accepted as a synonym for 'Nf2_1' while 'Nx' and 'Bfx' are 21414 accepted as synonyms for 'Nf1_1'. 21415 21416 GCC defines two macros based on the value of this option. The 21417 first is '_MIPS_ARCH', which gives the name of target architecture, 21418 as a string. The second has the form '_MIPS_ARCH_FOO', where FOO 21419 is the capitalized value of '_MIPS_ARCH'. For example, 21420 '-march=r2000' sets '_MIPS_ARCH' to '"r2000"' and defines the macro 21421 '_MIPS_ARCH_R2000'. 21422 21423 Note that the '_MIPS_ARCH' macro uses the processor names given 21424 above. In other words, it has the full prefix and does not 21425 abbreviate '000' as 'k'. In the case of 'from-abi', the macro 21426 names the resolved architecture (either '"mips1"' or '"mips3"'). 21427 It names the default architecture when no '-march' option is given. 21428 21429'-mtune=ARCH' 21430 Optimize for ARCH. Among other things, this option controls the 21431 way instructions are scheduled, and the perceived cost of 21432 arithmetic operations. The list of ARCH values is the same as for 21433 '-march'. 21434 21435 When this option is not used, GCC optimizes for the processor 21436 specified by '-march'. By using '-march' and '-mtune' together, it 21437 is possible to generate code that runs on a family of processors, 21438 but optimize the code for one particular member of that family. 21439 21440 '-mtune' defines the macros '_MIPS_TUNE' and '_MIPS_TUNE_FOO', 21441 which work in the same way as the '-march' ones described above. 21442 21443'-mips1' 21444 Equivalent to '-march=mips1'. 21445 21446'-mips2' 21447 Equivalent to '-march=mips2'. 21448 21449'-mips3' 21450 Equivalent to '-march=mips3'. 21451 21452'-mips4' 21453 Equivalent to '-march=mips4'. 21454 21455'-mips32' 21456 Equivalent to '-march=mips32'. 21457 21458'-mips32r3' 21459 Equivalent to '-march=mips32r3'. 21460 21461'-mips32r5' 21462 Equivalent to '-march=mips32r5'. 21463 21464'-mips32r6' 21465 Equivalent to '-march=mips32r6'. 21466 21467'-mips64' 21468 Equivalent to '-march=mips64'. 21469 21470'-mips64r2' 21471 Equivalent to '-march=mips64r2'. 21472 21473'-mips64r3' 21474 Equivalent to '-march=mips64r3'. 21475 21476'-mips64r5' 21477 Equivalent to '-march=mips64r5'. 21478 21479'-mips64r6' 21480 Equivalent to '-march=mips64r6'. 21481 21482'-mips16' 21483'-mno-mips16' 21484 Generate (do not generate) MIPS16 code. If GCC is targeting a 21485 MIPS32 or MIPS64 architecture, it makes use of the MIPS16e ASE. 21486 21487 MIPS16 code generation can also be controlled on a per-function 21488 basis by means of 'mips16' and 'nomips16' attributes. *Note 21489 Function Attributes::, for more information. 21490 21491'-mflip-mips16' 21492 Generate MIPS16 code on alternating functions. This option is 21493 provided for regression testing of mixed MIPS16/non-MIPS16 code 21494 generation, and is not intended for ordinary use in compiling user 21495 code. 21496 21497'-minterlink-compressed' 21498'-mno-interlink-compressed' 21499 Require (do not require) that code using the standard 21500 (uncompressed) MIPS ISA be link-compatible with MIPS16 and 21501 microMIPS code, and vice versa. 21502 21503 For example, code using the standard ISA encoding cannot jump 21504 directly to MIPS16 or microMIPS code; it must either use a call or 21505 an indirect jump. '-minterlink-compressed' therefore disables 21506 direct jumps unless GCC knows that the target of the jump is not 21507 compressed. 21508 21509'-minterlink-mips16' 21510'-mno-interlink-mips16' 21511 Aliases of '-minterlink-compressed' and 21512 '-mno-interlink-compressed'. These options predate the microMIPS 21513 ASE and are retained for backwards compatibility. 21514 21515'-mabi=32' 21516'-mabi=o64' 21517'-mabi=n32' 21518'-mabi=64' 21519'-mabi=eabi' 21520 Generate code for the given ABI. 21521 21522 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally 21523 generates 64-bit code when you select a 64-bit architecture, but 21524 you can use '-mgp32' to get 32-bit code instead. 21525 21526 For information about the O64 ABI, see 21527 <http://gcc.gnu.org/projects/mipso64-abi.html>. 21528 21529 GCC supports a variant of the o32 ABI in which floating-point 21530 registers are 64 rather than 32 bits wide. You can select this 21531 combination with '-mabi=32' '-mfp64'. This ABI relies on the 21532 'mthc1' and 'mfhc1' instructions and is therefore only supported 21533 for MIPS32R2, MIPS32R3 and MIPS32R5 processors. 21534 21535 The register assignments for arguments and return values remain the 21536 same, but each scalar value is passed in a single 64-bit register 21537 rather than a pair of 32-bit registers. For example, scalar 21538 floating-point values are returned in '$f0' only, not a '$f0'/'$f1' 21539 pair. The set of call-saved registers also remains the same in 21540 that the even-numbered double-precision registers are saved. 21541 21542 Two additional variants of the o32 ABI are supported to enable a 21543 transition from 32-bit to 64-bit registers. These are FPXX 21544 ('-mfpxx') and FP64A ('-mfp64' '-mno-odd-spreg'). The FPXX 21545 extension mandates that all code must execute correctly when run 21546 using 32-bit or 64-bit registers. The code can be interlinked with 21547 either FP32 or FP64, but not both. The FP64A extension is similar 21548 to the FP64 extension but forbids the use of odd-numbered 21549 single-precision registers. This can be used in conjunction with 21550 the 'FRE' mode of FPUs in MIPS32R5 processors and allows both FP32 21551 and FP64A code to interlink and run in the same process without 21552 changing FPU modes. 21553 21554'-mabicalls' 21555'-mno-abicalls' 21556 Generate (do not generate) code that is suitable for SVR4-style 21557 dynamic objects. '-mabicalls' is the default for SVR4-based 21558 systems. 21559 21560'-mshared' 21561'-mno-shared' 21562 Generate (do not generate) code that is fully position-independent, 21563 and that can therefore be linked into shared libraries. This 21564 option only affects '-mabicalls'. 21565 21566 All '-mabicalls' code has traditionally been position-independent, 21567 regardless of options like '-fPIC' and '-fpic'. However, as an 21568 extension, the GNU toolchain allows executables to use absolute 21569 accesses for locally-binding symbols. It can also use shorter GP 21570 initialization sequences and generate direct calls to 21571 locally-defined functions. This mode is selected by '-mno-shared'. 21572 21573 '-mno-shared' depends on binutils 2.16 or higher and generates 21574 objects that can only be linked by the GNU linker. However, the 21575 option does not affect the ABI of the final executable; it only 21576 affects the ABI of relocatable objects. Using '-mno-shared' 21577 generally makes executables both smaller and quicker. 21578 21579 '-mshared' is the default. 21580 21581'-mplt' 21582'-mno-plt' 21583 Assume (do not assume) that the static and dynamic linkers support 21584 PLTs and copy relocations. This option only affects '-mno-shared 21585 -mabicalls'. For the n64 ABI, this option has no effect without 21586 '-msym32'. 21587 21588 You can make '-mplt' the default by configuring GCC with 21589 '--with-mips-plt'. The default is '-mno-plt' otherwise. 21590 21591'-mxgot' 21592'-mno-xgot' 21593 Lift (do not lift) the usual restrictions on the size of the global 21594 offset table. 21595 21596 GCC normally uses a single instruction to load values from the GOT. 21597 While this is relatively efficient, it only works if the GOT is 21598 smaller than about 64k. Anything larger causes the linker to 21599 report an error such as: 21600 21601 relocation truncated to fit: R_MIPS_GOT16 foobar 21602 21603 If this happens, you should recompile your code with '-mxgot'. 21604 This works with very large GOTs, although the code is also less 21605 efficient, since it takes three instructions to fetch the value of 21606 a global symbol. 21607 21608 Note that some linkers can create multiple GOTs. If you have such 21609 a linker, you should only need to use '-mxgot' when a single object 21610 file accesses more than 64k's worth of GOT entries. Very few do. 21611 21612 These options have no effect unless GCC is generating position 21613 independent code. 21614 21615'-mgp32' 21616 Assume that general-purpose registers are 32 bits wide. 21617 21618'-mgp64' 21619 Assume that general-purpose registers are 64 bits wide. 21620 21621'-mfp32' 21622 Assume that floating-point registers are 32 bits wide. 21623 21624'-mfp64' 21625 Assume that floating-point registers are 64 bits wide. 21626 21627'-mfpxx' 21628 Do not assume the width of floating-point registers. 21629 21630'-mhard-float' 21631 Use floating-point coprocessor instructions. 21632 21633'-msoft-float' 21634 Do not use floating-point coprocessor instructions. Implement 21635 floating-point calculations using library calls instead. 21636 21637'-mno-float' 21638 Equivalent to '-msoft-float', but additionally asserts that the 21639 program being compiled does not perform any floating-point 21640 operations. This option is presently supported only by some 21641 bare-metal MIPS configurations, where it may select a special set 21642 of libraries that lack all floating-point support (including, for 21643 example, the floating-point 'printf' formats). If code compiled 21644 with '-mno-float' accidentally contains floating-point operations, 21645 it is likely to suffer a link-time or run-time failure. 21646 21647'-msingle-float' 21648 Assume that the floating-point coprocessor only supports 21649 single-precision operations. 21650 21651'-mdouble-float' 21652 Assume that the floating-point coprocessor supports 21653 double-precision operations. This is the default. 21654 21655'-modd-spreg' 21656'-mno-odd-spreg' 21657 Enable the use of odd-numbered single-precision floating-point 21658 registers for the o32 ABI. This is the default for processors that 21659 are known to support these registers. When using the o32 FPXX ABI, 21660 '-mno-odd-spreg' is set by default. 21661 21662'-mabs=2008' 21663'-mabs=legacy' 21664 These options control the treatment of the special not-a-number 21665 (NaN) IEEE 754 floating-point data with the 'abs.fmt' and 'neg.fmt' 21666 machine instructions. 21667 21668 By default or when '-mabs=legacy' is used the legacy treatment is 21669 selected. In this case these instructions are considered 21670 arithmetic and avoided where correct operation is required and the 21671 input operand might be a NaN. A longer sequence of instructions 21672 that manipulate the sign bit of floating-point datum manually is 21673 used instead unless the '-ffinite-math-only' option has also been 21674 specified. 21675 21676 The '-mabs=2008' option selects the IEEE 754-2008 treatment. In 21677 this case these instructions are considered non-arithmetic and 21678 therefore operating correctly in all cases, including in particular 21679 where the input operand is a NaN. These instructions are therefore 21680 always used for the respective operations. 21681 21682'-mnan=2008' 21683'-mnan=legacy' 21684 These options control the encoding of the special not-a-number 21685 (NaN) IEEE 754 floating-point data. 21686 21687 The '-mnan=legacy' option selects the legacy encoding. In this 21688 case quiet NaNs (qNaNs) are denoted by the first bit of their 21689 trailing significand field being 0, whereas signaling NaNs (sNaNs) 21690 are denoted by the first bit of their trailing significand field 21691 being 1. 21692 21693 The '-mnan=2008' option selects the IEEE 754-2008 encoding. In 21694 this case qNaNs are denoted by the first bit of their trailing 21695 significand field being 1, whereas sNaNs are denoted by the first 21696 bit of their trailing significand field being 0. 21697 21698 The default is '-mnan=legacy' unless GCC has been configured with 21699 '--with-nan=2008'. 21700 21701'-mllsc' 21702'-mno-llsc' 21703 Use (do not use) 'll', 'sc', and 'sync' instructions to implement 21704 atomic memory built-in functions. When neither option is 21705 specified, GCC uses the instructions if the target architecture 21706 supports them. 21707 21708 '-mllsc' is useful if the runtime environment can emulate the 21709 instructions and '-mno-llsc' can be useful when compiling for 21710 nonstandard ISAs. You can make either option the default by 21711 configuring GCC with '--with-llsc' and '--without-llsc' 21712 respectively. '--with-llsc' is the default for some 21713 configurations; see the installation documentation for details. 21714 21715'-mdsp' 21716'-mno-dsp' 21717 Use (do not use) revision 1 of the MIPS DSP ASE. *Note MIPS DSP 21718 Built-in Functions::. This option defines the preprocessor macro 21719 '__mips_dsp'. It also defines '__mips_dsp_rev' to 1. 21720 21721'-mdspr2' 21722'-mno-dspr2' 21723 Use (do not use) revision 2 of the MIPS DSP ASE. *Note MIPS DSP 21724 Built-in Functions::. This option defines the preprocessor macros 21725 '__mips_dsp' and '__mips_dspr2'. It also defines '__mips_dsp_rev' 21726 to 2. 21727 21728'-msmartmips' 21729'-mno-smartmips' 21730 Use (do not use) the MIPS SmartMIPS ASE. 21731 21732'-mpaired-single' 21733'-mno-paired-single' 21734 Use (do not use) paired-single floating-point instructions. *Note 21735 MIPS Paired-Single Support::. This option requires hardware 21736 floating-point support to be enabled. 21737 21738'-mdmx' 21739'-mno-mdmx' 21740 Use (do not use) MIPS Digital Media Extension instructions. This 21741 option can only be used when generating 64-bit code and requires 21742 hardware floating-point support to be enabled. 21743 21744'-mips3d' 21745'-mno-mips3d' 21746 Use (do not use) the MIPS-3D ASE. *Note MIPS-3D Built-in 21747 Functions::. The option '-mips3d' implies '-mpaired-single'. 21748 21749'-mmicromips' 21750'-mno-micromips' 21751 Generate (do not generate) microMIPS code. 21752 21753 MicroMIPS code generation can also be controlled on a per-function 21754 basis by means of 'micromips' and 'nomicromips' attributes. *Note 21755 Function Attributes::, for more information. 21756 21757'-mmt' 21758'-mno-mt' 21759 Use (do not use) MT Multithreading instructions. 21760 21761'-mmcu' 21762'-mno-mcu' 21763 Use (do not use) the MIPS MCU ASE instructions. 21764 21765'-meva' 21766'-mno-eva' 21767 Use (do not use) the MIPS Enhanced Virtual Addressing instructions. 21768 21769'-mvirt' 21770'-mno-virt' 21771 Use (do not use) the MIPS Virtualization (VZ) instructions. 21772 21773'-mxpa' 21774'-mno-xpa' 21775 Use (do not use) the MIPS eXtended Physical Address (XPA) 21776 instructions. 21777 21778'-mcrc' 21779'-mno-crc' 21780 Use (do not use) the MIPS Cyclic Redundancy Check (CRC) 21781 instructions. 21782 21783'-mginv' 21784'-mno-ginv' 21785 Use (do not use) the MIPS Global INValidate (GINV) instructions. 21786 21787'-mloongson-mmi' 21788'-mno-loongson-mmi' 21789 Use (do not use) the MIPS Loongson MultiMedia extensions 21790 Instructions (MMI). 21791 21792'-mloongson-ext' 21793'-mno-loongson-ext' 21794 Use (do not use) the MIPS Loongson EXTensions (EXT) instructions. 21795 21796'-mloongson-ext2' 21797'-mno-loongson-ext2' 21798 Use (do not use) the MIPS Loongson EXTensions r2 (EXT2) 21799 instructions. 21800 21801'-mlong64' 21802 Force 'long' types to be 64 bits wide. See '-mlong32' for an 21803 explanation of the default and the way that the pointer size is 21804 determined. 21805 21806'-mlong32' 21807 Force 'long', 'int', and pointer types to be 32 bits wide. 21808 21809 The default size of 'int's, 'long's and pointers depends on the 21810 ABI. All the supported ABIs use 32-bit 'int's. The n64 ABI uses 21811 64-bit 'long's, as does the 64-bit EABI; the others use 32-bit 21812 'long's. Pointers are the same size as 'long's, or the same size 21813 as integer registers, whichever is smaller. 21814 21815'-msym32' 21816'-mno-sym32' 21817 Assume (do not assume) that all symbols have 32-bit values, 21818 regardless of the selected ABI. This option is useful in 21819 combination with '-mabi=64' and '-mno-abicalls' because it allows 21820 GCC to generate shorter and faster references to symbolic 21821 addresses. 21822 21823'-G NUM' 21824 Put definitions of externally-visible data in a small data section 21825 if that data is no bigger than NUM bytes. GCC can then generate 21826 more efficient accesses to the data; see '-mgpopt' for details. 21827 21828 The default '-G' option depends on the configuration. 21829 21830'-mlocal-sdata' 21831'-mno-local-sdata' 21832 Extend (do not extend) the '-G' behavior to local data too, such as 21833 to static variables in C. '-mlocal-sdata' is the default for all 21834 configurations. 21835 21836 If the linker complains that an application is using too much small 21837 data, you might want to try rebuilding the less 21838 performance-critical parts with '-mno-local-sdata'. You might also 21839 want to build large libraries with '-mno-local-sdata', so that the 21840 libraries leave more room for the main program. 21841 21842'-mextern-sdata' 21843'-mno-extern-sdata' 21844 Assume (do not assume) that externally-defined data is in a small 21845 data section if the size of that data is within the '-G' limit. 21846 '-mextern-sdata' is the default for all configurations. 21847 21848 If you compile a module MOD with '-mextern-sdata' '-G NUM' 21849 '-mgpopt', and MOD references a variable VAR that is no bigger than 21850 NUM bytes, you must make sure that VAR is placed in a small data 21851 section. If VAR is defined by another module, you must either 21852 compile that module with a high-enough '-G' setting or attach a 21853 'section' attribute to VAR's definition. If VAR is common, you 21854 must link the application with a high-enough '-G' setting. 21855 21856 The easiest way of satisfying these restrictions is to compile and 21857 link every module with the same '-G' option. However, you may wish 21858 to build a library that supports several different small data 21859 limits. You can do this by compiling the library with the highest 21860 supported '-G' setting and additionally using '-mno-extern-sdata' 21861 to stop the library from making assumptions about 21862 externally-defined data. 21863 21864'-mgpopt' 21865'-mno-gpopt' 21866 Use (do not use) GP-relative accesses for symbols that are known to 21867 be in a small data section; see '-G', '-mlocal-sdata' and 21868 '-mextern-sdata'. '-mgpopt' is the default for all configurations. 21869 21870 '-mno-gpopt' is useful for cases where the '$gp' register might not 21871 hold the value of '_gp'. For example, if the code is part of a 21872 library that might be used in a boot monitor, programs that call 21873 boot monitor routines pass an unknown value in '$gp'. (In such 21874 situations, the boot monitor itself is usually compiled with 21875 '-G0'.) 21876 21877 '-mno-gpopt' implies '-mno-local-sdata' and '-mno-extern-sdata'. 21878 21879'-membedded-data' 21880'-mno-embedded-data' 21881 Allocate variables to the read-only data section first if possible, 21882 then next in the small data section if possible, otherwise in data. 21883 This gives slightly slower code than the default, but reduces the 21884 amount of RAM required when executing, and thus may be preferred 21885 for some embedded systems. 21886 21887'-muninit-const-in-rodata' 21888'-mno-uninit-const-in-rodata' 21889 Put uninitialized 'const' variables in the read-only data section. 21890 This option is only meaningful in conjunction with 21891 '-membedded-data'. 21892 21893'-mcode-readable=SETTING' 21894 Specify whether GCC may generate code that reads from executable 21895 sections. There are three possible settings: 21896 21897 '-mcode-readable=yes' 21898 Instructions may freely access executable sections. This is 21899 the default setting. 21900 21901 '-mcode-readable=pcrel' 21902 MIPS16 PC-relative load instructions can access executable 21903 sections, but other instructions must not do so. This option 21904 is useful on 4KSc and 4KSd processors when the code TLBs have 21905 the Read Inhibit bit set. It is also useful on processors 21906 that can be configured to have a dual instruction/data SRAM 21907 interface and that, like the M4K, automatically redirect 21908 PC-relative loads to the instruction RAM. 21909 21910 '-mcode-readable=no' 21911 Instructions must not access executable sections. This option 21912 can be useful on targets that are configured to have a dual 21913 instruction/data SRAM interface but that (unlike the M4K) do 21914 not automatically redirect PC-relative loads to the 21915 instruction RAM. 21916 21917'-msplit-addresses' 21918'-mno-split-addresses' 21919 Enable (disable) use of the '%hi()' and '%lo()' assembler 21920 relocation operators. This option has been superseded by 21921 '-mexplicit-relocs' but is retained for backwards compatibility. 21922 21923'-mexplicit-relocs' 21924'-mno-explicit-relocs' 21925 Use (do not use) assembler relocation operators when dealing with 21926 symbolic addresses. The alternative, selected by 21927 '-mno-explicit-relocs', is to use assembler macros instead. 21928 21929 '-mexplicit-relocs' is the default if GCC was configured to use an 21930 assembler that supports relocation operators. 21931 21932'-mcheck-zero-division' 21933'-mno-check-zero-division' 21934 Trap (do not trap) on integer division by zero. 21935 21936 The default is '-mcheck-zero-division'. 21937 21938'-mdivide-traps' 21939'-mdivide-breaks' 21940 MIPS systems check for division by zero by generating either a 21941 conditional trap or a break instruction. Using traps results in 21942 smaller code, but is only supported on MIPS II and later. Also, 21943 some versions of the Linux kernel have a bug that prevents trap 21944 from generating the proper signal ('SIGFPE'). Use '-mdivide-traps' 21945 to allow conditional traps on architectures that support them and 21946 '-mdivide-breaks' to force the use of breaks. 21947 21948 The default is usually '-mdivide-traps', but this can be overridden 21949 at configure time using '--with-divide=breaks'. Divide-by-zero 21950 checks can be completely disabled using '-mno-check-zero-division'. 21951 21952'-mload-store-pairs' 21953'-mno-load-store-pairs' 21954 Enable (disable) an optimization that pairs consecutive load or 21955 store instructions to enable load/store bonding. This option is 21956 enabled by default but only takes effect when the selected 21957 architecture is known to support bonding. 21958 21959'-mmemcpy' 21960'-mno-memcpy' 21961 Force (do not force) the use of 'memcpy' for non-trivial block 21962 moves. The default is '-mno-memcpy', which allows GCC to inline 21963 most constant-sized copies. 21964 21965'-mlong-calls' 21966'-mno-long-calls' 21967 Disable (do not disable) use of the 'jal' instruction. Calling 21968 functions using 'jal' is more efficient but requires the caller and 21969 callee to be in the same 256 megabyte segment. 21970 21971 This option has no effect on abicalls code. The default is 21972 '-mno-long-calls'. 21973 21974'-mmad' 21975'-mno-mad' 21976 Enable (disable) use of the 'mad', 'madu' and 'mul' instructions, 21977 as provided by the R4650 ISA. 21978 21979'-mimadd' 21980'-mno-imadd' 21981 Enable (disable) use of the 'madd' and 'msub' integer instructions. 21982 The default is '-mimadd' on architectures that support 'madd' and 21983 'msub' except for the 74k architecture where it was found to 21984 generate slower code. 21985 21986'-mfused-madd' 21987'-mno-fused-madd' 21988 Enable (disable) use of the floating-point multiply-accumulate 21989 instructions, when they are available. The default is 21990 '-mfused-madd'. 21991 21992 On the R8000 CPU when multiply-accumulate instructions are used, 21993 the intermediate product is calculated to infinite precision and is 21994 not subject to the FCSR Flush to Zero bit. This may be undesirable 21995 in some circumstances. On other processors the result is 21996 numerically identical to the equivalent computation using separate 21997 multiply, add, subtract and negate instructions. 21998 21999'-nocpp' 22000 Tell the MIPS assembler to not run its preprocessor over user 22001 assembler files (with a '.s' suffix) when assembling them. 22002 22003'-mfix-24k' 22004'-mno-fix-24k' 22005 Work around the 24K E48 (lost data on stores during refill) errata. 22006 The workarounds are implemented by the assembler rather than by 22007 GCC. 22008 22009'-mfix-r4000' 22010'-mno-fix-r4000' 22011 Work around certain R4000 CPU errata: 22012 - A double-word or a variable shift may give an incorrect result 22013 if executed immediately after starting an integer division. 22014 - A double-word or a variable shift may give an incorrect result 22015 if executed while an integer multiplication is in progress. 22016 - An integer division may give an incorrect result if started in 22017 a delay slot of a taken branch or a jump. 22018 22019'-mfix-r4400' 22020'-mno-fix-r4400' 22021 Work around certain R4400 CPU errata: 22022 - A double-word or a variable shift may give an incorrect result 22023 if executed immediately after starting an integer division. 22024 22025'-mfix-r10000' 22026'-mno-fix-r10000' 22027 Work around certain R10000 errata: 22028 - 'll'/'sc' sequences may not behave atomically on revisions 22029 prior to 3.0. They may deadlock on revisions 2.6 and earlier. 22030 22031 This option can only be used if the target architecture supports 22032 branch-likely instructions. '-mfix-r10000' is the default when 22033 '-march=r10000' is used; '-mno-fix-r10000' is the default 22034 otherwise. 22035 22036'-mfix-r5900' 22037'-mno-fix-r5900' 22038 Do not attempt to schedule the preceding instruction into the delay 22039 slot of a branch instruction placed at the end of a short loop of 22040 six instructions or fewer and always schedule a 'nop' instruction 22041 there instead. The short loop bug under certain conditions causes 22042 loops to execute only once or twice, due to a hardware bug in the 22043 R5900 chip. The workaround is implemented by the assembler rather 22044 than by GCC. 22045 22046'-mfix-rm7000' 22047'-mno-fix-rm7000' 22048 Work around the RM7000 'dmult'/'dmultu' errata. The workarounds 22049 are implemented by the assembler rather than by GCC. 22050 22051'-mfix-vr4120' 22052'-mno-fix-vr4120' 22053 Work around certain VR4120 errata: 22054 - 'dmultu' does not always produce the correct result. 22055 - 'div' and 'ddiv' do not always produce the correct result if 22056 one of the operands is negative. 22057 The workarounds for the division errata rely on special functions 22058 in 'libgcc.a'. At present, these functions are only provided by 22059 the 'mips64vr*-elf' configurations. 22060 22061 Other VR4120 errata require a NOP to be inserted between certain 22062 pairs of instructions. These errata are handled by the assembler, 22063 not by GCC itself. 22064 22065'-mfix-vr4130' 22066 Work around the VR4130 'mflo'/'mfhi' errata. The workarounds are 22067 implemented by the assembler rather than by GCC, although GCC 22068 avoids using 'mflo' and 'mfhi' if the VR4130 'macc', 'macchi', 22069 'dmacc' and 'dmacchi' instructions are available instead. 22070 22071'-mfix-sb1' 22072'-mno-fix-sb1' 22073 Work around certain SB-1 CPU core errata. (This flag currently 22074 works around the SB-1 revision 2 "F1" and "F2" floating-point 22075 errata.) 22076 22077'-mr10k-cache-barrier=SETTING' 22078 Specify whether GCC should insert cache barriers to avoid the side 22079 effects of speculation on R10K processors. 22080 22081 In common with many processors, the R10K tries to predict the 22082 outcome of a conditional branch and speculatively executes 22083 instructions from the "taken" branch. It later aborts these 22084 instructions if the predicted outcome is wrong. However, on the 22085 R10K, even aborted instructions can have side effects. 22086 22087 This problem only affects kernel stores and, depending on the 22088 system, kernel loads. As an example, a speculatively-executed 22089 store may load the target memory into cache and mark the cache line 22090 as dirty, even if the store itself is later aborted. If a DMA 22091 operation writes to the same area of memory before the "dirty" line 22092 is flushed, the cached data overwrites the DMA-ed data. See the 22093 R10K processor manual for a full description, including other 22094 potential problems. 22095 22096 One workaround is to insert cache barrier instructions before every 22097 memory access that might be speculatively executed and that might 22098 have side effects even if aborted. '-mr10k-cache-barrier=SETTING' 22099 controls GCC's implementation of this workaround. It assumes that 22100 aborted accesses to any byte in the following regions does not have 22101 side effects: 22102 22103 1. the memory occupied by the current function's stack frame; 22104 22105 2. the memory occupied by an incoming stack argument; 22106 22107 3. the memory occupied by an object with a link-time-constant 22108 address. 22109 22110 It is the kernel's responsibility to ensure that speculative 22111 accesses to these regions are indeed safe. 22112 22113 If the input program contains a function declaration such as: 22114 22115 void foo (void); 22116 22117 then the implementation of 'foo' must allow 'j foo' and 'jal foo' 22118 to be executed speculatively. GCC honors this restriction for 22119 functions it compiles itself. It expects non-GCC functions (such 22120 as hand-written assembly code) to do the same. 22121 22122 The option has three forms: 22123 22124 '-mr10k-cache-barrier=load-store' 22125 Insert a cache barrier before a load or store that might be 22126 speculatively executed and that might have side effects even 22127 if aborted. 22128 22129 '-mr10k-cache-barrier=store' 22130 Insert a cache barrier before a store that might be 22131 speculatively executed and that might have side effects even 22132 if aborted. 22133 22134 '-mr10k-cache-barrier=none' 22135 Disable the insertion of cache barriers. This is the default 22136 setting. 22137 22138'-mflush-func=FUNC' 22139'-mno-flush-func' 22140 Specifies the function to call to flush the I and D caches, or to 22141 not call any such function. If called, the function must take the 22142 same arguments as the common '_flush_func', that is, the address of 22143 the memory range for which the cache is being flushed, the size of 22144 the memory range, and the number 3 (to flush both caches). The 22145 default depends on the target GCC was configured for, but commonly 22146 is either '_flush_func' or '__cpu_flush'. 22147 22148'mbranch-cost=NUM' 22149 Set the cost of branches to roughly NUM "simple" instructions. 22150 This cost is only a heuristic and is not guaranteed to produce 22151 consistent results across releases. A zero cost redundantly 22152 selects the default, which is based on the '-mtune' setting. 22153 22154'-mbranch-likely' 22155'-mno-branch-likely' 22156 Enable or disable use of Branch Likely instructions, regardless of 22157 the default for the selected architecture. By default, Branch 22158 Likely instructions may be generated if they are supported by the 22159 selected architecture. An exception is for the MIPS32 and MIPS64 22160 architectures and processors that implement those architectures; 22161 for those, Branch Likely instructions are not be generated by 22162 default because the MIPS32 and MIPS64 architectures specifically 22163 deprecate their use. 22164 22165'-mcompact-branches=never' 22166'-mcompact-branches=optimal' 22167'-mcompact-branches=always' 22168 These options control which form of branches will be generated. 22169 The default is '-mcompact-branches=optimal'. 22170 22171 The '-mcompact-branches=never' option ensures that compact branch 22172 instructions will never be generated. 22173 22174 The '-mcompact-branches=always' option ensures that a compact 22175 branch instruction will be generated if available. If a compact 22176 branch instruction is not available, a delay slot form of the 22177 branch will be used instead. 22178 22179 This option is supported from MIPS Release 6 onwards. 22180 22181 The '-mcompact-branches=optimal' option will cause a delay slot 22182 branch to be used if one is available in the current ISA and the 22183 delay slot is successfully filled. If the delay slot is not 22184 filled, a compact branch will be chosen if one is available. 22185 22186'-mfp-exceptions' 22187'-mno-fp-exceptions' 22188 Specifies whether FP exceptions are enabled. This affects how FP 22189 instructions are scheduled for some processors. The default is 22190 that FP exceptions are enabled. 22191 22192 For instance, on the SB-1, if FP exceptions are disabled, and we 22193 are emitting 64-bit code, then we can use both FP pipes. 22194 Otherwise, we can only use one FP pipe. 22195 22196'-mvr4130-align' 22197'-mno-vr4130-align' 22198 The VR4130 pipeline is two-way superscalar, but can only issue two 22199 instructions together if the first one is 8-byte aligned. When 22200 this option is enabled, GCC aligns pairs of instructions that it 22201 thinks should execute in parallel. 22202 22203 This option only has an effect when optimizing for the VR4130. It 22204 normally makes code faster, but at the expense of making it bigger. 22205 It is enabled by default at optimization level '-O3'. 22206 22207'-msynci' 22208'-mno-synci' 22209 Enable (disable) generation of 'synci' instructions on 22210 architectures that support it. The 'synci' instructions (if 22211 enabled) are generated when '__builtin___clear_cache' is compiled. 22212 22213 This option defaults to '-mno-synci', but the default can be 22214 overridden by configuring GCC with '--with-synci'. 22215 22216 When compiling code for single processor systems, it is generally 22217 safe to use 'synci'. However, on many multi-core (SMP) systems, it 22218 does not invalidate the instruction caches on all cores and may 22219 lead to undefined behavior. 22220 22221'-mrelax-pic-calls' 22222'-mno-relax-pic-calls' 22223 Try to turn PIC calls that are normally dispatched via register 22224 '$25' into direct calls. This is only possible if the linker can 22225 resolve the destination at link time and if the destination is 22226 within range for a direct call. 22227 22228 '-mrelax-pic-calls' is the default if GCC was configured to use an 22229 assembler and a linker that support the '.reloc' assembly directive 22230 and '-mexplicit-relocs' is in effect. With '-mno-explicit-relocs', 22231 this optimization can be performed by the assembler and the linker 22232 alone without help from the compiler. 22233 22234'-mmcount-ra-address' 22235'-mno-mcount-ra-address' 22236 Emit (do not emit) code that allows '_mcount' to modify the calling 22237 function's return address. When enabled, this option extends the 22238 usual '_mcount' interface with a new RA-ADDRESS parameter, which 22239 has type 'intptr_t *' and is passed in register '$12'. '_mcount' 22240 can then modify the return address by doing both of the following: 22241 * Returning the new address in register '$31'. 22242 * Storing the new address in '*RA-ADDRESS', if RA-ADDRESS is 22243 nonnull. 22244 22245 The default is '-mno-mcount-ra-address'. 22246 22247'-mframe-header-opt' 22248'-mno-frame-header-opt' 22249 Enable (disable) frame header optimization in the o32 ABI. When 22250 using the o32 ABI, calling functions will allocate 16 bytes on the 22251 stack for the called function to write out register arguments. 22252 When enabled, this optimization will suppress the allocation of the 22253 frame header if it can be determined that it is unused. 22254 22255 This optimization is off by default at all optimization levels. 22256 22257'-mlxc1-sxc1' 22258'-mno-lxc1-sxc1' 22259 When applicable, enable (disable) the generation of 'lwxc1', 22260 'swxc1', 'ldxc1', 'sdxc1' instructions. Enabled by default. 22261 22262'-mmadd4' 22263'-mno-madd4' 22264 When applicable, enable (disable) the generation of 4-operand 22265 'madd.s', 'madd.d' and related instructions. Enabled by default. 22266 22267 22268File: gcc.info, Node: MMIX Options, Next: MN10300 Options, Prev: MIPS Options, Up: Submodel Options 22269 222703.19.30 MMIX Options 22271-------------------- 22272 22273These options are defined for the MMIX: 22274 22275'-mlibfuncs' 22276'-mno-libfuncs' 22277 Specify that intrinsic library functions are being compiled, 22278 passing all values in registers, no matter the size. 22279 22280'-mepsilon' 22281'-mno-epsilon' 22282 Generate floating-point comparison instructions that compare with 22283 respect to the 'rE' epsilon register. 22284 22285'-mabi=mmixware' 22286'-mabi=gnu' 22287 Generate code that passes function parameters and return values 22288 that (in the called function) are seen as registers '$0' and up, as 22289 opposed to the GNU ABI which uses global registers '$231' and up. 22290 22291'-mzero-extend' 22292'-mno-zero-extend' 22293 When reading data from memory in sizes shorter than 64 bits, use 22294 (do not use) zero-extending load instructions by default, rather 22295 than sign-extending ones. 22296 22297'-mknuthdiv' 22298'-mno-knuthdiv' 22299 Make the result of a division yielding a remainder have the same 22300 sign as the divisor. With the default, '-mno-knuthdiv', the sign 22301 of the remainder follows the sign of the dividend. Both methods 22302 are arithmetically valid, the latter being almost exclusively used. 22303 22304'-mtoplevel-symbols' 22305'-mno-toplevel-symbols' 22306 Prepend (do not prepend) a ':' to all global symbols, so the 22307 assembly code can be used with the 'PREFIX' assembly directive. 22308 22309'-melf' 22310 Generate an executable in the ELF format, rather than the default 22311 'mmo' format used by the 'mmix' simulator. 22312 22313'-mbranch-predict' 22314'-mno-branch-predict' 22315 Use (do not use) the probable-branch instructions, when static 22316 branch prediction indicates a probable branch. 22317 22318'-mbase-addresses' 22319'-mno-base-addresses' 22320 Generate (do not generate) code that uses _base addresses_. Using 22321 a base address automatically generates a request (handled by the 22322 assembler and the linker) for a constant to be set up in a global 22323 register. The register is used for one or more base address 22324 requests within the range 0 to 255 from the value held in the 22325 register. The generally leads to short and fast code, but the 22326 number of different data items that can be addressed is limited. 22327 This means that a program that uses lots of static data may require 22328 '-mno-base-addresses'. 22329 22330'-msingle-exit' 22331'-mno-single-exit' 22332 Force (do not force) generated code to have a single exit point in 22333 each function. 22334 22335 22336File: gcc.info, Node: MN10300 Options, Next: Moxie Options, Prev: MMIX Options, Up: Submodel Options 22337 223383.19.31 MN10300 Options 22339----------------------- 22340 22341These '-m' options are defined for Matsushita MN10300 architectures: 22342 22343'-mmult-bug' 22344 Generate code to avoid bugs in the multiply instructions for the 22345 MN10300 processors. This is the default. 22346 22347'-mno-mult-bug' 22348 Do not generate code to avoid bugs in the multiply instructions for 22349 the MN10300 processors. 22350 22351'-mam33' 22352 Generate code using features specific to the AM33 processor. 22353 22354'-mno-am33' 22355 Do not generate code using features specific to the AM33 processor. 22356 This is the default. 22357 22358'-mam33-2' 22359 Generate code using features specific to the AM33/2.0 processor. 22360 22361'-mam34' 22362 Generate code using features specific to the AM34 processor. 22363 22364'-mtune=CPU-TYPE' 22365 Use the timing characteristics of the indicated CPU type when 22366 scheduling instructions. This does not change the targeted 22367 processor type. The CPU type must be one of 'mn10300', 'am33', 22368 'am33-2' or 'am34'. 22369 22370'-mreturn-pointer-on-d0' 22371 When generating a function that returns a pointer, return the 22372 pointer in both 'a0' and 'd0'. Otherwise, the pointer is returned 22373 only in 'a0', and attempts to call such functions without a 22374 prototype result in errors. Note that this option is on by 22375 default; use '-mno-return-pointer-on-d0' to disable it. 22376 22377'-mno-crt0' 22378 Do not link in the C run-time initialization object file. 22379 22380'-mrelax' 22381 Indicate to the linker that it should perform a relaxation 22382 optimization pass to shorten branches, calls and absolute memory 22383 addresses. This option only has an effect when used on the command 22384 line for the final link step. 22385 22386 This option makes symbolic debugging impossible. 22387 22388'-mliw' 22389 Allow the compiler to generate _Long Instruction Word_ instructions 22390 if the target is the 'AM33' or later. This is the default. This 22391 option defines the preprocessor macro '__LIW__'. 22392 22393'-mno-liw' 22394 Do not allow the compiler to generate _Long Instruction Word_ 22395 instructions. This option defines the preprocessor macro 22396 '__NO_LIW__'. 22397 22398'-msetlb' 22399 Allow the compiler to generate the _SETLB_ and _Lcc_ instructions 22400 if the target is the 'AM33' or later. This is the default. This 22401 option defines the preprocessor macro '__SETLB__'. 22402 22403'-mno-setlb' 22404 Do not allow the compiler to generate _SETLB_ or _Lcc_ 22405 instructions. This option defines the preprocessor macro 22406 '__NO_SETLB__'. 22407 22408 22409File: gcc.info, Node: Moxie Options, Next: MSP430 Options, Prev: MN10300 Options, Up: Submodel Options 22410 224113.19.32 Moxie Options 22412--------------------- 22413 22414'-meb' 22415 Generate big-endian code. This is the default for 'moxie-*-*' 22416 configurations. 22417 22418'-mel' 22419 Generate little-endian code. 22420 22421'-mmul.x' 22422 Generate mul.x and umul.x instructions. This is the default for 22423 'moxiebox-*-*' configurations. 22424 22425'-mno-crt0' 22426 Do not link in the C run-time initialization object file. 22427 22428 22429File: gcc.info, Node: MSP430 Options, Next: NDS32 Options, Prev: Moxie Options, Up: Submodel Options 22430 224313.19.33 MSP430 Options 22432---------------------- 22433 22434These options are defined for the MSP430: 22435 22436'-masm-hex' 22437 Force assembly output to always use hex constants. Normally such 22438 constants are signed decimals, but this option is available for 22439 testsuite and/or aesthetic purposes. 22440 22441'-mmcu=' 22442 Select the MCU to target. This is used to create a C preprocessor 22443 symbol based upon the MCU name, converted to upper case and pre- 22444 and post-fixed with '__'. This in turn is used by the 'msp430.h' 22445 header file to select an MCU-specific supplementary header file. 22446 22447 The option also sets the ISA to use. If the MCU name is one that 22448 is known to only support the 430 ISA then that is selected, 22449 otherwise the 430X ISA is selected. A generic MCU name of 'msp430' 22450 can also be used to select the 430 ISA. Similarly the generic 22451 'msp430x' MCU name selects the 430X ISA. 22452 22453 In addition an MCU-specific linker script is added to the linker 22454 command line. The script's name is the name of the MCU with '.ld' 22455 appended. Thus specifying '-mmcu=xxx' on the 'gcc' command line 22456 defines the C preprocessor symbol '__XXX__' and cause the linker to 22457 search for a script called 'xxx.ld'. 22458 22459 The ISA and hardware multiply supported for the different MCUs is 22460 hard-coded into GCC. However, an external 'devices.csv' file can be 22461 used to extend device support beyond those that have been 22462 hard-coded. 22463 22464 GCC searches for the 'devices.csv' file using the following methods 22465 in the given precedence order, where the first method takes 22466 precendence over the second which takes precedence over the third. 22467 22468 Include path specified with '-I' and '-L' 22469 'devices.csv' will be searched for in each of the directories 22470 specified by include paths and linker library search paths. 22471 Path specified by the environment variable 'MSP430_GCC_INCLUDE_DIR' 22472 Define the value of the global environment variable 22473 'MSP430_GCC_INCLUDE_DIR' to the full path to the directory 22474 containing devices.csv, and GCC will search this directory for 22475 devices.csv. If devices.csv is found, this directory will 22476 also be registered as an include path, and linker library 22477 path. Header files and linker scripts in this directory can 22478 therefore be used without manually specifying '-I' and '-L' on 22479 the command line. 22480 The 'msp430-elf{,bare}/include/devices' directory 22481 Finally, GCC will examine 'msp430-elf{,bare}/include/devices' 22482 from the toolchain root directory. This directory does not 22483 exist in a default installation, but if the user has created 22484 it and copied 'devices.csv' there, then the MCU data will be 22485 read. As above, this directory will also be registered as an 22486 include path, and linker library path. 22487 22488 If none of the above search methods find 'devices.csv', then the 22489 hard-coded MCU data is used. 22490 22491'-mwarn-mcu' 22492'-mno-warn-mcu' 22493 This option enables or disables warnings about conflicts between 22494 the MCU name specified by the '-mmcu' option and the ISA set by the 22495 '-mcpu' option and/or the hardware multiply support set by the 22496 '-mhwmult' option. It also toggles warnings about unrecognized MCU 22497 names. This option is on by default. 22498 22499'-mcpu=' 22500 Specifies the ISA to use. Accepted values are 'msp430', 'msp430x' 22501 and 'msp430xv2'. This option is deprecated. The '-mmcu=' option 22502 should be used to select the ISA. 22503 22504'-msim' 22505 Link to the simulator runtime libraries and linker script. 22506 Overrides any scripts that would be selected by the '-mmcu=' 22507 option. 22508 22509'-mlarge' 22510 Use large-model addressing (20-bit pointers, 32-bit 'size_t'). 22511 22512'-msmall' 22513 Use small-model addressing (16-bit pointers, 16-bit 'size_t'). 22514 22515'-mrelax' 22516 This option is passed to the assembler and linker, and allows the 22517 linker to perform certain optimizations that cannot be done until 22518 the final link. 22519 22520'mhwmult=' 22521 Describes the type of hardware multiply supported by the target. 22522 Accepted values are 'none' for no hardware multiply, '16bit' for 22523 the original 16-bit-only multiply supported by early MCUs. '32bit' 22524 for the 16/32-bit multiply supported by later MCUs and 'f5series' 22525 for the 16/32-bit multiply supported by F5-series MCUs. A value of 22526 'auto' can also be given. This tells GCC to deduce the hardware 22527 multiply support based upon the MCU name provided by the '-mmcu' 22528 option. If no '-mmcu' option is specified or if the MCU name is 22529 not recognized then no hardware multiply support is assumed. 22530 'auto' is the default setting. 22531 22532 Hardware multiplies are normally performed by calling a library 22533 routine. This saves space in the generated code. When compiling 22534 at '-O3' or higher however the hardware multiplier is invoked 22535 inline. This makes for bigger, but faster code. 22536 22537 The hardware multiply routines disable interrupts whilst running 22538 and restore the previous interrupt state when they finish. This 22539 makes them safe to use inside interrupt handlers as well as in 22540 normal code. 22541 22542'-minrt' 22543 Enable the use of a minimum runtime environment - no static 22544 initializers or constructors. This is intended for 22545 memory-constrained devices. The compiler includes special symbols 22546 in some objects that tell the linker and runtime which code 22547 fragments are required. 22548 22549'-mtiny-printf' 22550 Enable reduced code size 'printf' and 'puts' library functions. 22551 The 'tiny' implementations of these functions are not reentrant, so 22552 must be used with caution in multi-threaded applications. 22553 22554 Support for streams has been removed and the string to be printed 22555 will always be sent to stdout via the 'write' syscall. The string 22556 is not buffered before it is sent to write. 22557 22558 This option requires Newlib Nano IO, so GCC must be configured with 22559 '--enable-newlib-nano-formatted-io'. 22560 22561'-mcode-region=' 22562'-mdata-region=' 22563 These options tell the compiler where to place functions and data 22564 that do not have one of the 'lower', 'upper', 'either' or 'section' 22565 attributes. Possible values are 'lower', 'upper', 'either' or 22566 'any'. The first three behave like the corresponding attribute. 22567 The fourth possible value - 'any' - is the default. It leaves 22568 placement entirely up to the linker script and how it assigns the 22569 standard sections ('.text', '.data', etc) to the memory regions. 22570 22571'-msilicon-errata=' 22572 This option passes on a request to assembler to enable the fixes 22573 for the named silicon errata. 22574 22575'-msilicon-errata-warn=' 22576 This option passes on a request to the assembler to enable warning 22577 messages when a silicon errata might need to be applied. 22578 22579'-mwarn-devices-csv' 22580'-mno-warn-devices-csv' 22581 Warn if 'devices.csv' is not found or there are problem parsing it 22582 (default: on). 22583 22584 22585File: gcc.info, Node: NDS32 Options, Next: Nios II Options, Prev: MSP430 Options, Up: Submodel Options 22586 225873.19.34 NDS32 Options 22588--------------------- 22589 22590These options are defined for NDS32 implementations: 22591 22592'-mbig-endian' 22593 Generate code in big-endian mode. 22594 22595'-mlittle-endian' 22596 Generate code in little-endian mode. 22597 22598'-mreduced-regs' 22599 Use reduced-set registers for register allocation. 22600 22601'-mfull-regs' 22602 Use full-set registers for register allocation. 22603 22604'-mcmov' 22605 Generate conditional move instructions. 22606 22607'-mno-cmov' 22608 Do not generate conditional move instructions. 22609 22610'-mext-perf' 22611 Generate performance extension instructions. 22612 22613'-mno-ext-perf' 22614 Do not generate performance extension instructions. 22615 22616'-mext-perf2' 22617 Generate performance extension 2 instructions. 22618 22619'-mno-ext-perf2' 22620 Do not generate performance extension 2 instructions. 22621 22622'-mext-string' 22623 Generate string extension instructions. 22624 22625'-mno-ext-string' 22626 Do not generate string extension instructions. 22627 22628'-mv3push' 22629 Generate v3 push25/pop25 instructions. 22630 22631'-mno-v3push' 22632 Do not generate v3 push25/pop25 instructions. 22633 22634'-m16-bit' 22635 Generate 16-bit instructions. 22636 22637'-mno-16-bit' 22638 Do not generate 16-bit instructions. 22639 22640'-misr-vector-size=NUM' 22641 Specify the size of each interrupt vector, which must be 4 or 16. 22642 22643'-mcache-block-size=NUM' 22644 Specify the size of each cache block, which must be a power of 2 22645 between 4 and 512. 22646 22647'-march=ARCH' 22648 Specify the name of the target architecture. 22649 22650'-mcmodel=CODE-MODEL' 22651 Set the code model to one of 22652 'small' 22653 All the data and read-only data segments must be within 512KB 22654 addressing space. The text segment must be within 16MB 22655 addressing space. 22656 'medium' 22657 The data segment must be within 512KB while the read-only data 22658 segment can be within 4GB addressing space. The text segment 22659 should be still within 16MB addressing space. 22660 'large' 22661 All the text and data segments can be within 4GB addressing 22662 space. 22663 22664'-mctor-dtor' 22665 Enable constructor/destructor feature. 22666 22667'-mrelax' 22668 Guide linker to relax instructions. 22669 22670 22671File: gcc.info, Node: Nios II Options, Next: Nvidia PTX Options, Prev: NDS32 Options, Up: Submodel Options 22672 226733.19.35 Nios II Options 22674----------------------- 22675 22676These are the options defined for the Altera Nios II processor. 22677 22678'-G NUM' 22679 Put global and static objects less than or equal to NUM bytes into 22680 the small data or BSS sections instead of the normal data or BSS 22681 sections. The default value of NUM is 8. 22682 22683'-mgpopt=OPTION' 22684'-mgpopt' 22685'-mno-gpopt' 22686 Generate (do not generate) GP-relative accesses. The following 22687 OPTION names are recognized: 22688 22689 'none' 22690 Do not generate GP-relative accesses. 22691 22692 'local' 22693 Generate GP-relative accesses for small data objects that are 22694 not external, weak, or uninitialized common symbols. Also use 22695 GP-relative addressing for objects that have been explicitly 22696 placed in a small data section via a 'section' attribute. 22697 22698 'global' 22699 As for 'local', but also generate GP-relative accesses for 22700 small data objects that are external, weak, or common. If you 22701 use this option, you must ensure that all parts of your 22702 program (including libraries) are compiled with the same '-G' 22703 setting. 22704 22705 'data' 22706 Generate GP-relative accesses for all data objects in the 22707 program. If you use this option, the entire data and BSS 22708 segments of your program must fit in 64K of memory and you 22709 must use an appropriate linker script to allocate them within 22710 the addressable range of the global pointer. 22711 22712 'all' 22713 Generate GP-relative addresses for function pointers as well 22714 as data pointers. If you use this option, the entire text, 22715 data, and BSS segments of your program must fit in 64K of 22716 memory and you must use an appropriate linker script to 22717 allocate them within the addressable range of the global 22718 pointer. 22719 22720 '-mgpopt' is equivalent to '-mgpopt=local', and '-mno-gpopt' is 22721 equivalent to '-mgpopt=none'. 22722 22723 The default is '-mgpopt' except when '-fpic' or '-fPIC' is 22724 specified to generate position-independent code. Note that the 22725 Nios II ABI does not permit GP-relative accesses from shared 22726 libraries. 22727 22728 You may need to specify '-mno-gpopt' explicitly when building 22729 programs that include large amounts of small data, including large 22730 GOT data sections. In this case, the 16-bit offset for GP-relative 22731 addressing may not be large enough to allow access to the entire 22732 small data section. 22733 22734'-mgprel-sec=REGEXP' 22735 This option specifies additional section names that can be accessed 22736 via GP-relative addressing. It is most useful in conjunction with 22737 'section' attributes on variable declarations (*note Common 22738 Variable Attributes::) and a custom linker script. The REGEXP is a 22739 POSIX Extended Regular Expression. 22740 22741 This option does not affect the behavior of the '-G' option, and 22742 the specified sections are in addition to the standard '.sdata' and 22743 '.sbss' small-data sections that are recognized by '-mgpopt'. 22744 22745'-mr0rel-sec=REGEXP' 22746 This option specifies names of sections that can be accessed via a 22747 16-bit offset from 'r0'; that is, in the low 32K or high 32K of the 22748 32-bit address space. It is most useful in conjunction with 22749 'section' attributes on variable declarations (*note Common 22750 Variable Attributes::) and a custom linker script. The REGEXP is a 22751 POSIX Extended Regular Expression. 22752 22753 In contrast to the use of GP-relative addressing for small data, 22754 zero-based addressing is never generated by default and there are 22755 no conventional section names used in standard linker scripts for 22756 sections in the low or high areas of memory. 22757 22758'-mel' 22759'-meb' 22760 Generate little-endian (default) or big-endian (experimental) code, 22761 respectively. 22762 22763'-march=ARCH' 22764 This specifies the name of the target Nios II architecture. GCC 22765 uses this name to determine what kind of instructions it can emit 22766 when generating assembly code. Permissible names are: 'r1', 'r2'. 22767 22768 The preprocessor macro '__nios2_arch__' is available to programs, 22769 with value 1 or 2, indicating the targeted ISA level. 22770 22771'-mbypass-cache' 22772'-mno-bypass-cache' 22773 Force all load and store instructions to always bypass cache by 22774 using I/O variants of the instructions. The default is not to 22775 bypass the cache. 22776 22777'-mno-cache-volatile' 22778'-mcache-volatile' 22779 Volatile memory access bypass the cache using the I/O variants of 22780 the load and store instructions. The default is not to bypass the 22781 cache. 22782 22783'-mno-fast-sw-div' 22784'-mfast-sw-div' 22785 Do not use table-based fast divide for small numbers. The default 22786 is to use the fast divide at '-O3' and above. 22787 22788'-mno-hw-mul' 22789'-mhw-mul' 22790'-mno-hw-mulx' 22791'-mhw-mulx' 22792'-mno-hw-div' 22793'-mhw-div' 22794 Enable or disable emitting 'mul', 'mulx' and 'div' family of 22795 instructions by the compiler. The default is to emit 'mul' and not 22796 emit 'div' and 'mulx'. 22797 22798'-mbmx' 22799'-mno-bmx' 22800'-mcdx' 22801'-mno-cdx' 22802 Enable or disable generation of Nios II R2 BMX (bit manipulation) 22803 and CDX (code density) instructions. Enabling these instructions 22804 also requires '-march=r2'. Since these instructions are optional 22805 extensions to the R2 architecture, the default is not to emit them. 22806 22807'-mcustom-INSN=N' 22808'-mno-custom-INSN' 22809 Each '-mcustom-INSN=N' option enables use of a custom instruction 22810 with encoding N when generating code that uses INSN. For example, 22811 '-mcustom-fadds=253' generates custom instruction 253 for 22812 single-precision floating-point add operations instead of the 22813 default behavior of using a library call. 22814 22815 The following values of INSN are supported. Except as otherwise 22816 noted, floating-point operations are expected to be implemented 22817 with normal IEEE 754 semantics and correspond directly to the C 22818 operators or the equivalent GCC built-in functions (*note Other 22819 Builtins::). 22820 22821 Single-precision floating point: 22822 22823 'fadds', 'fsubs', 'fdivs', 'fmuls' 22824 Binary arithmetic operations. 22825 22826 'fnegs' 22827 Unary negation. 22828 22829 'fabss' 22830 Unary absolute value. 22831 22832 'fcmpeqs', 'fcmpges', 'fcmpgts', 'fcmples', 'fcmplts', 'fcmpnes' 22833 Comparison operations. 22834 22835 'fmins', 'fmaxs' 22836 Floating-point minimum and maximum. These instructions are 22837 only generated if '-ffinite-math-only' is specified. 22838 22839 'fsqrts' 22840 Unary square root operation. 22841 22842 'fcoss', 'fsins', 'ftans', 'fatans', 'fexps', 'flogs' 22843 Floating-point trigonometric and exponential functions. These 22844 instructions are only generated if 22845 '-funsafe-math-optimizations' is also specified. 22846 22847 Double-precision floating point: 22848 22849 'faddd', 'fsubd', 'fdivd', 'fmuld' 22850 Binary arithmetic operations. 22851 22852 'fnegd' 22853 Unary negation. 22854 22855 'fabsd' 22856 Unary absolute value. 22857 22858 'fcmpeqd', 'fcmpged', 'fcmpgtd', 'fcmpled', 'fcmpltd', 'fcmpned' 22859 Comparison operations. 22860 22861 'fmind', 'fmaxd' 22862 Double-precision minimum and maximum. These instructions are 22863 only generated if '-ffinite-math-only' is specified. 22864 22865 'fsqrtd' 22866 Unary square root operation. 22867 22868 'fcosd', 'fsind', 'ftand', 'fatand', 'fexpd', 'flogd' 22869 Double-precision trigonometric and exponential functions. 22870 These instructions are only generated if 22871 '-funsafe-math-optimizations' is also specified. 22872 22873 Conversions: 22874 'fextsd' 22875 Conversion from single precision to double precision. 22876 22877 'ftruncds' 22878 Conversion from double precision to single precision. 22879 22880 'fixsi', 'fixsu', 'fixdi', 'fixdu' 22881 Conversion from floating point to signed or unsigned integer 22882 types, with truncation towards zero. 22883 22884 'round' 22885 Conversion from single-precision floating point to signed 22886 integer, rounding to the nearest integer and ties away from 22887 zero. This corresponds to the '__builtin_lroundf' function 22888 when '-fno-math-errno' is used. 22889 22890 'floatis', 'floatus', 'floatid', 'floatud' 22891 Conversion from signed or unsigned integer types to 22892 floating-point types. 22893 22894 In addition, all of the following transfer instructions for 22895 internal registers X and Y must be provided to use any of the 22896 double-precision floating-point instructions. Custom instructions 22897 taking two double-precision source operands expect the first 22898 operand in the 64-bit register X. The other operand (or only 22899 operand of a unary operation) is given to the custom arithmetic 22900 instruction with the least significant half in source register SRC1 22901 and the most significant half in SRC2. A custom instruction that 22902 returns a double-precision result returns the most significant 32 22903 bits in the destination register and the other half in 32-bit 22904 register Y. GCC automatically generates the necessary code 22905 sequences to write register X and/or read register Y when 22906 double-precision floating-point instructions are used. 22907 22908 'fwrx' 22909 Write SRC1 into the least significant half of X and SRC2 into 22910 the most significant half of X. 22911 22912 'fwry' 22913 Write SRC1 into Y. 22914 22915 'frdxhi', 'frdxlo' 22916 Read the most or least (respectively) significant half of X 22917 and store it in DEST. 22918 22919 'frdy' 22920 Read the value of Y and store it into DEST. 22921 22922 Note that you can gain more local control over generation of Nios 22923 II custom instructions by using the 'target("custom-INSN=N")' and 22924 'target("no-custom-INSN")' function attributes (*note Function 22925 Attributes::) or pragmas (*note Function Specific Option 22926 Pragmas::). 22927 22928'-mcustom-fpu-cfg=NAME' 22929 22930 This option enables a predefined, named set of custom instruction 22931 encodings (see '-mcustom-INSN' above). Currently, the following 22932 sets are defined: 22933 22934 '-mcustom-fpu-cfg=60-1' is equivalent to: 22935 -mcustom-fmuls=252 22936 -mcustom-fadds=253 22937 -mcustom-fsubs=254 22938 -fsingle-precision-constant 22939 22940 '-mcustom-fpu-cfg=60-2' is equivalent to: 22941 -mcustom-fmuls=252 22942 -mcustom-fadds=253 22943 -mcustom-fsubs=254 22944 -mcustom-fdivs=255 22945 -fsingle-precision-constant 22946 22947 '-mcustom-fpu-cfg=72-3' is equivalent to: 22948 -mcustom-floatus=243 22949 -mcustom-fixsi=244 22950 -mcustom-floatis=245 22951 -mcustom-fcmpgts=246 22952 -mcustom-fcmples=249 22953 -mcustom-fcmpeqs=250 22954 -mcustom-fcmpnes=251 22955 -mcustom-fmuls=252 22956 -mcustom-fadds=253 22957 -mcustom-fsubs=254 22958 -mcustom-fdivs=255 22959 -fsingle-precision-constant 22960 22961 Custom instruction assignments given by individual '-mcustom-INSN=' 22962 options override those given by '-mcustom-fpu-cfg=', regardless of 22963 the order of the options on the command line. 22964 22965 Note that you can gain more local control over selection of a FPU 22966 configuration by using the 'target("custom-fpu-cfg=NAME")' function 22967 attribute (*note Function Attributes::) or pragma (*note Function 22968 Specific Option Pragmas::). 22969 22970 These additional '-m' options are available for the Altera Nios II ELF 22971(bare-metal) target: 22972 22973'-mhal' 22974 Link with HAL BSP. This suppresses linking with the GCC-provided C 22975 runtime startup and termination code, and is typically used in 22976 conjunction with '-msys-crt0=' to specify the location of the 22977 alternate startup code provided by the HAL BSP. 22978 22979'-msmallc' 22980 Link with a limited version of the C library, '-lsmallc', rather 22981 than Newlib. 22982 22983'-msys-crt0=STARTFILE' 22984 STARTFILE is the file name of the startfile (crt0) to use when 22985 linking. This option is only useful in conjunction with '-mhal'. 22986 22987'-msys-lib=SYSTEMLIB' 22988 SYSTEMLIB is the library name of the library that provides 22989 low-level system calls required by the C library, e.g. 'read' and 22990 'write'. This option is typically used to link with a library 22991 provided by a HAL BSP. 22992 22993 22994File: gcc.info, Node: Nvidia PTX Options, Next: OpenRISC Options, Prev: Nios II Options, Up: Submodel Options 22995 229963.19.36 Nvidia PTX Options 22997-------------------------- 22998 22999These options are defined for Nvidia PTX: 23000 23001'-m32' 23002'-m64' 23003 Generate code for 32-bit or 64-bit ABI. 23004 23005'-misa=ISA-STRING' 23006 Generate code for given the specified PTX ISA (e.g. 'sm_35'). ISA 23007 strings must be lower-case. Valid ISA strings include 'sm_30' and 23008 'sm_35'. The default ISA is sm_30. 23009 23010'-mmainkernel' 23011 Link in code for a __main kernel. This is for stand-alone instead 23012 of offloading execution. 23013 23014'-moptimize' 23015 Apply partitioned execution optimizations. This is the default 23016 when any level of optimization is selected. 23017 23018'-msoft-stack' 23019 Generate code that does not use '.local' memory directly for stack 23020 storage. Instead, a per-warp stack pointer is maintained 23021 explicitly. This enables variable-length stack allocation (with 23022 variable-length arrays or 'alloca'), and when global memory is used 23023 for underlying storage, makes it possible to access automatic 23024 variables from other threads, or with atomic instructions. This 23025 code generation variant is used for OpenMP offloading, but the 23026 option is exposed on its own for the purpose of testing the 23027 compiler; to generate code suitable for linking into programs using 23028 OpenMP offloading, use option '-mgomp'. 23029 23030'-muniform-simt' 23031 Switch to code generation variant that allows to execute all 23032 threads in each warp, while maintaining memory state and side 23033 effects as if only one thread in each warp was active outside of 23034 OpenMP SIMD regions. All atomic operations and calls to runtime 23035 (malloc, free, vprintf) are conditionally executed (iff current 23036 lane index equals the master lane index), and the register being 23037 assigned is copied via a shuffle instruction from the master lane. 23038 Outside of SIMD regions lane 0 is the master; inside, each thread 23039 sees itself as the master. Shared memory array 'int __nvptx_uni[]' 23040 stores all-zeros or all-ones bitmasks for each warp, indicating 23041 current mode (0 outside of SIMD regions). Each thread can 23042 bitwise-and the bitmask at position 'tid.y' with current lane index 23043 to compute the master lane index. 23044 23045'-mgomp' 23046 Generate code for use in OpenMP offloading: enables '-msoft-stack' 23047 and '-muniform-simt' options, and selects corresponding multilib 23048 variant. 23049 23050 23051File: gcc.info, Node: OpenRISC Options, Next: PDP-11 Options, Prev: Nvidia PTX Options, Up: Submodel Options 23052 230533.19.37 OpenRISC Options 23054------------------------ 23055 23056These options are defined for OpenRISC: 23057 23058'-mboard=NAME' 23059 Configure a board specific runtime. This will be passed to the 23060 linker for newlib board library linking. The default is 'or1ksim'. 23061 23062'-mnewlib' 23063 This option is ignored; it is for compatibility purposes only. 23064 This used to select linker and preprocessor options for use with 23065 newlib. 23066 23067'-msoft-div' 23068'-mhard-div' 23069 Select software or hardware divide ('l.div', 'l.divu') 23070 instructions. This default is hardware divide. 23071 23072'-msoft-mul' 23073'-mhard-mul' 23074 Select software or hardware multiply ('l.mul', 'l.muli') 23075 instructions. This default is hardware multiply. 23076 23077'-msoft-float' 23078'-mhard-float' 23079 Select software or hardware for floating point operations. The 23080 default is software. 23081 23082'-mdouble-float' 23083 When '-mhard-float' is selected, enables generation of 23084 double-precision floating point instructions. By default functions 23085 from 'libgcc' are used to perform double-precision floating point 23086 operations. 23087 23088'-munordered-float' 23089 When '-mhard-float' is selected, enables generation of unordered 23090 floating point compare and set flag ('lf.sfun*') instructions. By 23091 default functions from 'libgcc' are used to perform unordered 23092 floating point compare and set flag operations. 23093 23094'-mcmov' 23095 Enable generation of conditional move ('l.cmov') instructions. By 23096 default the equivalent will be generated using set and branch. 23097 23098'-mror' 23099 Enable generation of rotate right ('l.ror') instructions. By 23100 default functions from 'libgcc' are used to perform rotate right 23101 operations. 23102 23103'-mrori' 23104 Enable generation of rotate right with immediate ('l.rori') 23105 instructions. By default functions from 'libgcc' are used to 23106 perform rotate right with immediate operations. 23107 23108'-msext' 23109 Enable generation of sign extension ('l.ext*') instructions. By 23110 default memory loads are used to perform sign extension. 23111 23112'-msfimm' 23113 Enable generation of compare and set flag with immediate ('l.sf*i') 23114 instructions. By default extra instructions will be generated to 23115 store the immediate to a register first. 23116 23117'-mshftimm' 23118 Enable generation of shift with immediate ('l.srai', 'l.srli', 23119 'l.slli') instructions. By default extra instructions will be 23120 generated to store the immediate to a register first. 23121 23122 23123File: gcc.info, Node: PDP-11 Options, Next: picoChip Options, Prev: OpenRISC Options, Up: Submodel Options 23124 231253.19.38 PDP-11 Options 23126---------------------- 23127 23128These options are defined for the PDP-11: 23129 23130'-mfpu' 23131 Use hardware FPP floating point. This is the default. (FIS 23132 floating point on the PDP-11/40 is not supported.) Implies -m45. 23133 23134'-msoft-float' 23135 Do not use hardware floating point. 23136 23137'-mac0' 23138 Return floating-point results in ac0 (fr0 in Unix assembler 23139 syntax). 23140 23141'-mno-ac0' 23142 Return floating-point results in memory. This is the default. 23143 23144'-m40' 23145 Generate code for a PDP-11/40. Implies -msoft-float -mno-split. 23146 23147'-m45' 23148 Generate code for a PDP-11/45. This is the default. 23149 23150'-m10' 23151 Generate code for a PDP-11/10. Implies -msoft-float -mno-split. 23152 23153'-mint16' 23154'-mno-int32' 23155 Use 16-bit 'int'. This is the default. 23156 23157'-mint32' 23158'-mno-int16' 23159 Use 32-bit 'int'. 23160 23161'-msplit' 23162 Target has split instruction and data space. Implies -m45. 23163 23164'-munix-asm' 23165 Use Unix assembler syntax. 23166 23167'-mdec-asm' 23168 Use DEC assembler syntax. 23169 23170'-mgnu-asm' 23171 Use GNU assembler syntax. This is the default. 23172 23173'-mlra' 23174 Use the new LRA register allocator. By default, the old "reload" 23175 allocator is used. 23176 23177 23178File: gcc.info, Node: picoChip Options, Next: PowerPC Options, Prev: PDP-11 Options, Up: Submodel Options 23179 231803.19.39 picoChip Options 23181------------------------ 23182 23183These '-m' options are defined for picoChip implementations: 23184 23185'-mae=AE_TYPE' 23186 Set the instruction set, register set, and instruction scheduling 23187 parameters for array element type AE_TYPE. Supported values for 23188 AE_TYPE are 'ANY', 'MUL', and 'MAC'. 23189 23190 '-mae=ANY' selects a completely generic AE type. Code generated 23191 with this option runs on any of the other AE types. The code is 23192 not as efficient as it would be if compiled for a specific AE type, 23193 and some types of operation (e.g., multiplication) do not work 23194 properly on all types of AE. 23195 23196 '-mae=MUL' selects a MUL AE type. This is the most useful AE type 23197 for compiled code, and is the default. 23198 23199 '-mae=MAC' selects a DSP-style MAC AE. Code compiled with this 23200 option may suffer from poor performance of byte (char) 23201 manipulation, since the DSP AE does not provide hardware support 23202 for byte load/stores. 23203 23204'-msymbol-as-address' 23205 Enable the compiler to directly use a symbol name as an address in 23206 a load/store instruction, without first loading it into a register. 23207 Typically, the use of this option generates larger programs, which 23208 run faster than when the option isn't used. However, the results 23209 vary from program to program, so it is left as a user option, 23210 rather than being permanently enabled. 23211 23212'-mno-inefficient-warnings' 23213 Disables warnings about the generation of inefficient code. These 23214 warnings can be generated, for example, when compiling code that 23215 performs byte-level memory operations on the MAC AE type. The MAC 23216 AE has no hardware support for byte-level memory operations, so all 23217 byte load/stores must be synthesized from word load/store 23218 operations. This is inefficient and a warning is generated to 23219 indicate that you should rewrite the code to avoid byte operations, 23220 or to target an AE type that has the necessary hardware support. 23221 This option disables these warnings. 23222 23223 23224File: gcc.info, Node: PowerPC Options, Next: PRU Options, Prev: picoChip Options, Up: Submodel Options 23225 232263.19.40 PowerPC Options 23227----------------------- 23228 23229These are listed under *Note RS/6000 and PowerPC Options::. 23230 23231 23232File: gcc.info, Node: PRU Options, Next: RISC-V Options, Prev: PowerPC Options, Up: Submodel Options 23233 232343.19.41 PRU Options 23235------------------- 23236 23237These command-line options are defined for PRU target: 23238 23239'-minrt' 23240 Link with a minimum runtime environment, with no support for static 23241 initializers and constructors. Using this option can significantly 23242 reduce the size of the final ELF binary. Beware that the compiler 23243 could still generate code with static initializers and 23244 constructors. It is up to the programmer to ensure that the source 23245 program will not use those features. 23246 23247'-mmcu=MCU' 23248 Specify the PRU MCU variant to use. Check Newlib for the exact 23249 list of supported MCUs. 23250 23251'-mno-relax' 23252 Make GCC pass the '--no-relax' command-line option to the linker 23253 instead of the '--relax' option. 23254 23255'-mloop' 23256 Allow (or do not allow) GCC to use the LOOP instruction. 23257 23258'-mabi=VARIANT' 23259 Specify the ABI variant to output code for. '-mabi=ti' selects the 23260 unmodified TI ABI while '-mabi=gnu' selects a GNU variant that 23261 copes more naturally with certain GCC assumptions. These are the 23262 differences: 23263 23264 'Function Pointer Size' 23265 TI ABI specifies that function (code) pointers are 16-bit, 23266 whereas GNU supports only 32-bit data and code pointers. 23267 23268 'Optional Return Value Pointer' 23269 Function return values larger than 64 bits are passed by using 23270 a hidden pointer as the first argument of the function. TI 23271 ABI, though, mandates that the pointer can be NULL in case the 23272 caller is not using the returned value. GNU always passes and 23273 expects a valid return value pointer. 23274 23275 The current '-mabi=ti' implementation simply raises a compile error 23276 when any of the above code constructs is detected. As a 23277 consequence the standard C library cannot be built and it is 23278 omitted when linking with '-mabi=ti'. 23279 23280 Relaxation is a GNU feature and for safety reasons is disabled when 23281 using '-mabi=ti'. The TI toolchain does not emit relocations for 23282 QBBx instructions, so the GNU linker cannot adjust them when 23283 shortening adjacent LDI32 pseudo instructions. 23284 23285 23286File: gcc.info, Node: RISC-V Options, Next: RL78 Options, Prev: PRU Options, Up: Submodel Options 23287 232883.19.42 RISC-V Options 23289---------------------- 23290 23291These command-line options are defined for RISC-V targets: 23292 23293'-mbranch-cost=N' 23294 Set the cost of branches to roughly N instructions. 23295 23296'-mplt' 23297'-mno-plt' 23298 When generating PIC code, do or don't allow the use of PLTs. 23299 Ignored for non-PIC. The default is '-mplt'. 23300 23301'-mabi=ABI-STRING' 23302 Specify integer and floating-point calling convention. ABI-STRING 23303 contains two parts: the size of integer types and the registers 23304 used for floating-point types. For example '-march=rv64ifd 23305 -mabi=lp64d' means that 'long' and pointers are 64-bit (implicitly 23306 defining 'int' to be 32-bit), and that floating-point values up to 23307 64 bits wide are passed in F registers. Contrast this with 23308 '-march=rv64ifd -mabi=lp64f', which still allows the compiler to 23309 generate code that uses the F and D extensions but only allows 23310 floating-point values up to 32 bits long to be passed in registers; 23311 or '-march=rv64ifd -mabi=lp64', in which no floating-point 23312 arguments will be passed in registers. 23313 23314 The default for this argument is system dependent, users who want a 23315 specific calling convention should specify one explicitly. The 23316 valid calling conventions are: 'ilp32', 'ilp32f', 'ilp32d', 'lp64', 23317 'lp64f', and 'lp64d'. Some calling conventions are impossible to 23318 implement on some ISAs: for example, '-march=rv32if -mabi=ilp32d' 23319 is invalid because the ABI requires 64-bit values be passed in F 23320 registers, but F registers are only 32 bits wide. There is also 23321 the 'ilp32e' ABI that can only be used with the 'rv32e' 23322 architecture. This ABI is not well specified at present, and is 23323 subject to change. 23324 23325'-mfdiv' 23326'-mno-fdiv' 23327 Do or don't use hardware floating-point divide and square root 23328 instructions. This requires the F or D extensions for 23329 floating-point registers. The default is to use them if the 23330 specified architecture has these instructions. 23331 23332'-mdiv' 23333'-mno-div' 23334 Do or don't use hardware instructions for integer division. This 23335 requires the M extension. The default is to use them if the 23336 specified architecture has these instructions. 23337 23338'-march=ISA-STRING' 23339 Generate code for given RISC-V ISA (e.g. 'rv64im'). ISA strings 23340 must be lower-case. Examples include 'rv64i', 'rv32g', 'rv32e', 23341 and 'rv32imaf'. 23342 23343'-mtune=PROCESSOR-STRING' 23344 Optimize the output for the given processor, specified by 23345 microarchitecture name. Permissible values for this option are: 23346 'rocket', 'sifive-3-series', 'sifive-5-series', 'sifive-7-series', 23347 and 'size'. 23348 23349 When '-mtune=' is not specified, the default is 'rocket'. 23350 23351 The 'size' choice is not intended for use by end-users. This is 23352 used when '-Os' is specified. It overrides the instruction cost 23353 info provided by '-mtune=', but does not override the pipeline 23354 info. This helps reduce code size while still giving good 23355 performance. 23356 23357'-mpreferred-stack-boundary=NUM' 23358 Attempt to keep the stack boundary aligned to a 2 raised to NUM 23359 byte boundary. If '-mpreferred-stack-boundary' is not specified, 23360 the default is 4 (16 bytes or 128-bits). 23361 23362 *Warning:* If you use this switch, then you must build all modules 23363 with the same value, including any libraries. This includes the 23364 system libraries and startup modules. 23365 23366'-msmall-data-limit=N' 23367 Put global and static data smaller than N bytes into a special 23368 section (on some targets). 23369 23370'-msave-restore' 23371'-mno-save-restore' 23372 Do or don't use smaller but slower prologue and epilogue code that 23373 uses library function calls. The default is to use fast inline 23374 prologues and epilogues. 23375 23376'-mstrict-align' 23377'-mno-strict-align' 23378 Do not or do generate unaligned memory accesses. The default is 23379 set depending on whether the processor we are optimizing for 23380 supports fast unaligned access or not. 23381 23382'-mcmodel=medlow' 23383 Generate code for the medium-low code model. The program and its 23384 statically defined symbols must lie within a single 2 GiB address 23385 range and must lie between absolute addresses -2 GiB and +2 GiB. 23386 Programs can be statically or dynamically linked. This is the 23387 default code model. 23388 23389'-mcmodel=medany' 23390 Generate code for the medium-any code model. The program and its 23391 statically defined symbols must be within any single 2 GiB address 23392 range. Programs can be statically or dynamically linked. 23393 23394'-mexplicit-relocs' 23395'-mno-exlicit-relocs' 23396 Use or do not use assembler relocation operators when dealing with 23397 symbolic addresses. The alternative is to use assembler macros 23398 instead, which may limit optimization. 23399 23400'-mrelax' 23401'-mno-relax' 23402 Take advantage of linker relaxations to reduce the number of 23403 instructions required to materialize symbol addresses. The default 23404 is to take advantage of linker relaxations. 23405 23406'-memit-attribute' 23407'-mno-emit-attribute' 23408 Emit (do not emit) RISC-V attribute to record extra information 23409 into ELF objects. This feature requires at least binutils 2.32. 23410 23411'-malign-data=TYPE' 23412 Control how GCC aligns variables and constants of array, structure, 23413 or union types. Supported values for TYPE are 'xlen' which uses x 23414 register width as the alignment value, and 'natural' which uses 23415 natural alignment. 'xlen' is the default. 23416 23417 23418File: gcc.info, Node: RL78 Options, Next: RS/6000 and PowerPC Options, Prev: RISC-V Options, Up: Submodel Options 23419 234203.19.43 RL78 Options 23421-------------------- 23422 23423'-msim' 23424 Links in additional target libraries to support operation within a 23425 simulator. 23426 23427'-mmul=none' 23428'-mmul=g10' 23429'-mmul=g13' 23430'-mmul=g14' 23431'-mmul=rl78' 23432 Specifies the type of hardware multiplication and division support 23433 to be used. The simplest is 'none', which uses software for both 23434 multiplication and division. This is the default. The 'g13' value 23435 is for the hardware multiply/divide peripheral found on the 23436 RL78/G13 (S2 core) targets. The 'g14' value selects the use of the 23437 multiplication and division instructions supported by the RL78/G14 23438 (S3 core) parts. The value 'rl78' is an alias for 'g14' and the 23439 value 'mg10' is an alias for 'none'. 23440 23441 In addition a C preprocessor macro is defined, based upon the 23442 setting of this option. Possible values are: '__RL78_MUL_NONE__', 23443 '__RL78_MUL_G13__' or '__RL78_MUL_G14__'. 23444 23445'-mcpu=g10' 23446'-mcpu=g13' 23447'-mcpu=g14' 23448'-mcpu=rl78' 23449 Specifies the RL78 core to target. The default is the G14 core, 23450 also known as an S3 core or just RL78. The G13 or S2 core does not 23451 have multiply or divide instructions, instead it uses a hardware 23452 peripheral for these operations. The G10 or S1 core does not have 23453 register banks, so it uses a different calling convention. 23454 23455 If this option is set it also selects the type of hardware multiply 23456 support to use, unless this is overridden by an explicit 23457 '-mmul=none' option on the command line. Thus specifying 23458 '-mcpu=g13' enables the use of the G13 hardware multiply peripheral 23459 and specifying '-mcpu=g10' disables the use of hardware 23460 multiplications altogether. 23461 23462 Note, although the RL78/G14 core is the default target, specifying 23463 '-mcpu=g14' or '-mcpu=rl78' on the command line does change the 23464 behavior of the toolchain since it also enables G14 hardware 23465 multiply support. If these options are not specified on the 23466 command line then software multiplication routines will be used 23467 even though the code targets the RL78 core. This is for backwards 23468 compatibility with older toolchains which did not have hardware 23469 multiply and divide support. 23470 23471 In addition a C preprocessor macro is defined, based upon the 23472 setting of this option. Possible values are: '__RL78_G10__', 23473 '__RL78_G13__' or '__RL78_G14__'. 23474 23475'-mg10' 23476'-mg13' 23477'-mg14' 23478'-mrl78' 23479 These are aliases for the corresponding '-mcpu=' option. They are 23480 provided for backwards compatibility. 23481 23482'-mallregs' 23483 Allow the compiler to use all of the available registers. By 23484 default registers 'r24..r31' are reserved for use in interrupt 23485 handlers. With this option enabled these registers can be used in 23486 ordinary functions as well. 23487 23488'-m64bit-doubles' 23489'-m32bit-doubles' 23490 Make the 'double' data type be 64 bits ('-m64bit-doubles') or 32 23491 bits ('-m32bit-doubles') in size. The default is 23492 '-m32bit-doubles'. 23493 23494'-msave-mduc-in-interrupts' 23495'-mno-save-mduc-in-interrupts' 23496 Specifies that interrupt handler functions should preserve the MDUC 23497 registers. This is only necessary if normal code might use the 23498 MDUC registers, for example because it performs multiplication and 23499 division operations. The default is to ignore the MDUC registers 23500 as this makes the interrupt handlers faster. The target option 23501 -mg13 needs to be passed for this to work as this feature is only 23502 available on the G13 target (S2 core). The MDUC registers will 23503 only be saved if the interrupt handler performs a multiplication or 23504 division operation or it calls another function. 23505 23506 23507File: gcc.info, Node: RS/6000 and PowerPC Options, Next: RX Options, Prev: RL78 Options, Up: Submodel Options 23508 235093.19.44 IBM RS/6000 and PowerPC Options 23510--------------------------------------- 23511 23512These '-m' options are defined for the IBM RS/6000 and PowerPC: 23513'-mpowerpc-gpopt' 23514'-mno-powerpc-gpopt' 23515'-mpowerpc-gfxopt' 23516'-mno-powerpc-gfxopt' 23517'-mpowerpc64' 23518'-mno-powerpc64' 23519'-mmfcrf' 23520'-mno-mfcrf' 23521'-mpopcntb' 23522'-mno-popcntb' 23523'-mpopcntd' 23524'-mno-popcntd' 23525'-mfprnd' 23526'-mno-fprnd' 23527'-mcmpb' 23528'-mno-cmpb' 23529'-mhard-dfp' 23530'-mno-hard-dfp' 23531 You use these options to specify which instructions are available 23532 on the processor you are using. The default value of these options 23533 is determined when configuring GCC. Specifying the 23534 '-mcpu=CPU_TYPE' overrides the specification of these options. We 23535 recommend you use the '-mcpu=CPU_TYPE' option rather than the 23536 options listed above. 23537 23538 Specifying '-mpowerpc-gpopt' allows GCC to use the optional PowerPC 23539 architecture instructions in the General Purpose group, including 23540 floating-point square root. Specifying '-mpowerpc-gfxopt' allows 23541 GCC to use the optional PowerPC architecture instructions in the 23542 Graphics group, including floating-point select. 23543 23544 The '-mmfcrf' option allows GCC to generate the move from condition 23545 register field instruction implemented on the POWER4 processor and 23546 other processors that support the PowerPC V2.01 architecture. The 23547 '-mpopcntb' option allows GCC to generate the popcount and 23548 double-precision FP reciprocal estimate instruction implemented on 23549 the POWER5 processor and other processors that support the PowerPC 23550 V2.02 architecture. The '-mpopcntd' option allows GCC to generate 23551 the popcount instruction implemented on the POWER7 processor and 23552 other processors that support the PowerPC V2.06 architecture. The 23553 '-mfprnd' option allows GCC to generate the FP round to integer 23554 instructions implemented on the POWER5+ processor and other 23555 processors that support the PowerPC V2.03 architecture. The 23556 '-mcmpb' option allows GCC to generate the compare bytes 23557 instruction implemented on the POWER6 processor and other 23558 processors that support the PowerPC V2.05 architecture. The 23559 '-mhard-dfp' option allows GCC to generate the decimal 23560 floating-point instructions implemented on some POWER processors. 23561 23562 The '-mpowerpc64' option allows GCC to generate the additional 23563 64-bit instructions that are found in the full PowerPC64 23564 architecture and to treat GPRs as 64-bit, doubleword quantities. 23565 GCC defaults to '-mno-powerpc64'. 23566 23567'-mcpu=CPU_TYPE' 23568 Set architecture type, register usage, and instruction scheduling 23569 parameters for machine type CPU_TYPE. Supported values for 23570 CPU_TYPE are '401', '403', '405', '405fp', '440', '440fp', '464', 23571 '464fp', '476', '476fp', '505', '601', '602', '603', '603e', '604', 23572 '604e', '620', '630', '740', '7400', '7450', '750', '801', '821', 23573 '823', '860', '970', '8540', 'a2', 'e300c2', 'e300c3', 'e500mc', 23574 'e500mc64', 'e5500', 'e6500', 'ec603e', 'G3', 'G4', 'G5', 'titan', 23575 'power3', 'power4', 'power5', 'power5+', 'power6', 'power6x', 23576 'power7', 'power8', 'power9', 'power10', 'powerpc', 'powerpc64', 23577 'powerpc64le', 'rs64', and 'native'. 23578 23579 '-mcpu=powerpc', '-mcpu=powerpc64', and '-mcpu=powerpc64le' specify 23580 pure 32-bit PowerPC (either endian), 64-bit big endian PowerPC and 23581 64-bit little endian PowerPC architecture machine types, with an 23582 appropriate, generic processor model assumed for scheduling 23583 purposes. 23584 23585 Specifying 'native' as cpu type detects and selects the 23586 architecture option that corresponds to the host processor of the 23587 system performing the compilation. '-mcpu=native' has no effect if 23588 GCC does not recognize the processor. 23589 23590 The other options specify a specific processor. Code generated 23591 under those options runs best on that processor, and may not run at 23592 all on others. 23593 23594 The '-mcpu' options automatically enable or disable the following 23595 options: 23596 23597 -maltivec -mfprnd -mhard-float -mmfcrf -mmultiple 23598 -mpopcntb -mpopcntd -mpowerpc64 23599 -mpowerpc-gpopt -mpowerpc-gfxopt 23600 -mmulhw -mdlmzb -mmfpgpr -mvsx 23601 -mcrypto -mhtm -mpower8-fusion -mpower8-vector 23602 -mquad-memory -mquad-memory-atomic -mfloat128 23603 -mfloat128-hardware -mprefixed -mpcrel -mmma 23604 23605 The particular options set for any particular CPU varies between 23606 compiler versions, depending on what setting seems to produce 23607 optimal code for that CPU; it doesn't necessarily reflect the 23608 actual hardware's capabilities. If you wish to set an individual 23609 option to a particular value, you may specify it after the '-mcpu' 23610 option, like '-mcpu=970 -mno-altivec'. 23611 23612 On AIX, the '-maltivec' and '-mpowerpc64' options are not enabled 23613 or disabled by the '-mcpu' option at present because AIX does not 23614 have full support for these options. You may still enable or 23615 disable them individually if you're sure it'll work in your 23616 environment. 23617 23618'-mtune=CPU_TYPE' 23619 Set the instruction scheduling parameters for machine type 23620 CPU_TYPE, but do not set the architecture type or register usage, 23621 as '-mcpu=CPU_TYPE' does. The same values for CPU_TYPE are used 23622 for '-mtune' as for '-mcpu'. If both are specified, the code 23623 generated uses the architecture and registers set by '-mcpu', but 23624 the scheduling parameters set by '-mtune'. 23625 23626'-mcmodel=small' 23627 Generate PowerPC64 code for the small model: The TOC is limited to 23628 64k. 23629 23630'-mcmodel=medium' 23631 Generate PowerPC64 code for the medium model: The TOC and other 23632 static data may be up to a total of 4G in size. This is the 23633 default for 64-bit Linux. 23634 23635'-mcmodel=large' 23636 Generate PowerPC64 code for the large model: The TOC may be up to 23637 4G in size. Other data and code is only limited by the 64-bit 23638 address space. 23639 23640'-maltivec' 23641'-mno-altivec' 23642 Generate code that uses (does not use) AltiVec instructions, and 23643 also enable the use of built-in functions that allow more direct 23644 access to the AltiVec instruction set. You may also need to set 23645 '-mabi=altivec' to adjust the current ABI with AltiVec ABI 23646 enhancements. 23647 23648 When '-maltivec' is used, the element order for AltiVec intrinsics 23649 such as 'vec_splat', 'vec_extract', and 'vec_insert' match array 23650 element order corresponding to the endianness of the target. That 23651 is, element zero identifies the leftmost element in a vector 23652 register when targeting a big-endian platform, and identifies the 23653 rightmost element in a vector register when targeting a 23654 little-endian platform. 23655 23656'-mvrsave' 23657'-mno-vrsave' 23658 Generate VRSAVE instructions when generating AltiVec code. 23659 23660'-msecure-plt' 23661 Generate code that allows 'ld' and 'ld.so' to build executables and 23662 shared libraries with non-executable '.plt' and '.got' sections. 23663 This is a PowerPC 32-bit SYSV ABI option. 23664 23665'-mbss-plt' 23666 Generate code that uses a BSS '.plt' section that 'ld.so' fills in, 23667 and requires '.plt' and '.got' sections that are both writable and 23668 executable. This is a PowerPC 32-bit SYSV ABI option. 23669 23670'-misel' 23671'-mno-isel' 23672 This switch enables or disables the generation of ISEL 23673 instructions. 23674 23675'-mvsx' 23676'-mno-vsx' 23677 Generate code that uses (does not use) vector/scalar (VSX) 23678 instructions, and also enable the use of built-in functions that 23679 allow more direct access to the VSX instruction set. 23680 23681'-mcrypto' 23682'-mno-crypto' 23683 Enable the use (disable) of the built-in functions that allow 23684 direct access to the cryptographic instructions that were added in 23685 version 2.07 of the PowerPC ISA. 23686 23687'-mhtm' 23688'-mno-htm' 23689 Enable (disable) the use of the built-in functions that allow 23690 direct access to the Hardware Transactional Memory (HTM) 23691 instructions that were added in version 2.07 of the PowerPC ISA. 23692 23693'-mpower8-fusion' 23694'-mno-power8-fusion' 23695 Generate code that keeps (does not keeps) some integer operations 23696 adjacent so that the instructions can be fused together on power8 23697 and later processors. 23698 23699'-mpower8-vector' 23700'-mno-power8-vector' 23701 Generate code that uses (does not use) the vector and scalar 23702 instructions that were added in version 2.07 of the PowerPC ISA. 23703 Also enable the use of built-in functions that allow more direct 23704 access to the vector instructions. 23705 23706'-mquad-memory' 23707'-mno-quad-memory' 23708 Generate code that uses (does not use) the non-atomic quad word 23709 memory instructions. The '-mquad-memory' option requires use of 23710 64-bit mode. 23711 23712'-mquad-memory-atomic' 23713'-mno-quad-memory-atomic' 23714 Generate code that uses (does not use) the atomic quad word memory 23715 instructions. The '-mquad-memory-atomic' option requires use of 23716 64-bit mode. 23717 23718'-mfloat128' 23719'-mno-float128' 23720 Enable/disable the __FLOAT128 keyword for IEEE 128-bit floating 23721 point and use either software emulation for IEEE 128-bit floating 23722 point or hardware instructions. 23723 23724 The VSX instruction set ('-mvsx') must be enabled to use the IEEE 23725 128-bit floating point support. The IEEE 128-bit floating point is 23726 only supported on Linux. 23727 23728 The default for '-mfloat128' is enabled on PowerPC Linux systems 23729 using the VSX instruction set, and disabled on other systems. 23730 23731 If you use the ISA 3.0 instruction set ('-mpower9-vector' or 23732 '-mcpu=power9') on a 64-bit system, the IEEE 128-bit floating point 23733 support will also enable the generation of ISA 3.0 IEEE 128-bit 23734 floating point instructions. Otherwise, if you do not specify to 23735 generate ISA 3.0 instructions or you are targeting a 32-bit big 23736 endian system, IEEE 128-bit floating point will be done with 23737 software emulation. 23738 23739'-mfloat128-hardware' 23740'-mno-float128-hardware' 23741 Enable/disable using ISA 3.0 hardware instructions to support the 23742 __FLOAT128 data type. 23743 23744 The default for '-mfloat128-hardware' is enabled on PowerPC Linux 23745 systems using the ISA 3.0 instruction set, and disabled on other 23746 systems. 23747 23748'-m32' 23749'-m64' 23750 Generate code for 32-bit or 64-bit environments of Darwin and SVR4 23751 targets (including GNU/Linux). The 32-bit environment sets int, 23752 long and pointer to 32 bits and generates code that runs on any 23753 PowerPC variant. The 64-bit environment sets int to 32 bits and 23754 long and pointer to 64 bits, and generates code for PowerPC64, as 23755 for '-mpowerpc64'. 23756 23757'-mfull-toc' 23758'-mno-fp-in-toc' 23759'-mno-sum-in-toc' 23760'-mminimal-toc' 23761 Modify generation of the TOC (Table Of Contents), which is created 23762 for every executable file. The '-mfull-toc' option is selected by 23763 default. In that case, GCC allocates at least one TOC entry for 23764 each unique non-automatic variable reference in your program. GCC 23765 also places floating-point constants in the TOC. However, only 23766 16,384 entries are available in the TOC. 23767 23768 If you receive a linker error message that saying you have 23769 overflowed the available TOC space, you can reduce the amount of 23770 TOC space used with the '-mno-fp-in-toc' and '-mno-sum-in-toc' 23771 options. '-mno-fp-in-toc' prevents GCC from putting floating-point 23772 constants in the TOC and '-mno-sum-in-toc' forces GCC to generate 23773 code to calculate the sum of an address and a constant at run time 23774 instead of putting that sum into the TOC. You may specify one or 23775 both of these options. Each causes GCC to produce very slightly 23776 slower and larger code at the expense of conserving TOC space. 23777 23778 If you still run out of space in the TOC even when you specify both 23779 of these options, specify '-mminimal-toc' instead. This option 23780 causes GCC to make only one TOC entry for every file. When you 23781 specify this option, GCC produces code that is slower and larger 23782 but which uses extremely little TOC space. You may wish to use 23783 this option only on files that contain less frequently-executed 23784 code. 23785 23786'-maix64' 23787'-maix32' 23788 Enable 64-bit AIX ABI and calling convention: 64-bit pointers, 23789 64-bit 'long' type, and the infrastructure needed to support them. 23790 Specifying '-maix64' implies '-mpowerpc64', while '-maix32' 23791 disables the 64-bit ABI and implies '-mno-powerpc64'. GCC defaults 23792 to '-maix32'. 23793 23794'-mxl-compat' 23795'-mno-xl-compat' 23796 Produce code that conforms more closely to IBM XL compiler 23797 semantics when using AIX-compatible ABI. Pass floating-point 23798 arguments to prototyped functions beyond the register save area 23799 (RSA) on the stack in addition to argument FPRs. Do not assume 23800 that most significant double in 128-bit long double value is 23801 properly rounded when comparing values and converting to double. 23802 Use XL symbol names for long double support routines. 23803 23804 The AIX calling convention was extended but not initially 23805 documented to handle an obscure K&R C case of calling a function 23806 that takes the address of its arguments with fewer arguments than 23807 declared. IBM XL compilers access floating-point arguments that do 23808 not fit in the RSA from the stack when a subroutine is compiled 23809 without optimization. Because always storing floating-point 23810 arguments on the stack is inefficient and rarely needed, this 23811 option is not enabled by default and only is necessary when calling 23812 subroutines compiled by IBM XL compilers without optimization. 23813 23814'-mpe' 23815 Support "IBM RS/6000 SP" "Parallel Environment" (PE). Link an 23816 application written to use message passing with special startup 23817 code to enable the application to run. The system must have PE 23818 installed in the standard location ('/usr/lpp/ppe.poe/'), or the 23819 'specs' file must be overridden with the '-specs=' option to 23820 specify the appropriate directory location. The Parallel 23821 Environment does not support threads, so the '-mpe' option and the 23822 '-pthread' option are incompatible. 23823 23824'-malign-natural' 23825'-malign-power' 23826 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option 23827 '-malign-natural' overrides the ABI-defined alignment of larger 23828 types, such as floating-point doubles, on their natural size-based 23829 boundary. The option '-malign-power' instructs GCC to follow the 23830 ABI-specified alignment rules. GCC defaults to the standard 23831 alignment defined in the ABI. 23832 23833 On 64-bit Darwin, natural alignment is the default, and 23834 '-malign-power' is not supported. 23835 23836'-msoft-float' 23837'-mhard-float' 23838 Generate code that does not use (uses) the floating-point register 23839 set. Software floating-point emulation is provided if you use the 23840 '-msoft-float' option, and pass the option to GCC when linking. 23841 23842'-mmultiple' 23843'-mno-multiple' 23844 Generate code that uses (does not use) the load multiple word 23845 instructions and the store multiple word instructions. These 23846 instructions are generated by default on POWER systems, and not 23847 generated on PowerPC systems. Do not use '-mmultiple' on 23848 little-endian PowerPC systems, since those instructions do not work 23849 when the processor is in little-endian mode. The exceptions are 23850 PPC740 and PPC750 which permit these instructions in little-endian 23851 mode. 23852 23853'-mupdate' 23854'-mno-update' 23855 Generate code that uses (does not use) the load or store 23856 instructions that update the base register to the address of the 23857 calculated memory location. These instructions are generated by 23858 default. If you use '-mno-update', there is a small window between 23859 the time that the stack pointer is updated and the address of the 23860 previous frame is stored, which means code that walks the stack 23861 frame across interrupts or signals may get corrupted data. 23862 23863'-mavoid-indexed-addresses' 23864'-mno-avoid-indexed-addresses' 23865 Generate code that tries to avoid (not avoid) the use of indexed 23866 load or store instructions. These instructions can incur a 23867 performance penalty on Power6 processors in certain situations, 23868 such as when stepping through large arrays that cross a 16M 23869 boundary. This option is enabled by default when targeting Power6 23870 and disabled otherwise. 23871 23872'-mfused-madd' 23873'-mno-fused-madd' 23874 Generate code that uses (does not use) the floating-point multiply 23875 and accumulate instructions. These instructions are generated by 23876 default if hardware floating point is used. The machine-dependent 23877 '-mfused-madd' option is now mapped to the machine-independent 23878 '-ffp-contract=fast' option, and '-mno-fused-madd' is mapped to 23879 '-ffp-contract=off'. 23880 23881'-mmulhw' 23882'-mno-mulhw' 23883 Generate code that uses (does not use) the half-word multiply and 23884 multiply-accumulate instructions on the IBM 405, 440, 464 and 476 23885 processors. These instructions are generated by default when 23886 targeting those processors. 23887 23888'-mdlmzb' 23889'-mno-dlmzb' 23890 Generate code that uses (does not use) the string-search 'dlmzb' 23891 instruction on the IBM 405, 440, 464 and 476 processors. This 23892 instruction is generated by default when targeting those 23893 processors. 23894 23895'-mno-bit-align' 23896'-mbit-align' 23897 On System V.4 and embedded PowerPC systems do not (do) force 23898 structures and unions that contain bit-fields to be aligned to the 23899 base type of the bit-field. 23900 23901 For example, by default a structure containing nothing but 8 23902 'unsigned' bit-fields of length 1 is aligned to a 4-byte boundary 23903 and has a size of 4 bytes. By using '-mno-bit-align', the 23904 structure is aligned to a 1-byte boundary and is 1 byte in size. 23905 23906'-mno-strict-align' 23907'-mstrict-align' 23908 On System V.4 and embedded PowerPC systems do not (do) assume that 23909 unaligned memory references are handled by the system. 23910 23911'-mrelocatable' 23912'-mno-relocatable' 23913 Generate code that allows (does not allow) a static executable to 23914 be relocated to a different address at run time. A simple embedded 23915 PowerPC system loader should relocate the entire contents of 23916 '.got2' and 4-byte locations listed in the '.fixup' section, a 23917 table of 32-bit addresses generated by this option. For this to 23918 work, all objects linked together must be compiled with 23919 '-mrelocatable' or '-mrelocatable-lib'. '-mrelocatable' code 23920 aligns the stack to an 8-byte boundary. 23921 23922'-mrelocatable-lib' 23923'-mno-relocatable-lib' 23924 Like '-mrelocatable', '-mrelocatable-lib' generates a '.fixup' 23925 section to allow static executables to be relocated at run time, 23926 but '-mrelocatable-lib' does not use the smaller stack alignment of 23927 '-mrelocatable'. Objects compiled with '-mrelocatable-lib' may be 23928 linked with objects compiled with any combination of the 23929 '-mrelocatable' options. 23930 23931'-mno-toc' 23932'-mtoc' 23933 On System V.4 and embedded PowerPC systems do not (do) assume that 23934 register 2 contains a pointer to a global area pointing to the 23935 addresses used in the program. 23936 23937'-mlittle' 23938'-mlittle-endian' 23939 On System V.4 and embedded PowerPC systems compile code for the 23940 processor in little-endian mode. The '-mlittle-endian' option is 23941 the same as '-mlittle'. 23942 23943'-mbig' 23944'-mbig-endian' 23945 On System V.4 and embedded PowerPC systems compile code for the 23946 processor in big-endian mode. The '-mbig-endian' option is the 23947 same as '-mbig'. 23948 23949'-mdynamic-no-pic' 23950 On Darwin and Mac OS X systems, compile code so that it is not 23951 relocatable, but that its external references are relocatable. The 23952 resulting code is suitable for applications, but not shared 23953 libraries. 23954 23955'-msingle-pic-base' 23956 Treat the register used for PIC addressing as read-only, rather 23957 than loading it in the prologue for each function. The runtime 23958 system is responsible for initializing this register with an 23959 appropriate value before execution begins. 23960 23961'-mprioritize-restricted-insns=PRIORITY' 23962 This option controls the priority that is assigned to dispatch-slot 23963 restricted instructions during the second scheduling pass. The 23964 argument PRIORITY takes the value '0', '1', or '2' to assign no, 23965 highest, or second-highest (respectively) priority to dispatch-slot 23966 restricted instructions. 23967 23968'-msched-costly-dep=DEPENDENCE_TYPE' 23969 This option controls which dependences are considered costly by the 23970 target during instruction scheduling. The argument DEPENDENCE_TYPE 23971 takes one of the following values: 23972 23973 'no' 23974 No dependence is costly. 23975 23976 'all' 23977 All dependences are costly. 23978 23979 'true_store_to_load' 23980 A true dependence from store to load is costly. 23981 23982 'store_to_load' 23983 Any dependence from store to load is costly. 23984 23985 NUMBER 23986 Any dependence for which the latency is greater than or equal 23987 to NUMBER is costly. 23988 23989'-minsert-sched-nops=SCHEME' 23990 This option controls which NOP insertion scheme is used during the 23991 second scheduling pass. The argument SCHEME takes one of the 23992 following values: 23993 23994 'no' 23995 Don't insert NOPs. 23996 23997 'pad' 23998 Pad with NOPs any dispatch group that has vacant issue slots, 23999 according to the scheduler's grouping. 24000 24001 'regroup_exact' 24002 Insert NOPs to force costly dependent insns into separate 24003 groups. Insert exactly as many NOPs as needed to force an 24004 insn to a new group, according to the estimated processor 24005 grouping. 24006 24007 NUMBER 24008 Insert NOPs to force costly dependent insns into separate 24009 groups. Insert NUMBER NOPs to force an insn to a new group. 24010 24011'-mcall-sysv' 24012 On System V.4 and embedded PowerPC systems compile code using 24013 calling conventions that adhere to the March 1995 draft of the 24014 System V Application Binary Interface, PowerPC processor 24015 supplement. This is the default unless you configured GCC using 24016 'powerpc-*-eabiaix'. 24017 24018'-mcall-sysv-eabi' 24019'-mcall-eabi' 24020 Specify both '-mcall-sysv' and '-meabi' options. 24021 24022'-mcall-sysv-noeabi' 24023 Specify both '-mcall-sysv' and '-mno-eabi' options. 24024 24025'-mcall-aixdesc' 24026 On System V.4 and embedded PowerPC systems compile code for the AIX 24027 operating system. 24028 24029'-mcall-linux' 24030 On System V.4 and embedded PowerPC systems compile code for the 24031 Linux-based GNU system. 24032 24033'-mcall-freebsd' 24034 On System V.4 and embedded PowerPC systems compile code for the 24035 FreeBSD operating system. 24036 24037'-mcall-netbsd' 24038 On System V.4 and embedded PowerPC systems compile code for the 24039 NetBSD operating system. 24040 24041'-mcall-openbsd' 24042 On System V.4 and embedded PowerPC systems compile code for the 24043 OpenBSD operating system. 24044 24045'-mtraceback=TRACEBACK_TYPE' 24046 Select the type of traceback table. Valid values for 24047 TRACEBACK_TYPE are 'full', 'part', and 'no'. 24048 24049'-maix-struct-return' 24050 Return all structures in memory (as specified by the AIX ABI). 24051 24052'-msvr4-struct-return' 24053 Return structures smaller than 8 bytes in registers (as specified 24054 by the SVR4 ABI). 24055 24056'-mabi=ABI-TYPE' 24057 Extend the current ABI with a particular extension, or remove such 24058 extension. Valid values are: 'altivec', 'no-altivec', 24059 'ibmlongdouble', 'ieeelongdouble', 'elfv1', 'elfv2', and for AIX: 24060 'vec-extabi', 'vec-default'. 24061 24062'-mabi=ibmlongdouble' 24063 Change the current ABI to use IBM extended-precision long double. 24064 This is not likely to work if your system defaults to using IEEE 24065 extended-precision long double. If you change the long double type 24066 from IEEE extended-precision, the compiler will issue a warning 24067 unless you use the '-Wno-psabi' option. Requires 24068 '-mlong-double-128' to be enabled. 24069 24070'-mabi=ieeelongdouble' 24071 Change the current ABI to use IEEE extended-precision long double. 24072 This is not likely to work if your system defaults to using IBM 24073 extended-precision long double. If you change the long double type 24074 from IBM extended-precision, the compiler will issue a warning 24075 unless you use the '-Wno-psabi' option. Requires 24076 '-mlong-double-128' to be enabled. 24077 24078'-mabi=elfv1' 24079 Change the current ABI to use the ELFv1 ABI. This is the default 24080 ABI for big-endian PowerPC 64-bit Linux. Overriding the default 24081 ABI requires special system support and is likely to fail in 24082 spectacular ways. 24083 24084'-mabi=elfv2' 24085 Change the current ABI to use the ELFv2 ABI. This is the default 24086 ABI for little-endian PowerPC 64-bit Linux. Overriding the default 24087 ABI requires special system support and is likely to fail in 24088 spectacular ways. 24089 24090'-mgnu-attribute' 24091'-mno-gnu-attribute' 24092 Emit .gnu_attribute assembly directives to set tag/value pairs in a 24093 .gnu.attributes section that specify ABI variations in function 24094 parameters or return values. 24095 24096'-mprototype' 24097'-mno-prototype' 24098 On System V.4 and embedded PowerPC systems assume that all calls to 24099 variable argument functions are properly prototyped. Otherwise, 24100 the compiler must insert an instruction before every non-prototyped 24101 call to set or clear bit 6 of the condition code register ('CR') to 24102 indicate whether floating-point values are passed in the 24103 floating-point registers in case the function takes variable 24104 arguments. With '-mprototype', only calls to prototyped variable 24105 argument functions set or clear the bit. 24106 24107'-msim' 24108 On embedded PowerPC systems, assume that the startup module is 24109 called 'sim-crt0.o' and that the standard C libraries are 24110 'libsim.a' and 'libc.a'. This is the default for 24111 'powerpc-*-eabisim' configurations. 24112 24113'-mmvme' 24114 On embedded PowerPC systems, assume that the startup module is 24115 called 'crt0.o' and the standard C libraries are 'libmvme.a' and 24116 'libc.a'. 24117 24118'-mads' 24119 On embedded PowerPC systems, assume that the startup module is 24120 called 'crt0.o' and the standard C libraries are 'libads.a' and 24121 'libc.a'. 24122 24123'-myellowknife' 24124 On embedded PowerPC systems, assume that the startup module is 24125 called 'crt0.o' and the standard C libraries are 'libyk.a' and 24126 'libc.a'. 24127 24128'-mvxworks' 24129 On System V.4 and embedded PowerPC systems, specify that you are 24130 compiling for a VxWorks system. 24131 24132'-memb' 24133 On embedded PowerPC systems, set the 'PPC_EMB' bit in the ELF flags 24134 header to indicate that 'eabi' extended relocations are used. 24135 24136'-meabi' 24137'-mno-eabi' 24138 On System V.4 and embedded PowerPC systems do (do not) adhere to 24139 the Embedded Applications Binary Interface (EABI), which is a set 24140 of modifications to the System V.4 specifications. Selecting 24141 '-meabi' means that the stack is aligned to an 8-byte boundary, a 24142 function '__eabi' is called from 'main' to set up the EABI 24143 environment, and the '-msdata' option can use both 'r2' and 'r13' 24144 to point to two separate small data areas. Selecting '-mno-eabi' 24145 means that the stack is aligned to a 16-byte boundary, no EABI 24146 initialization function is called from 'main', and the '-msdata' 24147 option only uses 'r13' to point to a single small data area. The 24148 '-meabi' option is on by default if you configured GCC using one of 24149 the 'powerpc*-*-eabi*' options. 24150 24151'-msdata=eabi' 24152 On System V.4 and embedded PowerPC systems, put small initialized 24153 'const' global and static data in the '.sdata2' section, which is 24154 pointed to by register 'r2'. Put small initialized non-'const' 24155 global and static data in the '.sdata' section, which is pointed to 24156 by register 'r13'. Put small uninitialized global and static data 24157 in the '.sbss' section, which is adjacent to the '.sdata' section. 24158 The '-msdata=eabi' option is incompatible with the '-mrelocatable' 24159 option. The '-msdata=eabi' option also sets the '-memb' option. 24160 24161'-msdata=sysv' 24162 On System V.4 and embedded PowerPC systems, put small global and 24163 static data in the '.sdata' section, which is pointed to by 24164 register 'r13'. Put small uninitialized global and static data in 24165 the '.sbss' section, which is adjacent to the '.sdata' section. 24166 The '-msdata=sysv' option is incompatible with the '-mrelocatable' 24167 option. 24168 24169'-msdata=default' 24170'-msdata' 24171 On System V.4 and embedded PowerPC systems, if '-meabi' is used, 24172 compile code the same as '-msdata=eabi', otherwise compile code the 24173 same as '-msdata=sysv'. 24174 24175'-msdata=data' 24176 On System V.4 and embedded PowerPC systems, put small global data 24177 in the '.sdata' section. Put small uninitialized global data in 24178 the '.sbss' section. Do not use register 'r13' to address small 24179 data however. This is the default behavior unless other '-msdata' 24180 options are used. 24181 24182'-msdata=none' 24183'-mno-sdata' 24184 On embedded PowerPC systems, put all initialized global and static 24185 data in the '.data' section, and all uninitialized data in the 24186 '.bss' section. 24187 24188'-mreadonly-in-sdata' 24189 Put read-only objects in the '.sdata' section as well. This is the 24190 default. 24191 24192'-mblock-move-inline-limit=NUM' 24193 Inline all block moves (such as calls to 'memcpy' or structure 24194 copies) less than or equal to NUM bytes. The minimum value for NUM 24195 is 32 bytes on 32-bit targets and 64 bytes on 64-bit targets. The 24196 default value is target-specific. 24197 24198'-mblock-compare-inline-limit=NUM' 24199 Generate non-looping inline code for all block compares (such as 24200 calls to 'memcmp' or structure compares) less than or equal to NUM 24201 bytes. If NUM is 0, all inline expansion (non-loop and loop) of 24202 block compare is disabled. The default value is target-specific. 24203 24204'-mblock-compare-inline-loop-limit=NUM' 24205 Generate an inline expansion using loop code for all block compares 24206 that are less than or equal to NUM bytes, but greater than the 24207 limit for non-loop inline block compare expansion. If the block 24208 length is not constant, at most NUM bytes will be compared before 24209 'memcmp' is called to compare the remainder of the block. The 24210 default value is target-specific. 24211 24212'-mstring-compare-inline-limit=NUM' 24213 Compare at most NUM string bytes with inline code. If the 24214 difference or end of string is not found at the end of the inline 24215 compare a call to 'strcmp' or 'strncmp' will take care of the rest 24216 of the comparison. The default is 64 bytes. 24217 24218'-G NUM' 24219 On embedded PowerPC systems, put global and static items less than 24220 or equal to NUM bytes into the small data or BSS sections instead 24221 of the normal data or BSS section. By default, NUM is 8. The '-G 24222 NUM' switch is also passed to the linker. All modules should be 24223 compiled with the same '-G NUM' value. 24224 24225'-mregnames' 24226'-mno-regnames' 24227 On System V.4 and embedded PowerPC systems do (do not) emit 24228 register names in the assembly language output using symbolic 24229 forms. 24230 24231'-mlongcall' 24232'-mno-longcall' 24233 By default assume that all calls are far away so that a longer and 24234 more expensive calling sequence is required. This is required for 24235 calls farther than 32 megabytes (33,554,432 bytes) from the current 24236 location. A short call is generated if the compiler knows the call 24237 cannot be that far away. This setting can be overridden by the 24238 'shortcall' function attribute, or by '#pragma longcall(0)'. 24239 24240 Some linkers are capable of detecting out-of-range calls and 24241 generating glue code on the fly. On these systems, long calls are 24242 unnecessary and generate slower code. As of this writing, the AIX 24243 linker can do this, as can the GNU linker for PowerPC/64. It is 24244 planned to add this feature to the GNU linker for 32-bit PowerPC 24245 systems as well. 24246 24247 On PowerPC64 ELFv2 and 32-bit PowerPC systems with newer GNU 24248 linkers, GCC can generate long calls using an inline PLT call 24249 sequence (see '-mpltseq'). PowerPC with '-mbss-plt' and PowerPC64 24250 ELFv1 (big-endian) do not support inline PLT calls. 24251 24252 On Darwin/PPC systems, '#pragma longcall' generates 'jbsr callee, 24253 L42', plus a "branch island" (glue code). The two target addresses 24254 represent the callee and the branch island. The Darwin/PPC linker 24255 prefers the first address and generates a 'bl callee' if the PPC 24256 'bl' instruction reaches the callee directly; otherwise, the linker 24257 generates 'bl L42' to call the branch island. The branch island is 24258 appended to the body of the calling function; it computes the full 24259 32-bit address of the callee and jumps to it. 24260 24261 On Mach-O (Darwin) systems, this option directs the compiler emit 24262 to the glue for every direct call, and the Darwin linker decides 24263 whether to use or discard it. 24264 24265 In the future, GCC may ignore all longcall specifications when the 24266 linker is known to generate glue. 24267 24268'-mpltseq' 24269'-mno-pltseq' 24270 Implement (do not implement) -fno-plt and long calls using an 24271 inline PLT call sequence that supports lazy linking and long calls 24272 to functions in dlopen'd shared libraries. Inline PLT calls are 24273 only supported on PowerPC64 ELFv2 and 32-bit PowerPC systems with 24274 newer GNU linkers, and are enabled by default if the support is 24275 detected when configuring GCC, and, in the case of 32-bit PowerPC, 24276 if GCC is configured with '--enable-secureplt'. '-mpltseq' code 24277 and '-mbss-plt' 32-bit PowerPC relocatable objects may not be 24278 linked together. 24279 24280'-mtls-markers' 24281'-mno-tls-markers' 24282 Mark (do not mark) calls to '__tls_get_addr' with a relocation 24283 specifying the function argument. The relocation allows the linker 24284 to reliably associate function call with argument setup 24285 instructions for TLS optimization, which in turn allows GCC to 24286 better schedule the sequence. 24287 24288'-mrecip' 24289'-mno-recip' 24290 This option enables use of the reciprocal estimate and reciprocal 24291 square root estimate instructions with additional Newton-Raphson 24292 steps to increase precision instead of doing a divide or square 24293 root and divide for floating-point arguments. You should use the 24294 '-ffast-math' option when using '-mrecip' (or at least 24295 '-funsafe-math-optimizations', '-ffinite-math-only', 24296 '-freciprocal-math' and '-fno-trapping-math'). Note that while the 24297 throughput of the sequence is generally higher than the throughput 24298 of the non-reciprocal instruction, the precision of the sequence 24299 can be decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 24300 0.99999994) for reciprocal square roots. 24301 24302'-mrecip=OPT' 24303 This option controls which reciprocal estimate instructions may be 24304 used. OPT is a comma-separated list of options, which may be 24305 preceded by a '!' to invert the option: 24306 24307 'all' 24308 Enable all estimate instructions. 24309 24310 'default' 24311 Enable the default instructions, equivalent to '-mrecip'. 24312 24313 'none' 24314 Disable all estimate instructions, equivalent to '-mno-recip'. 24315 24316 'div' 24317 Enable the reciprocal approximation instructions for both 24318 single and double precision. 24319 24320 'divf' 24321 Enable the single-precision reciprocal approximation 24322 instructions. 24323 24324 'divd' 24325 Enable the double-precision reciprocal approximation 24326 instructions. 24327 24328 'rsqrt' 24329 Enable the reciprocal square root approximation instructions 24330 for both single and double precision. 24331 24332 'rsqrtf' 24333 Enable the single-precision reciprocal square root 24334 approximation instructions. 24335 24336 'rsqrtd' 24337 Enable the double-precision reciprocal square root 24338 approximation instructions. 24339 24340 So, for example, '-mrecip=all,!rsqrtd' enables all of the 24341 reciprocal estimate instructions, except for the 'FRSQRTE', 24342 'XSRSQRTEDP', and 'XVRSQRTEDP' instructions which handle the 24343 double-precision reciprocal square root calculations. 24344 24345'-mrecip-precision' 24346'-mno-recip-precision' 24347 Assume (do not assume) that the reciprocal estimate instructions 24348 provide higher-precision estimates than is mandated by the PowerPC 24349 ABI. Selecting '-mcpu=power6', '-mcpu=power7' or '-mcpu=power8' 24350 automatically selects '-mrecip-precision'. The double-precision 24351 square root estimate instructions are not generated by default on 24352 low-precision machines, since they do not provide an estimate that 24353 converges after three steps. 24354 24355'-mveclibabi=TYPE' 24356 Specifies the ABI type to use for vectorizing intrinsics using an 24357 external library. The only type supported at present is 'mass', 24358 which specifies to use IBM's Mathematical Acceleration Subsystem 24359 (MASS) libraries for vectorizing intrinsics using external 24360 libraries. GCC currently emits calls to 'acosd2', 'acosf4', 24361 'acoshd2', 'acoshf4', 'asind2', 'asinf4', 'asinhd2', 'asinhf4', 24362 'atan2d2', 'atan2f4', 'atand2', 'atanf4', 'atanhd2', 'atanhf4', 24363 'cbrtd2', 'cbrtf4', 'cosd2', 'cosf4', 'coshd2', 'coshf4', 'erfcd2', 24364 'erfcf4', 'erfd2', 'erff4', 'exp2d2', 'exp2f4', 'expd2', 'expf4', 24365 'expm1d2', 'expm1f4', 'hypotd2', 'hypotf4', 'lgammad2', 'lgammaf4', 24366 'log10d2', 'log10f4', 'log1pd2', 'log1pf4', 'log2d2', 'log2f4', 24367 'logd2', 'logf4', 'powd2', 'powf4', 'sind2', 'sinf4', 'sinhd2', 24368 'sinhf4', 'sqrtd2', 'sqrtf4', 'tand2', 'tanf4', 'tanhd2', and 24369 'tanhf4' when generating code for power7. Both '-ftree-vectorize' 24370 and '-funsafe-math-optimizations' must also be enabled. The MASS 24371 libraries must be specified at link time. 24372 24373'-mfriz' 24374'-mno-friz' 24375 Generate (do not generate) the 'friz' instruction when the 24376 '-funsafe-math-optimizations' option is used to optimize rounding 24377 of floating-point values to 64-bit integer and back to floating 24378 point. The 'friz' instruction does not return the same value if 24379 the floating-point number is too large to fit in an integer. 24380 24381'-mpointers-to-nested-functions' 24382'-mno-pointers-to-nested-functions' 24383 Generate (do not generate) code to load up the static chain 24384 register ('r11') when calling through a pointer on AIX and 64-bit 24385 Linux systems where a function pointer points to a 3-word 24386 descriptor giving the function address, TOC value to be loaded in 24387 register 'r2', and static chain value to be loaded in register 24388 'r11'. The '-mpointers-to-nested-functions' is on by default. You 24389 cannot call through pointers to nested functions or pointers to 24390 functions compiled in other languages that use the static chain if 24391 you use '-mno-pointers-to-nested-functions'. 24392 24393'-msave-toc-indirect' 24394'-mno-save-toc-indirect' 24395 Generate (do not generate) code to save the TOC value in the 24396 reserved stack location in the function prologue if the function 24397 calls through a pointer on AIX and 64-bit Linux systems. If the 24398 TOC value is not saved in the prologue, it is saved just before the 24399 call through the pointer. The '-mno-save-toc-indirect' option is 24400 the default. 24401 24402'-mcompat-align-parm' 24403'-mno-compat-align-parm' 24404 Generate (do not generate) code to pass structure parameters with a 24405 maximum alignment of 64 bits, for compatibility with older versions 24406 of GCC. 24407 24408 Older versions of GCC (prior to 4.9.0) incorrectly did not align a 24409 structure parameter on a 128-bit boundary when that structure 24410 contained a member requiring 128-bit alignment. This is corrected 24411 in more recent versions of GCC. This option may be used to generate 24412 code that is compatible with functions compiled with older versions 24413 of GCC. 24414 24415 The '-mno-compat-align-parm' option is the default. 24416 24417'-mstack-protector-guard=GUARD' 24418'-mstack-protector-guard-reg=REG' 24419'-mstack-protector-guard-offset=OFFSET' 24420'-mstack-protector-guard-symbol=SYMBOL' 24421 Generate stack protection code using canary at GUARD. Supported 24422 locations are 'global' for global canary or 'tls' for per-thread 24423 canary in the TLS block (the default with GNU libc version 2.4 or 24424 later). 24425 24426 With the latter choice the options 24427 '-mstack-protector-guard-reg=REG' and 24428 '-mstack-protector-guard-offset=OFFSET' furthermore specify which 24429 register to use as base register for reading the canary, and from 24430 what offset from that base register. The default for those is as 24431 specified in the relevant ABI. 24432 '-mstack-protector-guard-symbol=SYMBOL' overrides the offset with a 24433 symbol reference to a canary in the TLS block. 24434 24435'-mpcrel' 24436'-mno-pcrel' 24437 Generate (do not generate) pc-relative addressing. The '-mpcrel' 24438 option requires that the medium code model ('-mcmodel=medium') and 24439 prefixed addressing ('-mprefixed') options are enabled. 24440 24441'-mprefixed' 24442'-mno-prefixed' 24443 Generate (do not generate) addressing modes using prefixed load and 24444 store instructions. The '-mprefixed' option requires that the 24445 option '-mcpu=power10' (or later) is enabled. 24446 24447'-mmma' 24448'-mno-mma' 24449 Generate (do not generate) the MMA instructions. The '-mma' option 24450 requires that the option '-mcpu=power10' (or later) is enabled. 24451 24452 24453File: gcc.info, Node: RX Options, Next: S/390 and zSeries Options, Prev: RS/6000 and PowerPC Options, Up: Submodel Options 24454 244553.19.45 RX Options 24456------------------ 24457 24458These command-line options are defined for RX targets: 24459 24460'-m64bit-doubles' 24461'-m32bit-doubles' 24462 Make the 'double' data type be 64 bits ('-m64bit-doubles') or 32 24463 bits ('-m32bit-doubles') in size. The default is 24464 '-m32bit-doubles'. _Note_ RX floating-point hardware only works on 24465 32-bit values, which is why the default is '-m32bit-doubles'. 24466 24467'-fpu' 24468'-nofpu' 24469 Enables ('-fpu') or disables ('-nofpu') the use of RX 24470 floating-point hardware. The default is enabled for the RX600 24471 series and disabled for the RX200 series. 24472 24473 Floating-point instructions are only generated for 32-bit 24474 floating-point values, however, so the FPU hardware is not used for 24475 doubles if the '-m64bit-doubles' option is used. 24476 24477 _Note_ If the '-fpu' option is enabled then 24478 '-funsafe-math-optimizations' is also enabled automatically. This 24479 is because the RX FPU instructions are themselves unsafe. 24480 24481'-mcpu=NAME' 24482 Selects the type of RX CPU to be targeted. Currently three types 24483 are supported, the generic 'RX600' and 'RX200' series hardware and 24484 the specific 'RX610' CPU. The default is 'RX600'. 24485 24486 The only difference between 'RX600' and 'RX610' is that the 'RX610' 24487 does not support the 'MVTIPL' instruction. 24488 24489 The 'RX200' series does not have a hardware floating-point unit and 24490 so '-nofpu' is enabled by default when this type is selected. 24491 24492'-mbig-endian-data' 24493'-mlittle-endian-data' 24494 Store data (but not code) in the big-endian format. The default is 24495 '-mlittle-endian-data', i.e. to store data in the little-endian 24496 format. 24497 24498'-msmall-data-limit=N' 24499 Specifies the maximum size in bytes of global and static variables 24500 which can be placed into the small data area. Using the small data 24501 area can lead to smaller and faster code, but the size of area is 24502 limited and it is up to the programmer to ensure that the area does 24503 not overflow. Also when the small data area is used one of the 24504 RX's registers (usually 'r13') is reserved for use pointing to this 24505 area, so it is no longer available for use by the compiler. This 24506 could result in slower and/or larger code if variables are pushed 24507 onto the stack instead of being held in this register. 24508 24509 Note, common variables (variables that have not been initialized) 24510 and constants are not placed into the small data area as they are 24511 assigned to other sections in the output executable. 24512 24513 The default value is zero, which disables this feature. Note, this 24514 feature is not enabled by default with higher optimization levels 24515 ('-O2' etc) because of the potentially detrimental effects of 24516 reserving a register. It is up to the programmer to experiment and 24517 discover whether this feature is of benefit to their program. See 24518 the description of the '-mpid' option for a description of how the 24519 actual register to hold the small data area pointer is chosen. 24520 24521'-msim' 24522'-mno-sim' 24523 Use the simulator runtime. The default is to use the libgloss 24524 board-specific runtime. 24525 24526'-mas100-syntax' 24527'-mno-as100-syntax' 24528 When generating assembler output use a syntax that is compatible 24529 with Renesas's AS100 assembler. This syntax can also be handled by 24530 the GAS assembler, but it has some restrictions so it is not 24531 generated by default. 24532 24533'-mmax-constant-size=N' 24534 Specifies the maximum size, in bytes, of a constant that can be 24535 used as an operand in a RX instruction. Although the RX 24536 instruction set does allow constants of up to 4 bytes in length to 24537 be used in instructions, a longer value equates to a longer 24538 instruction. Thus in some circumstances it can be beneficial to 24539 restrict the size of constants that are used in instructions. 24540 Constants that are too big are instead placed into a constant pool 24541 and referenced via register indirection. 24542 24543 The value N can be between 0 and 4. A value of 0 (the default) or 24544 4 means that constants of any size are allowed. 24545 24546'-mrelax' 24547 Enable linker relaxation. Linker relaxation is a process whereby 24548 the linker attempts to reduce the size of a program by finding 24549 shorter versions of various instructions. Disabled by default. 24550 24551'-mint-register=N' 24552 Specify the number of registers to reserve for fast interrupt 24553 handler functions. The value N can be between 0 and 4. A value of 24554 1 means that register 'r13' is reserved for the exclusive use of 24555 fast interrupt handlers. A value of 2 reserves 'r13' and 'r12'. A 24556 value of 3 reserves 'r13', 'r12' and 'r11', and a value of 4 24557 reserves 'r13' through 'r10'. A value of 0, the default, does not 24558 reserve any registers. 24559 24560'-msave-acc-in-interrupts' 24561 Specifies that interrupt handler functions should preserve the 24562 accumulator register. This is only necessary if normal code might 24563 use the accumulator register, for example because it performs 24564 64-bit multiplications. The default is to ignore the accumulator 24565 as this makes the interrupt handlers faster. 24566 24567'-mpid' 24568'-mno-pid' 24569 Enables the generation of position independent data. When enabled 24570 any access to constant data is done via an offset from a base 24571 address held in a register. This allows the location of constant 24572 data to be determined at run time without requiring the executable 24573 to be relocated, which is a benefit to embedded applications with 24574 tight memory constraints. Data that can be modified is not 24575 affected by this option. 24576 24577 Note, using this feature reserves a register, usually 'r13', for 24578 the constant data base address. This can result in slower and/or 24579 larger code, especially in complicated functions. 24580 24581 The actual register chosen to hold the constant data base address 24582 depends upon whether the '-msmall-data-limit' and/or the 24583 '-mint-register' command-line options are enabled. Starting with 24584 register 'r13' and proceeding downwards, registers are allocated 24585 first to satisfy the requirements of '-mint-register', then '-mpid' 24586 and finally '-msmall-data-limit'. Thus it is possible for the 24587 small data area register to be 'r8' if both '-mint-register=4' and 24588 '-mpid' are specified on the command line. 24589 24590 By default this feature is not enabled. The default can be 24591 restored via the '-mno-pid' command-line option. 24592 24593'-mno-warn-multiple-fast-interrupts' 24594'-mwarn-multiple-fast-interrupts' 24595 Prevents GCC from issuing a warning message if it finds more than 24596 one fast interrupt handler when it is compiling a file. The 24597 default is to issue a warning for each extra fast interrupt handler 24598 found, as the RX only supports one such interrupt. 24599 24600'-mallow-string-insns' 24601'-mno-allow-string-insns' 24602 Enables or disables the use of the string manipulation instructions 24603 'SMOVF', 'SCMPU', 'SMOVB', 'SMOVU', 'SUNTIL' 'SWHILE' and also the 24604 'RMPA' instruction. These instructions may prefetch data, which is 24605 not safe to do if accessing an I/O register. (See section 12.2.7 24606 of the RX62N Group User's Manual for more information). 24607 24608 The default is to allow these instructions, but it is not possible 24609 for GCC to reliably detect all circumstances where a string 24610 instruction might be used to access an I/O register, so their use 24611 cannot be disabled automatically. Instead it is reliant upon the 24612 programmer to use the '-mno-allow-string-insns' option if their 24613 program accesses I/O space. 24614 24615 When the instructions are enabled GCC defines the C preprocessor 24616 symbol '__RX_ALLOW_STRING_INSNS__', otherwise it defines the symbol 24617 '__RX_DISALLOW_STRING_INSNS__'. 24618 24619'-mjsr' 24620'-mno-jsr' 24621 Use only (or not only) 'JSR' instructions to access functions. 24622 This option can be used when code size exceeds the range of 'BSR' 24623 instructions. Note that '-mno-jsr' does not mean to not use 'JSR' 24624 but instead means that any type of branch may be used. 24625 24626 _Note:_ The generic GCC command-line option '-ffixed-REG' has special 24627significance to the RX port when used with the 'interrupt' function 24628attribute. This attribute indicates a function intended to process fast 24629interrupts. GCC ensures that it only uses the registers 'r10', 'r11', 24630'r12' and/or 'r13' and only provided that the normal use of the 24631corresponding registers have been restricted via the '-ffixed-REG' or 24632'-mint-register' command-line options. 24633 24634 24635File: gcc.info, Node: S/390 and zSeries Options, Next: Score Options, Prev: RX Options, Up: Submodel Options 24636 246373.19.46 S/390 and zSeries Options 24638--------------------------------- 24639 24640These are the '-m' options defined for the S/390 and zSeries 24641architecture. 24642 24643'-mhard-float' 24644'-msoft-float' 24645 Use (do not use) the hardware floating-point instructions and 24646 registers for floating-point operations. When '-msoft-float' is 24647 specified, functions in 'libgcc.a' are used to perform 24648 floating-point operations. When '-mhard-float' is specified, the 24649 compiler generates IEEE floating-point instructions. This is the 24650 default. 24651 24652'-mhard-dfp' 24653'-mno-hard-dfp' 24654 Use (do not use) the hardware decimal-floating-point instructions 24655 for decimal-floating-point operations. When '-mno-hard-dfp' is 24656 specified, functions in 'libgcc.a' are used to perform 24657 decimal-floating-point operations. When '-mhard-dfp' is specified, 24658 the compiler generates decimal-floating-point hardware 24659 instructions. This is the default for '-march=z9-ec' or higher. 24660 24661'-mlong-double-64' 24662'-mlong-double-128' 24663 These switches control the size of 'long double' type. A size of 24664 64 bits makes the 'long double' type equivalent to the 'double' 24665 type. This is the default. 24666 24667'-mbackchain' 24668'-mno-backchain' 24669 Store (do not store) the address of the caller's frame as backchain 24670 pointer into the callee's stack frame. A backchain may be needed 24671 to allow debugging using tools that do not understand DWARF call 24672 frame information. When '-mno-packed-stack' is in effect, the 24673 backchain pointer is stored at the bottom of the stack frame; when 24674 '-mpacked-stack' is in effect, the backchain is placed into the 24675 topmost word of the 96/160 byte register save area. 24676 24677 In general, code compiled with '-mbackchain' is call-compatible 24678 with code compiled with '-mmo-backchain'; however, use of the 24679 backchain for debugging purposes usually requires that the whole 24680 binary is built with '-mbackchain'. Note that the combination of 24681 '-mbackchain', '-mpacked-stack' and '-mhard-float' is not 24682 supported. In order to build a linux kernel use '-msoft-float'. 24683 24684 The default is to not maintain the backchain. 24685 24686'-mpacked-stack' 24687'-mno-packed-stack' 24688 Use (do not use) the packed stack layout. When '-mno-packed-stack' 24689 is specified, the compiler uses the all fields of the 96/160 byte 24690 register save area only for their default purpose; unused fields 24691 still take up stack space. When '-mpacked-stack' is specified, 24692 register save slots are densely packed at the top of the register 24693 save area; unused space is reused for other purposes, allowing for 24694 more efficient use of the available stack space. However, when 24695 '-mbackchain' is also in effect, the topmost word of the save area 24696 is always used to store the backchain, and the return address 24697 register is always saved two words below the backchain. 24698 24699 As long as the stack frame backchain is not used, code generated 24700 with '-mpacked-stack' is call-compatible with code generated with 24701 '-mno-packed-stack'. Note that some non-FSF releases of GCC 2.95 24702 for S/390 or zSeries generated code that uses the stack frame 24703 backchain at run time, not just for debugging purposes. Such code 24704 is not call-compatible with code compiled with '-mpacked-stack'. 24705 Also, note that the combination of '-mbackchain', '-mpacked-stack' 24706 and '-mhard-float' is not supported. In order to build a linux 24707 kernel use '-msoft-float'. 24708 24709 The default is to not use the packed stack layout. 24710 24711'-msmall-exec' 24712'-mno-small-exec' 24713 Generate (or do not generate) code using the 'bras' instruction to 24714 do subroutine calls. This only works reliably if the total 24715 executable size does not exceed 64k. The default is to use the 24716 'basr' instruction instead, which does not have this limitation. 24717 24718'-m64' 24719'-m31' 24720 When '-m31' is specified, generate code compliant to the GNU/Linux 24721 for S/390 ABI. When '-m64' is specified, generate code compliant 24722 to the GNU/Linux for zSeries ABI. This allows GCC in particular to 24723 generate 64-bit instructions. For the 's390' targets, the default 24724 is '-m31', while the 's390x' targets default to '-m64'. 24725 24726'-mzarch' 24727'-mesa' 24728 When '-mzarch' is specified, generate code using the instructions 24729 available on z/Architecture. When '-mesa' is specified, generate 24730 code using the instructions available on ESA/390. Note that 24731 '-mesa' is not possible with '-m64'. When generating code 24732 compliant to the GNU/Linux for S/390 ABI, the default is '-mesa'. 24733 When generating code compliant to the GNU/Linux for zSeries ABI, 24734 the default is '-mzarch'. 24735 24736'-mhtm' 24737'-mno-htm' 24738 The '-mhtm' option enables a set of builtins making use of 24739 instructions available with the transactional execution facility 24740 introduced with the IBM zEnterprise EC12 machine generation *note 24741 S/390 System z Built-in Functions::. '-mhtm' is enabled by default 24742 when using '-march=zEC12'. 24743 24744'-mvx' 24745'-mno-vx' 24746 When '-mvx' is specified, generate code using the instructions 24747 available with the vector extension facility introduced with the 24748 IBM z13 machine generation. This option changes the ABI for some 24749 vector type values with regard to alignment and calling 24750 conventions. In case vector type values are being used in an 24751 ABI-relevant context a GAS '.gnu_attribute' command will be added 24752 to mark the resulting binary with the ABI used. '-mvx' is enabled 24753 by default when using '-march=z13'. 24754 24755'-mzvector' 24756'-mno-zvector' 24757 The '-mzvector' option enables vector language extensions and 24758 builtins using instructions available with the vector extension 24759 facility introduced with the IBM z13 machine generation. This 24760 option adds support for 'vector' to be used as a keyword to define 24761 vector type variables and arguments. 'vector' is only available 24762 when GNU extensions are enabled. It will not be expanded when 24763 requesting strict standard compliance e.g. with '-std=c99'. In 24764 addition to the GCC low-level builtins '-mzvector' enables a set of 24765 builtins added for compatibility with AltiVec-style implementations 24766 like Power and Cell. In order to make use of these builtins the 24767 header file 'vecintrin.h' needs to be included. '-mzvector' is 24768 disabled by default. 24769 24770'-mmvcle' 24771'-mno-mvcle' 24772 Generate (or do not generate) code using the 'mvcle' instruction to 24773 perform block moves. When '-mno-mvcle' is specified, use a 'mvc' 24774 loop instead. This is the default unless optimizing for size. 24775 24776'-mdebug' 24777'-mno-debug' 24778 Print (or do not print) additional debug information when 24779 compiling. The default is to not print debug information. 24780 24781'-march=CPU-TYPE' 24782 Generate code that runs on CPU-TYPE, which is the name of a system 24783 representing a certain processor type. Possible values for 24784 CPU-TYPE are 'z900'/'arch5', 'z990'/'arch6', 'z9-109', 24785 'z9-ec'/'arch7', 'z10'/'arch8', 'z196'/'arch9', 'zEC12', 24786 'z13'/'arch11', 'z14'/'arch12', 'z15'/'arch13', and 'native'. 24787 24788 The default is '-march=z900'. 24789 24790 Specifying 'native' as cpu type can be used to select the best 24791 architecture option for the host processor. '-march=native' has no 24792 effect if GCC does not recognize the processor. 24793 24794'-mtune=CPU-TYPE' 24795 Tune to CPU-TYPE everything applicable about the generated code, 24796 except for the ABI and the set of available instructions. The list 24797 of CPU-TYPE values is the same as for '-march'. The default is the 24798 value used for '-march'. 24799 24800'-mtpf-trace' 24801'-mno-tpf-trace' 24802 Generate code that adds (does not add) in TPF OS specific branches 24803 to trace routines in the operating system. This option is off by 24804 default, even when compiling for the TPF OS. 24805 24806'-mtpf-trace-skip' 24807'-mno-tpf-trace-skip' 24808 Generate code that changes (does not change) the default branch 24809 targets enabled by '-mtpf-trace' to point to specialized trace 24810 routines providing the ability of selectively skipping function 24811 trace entries for the TPF OS. This option is off by default, even 24812 when compiling for the TPF OS and specifying '-mtpf-trace'. 24813 24814'-mfused-madd' 24815'-mno-fused-madd' 24816 Generate code that uses (does not use) the floating-point multiply 24817 and accumulate instructions. These instructions are generated by 24818 default if hardware floating point is used. 24819 24820'-mwarn-framesize=FRAMESIZE' 24821 Emit a warning if the current function exceeds the given frame 24822 size. Because this is a compile-time check it doesn't need to be a 24823 real problem when the program runs. It is intended to identify 24824 functions that most probably cause a stack overflow. It is useful 24825 to be used in an environment with limited stack size e.g. the linux 24826 kernel. 24827 24828'-mwarn-dynamicstack' 24829 Emit a warning if the function calls 'alloca' or uses 24830 dynamically-sized arrays. This is generally a bad idea with a 24831 limited stack size. 24832 24833'-mstack-guard=STACK-GUARD' 24834'-mstack-size=STACK-SIZE' 24835 If these options are provided the S/390 back end emits additional 24836 instructions in the function prologue that trigger a trap if the 24837 stack size is STACK-GUARD bytes above the STACK-SIZE (remember that 24838 the stack on S/390 grows downward). If the STACK-GUARD option is 24839 omitted the smallest power of 2 larger than the frame size of the 24840 compiled function is chosen. These options are intended to be used 24841 to help debugging stack overflow problems. The additionally 24842 emitted code causes only little overhead and hence can also be used 24843 in production-like systems without greater performance degradation. 24844 The given values have to be exact powers of 2 and STACK-SIZE has to 24845 be greater than STACK-GUARD without exceeding 64k. In order to be 24846 efficient the extra code makes the assumption that the stack starts 24847 at an address aligned to the value given by STACK-SIZE. The 24848 STACK-GUARD option can only be used in conjunction with STACK-SIZE. 24849 24850'-mhotpatch=PRE-HALFWORDS,POST-HALFWORDS' 24851 If the hotpatch option is enabled, a "hot-patching" function 24852 prologue is generated for all functions in the compilation unit. 24853 The funtion label is prepended with the given number of two-byte 24854 NOP instructions (PRE-HALFWORDS, maximum 1000000). After the 24855 label, 2 * POST-HALFWORDS bytes are appended, using the largest NOP 24856 like instructions the architecture allows (maximum 1000000). 24857 24858 If both arguments are zero, hotpatching is disabled. 24859 24860 This option can be overridden for individual functions with the 24861 'hotpatch' attribute. 24862 24863 24864File: gcc.info, Node: Score Options, Next: SH Options, Prev: S/390 and zSeries Options, Up: Submodel Options 24865 248663.19.47 Score Options 24867--------------------- 24868 24869These options are defined for Score implementations: 24870 24871'-meb' 24872 Compile code for big-endian mode. This is the default. 24873 24874'-mel' 24875 Compile code for little-endian mode. 24876 24877'-mnhwloop' 24878 Disable generation of 'bcnz' instructions. 24879 24880'-muls' 24881 Enable generation of unaligned load and store instructions. 24882 24883'-mmac' 24884 Enable the use of multiply-accumulate instructions. Disabled by 24885 default. 24886 24887'-mscore5' 24888 Specify the SCORE5 as the target architecture. 24889 24890'-mscore5u' 24891 Specify the SCORE5U of the target architecture. 24892 24893'-mscore7' 24894 Specify the SCORE7 as the target architecture. This is the 24895 default. 24896 24897'-mscore7d' 24898 Specify the SCORE7D as the target architecture. 24899 24900 24901File: gcc.info, Node: SH Options, Next: Solaris 2 Options, Prev: Score Options, Up: Submodel Options 24902 249033.19.48 SH Options 24904------------------ 24905 24906These '-m' options are defined for the SH implementations: 24907 24908'-m1' 24909 Generate code for the SH1. 24910 24911'-m2' 24912 Generate code for the SH2. 24913 24914'-m2e' 24915 Generate code for the SH2e. 24916 24917'-m2a-nofpu' 24918 Generate code for the SH2a without FPU, or for a SH2a-FPU in such a 24919 way that the floating-point unit is not used. 24920 24921'-m2a-single-only' 24922 Generate code for the SH2a-FPU, in such a way that no 24923 double-precision floating-point operations are used. 24924 24925'-m2a-single' 24926 Generate code for the SH2a-FPU assuming the floating-point unit is 24927 in single-precision mode by default. 24928 24929'-m2a' 24930 Generate code for the SH2a-FPU assuming the floating-point unit is 24931 in double-precision mode by default. 24932 24933'-m3' 24934 Generate code for the SH3. 24935 24936'-m3e' 24937 Generate code for the SH3e. 24938 24939'-m4-nofpu' 24940 Generate code for the SH4 without a floating-point unit. 24941 24942'-m4-single-only' 24943 Generate code for the SH4 with a floating-point unit that only 24944 supports single-precision arithmetic. 24945 24946'-m4-single' 24947 Generate code for the SH4 assuming the floating-point unit is in 24948 single-precision mode by default. 24949 24950'-m4' 24951 Generate code for the SH4. 24952 24953'-m4-100' 24954 Generate code for SH4-100. 24955 24956'-m4-100-nofpu' 24957 Generate code for SH4-100 in such a way that the floating-point 24958 unit is not used. 24959 24960'-m4-100-single' 24961 Generate code for SH4-100 assuming the floating-point unit is in 24962 single-precision mode by default. 24963 24964'-m4-100-single-only' 24965 Generate code for SH4-100 in such a way that no double-precision 24966 floating-point operations are used. 24967 24968'-m4-200' 24969 Generate code for SH4-200. 24970 24971'-m4-200-nofpu' 24972 Generate code for SH4-200 without in such a way that the 24973 floating-point unit is not used. 24974 24975'-m4-200-single' 24976 Generate code for SH4-200 assuming the floating-point unit is in 24977 single-precision mode by default. 24978 24979'-m4-200-single-only' 24980 Generate code for SH4-200 in such a way that no double-precision 24981 floating-point operations are used. 24982 24983'-m4-300' 24984 Generate code for SH4-300. 24985 24986'-m4-300-nofpu' 24987 Generate code for SH4-300 without in such a way that the 24988 floating-point unit is not used. 24989 24990'-m4-300-single' 24991 Generate code for SH4-300 in such a way that no double-precision 24992 floating-point operations are used. 24993 24994'-m4-300-single-only' 24995 Generate code for SH4-300 in such a way that no double-precision 24996 floating-point operations are used. 24997 24998'-m4-340' 24999 Generate code for SH4-340 (no MMU, no FPU). 25000 25001'-m4-500' 25002 Generate code for SH4-500 (no FPU). Passes '-isa=sh4-nofpu' to the 25003 assembler. 25004 25005'-m4a-nofpu' 25006 Generate code for the SH4al-dsp, or for a SH4a in such a way that 25007 the floating-point unit is not used. 25008 25009'-m4a-single-only' 25010 Generate code for the SH4a, in such a way that no double-precision 25011 floating-point operations are used. 25012 25013'-m4a-single' 25014 Generate code for the SH4a assuming the floating-point unit is in 25015 single-precision mode by default. 25016 25017'-m4a' 25018 Generate code for the SH4a. 25019 25020'-m4al' 25021 Same as '-m4a-nofpu', except that it implicitly passes '-dsp' to 25022 the assembler. GCC doesn't generate any DSP instructions at the 25023 moment. 25024 25025'-mb' 25026 Compile code for the processor in big-endian mode. 25027 25028'-ml' 25029 Compile code for the processor in little-endian mode. 25030 25031'-mdalign' 25032 Align doubles at 64-bit boundaries. Note that this changes the 25033 calling conventions, and thus some functions from the standard C 25034 library do not work unless you recompile it first with '-mdalign'. 25035 25036'-mrelax' 25037 Shorten some address references at link time, when possible; uses 25038 the linker option '-relax'. 25039 25040'-mbigtable' 25041 Use 32-bit offsets in 'switch' tables. The default is to use 25042 16-bit offsets. 25043 25044'-mbitops' 25045 Enable the use of bit manipulation instructions on SH2A. 25046 25047'-mfmovd' 25048 Enable the use of the instruction 'fmovd'. Check '-mdalign' for 25049 alignment constraints. 25050 25051'-mrenesas' 25052 Comply with the calling conventions defined by Renesas. 25053 25054'-mno-renesas' 25055 Comply with the calling conventions defined for GCC before the 25056 Renesas conventions were available. This option is the default for 25057 all targets of the SH toolchain. 25058 25059'-mnomacsave' 25060 Mark the 'MAC' register as call-clobbered, even if '-mrenesas' is 25061 given. 25062 25063'-mieee' 25064'-mno-ieee' 25065 Control the IEEE compliance of floating-point comparisons, which 25066 affects the handling of cases where the result of a comparison is 25067 unordered. By default '-mieee' is implicitly enabled. If 25068 '-ffinite-math-only' is enabled '-mno-ieee' is implicitly set, 25069 which results in faster floating-point greater-equal and less-equal 25070 comparisons. The implicit settings can be overridden by specifying 25071 either '-mieee' or '-mno-ieee'. 25072 25073'-minline-ic_invalidate' 25074 Inline code to invalidate instruction cache entries after setting 25075 up nested function trampolines. This option has no effect if 25076 '-musermode' is in effect and the selected code generation option 25077 (e.g. '-m4') does not allow the use of the 'icbi' instruction. If 25078 the selected code generation option does not allow the use of the 25079 'icbi' instruction, and '-musermode' is not in effect, the inlined 25080 code manipulates the instruction cache address array directly with 25081 an associative write. This not only requires privileged mode at 25082 run time, but it also fails if the cache line had been mapped via 25083 the TLB and has become unmapped. 25084 25085'-misize' 25086 Dump instruction size and location in the assembly code. 25087 25088'-mpadstruct' 25089 This option is deprecated. It pads structures to multiple of 4 25090 bytes, which is incompatible with the SH ABI. 25091 25092'-matomic-model=MODEL' 25093 Sets the model of atomic operations and additional parameters as a 25094 comma separated list. For details on the atomic built-in functions 25095 see *note __atomic Builtins::. The following models and parameters 25096 are supported: 25097 25098 'none' 25099 Disable compiler generated atomic sequences and emit library 25100 calls for atomic operations. This is the default if the 25101 target is not 'sh*-*-linux*'. 25102 25103 'soft-gusa' 25104 Generate GNU/Linux compatible gUSA software atomic sequences 25105 for the atomic built-in functions. The generated atomic 25106 sequences require additional support from the 25107 interrupt/exception handling code of the system and are only 25108 suitable for SH3* and SH4* single-core systems. This option 25109 is enabled by default when the target is 'sh*-*-linux*' and 25110 SH3* or SH4*. When the target is SH4A, this option also 25111 partially utilizes the hardware atomic instructions 'movli.l' 25112 and 'movco.l' to create more efficient code, unless 'strict' 25113 is specified. 25114 25115 'soft-tcb' 25116 Generate software atomic sequences that use a variable in the 25117 thread control block. This is a variation of the gUSA 25118 sequences which can also be used on SH1* and SH2* targets. 25119 The generated atomic sequences require additional support from 25120 the interrupt/exception handling code of the system and are 25121 only suitable for single-core systems. When using this model, 25122 the 'gbr-offset=' parameter has to be specified as well. 25123 25124 'soft-imask' 25125 Generate software atomic sequences that temporarily disable 25126 interrupts by setting 'SR.IMASK = 1111'. This model works 25127 only when the program runs in privileged mode and is only 25128 suitable for single-core systems. Additional support from the 25129 interrupt/exception handling code of the system is not 25130 required. This model is enabled by default when the target is 25131 'sh*-*-linux*' and SH1* or SH2*. 25132 25133 'hard-llcs' 25134 Generate hardware atomic sequences using the 'movli.l' and 25135 'movco.l' instructions only. This is only available on SH4A 25136 and is suitable for multi-core systems. Since the hardware 25137 instructions support only 32 bit atomic variables access to 8 25138 or 16 bit variables is emulated with 32 bit accesses. Code 25139 compiled with this option is also compatible with other 25140 software atomic model interrupt/exception handling systems if 25141 executed on an SH4A system. Additional support from the 25142 interrupt/exception handling code of the system is not 25143 required for this model. 25144 25145 'gbr-offset=' 25146 This parameter specifies the offset in bytes of the variable 25147 in the thread control block structure that should be used by 25148 the generated atomic sequences when the 'soft-tcb' model has 25149 been selected. For other models this parameter is ignored. 25150 The specified value must be an integer multiple of four and in 25151 the range 0-1020. 25152 25153 'strict' 25154 This parameter prevents mixed usage of multiple atomic models, 25155 even if they are compatible, and makes the compiler generate 25156 atomic sequences of the specified model only. 25157 25158'-mtas' 25159 Generate the 'tas.b' opcode for '__atomic_test_and_set'. Notice 25160 that depending on the particular hardware and software 25161 configuration this can degrade overall performance due to the 25162 operand cache line flushes that are implied by the 'tas.b' 25163 instruction. On multi-core SH4A processors the 'tas.b' instruction 25164 must be used with caution since it can result in data corruption 25165 for certain cache configurations. 25166 25167'-mprefergot' 25168 When generating position-independent code, emit function calls 25169 using the Global Offset Table instead of the Procedure Linkage 25170 Table. 25171 25172'-musermode' 25173'-mno-usermode' 25174 Don't allow (allow) the compiler generating privileged mode code. 25175 Specifying '-musermode' also implies '-mno-inline-ic_invalidate' if 25176 the inlined code would not work in user mode. '-musermode' is the 25177 default when the target is 'sh*-*-linux*'. If the target is SH1* 25178 or SH2* '-musermode' has no effect, since there is no user mode. 25179 25180'-multcost=NUMBER' 25181 Set the cost to assume for a multiply insn. 25182 25183'-mdiv=STRATEGY' 25184 Set the division strategy to be used for integer division 25185 operations. STRATEGY can be one of: 25186 25187 'call-div1' 25188 Calls a library function that uses the single-step division 25189 instruction 'div1' to perform the operation. Division by zero 25190 calculates an unspecified result and does not trap. This is 25191 the default except for SH4, SH2A and SHcompact. 25192 25193 'call-fp' 25194 Calls a library function that performs the operation in double 25195 precision floating point. Division by zero causes a 25196 floating-point exception. This is the default for SHcompact 25197 with FPU. Specifying this for targets that do not have a 25198 double precision FPU defaults to 'call-div1'. 25199 25200 'call-table' 25201 Calls a library function that uses a lookup table for small 25202 divisors and the 'div1' instruction with case distinction for 25203 larger divisors. Division by zero calculates an unspecified 25204 result and does not trap. This is the default for SH4. 25205 Specifying this for targets that do not have dynamic shift 25206 instructions defaults to 'call-div1'. 25207 25208 When a division strategy has not been specified the default 25209 strategy is selected based on the current target. For SH2A the 25210 default strategy is to use the 'divs' and 'divu' instructions 25211 instead of library function calls. 25212 25213'-maccumulate-outgoing-args' 25214 Reserve space once for outgoing arguments in the function prologue 25215 rather than around each call. Generally beneficial for performance 25216 and size. Also needed for unwinding to avoid changing the stack 25217 frame around conditional code. 25218 25219'-mdivsi3_libfunc=NAME' 25220 Set the name of the library function used for 32-bit signed 25221 division to NAME. This only affects the name used in the 'call' 25222 division strategies, and the compiler still expects the same sets 25223 of input/output/clobbered registers as if this option were not 25224 present. 25225 25226'-mfixed-range=REGISTER-RANGE' 25227 Generate code treating the given register range as fixed registers. 25228 A fixed register is one that the register allocator cannot use. 25229 This is useful when compiling kernel code. A register range is 25230 specified as two registers separated by a dash. Multiple register 25231 ranges can be specified separated by a comma. 25232 25233'-mbranch-cost=NUM' 25234 Assume NUM to be the cost for a branch instruction. Higher numbers 25235 make the compiler try to generate more branch-free code if 25236 possible. If not specified the value is selected depending on the 25237 processor type that is being compiled for. 25238 25239'-mzdcbranch' 25240'-mno-zdcbranch' 25241 Assume (do not assume) that zero displacement conditional branch 25242 instructions 'bt' and 'bf' are fast. If '-mzdcbranch' is 25243 specified, the compiler prefers zero displacement branch code 25244 sequences. This is enabled by default when generating code for SH4 25245 and SH4A. It can be explicitly disabled by specifying 25246 '-mno-zdcbranch'. 25247 25248'-mcbranch-force-delay-slot' 25249 Force the usage of delay slots for conditional branches, which 25250 stuffs the delay slot with a 'nop' if a suitable instruction cannot 25251 be found. By default this option is disabled. It can be enabled 25252 to work around hardware bugs as found in the original SH7055. 25253 25254'-mfused-madd' 25255'-mno-fused-madd' 25256 Generate code that uses (does not use) the floating-point multiply 25257 and accumulate instructions. These instructions are generated by 25258 default if hardware floating point is used. The machine-dependent 25259 '-mfused-madd' option is now mapped to the machine-independent 25260 '-ffp-contract=fast' option, and '-mno-fused-madd' is mapped to 25261 '-ffp-contract=off'. 25262 25263'-mfsca' 25264'-mno-fsca' 25265 Allow or disallow the compiler to emit the 'fsca' instruction for 25266 sine and cosine approximations. The option '-mfsca' must be used 25267 in combination with '-funsafe-math-optimizations'. It is enabled 25268 by default when generating code for SH4A. Using '-mno-fsca' 25269 disables sine and cosine approximations even if 25270 '-funsafe-math-optimizations' is in effect. 25271 25272'-mfsrra' 25273'-mno-fsrra' 25274 Allow or disallow the compiler to emit the 'fsrra' instruction for 25275 reciprocal square root approximations. The option '-mfsrra' must 25276 be used in combination with '-funsafe-math-optimizations' and 25277 '-ffinite-math-only'. It is enabled by default when generating 25278 code for SH4A. Using '-mno-fsrra' disables reciprocal square root 25279 approximations even if '-funsafe-math-optimizations' and 25280 '-ffinite-math-only' are in effect. 25281 25282'-mpretend-cmove' 25283 Prefer zero-displacement conditional branches for conditional move 25284 instruction patterns. This can result in faster code on the SH4 25285 processor. 25286 25287'-mfdpic' 25288 Generate code using the FDPIC ABI. 25289 25290 25291File: gcc.info, Node: Solaris 2 Options, Next: SPARC Options, Prev: SH Options, Up: Submodel Options 25292 252933.19.49 Solaris 2 Options 25294------------------------- 25295 25296These '-m' options are supported on Solaris 2: 25297 25298'-mclear-hwcap' 25299 '-mclear-hwcap' tells the compiler to remove the hardware 25300 capabilities generated by the Solaris assembler. This is only 25301 necessary when object files use ISA extensions not supported by the 25302 current machine, but check at runtime whether or not to use them. 25303 25304'-mimpure-text' 25305 '-mimpure-text', used in addition to '-shared', tells the compiler 25306 to not pass '-z text' to the linker when linking a shared object. 25307 Using this option, you can link position-dependent code into a 25308 shared object. 25309 25310 '-mimpure-text' suppresses the "relocations remain against 25311 allocatable but non-writable sections" linker error message. 25312 However, the necessary relocations trigger copy-on-write, and the 25313 shared object is not actually shared across processes. Instead of 25314 using '-mimpure-text', you should compile all source code with 25315 '-fpic' or '-fPIC'. 25316 25317 These switches are supported in addition to the above on Solaris 2: 25318 25319'-pthreads' 25320 This is a synonym for '-pthread'. 25321 25322 25323File: gcc.info, Node: SPARC Options, Next: System V Options, Prev: Solaris 2 Options, Up: Submodel Options 25324 253253.19.50 SPARC Options 25326--------------------- 25327 25328These '-m' options are supported on the SPARC: 25329 25330'-mno-app-regs' 25331'-mapp-regs' 25332 Specify '-mapp-regs' to generate output using the global registers 25333 2 through 4, which the SPARC SVR4 ABI reserves for applications. 25334 Like the global register 1, each global register 2 through 4 is 25335 then treated as an allocable register that is clobbered by function 25336 calls. This is the default. 25337 25338 To be fully SVR4 ABI-compliant at the cost of some performance 25339 loss, specify '-mno-app-regs'. You should compile libraries and 25340 system software with this option. 25341 25342'-mflat' 25343'-mno-flat' 25344 With '-mflat', the compiler does not generate save/restore 25345 instructions and uses a "flat" or single register window model. 25346 This model is compatible with the regular register window model. 25347 The local registers and the input registers (0-5) are still treated 25348 as "call-saved" registers and are saved on the stack as needed. 25349 25350 With '-mno-flat' (the default), the compiler generates save/restore 25351 instructions (except for leaf functions). This is the normal 25352 operating mode. 25353 25354'-mfpu' 25355'-mhard-float' 25356 Generate output containing floating-point instructions. This is 25357 the default. 25358 25359'-mno-fpu' 25360'-msoft-float' 25361 Generate output containing library calls for floating point. 25362 *Warning:* the requisite libraries are not available for all SPARC 25363 targets. Normally the facilities of the machine's usual C compiler 25364 are used, but this cannot be done directly in cross-compilation. 25365 You must make your own arrangements to provide suitable library 25366 functions for cross-compilation. The embedded targets 25367 'sparc-*-aout' and 'sparclite-*-*' do provide software 25368 floating-point support. 25369 25370 '-msoft-float' changes the calling convention in the output file; 25371 therefore, it is only useful if you compile _all_ of a program with 25372 this option. In particular, you need to compile 'libgcc.a', the 25373 library that comes with GCC, with '-msoft-float' in order for this 25374 to work. 25375 25376'-mhard-quad-float' 25377 Generate output containing quad-word (long double) floating-point 25378 instructions. 25379 25380'-msoft-quad-float' 25381 Generate output containing library calls for quad-word (long 25382 double) floating-point instructions. The functions called are 25383 those specified in the SPARC ABI. This is the default. 25384 25385 As of this writing, there are no SPARC implementations that have 25386 hardware support for the quad-word floating-point instructions. 25387 They all invoke a trap handler for one of these instructions, and 25388 then the trap handler emulates the effect of the instruction. 25389 Because of the trap handler overhead, this is much slower than 25390 calling the ABI library routines. Thus the '-msoft-quad-float' 25391 option is the default. 25392 25393'-mno-unaligned-doubles' 25394'-munaligned-doubles' 25395 Assume that doubles have 8-byte alignment. This is the default. 25396 25397 With '-munaligned-doubles', GCC assumes that doubles have 8-byte 25398 alignment only if they are contained in another type, or if they 25399 have an absolute address. Otherwise, it assumes they have 4-byte 25400 alignment. Specifying this option avoids some rare compatibility 25401 problems with code generated by other compilers. It is not the 25402 default because it results in a performance loss, especially for 25403 floating-point code. 25404 25405'-muser-mode' 25406'-mno-user-mode' 25407 Do not generate code that can only run in supervisor mode. This is 25408 relevant only for the 'casa' instruction emitted for the LEON3 25409 processor. This is the default. 25410 25411'-mfaster-structs' 25412'-mno-faster-structs' 25413 With '-mfaster-structs', the compiler assumes that structures 25414 should have 8-byte alignment. This enables the use of pairs of 25415 'ldd' and 'std' instructions for copies in structure assignment, in 25416 place of twice as many 'ld' and 'st' pairs. However, the use of 25417 this changed alignment directly violates the SPARC ABI. Thus, it's 25418 intended only for use on targets where the developer acknowledges 25419 that their resulting code is not directly in line with the rules of 25420 the ABI. 25421 25422'-mstd-struct-return' 25423'-mno-std-struct-return' 25424 With '-mstd-struct-return', the compiler generates checking code in 25425 functions returning structures or unions to detect size mismatches 25426 between the two sides of function calls, as per the 32-bit ABI. 25427 25428 The default is '-mno-std-struct-return'. This option has no effect 25429 in 64-bit mode. 25430 25431'-mlra' 25432'-mno-lra' 25433 Enable Local Register Allocation. This is the default for SPARC 25434 since GCC 7 so '-mno-lra' needs to be passed to get old Reload. 25435 25436'-mcpu=CPU_TYPE' 25437 Set the instruction set, register set, and instruction scheduling 25438 parameters for machine type CPU_TYPE. Supported values for 25439 CPU_TYPE are 'v7', 'cypress', 'v8', 'supersparc', 'hypersparc', 25440 'leon', 'leon3', 'leon3v7', 'leon5', 'sparclite', 'f930', 'f934', 25441 'sparclite86x', 'sparclet', 'tsc701', 'v9', 'ultrasparc', 25442 'ultrasparc3', 'niagara', 'niagara2', 'niagara3', 'niagara4', 25443 'niagara7' and 'm8'. 25444 25445 Native Solaris and GNU/Linux toolchains also support the value 25446 'native', which selects the best architecture option for the host 25447 processor. '-mcpu=native' has no effect if GCC does not recognize 25448 the processor. 25449 25450 Default instruction scheduling parameters are used for values that 25451 select an architecture and not an implementation. These are 'v7', 25452 'v8', 'sparclite', 'sparclet', 'v9'. 25453 25454 Here is a list of each supported architecture and their supported 25455 implementations. 25456 25457 v7 25458 cypress, leon3v7 25459 25460 v8 25461 supersparc, hypersparc, leon, leon3, leon5 25462 25463 sparclite 25464 f930, f934, sparclite86x 25465 25466 sparclet 25467 tsc701 25468 25469 v9 25470 ultrasparc, ultrasparc3, niagara, niagara2, niagara3, 25471 niagara4, niagara7, m8 25472 25473 By default (unless configured otherwise), GCC generates code for 25474 the V7 variant of the SPARC architecture. With '-mcpu=cypress', 25475 the compiler additionally optimizes it for the Cypress CY7C602 25476 chip, as used in the SPARCStation/SPARCServer 3xx series. This is 25477 also appropriate for the older SPARCStation 1, 2, IPX etc. 25478 25479 With '-mcpu=v8', GCC generates code for the V8 variant of the SPARC 25480 architecture. The only difference from V7 code is that the 25481 compiler emits the integer multiply and integer divide instructions 25482 which exist in SPARC-V8 but not in SPARC-V7. With 25483 '-mcpu=supersparc', the compiler additionally optimizes it for the 25484 SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000 25485 series. 25486 25487 With '-mcpu=sparclite', GCC generates code for the SPARClite 25488 variant of the SPARC architecture. This adds the integer multiply, 25489 integer divide step and scan ('ffs') instructions which exist in 25490 SPARClite but not in SPARC-V7. With '-mcpu=f930', the compiler 25491 additionally optimizes it for the Fujitsu MB86930 chip, which is 25492 the original SPARClite, with no FPU. With '-mcpu=f934', the 25493 compiler additionally optimizes it for the Fujitsu MB86934 chip, 25494 which is the more recent SPARClite with FPU. 25495 25496 With '-mcpu=sparclet', GCC generates code for the SPARClet variant 25497 of the SPARC architecture. This adds the integer multiply, 25498 multiply/accumulate, integer divide step and scan ('ffs') 25499 instructions which exist in SPARClet but not in SPARC-V7. With 25500 '-mcpu=tsc701', the compiler additionally optimizes it for the 25501 TEMIC SPARClet chip. 25502 25503 With '-mcpu=v9', GCC generates code for the V9 variant of the SPARC 25504 architecture. This adds 64-bit integer and floating-point move 25505 instructions, 3 additional floating-point condition code registers 25506 and conditional move instructions. With '-mcpu=ultrasparc', the 25507 compiler additionally optimizes it for the Sun UltraSPARC I/II/IIi 25508 chips. With '-mcpu=ultrasparc3', the compiler additionally 25509 optimizes it for the Sun UltraSPARC III/III+/IIIi/IIIi+/IV/IV+ 25510 chips. With '-mcpu=niagara', the compiler additionally optimizes 25511 it for Sun UltraSPARC T1 chips. With '-mcpu=niagara2', the 25512 compiler additionally optimizes it for Sun UltraSPARC T2 chips. 25513 With '-mcpu=niagara3', the compiler additionally optimizes it for 25514 Sun UltraSPARC T3 chips. With '-mcpu=niagara4', the compiler 25515 additionally optimizes it for Sun UltraSPARC T4 chips. With 25516 '-mcpu=niagara7', the compiler additionally optimizes it for Oracle 25517 SPARC M7 chips. With '-mcpu=m8', the compiler additionally 25518 optimizes it for Oracle M8 chips. 25519 25520'-mtune=CPU_TYPE' 25521 Set the instruction scheduling parameters for machine type 25522 CPU_TYPE, but do not set the instruction set or register set that 25523 the option '-mcpu=CPU_TYPE' does. 25524 25525 The same values for '-mcpu=CPU_TYPE' can be used for 25526 '-mtune=CPU_TYPE', but the only useful values are those that select 25527 a particular CPU implementation. Those are 'cypress', 25528 'supersparc', 'hypersparc', 'leon', 'leon3', 'leon3v7', 'leon5', 25529 'f930', 'f934', 'sparclite86x', 'tsc701', 'ultrasparc', 25530 'ultrasparc3', 'niagara', 'niagara2', 'niagara3', 'niagara4', 25531 'niagara7' and 'm8'. With native Solaris and GNU/Linux toolchains, 25532 'native' can also be used. 25533 25534'-mv8plus' 25535'-mno-v8plus' 25536 With '-mv8plus', GCC generates code for the SPARC-V8+ ABI. The 25537 difference from the V8 ABI is that the global and out registers are 25538 considered 64 bits wide. This is enabled by default on Solaris in 25539 32-bit mode for all SPARC-V9 processors. 25540 25541'-mvis' 25542'-mno-vis' 25543 With '-mvis', GCC generates code that takes advantage of the 25544 UltraSPARC Visual Instruction Set extensions. The default is 25545 '-mno-vis'. 25546 25547'-mvis2' 25548'-mno-vis2' 25549 With '-mvis2', GCC generates code that takes advantage of version 25550 2.0 of the UltraSPARC Visual Instruction Set extensions. The 25551 default is '-mvis2' when targeting a cpu that supports such 25552 instructions, such as UltraSPARC-III and later. Setting '-mvis2' 25553 also sets '-mvis'. 25554 25555'-mvis3' 25556'-mno-vis3' 25557 With '-mvis3', GCC generates code that takes advantage of version 25558 3.0 of the UltraSPARC Visual Instruction Set extensions. The 25559 default is '-mvis3' when targeting a cpu that supports such 25560 instructions, such as niagara-3 and later. Setting '-mvis3' also 25561 sets '-mvis2' and '-mvis'. 25562 25563'-mvis4' 25564'-mno-vis4' 25565 With '-mvis4', GCC generates code that takes advantage of version 25566 4.0 of the UltraSPARC Visual Instruction Set extensions. The 25567 default is '-mvis4' when targeting a cpu that supports such 25568 instructions, such as niagara-7 and later. Setting '-mvis4' also 25569 sets '-mvis3', '-mvis2' and '-mvis'. 25570 25571'-mvis4b' 25572'-mno-vis4b' 25573 With '-mvis4b', GCC generates code that takes advantage of version 25574 4.0 of the UltraSPARC Visual Instruction Set extensions, plus the 25575 additional VIS instructions introduced in the Oracle SPARC 25576 Architecture 2017. The default is '-mvis4b' when targeting a cpu 25577 that supports such instructions, such as m8 and later. Setting 25578 '-mvis4b' also sets '-mvis4', '-mvis3', '-mvis2' and '-mvis'. 25579 25580'-mcbcond' 25581'-mno-cbcond' 25582 With '-mcbcond', GCC generates code that takes advantage of the 25583 UltraSPARC Compare-and-Branch-on-Condition instructions. The 25584 default is '-mcbcond' when targeting a CPU that supports such 25585 instructions, such as Niagara-4 and later. 25586 25587'-mfmaf' 25588'-mno-fmaf' 25589 With '-mfmaf', GCC generates code that takes advantage of the 25590 UltraSPARC Fused Multiply-Add Floating-point instructions. The 25591 default is '-mfmaf' when targeting a CPU that supports such 25592 instructions, such as Niagara-3 and later. 25593 25594'-mfsmuld' 25595'-mno-fsmuld' 25596 With '-mfsmuld', GCC generates code that takes advantage of the 25597 Floating-point Multiply Single to Double (FsMULd) instruction. The 25598 default is '-mfsmuld' when targeting a CPU supporting the 25599 architecture versions V8 or V9 with FPU except '-mcpu=leon'. 25600 25601'-mpopc' 25602'-mno-popc' 25603 With '-mpopc', GCC generates code that takes advantage of the 25604 UltraSPARC Population Count instruction. The default is '-mpopc' 25605 when targeting a CPU that supports such an instruction, such as 25606 Niagara-2 and later. 25607 25608'-msubxc' 25609'-mno-subxc' 25610 With '-msubxc', GCC generates code that takes advantage of the 25611 UltraSPARC Subtract-Extended-with-Carry instruction. The default 25612 is '-msubxc' when targeting a CPU that supports such an 25613 instruction, such as Niagara-7 and later. 25614 25615'-mfix-at697f' 25616 Enable the documented workaround for the single erratum of the 25617 Atmel AT697F processor (which corresponds to erratum #13 of the 25618 AT697E processor). 25619 25620'-mfix-ut699' 25621 Enable the documented workarounds for the floating-point errata and 25622 the data cache nullify errata of the UT699 processor. 25623 25624'-mfix-ut700' 25625 Enable the documented workaround for the back-to-back store errata 25626 of the UT699E/UT700 processor. 25627 25628'-mfix-gr712rc' 25629 Enable the documented workaround for the back-to-back store errata 25630 of the GR712RC processor. 25631 25632 These '-m' options are supported in addition to the above on SPARC-V9 25633processors in 64-bit environments: 25634 25635'-m32' 25636'-m64' 25637 Generate code for a 32-bit or 64-bit environment. The 32-bit 25638 environment sets int, long and pointer to 32 bits. The 64-bit 25639 environment sets int to 32 bits and long and pointer to 64 bits. 25640 25641'-mcmodel=WHICH' 25642 Set the code model to one of 25643 25644 'medlow' 25645 The Medium/Low code model: 64-bit addresses, programs must be 25646 linked in the low 32 bits of memory. Programs can be 25647 statically or dynamically linked. 25648 25649 'medmid' 25650 The Medium/Middle code model: 64-bit addresses, programs must 25651 be linked in the low 44 bits of memory, the text and data 25652 segments must be less than 2GB in size and the data segment 25653 must be located within 2GB of the text segment. 25654 25655 'medany' 25656 The Medium/Anywhere code model: 64-bit addresses, programs may 25657 be linked anywhere in memory, the text and data segments must 25658 be less than 2GB in size and the data segment must be located 25659 within 2GB of the text segment. 25660 25661 'embmedany' 25662 The Medium/Anywhere code model for embedded systems: 64-bit 25663 addresses, the text and data segments must be less than 2GB in 25664 size, both starting anywhere in memory (determined at link 25665 time). The global register %g4 points to the base of the data 25666 segment. Programs are statically linked and PIC is not 25667 supported. 25668 25669'-mmemory-model=MEM-MODEL' 25670 Set the memory model in force on the processor to one of 25671 25672 'default' 25673 The default memory model for the processor and operating 25674 system. 25675 25676 'rmo' 25677 Relaxed Memory Order 25678 25679 'pso' 25680 Partial Store Order 25681 25682 'tso' 25683 Total Store Order 25684 25685 'sc' 25686 Sequential Consistency 25687 25688 These memory models are formally defined in Appendix D of the 25689 SPARC-V9 architecture manual, as set in the processor's 'PSTATE.MM' 25690 field. 25691 25692'-mstack-bias' 25693'-mno-stack-bias' 25694 With '-mstack-bias', GCC assumes that the stack pointer, and frame 25695 pointer if present, are offset by -2047 which must be added back 25696 when making stack frame references. This is the default in 64-bit 25697 mode. Otherwise, assume no such offset is present. 25698 25699 25700File: gcc.info, Node: System V Options, Next: TILE-Gx Options, Prev: SPARC Options, Up: Submodel Options 25701 257023.19.51 Options for System V 25703---------------------------- 25704 25705These additional options are available on System V Release 4 for 25706compatibility with other compilers on those systems: 25707 25708'-G' 25709 Create a shared object. It is recommended that '-symbolic' or 25710 '-shared' be used instead. 25711 25712'-Qy' 25713 Identify the versions of each tool used by the compiler, in a 25714 '.ident' assembler directive in the output. 25715 25716'-Qn' 25717 Refrain from adding '.ident' directives to the output file (this is 25718 the default). 25719 25720'-YP,DIRS' 25721 Search the directories DIRS, and no others, for libraries specified 25722 with '-l'. 25723 25724'-Ym,DIR' 25725 Look in the directory DIR to find the M4 preprocessor. The 25726 assembler uses this option. 25727 25728 25729File: gcc.info, Node: TILE-Gx Options, Next: TILEPro Options, Prev: System V Options, Up: Submodel Options 25730 257313.19.52 TILE-Gx Options 25732----------------------- 25733 25734These '-m' options are supported on the TILE-Gx: 25735 25736'-mcmodel=small' 25737 Generate code for the small model. The distance for direct calls 25738 is limited to 500M in either direction. PC-relative addresses are 25739 32 bits. Absolute addresses support the full address range. 25740 25741'-mcmodel=large' 25742 Generate code for the large model. There is no limitation on call 25743 distance, pc-relative addresses, or absolute addresses. 25744 25745'-mcpu=NAME' 25746 Selects the type of CPU to be targeted. Currently the only 25747 supported type is 'tilegx'. 25748 25749'-m32' 25750'-m64' 25751 Generate code for a 32-bit or 64-bit environment. The 32-bit 25752 environment sets int, long, and pointer to 32 bits. The 64-bit 25753 environment sets int to 32 bits and long and pointer to 64 bits. 25754 25755'-mbig-endian' 25756'-mlittle-endian' 25757 Generate code in big/little endian mode, respectively. 25758 25759 25760File: gcc.info, Node: TILEPro Options, Next: V850 Options, Prev: TILE-Gx Options, Up: Submodel Options 25761 257623.19.53 TILEPro Options 25763----------------------- 25764 25765These '-m' options are supported on the TILEPro: 25766 25767'-mcpu=NAME' 25768 Selects the type of CPU to be targeted. Currently the only 25769 supported type is 'tilepro'. 25770 25771'-m32' 25772 Generate code for a 32-bit environment, which sets int, long, and 25773 pointer to 32 bits. This is the only supported behavior so the 25774 flag is essentially ignored. 25775 25776 25777File: gcc.info, Node: V850 Options, Next: VAX Options, Prev: TILEPro Options, Up: Submodel Options 25778 257793.19.54 V850 Options 25780-------------------- 25781 25782These '-m' options are defined for V850 implementations: 25783 25784'-mlong-calls' 25785'-mno-long-calls' 25786 Treat all calls as being far away (near). If calls are assumed to 25787 be far away, the compiler always loads the function's address into 25788 a register, and calls indirect through the pointer. 25789 25790'-mno-ep' 25791'-mep' 25792 Do not optimize (do optimize) basic blocks that use the same index 25793 pointer 4 or more times to copy pointer into the 'ep' register, and 25794 use the shorter 'sld' and 'sst' instructions. The '-mep' option is 25795 on by default if you optimize. 25796 25797'-mno-prolog-function' 25798'-mprolog-function' 25799 Do not use (do use) external functions to save and restore 25800 registers at the prologue and epilogue of a function. The external 25801 functions are slower, but use less code space if more than one 25802 function saves the same number of registers. The 25803 '-mprolog-function' option is on by default if you optimize. 25804 25805'-mspace' 25806 Try to make the code as small as possible. At present, this just 25807 turns on the '-mep' and '-mprolog-function' options. 25808 25809'-mtda=N' 25810 Put static or global variables whose size is N bytes or less into 25811 the tiny data area that register 'ep' points to. The tiny data 25812 area can hold up to 256 bytes in total (128 bytes for byte 25813 references). 25814 25815'-msda=N' 25816 Put static or global variables whose size is N bytes or less into 25817 the small data area that register 'gp' points to. The small data 25818 area can hold up to 64 kilobytes. 25819 25820'-mzda=N' 25821 Put static or global variables whose size is N bytes or less into 25822 the first 32 kilobytes of memory. 25823 25824'-mv850' 25825 Specify that the target processor is the V850. 25826 25827'-mv850e3v5' 25828 Specify that the target processor is the V850E3V5. The 25829 preprocessor constant '__v850e3v5__' is defined if this option is 25830 used. 25831 25832'-mv850e2v4' 25833 Specify that the target processor is the V850E3V5. This is an 25834 alias for the '-mv850e3v5' option. 25835 25836'-mv850e2v3' 25837 Specify that the target processor is the V850E2V3. The 25838 preprocessor constant '__v850e2v3__' is defined if this option is 25839 used. 25840 25841'-mv850e2' 25842 Specify that the target processor is the V850E2. The preprocessor 25843 constant '__v850e2__' is defined if this option is used. 25844 25845'-mv850e1' 25846 Specify that the target processor is the V850E1. The preprocessor 25847 constants '__v850e1__' and '__v850e__' are defined if this option 25848 is used. 25849 25850'-mv850es' 25851 Specify that the target processor is the V850ES. This is an alias 25852 for the '-mv850e1' option. 25853 25854'-mv850e' 25855 Specify that the target processor is the V850E. The preprocessor 25856 constant '__v850e__' is defined if this option is used. 25857 25858 If neither '-mv850' nor '-mv850e' nor '-mv850e1' nor '-mv850e2' nor 25859 '-mv850e2v3' nor '-mv850e3v5' are defined then a default target 25860 processor is chosen and the relevant '__v850*__' preprocessor 25861 constant is defined. 25862 25863 The preprocessor constants '__v850' and '__v851__' are always 25864 defined, regardless of which processor variant is the target. 25865 25866'-mdisable-callt' 25867'-mno-disable-callt' 25868 This option suppresses generation of the 'CALLT' instruction for 25869 the v850e, v850e1, v850e2, v850e2v3 and v850e3v5 flavors of the 25870 v850 architecture. 25871 25872 This option is enabled by default when the RH850 ABI is in use (see 25873 '-mrh850-abi'), and disabled by default when the GCC ABI is in use. 25874 If 'CALLT' instructions are being generated then the C preprocessor 25875 symbol '__V850_CALLT__' is defined. 25876 25877'-mrelax' 25878'-mno-relax' 25879 Pass on (or do not pass on) the '-mrelax' command-line option to 25880 the assembler. 25881 25882'-mlong-jumps' 25883'-mno-long-jumps' 25884 Disable (or re-enable) the generation of PC-relative jump 25885 instructions. 25886 25887'-msoft-float' 25888'-mhard-float' 25889 Disable (or re-enable) the generation of hardware floating point 25890 instructions. This option is only significant when the target 25891 architecture is 'V850E2V3' or higher. If hardware floating point 25892 instructions are being generated then the C preprocessor symbol 25893 '__FPU_OK__' is defined, otherwise the symbol '__NO_FPU__' is 25894 defined. 25895 25896'-mloop' 25897 Enables the use of the e3v5 LOOP instruction. The use of this 25898 instruction is not enabled by default when the e3v5 architecture is 25899 selected because its use is still experimental. 25900 25901'-mrh850-abi' 25902'-mghs' 25903 Enables support for the RH850 version of the V850 ABI. This is the 25904 default. With this version of the ABI the following rules apply: 25905 25906 * Integer sized structures and unions are returned via a memory 25907 pointer rather than a register. 25908 25909 * Large structures and unions (more than 8 bytes in size) are 25910 passed by value. 25911 25912 * Functions are aligned to 16-bit boundaries. 25913 25914 * The '-m8byte-align' command-line option is supported. 25915 25916 * The '-mdisable-callt' command-line option is enabled by 25917 default. The '-mno-disable-callt' command-line option is not 25918 supported. 25919 25920 When this version of the ABI is enabled the C preprocessor symbol 25921 '__V850_RH850_ABI__' is defined. 25922 25923'-mgcc-abi' 25924 Enables support for the old GCC version of the V850 ABI. With this 25925 version of the ABI the following rules apply: 25926 25927 * Integer sized structures and unions are returned in register 25928 'r10'. 25929 25930 * Large structures and unions (more than 8 bytes in size) are 25931 passed by reference. 25932 25933 * Functions are aligned to 32-bit boundaries, unless optimizing 25934 for size. 25935 25936 * The '-m8byte-align' command-line option is not supported. 25937 25938 * The '-mdisable-callt' command-line option is supported but not 25939 enabled by default. 25940 25941 When this version of the ABI is enabled the C preprocessor symbol 25942 '__V850_GCC_ABI__' is defined. 25943 25944'-m8byte-align' 25945'-mno-8byte-align' 25946 Enables support for 'double' and 'long long' types to be aligned on 25947 8-byte boundaries. The default is to restrict the alignment of all 25948 objects to at most 4-bytes. When '-m8byte-align' is in effect the 25949 C preprocessor symbol '__V850_8BYTE_ALIGN__' is defined. 25950 25951'-mbig-switch' 25952 Generate code suitable for big switch tables. Use this option only 25953 if the assembler/linker complain about out of range branches within 25954 a switch table. 25955 25956'-mapp-regs' 25957 This option causes r2 and r5 to be used in the code generated by 25958 the compiler. This setting is the default. 25959 25960'-mno-app-regs' 25961 This option causes r2 and r5 to be treated as fixed registers. 25962 25963 25964File: gcc.info, Node: VAX Options, Next: Visium Options, Prev: V850 Options, Up: Submodel Options 25965 259663.19.55 VAX Options 25967------------------- 25968 25969These '-m' options are defined for the VAX: 25970 25971'-munix' 25972 Do not output certain jump instructions ('aobleq' and so on) that 25973 the Unix assembler for the VAX cannot handle across long ranges. 25974 25975'-mgnu' 25976 Do output those jump instructions, on the assumption that the GNU 25977 assembler is being used. 25978 25979'-mg' 25980 Output code for G-format floating-point numbers instead of 25981 D-format. 25982 25983 25984File: gcc.info, Node: Visium Options, Next: VMS Options, Prev: VAX Options, Up: Submodel Options 25985 259863.19.56 Visium Options 25987---------------------- 25988 25989'-mdebug' 25990 A program which performs file I/O and is destined to run on an MCM 25991 target should be linked with this option. It causes the libraries 25992 libc.a and libdebug.a to be linked. The program should be run on 25993 the target under the control of the GDB remote debugging stub. 25994 25995'-msim' 25996 A program which performs file I/O and is destined to run on the 25997 simulator should be linked with option. This causes libraries 25998 libc.a and libsim.a to be linked. 25999 26000'-mfpu' 26001'-mhard-float' 26002 Generate code containing floating-point instructions. This is the 26003 default. 26004 26005'-mno-fpu' 26006'-msoft-float' 26007 Generate code containing library calls for floating-point. 26008 26009 '-msoft-float' changes the calling convention in the output file; 26010 therefore, it is only useful if you compile _all_ of a program with 26011 this option. In particular, you need to compile 'libgcc.a', the 26012 library that comes with GCC, with '-msoft-float' in order for this 26013 to work. 26014 26015'-mcpu=CPU_TYPE' 26016 Set the instruction set, register set, and instruction scheduling 26017 parameters for machine type CPU_TYPE. Supported values for 26018 CPU_TYPE are 'mcm', 'gr5' and 'gr6'. 26019 26020 'mcm' is a synonym of 'gr5' present for backward compatibility. 26021 26022 By default (unless configured otherwise), GCC generates code for 26023 the GR5 variant of the Visium architecture. 26024 26025 With '-mcpu=gr6', GCC generates code for the GR6 variant of the 26026 Visium architecture. The only difference from GR5 code is that the 26027 compiler will generate block move instructions. 26028 26029'-mtune=CPU_TYPE' 26030 Set the instruction scheduling parameters for machine type 26031 CPU_TYPE, but do not set the instruction set or register set that 26032 the option '-mcpu=CPU_TYPE' would. 26033 26034'-msv-mode' 26035 Generate code for the supervisor mode, where there are no 26036 restrictions on the access to general registers. This is the 26037 default. 26038 26039'-muser-mode' 26040 Generate code for the user mode, where the access to some general 26041 registers is forbidden: on the GR5, registers r24 to r31 cannot be 26042 accessed in this mode; on the GR6, only registers r29 to r31 are 26043 affected. 26044 26045 26046File: gcc.info, Node: VMS Options, Next: VxWorks Options, Prev: Visium Options, Up: Submodel Options 26047 260483.19.57 VMS Options 26049------------------- 26050 26051These '-m' options are defined for the VMS implementations: 26052 26053'-mvms-return-codes' 26054 Return VMS condition codes from 'main'. The default is to return 26055 POSIX-style condition (e.g. error) codes. 26056 26057'-mdebug-main=PREFIX' 26058 Flag the first routine whose name starts with PREFIX as the main 26059 routine for the debugger. 26060 26061'-mmalloc64' 26062 Default to 64-bit memory allocation routines. 26063 26064'-mpointer-size=SIZE' 26065 Set the default size of pointers. Possible options for SIZE are 26066 '32' or 'short' for 32 bit pointers, '64' or 'long' for 64 bit 26067 pointers, and 'no' for supporting only 32 bit pointers. The later 26068 option disables 'pragma pointer_size'. 26069 26070 26071File: gcc.info, Node: VxWorks Options, Next: x86 Options, Prev: VMS Options, Up: Submodel Options 26072 260733.19.58 VxWorks Options 26074----------------------- 26075 26076The options in this section are defined for all VxWorks targets. 26077Options specific to the target hardware are listed with the other 26078options for that target. 26079 26080'-mrtp' 26081 GCC can generate code for both VxWorks kernels and real time 26082 processes (RTPs). This option switches from the former to the 26083 latter. It also defines the preprocessor macro '__RTP__'. 26084 26085'-non-static' 26086 Link an RTP executable against shared libraries rather than static 26087 libraries. The options '-static' and '-shared' can also be used 26088 for RTPs (*note Link Options::); '-static' is the default. 26089 26090'-Bstatic' 26091'-Bdynamic' 26092 These options are passed down to the linker. They are defined for 26093 compatibility with Diab. 26094 26095'-Xbind-lazy' 26096 Enable lazy binding of function calls. This option is equivalent 26097 to '-Wl,-z,now' and is defined for compatibility with Diab. 26098 26099'-Xbind-now' 26100 Disable lazy binding of function calls. This option is the default 26101 and is defined for compatibility with Diab. 26102 26103 26104File: gcc.info, Node: x86 Options, Next: x86 Windows Options, Prev: VxWorks Options, Up: Submodel Options 26105 261063.19.59 x86 Options 26107------------------- 26108 26109These '-m' options are defined for the x86 family of computers. 26110 26111'-march=CPU-TYPE' 26112 Generate instructions for the machine type CPU-TYPE. In contrast 26113 to '-mtune=CPU-TYPE', which merely tunes the generated code for the 26114 specified CPU-TYPE, '-march=CPU-TYPE' allows GCC to generate code 26115 that may not run at all on processors other than the one indicated. 26116 Specifying '-march=CPU-TYPE' implies '-mtune=CPU-TYPE'. 26117 26118 The choices for CPU-TYPE are: 26119 26120 'native' 26121 This selects the CPU to generate code for at compilation time 26122 by determining the processor type of the compiling machine. 26123 Using '-march=native' enables all instruction subsets 26124 supported by the local machine (hence the result might not run 26125 on different machines). Using '-mtune=native' produces code 26126 optimized for the local machine under the constraints of the 26127 selected instruction set. 26128 26129 'x86-64' 26130 A generic CPU with 64-bit extensions. 26131 26132 'i386' 26133 Original Intel i386 CPU. 26134 26135 'i486' 26136 Intel i486 CPU. (No scheduling is implemented for this chip.) 26137 26138 'i586' 26139 'pentium' 26140 Intel Pentium CPU with no MMX support. 26141 26142 'lakemont' 26143 Intel Lakemont MCU, based on Intel Pentium CPU. 26144 26145 'pentium-mmx' 26146 Intel Pentium MMX CPU, based on Pentium core with MMX 26147 instruction set support. 26148 26149 'pentiumpro' 26150 Intel Pentium Pro CPU. 26151 26152 'i686' 26153 When used with '-march', the Pentium Pro instruction set is 26154 used, so the code runs on all i686 family chips. When used 26155 with '-mtune', it has the same meaning as 'generic'. 26156 26157 'pentium2' 26158 Intel Pentium II CPU, based on Pentium Pro core with MMX 26159 instruction set support. 26160 26161 'pentium3' 26162 'pentium3m' 26163 Intel Pentium III CPU, based on Pentium Pro core with MMX and 26164 SSE instruction set support. 26165 26166 'pentium-m' 26167 Intel Pentium M; low-power version of Intel Pentium III CPU 26168 with MMX, SSE and SSE2 instruction set support. Used by 26169 Centrino notebooks. 26170 26171 'pentium4' 26172 'pentium4m' 26173 Intel Pentium 4 CPU with MMX, SSE and SSE2 instruction set 26174 support. 26175 26176 'prescott' 26177 Improved version of Intel Pentium 4 CPU with MMX, SSE, SSE2 26178 and SSE3 instruction set support. 26179 26180 'nocona' 26181 Improved version of Intel Pentium 4 CPU with 64-bit 26182 extensions, MMX, SSE, SSE2 and SSE3 instruction set support. 26183 26184 'core2' 26185 Intel Core 2 CPU with 64-bit extensions, MMX, SSE, SSE2, SSE3 26186 and SSSE3 instruction set support. 26187 26188 'nehalem' 26189 Intel Nehalem CPU with 64-bit extensions, MMX, SSE, SSE2, 26190 SSE3, SSSE3, SSE4.1, SSE4.2 and POPCNT instruction set 26191 support. 26192 26193 'westmere' 26194 Intel Westmere CPU with 64-bit extensions, MMX, SSE, SSE2, 26195 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES and PCLMUL 26196 instruction set support. 26197 26198 'sandybridge' 26199 Intel Sandy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, 26200 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AES and PCLMUL 26201 instruction set support. 26202 26203 'ivybridge' 26204 Intel Ivy Bridge CPU with 64-bit extensions, MMX, SSE, SSE2, 26205 SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AES, PCLMUL, 26206 FSGSBASE, RDRND and F16C instruction set support. 26207 26208 'haswell' 26209 Intel Haswell CPU with 64-bit extensions, MOVBE, MMX, SSE, 26210 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, 26211 PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2 and F16C instruction 26212 set support. 26213 26214 'broadwell' 26215 Intel Broadwell CPU with 64-bit extensions, MOVBE, MMX, SSE, 26216 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, 26217 PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED ADCX and 26218 PREFETCHW instruction set support. 26219 26220 'skylake' 26221 Intel Skylake CPU with 64-bit extensions, MOVBE, MMX, SSE, 26222 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, AES, 26223 PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, ADCX, 26224 PREFETCHW, CLFLUSHOPT, XSAVEC and XSAVES instruction set 26225 support. 26226 26227 'bonnell' 26228 Intel Bonnell CPU with 64-bit extensions, MOVBE, MMX, SSE, 26229 SSE2, SSE3 and SSSE3 instruction set support. 26230 26231 'silvermont' 26232 Intel Silvermont CPU with 64-bit extensions, MOVBE, MMX, SSE, 26233 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, 26234 PCLMUL and RDRND instruction set support. 26235 26236 'goldmont' 26237 Intel Goldmont CPU with 64-bit extensions, MOVBE, MMX, SSE, 26238 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, 26239 PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT and FSGSBASE 26240 instruction set support. 26241 26242 'goldmont-plus' 26243 Intel Goldmont Plus CPU with 64-bit extensions, MOVBE, MMX, 26244 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, 26245 PREFETCHW, PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, 26246 FSGSBASE, PTWRITE, RDPID, SGX and UMIP instruction set 26247 support. 26248 26249 'tremont' 26250 Intel Tremont CPU with 64-bit extensions, MOVBE, MMX, SSE, 26251 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AES, PREFETCHW, 26252 PCLMUL, RDRND, XSAVE, XSAVEC, XSAVES, XSAVEOPT, FSGSBASE, 26253 PTWRITE, RDPID, SGX, UMIP, GFNI-SSE, CLWB, MOVDIRI, MOVDIR64B, 26254 CLDEMOTE and WAITPKG instruction set support. 26255 26256 'knl' 26257 Intel Knight's Landing CPU with 64-bit extensions, MOVBE, MMX, 26258 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, 26259 AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, 26260 ADCX, PREFETCHW, PREFETCHWT1, AVX512F, AVX512PF, AVX512ER and 26261 AVX512CD instruction set support. 26262 26263 'knm' 26264 Intel Knights Mill CPU with 64-bit extensions, MOVBE, MMX, 26265 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, AVX, AVX2, 26266 AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, 26267 ADCX, PREFETCHW, PREFETCHWT1, AVX512F, AVX512PF, AVX512ER, 26268 AVX512CD, AVX5124VNNIW, AVX5124FMAPS and AVX512VPOPCNTDQ 26269 instruction set support. 26270 26271 'skylake-avx512' 26272 Intel Skylake Server CPU with 64-bit extensions, MOVBE, MMX, 26273 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, 26274 AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, 26275 RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 26276 CLWB, AVX512VL, AVX512BW, AVX512DQ and AVX512CD instruction 26277 set support. 26278 26279 'cannonlake' 26280 Intel Cannonlake Server CPU with 64-bit extensions, MOVBE, 26281 MMX, SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, 26282 AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, 26283 RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 26284 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 26285 AVX512IFMA, SHA and UMIP instruction set support. 26286 26287 'icelake-client' 26288 Intel Icelake Client CPU with 64-bit extensions, MOVBE, MMX, 26289 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, 26290 AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, 26291 RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 26292 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 26293 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, 26294 AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES 26295 instruction set support. 26296 26297 'icelake-server' 26298 Intel Icelake Server CPU with 64-bit extensions, MOVBE, MMX, 26299 SSE, SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, 26300 AVX2, AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, 26301 RDSEED, ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 26302 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 26303 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, 26304 AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, 26305 PCONFIG and WBNOINVD instruction set support. 26306 26307 'cascadelake' 26308 Intel Cascadelake CPU with 64-bit extensions, MOVBE, MMX, SSE, 26309 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, 26310 AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, 26311 ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB, 26312 AVX512VL, AVX512BW, AVX512DQ, AVX512CD and AVX512VNNI 26313 instruction set support. 26314 26315 'cooperlake' 26316 Intel cooperlake CPU with 64-bit extensions, MOVBE, MMX, SSE, 26317 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, 26318 AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, 26319 ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, CLWB, 26320 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VNNI and 26321 AVX512BF16 instruction set support. 26322 26323 'tigerlake' 26324 Intel Tigerlake CPU with 64-bit extensions, MOVBE, MMX, SSE, 26325 SSE2, SSE3, SSSE3, SSE4.1, SSE4.2, POPCNT, PKU, AVX, AVX2, 26326 AES, PCLMUL, FSGSBASE, RDRND, FMA, BMI, BMI2, F16C, RDSEED, 26327 ADCX, PREFETCHW, CLFLUSHOPT, XSAVEC, XSAVES, AVX512F, 26328 AVX512VL, AVX512BW, AVX512DQ, AVX512CD, AVX512VBMI, 26329 AVX512IFMA, SHA, CLWB, UMIP, RDPID, GFNI, AVX512VBMI2, 26330 AVX512VPOPCNTDQ, AVX512BITALG, AVX512VNNI, VPCLMULQDQ, VAES, 26331 PCONFIG, WBNOINVD, MOVDIRI, MOVDIR64B, AVX512VP2INTERSECT and 26332 KEYLOCKER instruction set support. 26333 26334 'k6' 26335 AMD K6 CPU with MMX instruction set support. 26336 26337 'k6-2' 26338 'k6-3' 26339 Improved versions of AMD K6 CPU with MMX and 3DNow! 26340 instruction set support. 26341 26342 'athlon' 26343 'athlon-tbird' 26344 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3DNow! and SSE 26345 prefetch instructions support. 26346 26347 'athlon-4' 26348 'athlon-xp' 26349 'athlon-mp' 26350 Improved AMD Athlon CPU with MMX, 3DNow!, enhanced 3DNow! and 26351 full SSE instruction set support. 26352 26353 'k8' 26354 'opteron' 26355 'athlon64' 26356 'athlon-fx' 26357 Processors based on the AMD K8 core with x86-64 instruction 26358 set support, including the AMD Opteron, Athlon 64, and Athlon 26359 64 FX processors. (This supersets MMX, SSE, SSE2, 3DNow!, 26360 enhanced 3DNow! and 64-bit instruction set extensions.) 26361 26362 'k8-sse3' 26363 'opteron-sse3' 26364 'athlon64-sse3' 26365 Improved versions of AMD K8 cores with SSE3 instruction set 26366 support. 26367 26368 'amdfam10' 26369 'barcelona' 26370 CPUs based on AMD Family 10h cores with x86-64 instruction set 26371 support. (This supersets MMX, SSE, SSE2, SSE3, SSE4A, 3DNow!, 26372 enhanced 3DNow!, ABM and 64-bit instruction set extensions.) 26373 26374 'bdver1' 26375 CPUs based on AMD Family 15h cores with x86-64 instruction set 26376 support. (This supersets FMA4, AVX, XOP, LWP, AES, PCLMUL, 26377 CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM 26378 and 64-bit instruction set extensions.) 26379 26380 'bdver2' 26381 AMD Family 15h core based CPUs with x86-64 instruction set 26382 support. (This supersets BMI, TBM, F16C, FMA, FMA4, AVX, XOP, 26383 LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, 26384 SSE4.1, SSE4.2, ABM and 64-bit instruction set extensions.) 26385 26386 'bdver3' 26387 AMD Family 15h core based CPUs with x86-64 instruction set 26388 support. (This supersets BMI, TBM, F16C, FMA, FMA4, FSGSBASE, 26389 AVX, XOP, LWP, AES, PCLMUL, CX16, MMX, SSE, SSE2, SSE3, SSE4A, 26390 SSSE3, SSE4.1, SSE4.2, ABM and 64-bit instruction set 26391 extensions.) 26392 26393 'bdver4' 26394 AMD Family 15h core based CPUs with x86-64 instruction set 26395 support. (This supersets BMI, BMI2, TBM, F16C, FMA, FMA4, 26396 FSGSBASE, AVX, AVX2, XOP, LWP, AES, PCLMUL, CX16, MOVBE, MMX, 26397 SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, ABM and 64-bit 26398 instruction set extensions.) 26399 26400 'znver1' 26401 AMD Family 17h core based CPUs with x86-64 instruction set 26402 support. (This supersets BMI, BMI2, F16C, FMA, FSGSBASE, AVX, 26403 AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, PCLMUL, CX16, 26404 MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, SSE4.1, SSE4.2, 26405 ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, and 64-bit 26406 instruction set extensions.) 26407 26408 'znver2' 26409 AMD Family 17h core based CPUs with x86-64 instruction set 26410 support. (This supersets BMI, BMI2, CLWB, F16C, FMA, 26411 FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, 26412 PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, 26413 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, 26414 RDPID, WBNOINVD, and 64-bit instruction set extensions.) 26415 26416 'znver3' 26417 AMD Family 19h core based CPUs with x86-64 instruction set 26418 support. (This supersets BMI, BMI2, CLWB, F16C, FMA, 26419 FSGSBASE, AVX, AVX2, ADCX, RDSEED, MWAITX, SHA, CLZERO, AES, 26420 PCLMUL, CX16, MOVBE, MMX, SSE, SSE2, SSE3, SSE4A, SSSE3, 26421 SSE4.1, SSE4.2, ABM, XSAVEC, XSAVES, CLFLUSHOPT, POPCNT, 26422 RDPID, WBNOINVD, PKU, VPCLMULQDQ, VAES, and 64-bit instruction 26423 set extensions.) 26424 26425 'btver1' 26426 CPUs based on AMD Family 14h cores with x86-64 instruction set 26427 support. (This supersets MMX, SSE, SSE2, SSE3, SSSE3, SSE4A, 26428 CX16, ABM and 64-bit instruction set extensions.) 26429 26430 'btver2' 26431 CPUs based on AMD Family 16h cores with x86-64 instruction set 26432 support. This includes MOVBE, F16C, BMI, AVX, PCLMUL, AES, 26433 SSE4.2, SSE4.1, CX16, ABM, SSE4A, SSSE3, SSE3, SSE2, SSE, MMX 26434 and 64-bit instruction set extensions. 26435 26436 'winchip-c6' 26437 IDT WinChip C6 CPU, dealt in same way as i486 with additional 26438 MMX instruction set support. 26439 26440 'winchip2' 26441 IDT WinChip 2 CPU, dealt in same way as i486 with additional 26442 MMX and 3DNow! instruction set support. 26443 26444 'c3' 26445 VIA C3 CPU with MMX and 3DNow! instruction set support. (No 26446 scheduling is implemented for this chip.) 26447 26448 'c3-2' 26449 VIA C3-2 (Nehemiah/C5XL) CPU with MMX and SSE instruction set 26450 support. (No scheduling is implemented for this chip.) 26451 26452 'c7' 26453 VIA C7 (Esther) CPU with MMX, SSE, SSE2 and SSE3 instruction 26454 set support. (No scheduling is implemented for this chip.) 26455 26456 'samuel-2' 26457 VIA Eden Samuel 2 CPU with MMX and 3DNow! instruction set 26458 support. (No scheduling is implemented for this chip.) 26459 26460 'nehemiah' 26461 VIA Eden Nehemiah CPU with MMX and SSE instruction set 26462 support. (No scheduling is implemented for this chip.) 26463 26464 'esther' 26465 VIA Eden Esther CPU with MMX, SSE, SSE2 and SSE3 instruction 26466 set support. (No scheduling is implemented for this chip.) 26467 26468 'eden-x2' 26469 VIA Eden X2 CPU with x86-64, MMX, SSE, SSE2 and SSE3 26470 instruction set support. (No scheduling is implemented for 26471 this chip.) 26472 26473 'eden-x4' 26474 VIA Eden X4 CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3, 26475 SSE4.1, SSE4.2, AVX and AVX2 instruction set support. (No 26476 scheduling is implemented for this chip.) 26477 26478 'nano' 26479 Generic VIA Nano CPU with x86-64, MMX, SSE, SSE2, SSE3 and 26480 SSSE3 instruction set support. (No scheduling is implemented 26481 for this chip.) 26482 26483 'nano-1000' 26484 VIA Nano 1xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 26485 instruction set support. (No scheduling is implemented for 26486 this chip.) 26487 26488 'nano-2000' 26489 VIA Nano 2xxx CPU with x86-64, MMX, SSE, SSE2, SSE3 and SSSE3 26490 instruction set support. (No scheduling is implemented for 26491 this chip.) 26492 26493 'nano-3000' 26494 VIA Nano 3xxx CPU with x86-64, MMX, SSE, SSE2, SSE3, SSSE3 and 26495 SSE4.1 instruction set support. (No scheduling is implemented 26496 for this chip.) 26497 26498 'nano-x2' 26499 VIA Nano Dual Core CPU with x86-64, MMX, SSE, SSE2, SSE3, 26500 SSSE3 and SSE4.1 instruction set support. (No scheduling is 26501 implemented for this chip.) 26502 26503 'nano-x4' 26504 VIA Nano Quad Core CPU with x86-64, MMX, SSE, SSE2, SSE3, 26505 SSSE3 and SSE4.1 instruction set support. (No scheduling is 26506 implemented for this chip.) 26507 26508 'geode' 26509 AMD Geode embedded processor with MMX and 3DNow! instruction 26510 set support. 26511 26512'-mtune=CPU-TYPE' 26513 Tune to CPU-TYPE everything applicable about the generated code, 26514 except for the ABI and the set of available instructions. While 26515 picking a specific CPU-TYPE schedules things appropriately for that 26516 particular chip, the compiler does not generate any code that 26517 cannot run on the default machine type unless you use a 26518 '-march=CPU-TYPE' option. For example, if GCC is configured for 26519 i686-pc-linux-gnu then '-mtune=pentium4' generates code that is 26520 tuned for Pentium 4 but still runs on i686 machines. 26521 26522 The choices for CPU-TYPE are the same as for '-march'. In 26523 addition, '-mtune' supports 2 extra choices for CPU-TYPE: 26524 26525 'generic' 26526 Produce code optimized for the most common IA32/AMD64/EM64T 26527 processors. If you know the CPU on which your code will run, 26528 then you should use the corresponding '-mtune' or '-march' 26529 option instead of '-mtune=generic'. But, if you do not know 26530 exactly what CPU users of your application will have, then you 26531 should use this option. 26532 26533 As new processors are deployed in the marketplace, the 26534 behavior of this option will change. Therefore, if you 26535 upgrade to a newer version of GCC, code generation controlled 26536 by this option will change to reflect the processors that are 26537 most common at the time that version of GCC is released. 26538 26539 There is no '-march=generic' option because '-march' indicates 26540 the instruction set the compiler can use, and there is no 26541 generic instruction set applicable to all processors. In 26542 contrast, '-mtune' indicates the processor (or, in this case, 26543 collection of processors) for which the code is optimized. 26544 26545 'intel' 26546 Produce code optimized for the most current Intel processors, 26547 which are Haswell and Silvermont for this version of GCC. If 26548 you know the CPU on which your code will run, then you should 26549 use the corresponding '-mtune' or '-march' option instead of 26550 '-mtune=intel'. But, if you want your application performs 26551 better on both Haswell and Silvermont, then you should use 26552 this option. 26553 26554 As new Intel processors are deployed in the marketplace, the 26555 behavior of this option will change. Therefore, if you 26556 upgrade to a newer version of GCC, code generation controlled 26557 by this option will change to reflect the most current Intel 26558 processors at the time that version of GCC is released. 26559 26560 There is no '-march=intel' option because '-march' indicates 26561 the instruction set the compiler can use, and there is no 26562 common instruction set applicable to all processors. In 26563 contrast, '-mtune' indicates the processor (or, in this case, 26564 collection of processors) for which the code is optimized. 26565 26566'-mcpu=CPU-TYPE' 26567 A deprecated synonym for '-mtune'. 26568 26569'-mfpmath=UNIT' 26570 Generate floating-point arithmetic for selected unit UNIT. The 26571 choices for UNIT are: 26572 26573 '387' 26574 Use the standard 387 floating-point coprocessor present on the 26575 majority of chips and emulated otherwise. Code compiled with 26576 this option runs almost everywhere. The temporary results are 26577 computed in 80-bit precision instead of the precision 26578 specified by the type, resulting in slightly different results 26579 compared to most of other chips. See '-ffloat-store' for more 26580 detailed description. 26581 26582 This is the default choice for non-Darwin x86-32 targets. 26583 26584 'sse' 26585 Use scalar floating-point instructions present in the SSE 26586 instruction set. This instruction set is supported by Pentium 26587 III and newer chips, and in the AMD line by Athlon-4, Athlon 26588 XP and Athlon MP chips. The earlier version of the SSE 26589 instruction set supports only single-precision arithmetic, 26590 thus the double and extended-precision arithmetic are still 26591 done using 387. A later version, present only in Pentium 4 26592 and AMD x86-64 chips, supports double-precision arithmetic 26593 too. 26594 26595 For the x86-32 compiler, you must use '-march=CPU-TYPE', 26596 '-msse' or '-msse2' switches to enable SSE extensions and make 26597 this option effective. For the x86-64 compiler, these 26598 extensions are enabled by default. 26599 26600 The resulting code should be considerably faster in the 26601 majority of cases and avoid the numerical instability problems 26602 of 387 code, but may break some existing code that expects 26603 temporaries to be 80 bits. 26604 26605 This is the default choice for the x86-64 compiler, Darwin 26606 x86-32 targets, and the default choice for x86-32 targets with 26607 the SSE2 instruction set when '-ffast-math' is enabled. 26608 26609 'sse,387' 26610 'sse+387' 26611 'both' 26612 Attempt to utilize both instruction sets at once. This 26613 effectively doubles the amount of available registers, and on 26614 chips with separate execution units for 387 and SSE the 26615 execution resources too. Use this option with care, as it is 26616 still experimental, because the GCC register allocator does 26617 not model separate functional units well, resulting in 26618 unstable performance. 26619 26620'-masm=DIALECT' 26621 Output assembly instructions using selected DIALECT. Also affects 26622 which dialect is used for basic 'asm' (*note Basic Asm::) and 26623 extended 'asm' (*note Extended Asm::). Supported choices (in 26624 dialect order) are 'att' or 'intel'. The default is 'att'. Darwin 26625 does not support 'intel'. 26626 26627'-mieee-fp' 26628'-mno-ieee-fp' 26629 Control whether or not the compiler uses IEEE floating-point 26630 comparisons. These correctly handle the case where the result of a 26631 comparison is unordered. 26632 26633'-m80387' 26634'-mhard-float' 26635 Generate output containing 80387 instructions for floating point. 26636 26637'-mno-80387' 26638'-msoft-float' 26639 Generate output containing library calls for floating point. 26640 26641 *Warning:* the requisite libraries are not part of GCC. Normally 26642 the facilities of the machine's usual C compiler are used, but this 26643 cannot be done directly in cross-compilation. You must make your 26644 own arrangements to provide suitable library functions for 26645 cross-compilation. 26646 26647 On machines where a function returns floating-point results in the 26648 80387 register stack, some floating-point opcodes may be emitted 26649 even if '-msoft-float' is used. 26650 26651'-mno-fp-ret-in-387' 26652 Do not use the FPU registers for return values of functions. 26653 26654 The usual calling convention has functions return values of types 26655 'float' and 'double' in an FPU register, even if there is no FPU. 26656 The idea is that the operating system should emulate an FPU. 26657 26658 The option '-mno-fp-ret-in-387' causes such values to be returned 26659 in ordinary CPU registers instead. 26660 26661'-mno-fancy-math-387' 26662 Some 387 emulators do not support the 'sin', 'cos' and 'sqrt' 26663 instructions for the 387. Specify this option to avoid generating 26664 those instructions. This option is overridden when '-march' 26665 indicates that the target CPU always has an FPU and so the 26666 instruction does not need emulation. These instructions are not 26667 generated unless you also use the '-funsafe-math-optimizations' 26668 switch. 26669 26670'-malign-double' 26671'-mno-align-double' 26672 Control whether GCC aligns 'double', 'long double', and 'long long' 26673 variables on a two-word boundary or a one-word boundary. Aligning 26674 'double' variables on a two-word boundary produces code that runs 26675 somewhat faster on a Pentium at the expense of more memory. 26676 26677 On x86-64, '-malign-double' is enabled by default. 26678 26679 *Warning:* if you use the '-malign-double' switch, structures 26680 containing the above types are aligned differently than the 26681 published application binary interface specifications for the 26682 x86-32 and are not binary compatible with structures in code 26683 compiled without that switch. 26684 26685'-m96bit-long-double' 26686'-m128bit-long-double' 26687 These switches control the size of 'long double' type. The x86-32 26688 application binary interface specifies the size to be 96 bits, so 26689 '-m96bit-long-double' is the default in 32-bit mode. 26690 26691 Modern architectures (Pentium and newer) prefer 'long double' to be 26692 aligned to an 8- or 16-byte boundary. In arrays or structures 26693 conforming to the ABI, this is not possible. So specifying 26694 '-m128bit-long-double' aligns 'long double' to a 16-byte boundary 26695 by padding the 'long double' with an additional 32-bit zero. 26696 26697 In the x86-64 compiler, '-m128bit-long-double' is the default 26698 choice as its ABI specifies that 'long double' is aligned on 26699 16-byte boundary. 26700 26701 Notice that neither of these options enable any extra precision 26702 over the x87 standard of 80 bits for a 'long double'. 26703 26704 *Warning:* if you override the default value for your target ABI, 26705 this changes the size of structures and arrays containing 'long 26706 double' variables, as well as modifying the function calling 26707 convention for functions taking 'long double'. Hence they are not 26708 binary-compatible with code compiled without that switch. 26709 26710'-mlong-double-64' 26711'-mlong-double-80' 26712'-mlong-double-128' 26713 These switches control the size of 'long double' type. A size of 26714 64 bits makes the 'long double' type equivalent to the 'double' 26715 type. This is the default for 32-bit Bionic C library. A size of 26716 128 bits makes the 'long double' type equivalent to the 26717 '__float128' type. This is the default for 64-bit Bionic C 26718 library. 26719 26720 *Warning:* if you override the default value for your target ABI, 26721 this changes the size of structures and arrays containing 'long 26722 double' variables, as well as modifying the function calling 26723 convention for functions taking 'long double'. Hence they are not 26724 binary-compatible with code compiled without that switch. 26725 26726'-malign-data=TYPE' 26727 Control how GCC aligns variables. Supported values for TYPE are 26728 'compat' uses increased alignment value compatible uses GCC 4.8 and 26729 earlier, 'abi' uses alignment value as specified by the psABI, and 26730 'cacheline' uses increased alignment value to match the cache line 26731 size. 'compat' is the default. 26732 26733'-mlarge-data-threshold=THRESHOLD' 26734 When '-mcmodel=medium' is specified, data objects larger than 26735 THRESHOLD are placed in the large data section. This value must be 26736 the same across all objects linked into the binary, and defaults to 26737 65535. 26738 26739'-mrtd' 26740 Use a different function-calling convention, in which functions 26741 that take a fixed number of arguments return with the 'ret NUM' 26742 instruction, which pops their arguments while returning. This 26743 saves one instruction in the caller since there is no need to pop 26744 the arguments there. 26745 26746 You can specify that an individual function is called with this 26747 calling sequence with the function attribute 'stdcall'. You can 26748 also override the '-mrtd' option by using the function attribute 26749 'cdecl'. *Note Function Attributes::. 26750 26751 *Warning:* this calling convention is incompatible with the one 26752 normally used on Unix, so you cannot use it if you need to call 26753 libraries compiled with the Unix compiler. 26754 26755 Also, you must provide function prototypes for all functions that 26756 take variable numbers of arguments (including 'printf'); otherwise 26757 incorrect code is generated for calls to those functions. 26758 26759 In addition, seriously incorrect code results if you call a 26760 function with too many arguments. (Normally, extra arguments are 26761 harmlessly ignored.) 26762 26763'-mregparm=NUM' 26764 Control how many registers are used to pass integer arguments. By 26765 default, no registers are used to pass arguments, and at most 3 26766 registers can be used. You can control this behavior for a 26767 specific function by using the function attribute 'regparm'. *Note 26768 Function Attributes::. 26769 26770 *Warning:* if you use this switch, and NUM is nonzero, then you 26771 must build all modules with the same value, including any 26772 libraries. This includes the system libraries and startup modules. 26773 26774'-msseregparm' 26775 Use SSE register passing conventions for float and double arguments 26776 and return values. You can control this behavior for a specific 26777 function by using the function attribute 'sseregparm'. *Note 26778 Function Attributes::. 26779 26780 *Warning:* if you use this switch then you must build all modules 26781 with the same value, including any libraries. This includes the 26782 system libraries and startup modules. 26783 26784'-mvect8-ret-in-mem' 26785 Return 8-byte vectors in memory instead of MMX registers. This is 26786 the default on VxWorks to match the ABI of the Sun Studio compilers 26787 until version 12. _Only_ use this option if you need to remain 26788 compatible with existing code produced by those previous compiler 26789 versions or older versions of GCC. 26790 26791'-mpc32' 26792'-mpc64' 26793'-mpc80' 26794 26795 Set 80387 floating-point precision to 32, 64 or 80 bits. When 26796 '-mpc32' is specified, the significands of results of 26797 floating-point operations are rounded to 24 bits (single 26798 precision); '-mpc64' rounds the significands of results of 26799 floating-point operations to 53 bits (double precision) and 26800 '-mpc80' rounds the significands of results of floating-point 26801 operations to 64 bits (extended double precision), which is the 26802 default. When this option is used, floating-point operations in 26803 higher precisions are not available to the programmer without 26804 setting the FPU control word explicitly. 26805 26806 Setting the rounding of floating-point operations to less than the 26807 default 80 bits can speed some programs by 2% or more. Note that 26808 some mathematical libraries assume that extended-precision (80-bit) 26809 floating-point operations are enabled by default; routines in such 26810 libraries could suffer significant loss of accuracy, typically 26811 through so-called "catastrophic cancellation", when this option is 26812 used to set the precision to less than extended precision. 26813 26814'-mstackrealign' 26815 Realign the stack at entry. On the x86, the '-mstackrealign' 26816 option generates an alternate prologue and epilogue that realigns 26817 the run-time stack if necessary. This supports mixing legacy codes 26818 that keep 4-byte stack alignment with modern codes that keep 26819 16-byte stack alignment for SSE compatibility. See also the 26820 attribute 'force_align_arg_pointer', applicable to individual 26821 functions. 26822 26823'-mpreferred-stack-boundary=NUM' 26824 Attempt to keep the stack boundary aligned to a 2 raised to NUM 26825 byte boundary. If '-mpreferred-stack-boundary' is not specified, 26826 the default is 4 (16 bytes or 128 bits). 26827 26828 *Warning:* When generating code for the x86-64 architecture with 26829 SSE extensions disabled, '-mpreferred-stack-boundary=3' can be used 26830 to keep the stack boundary aligned to 8 byte boundary. Since 26831 x86-64 ABI require 16 byte stack alignment, this is ABI 26832 incompatible and intended to be used in controlled environment 26833 where stack space is important limitation. This option leads to 26834 wrong code when functions compiled with 16 byte stack alignment 26835 (such as functions from a standard library) are called with 26836 misaligned stack. In this case, SSE instructions may lead to 26837 misaligned memory access traps. In addition, variable arguments 26838 are handled incorrectly for 16 byte aligned objects (including x87 26839 long double and __int128), leading to wrong results. You must 26840 build all modules with '-mpreferred-stack-boundary=3', including 26841 any libraries. This includes the system libraries and startup 26842 modules. 26843 26844'-mincoming-stack-boundary=NUM' 26845 Assume the incoming stack is aligned to a 2 raised to NUM byte 26846 boundary. If '-mincoming-stack-boundary' is not specified, the one 26847 specified by '-mpreferred-stack-boundary' is used. 26848 26849 On Pentium and Pentium Pro, 'double' and 'long double' values 26850 should be aligned to an 8-byte boundary (see '-malign-double') or 26851 suffer significant run time performance penalties. On Pentium III, 26852 the Streaming SIMD Extension (SSE) data type '__m128' may not work 26853 properly if it is not 16-byte aligned. 26854 26855 To ensure proper alignment of this values on the stack, the stack 26856 boundary must be as aligned as that required by any value stored on 26857 the stack. Further, every function must be generated such that it 26858 keeps the stack aligned. Thus calling a function compiled with a 26859 higher preferred stack boundary from a function compiled with a 26860 lower preferred stack boundary most likely misaligns the stack. It 26861 is recommended that libraries that use callbacks always use the 26862 default setting. 26863 26864 This extra alignment does consume extra stack space, and generally 26865 increases code size. Code that is sensitive to stack space usage, 26866 such as embedded systems and operating system kernels, may want to 26867 reduce the preferred alignment to '-mpreferred-stack-boundary=2'. 26868 26869'-mmmx' 26870'-msse' 26871'-msse2' 26872'-msse3' 26873'-mssse3' 26874'-msse4' 26875'-msse4a' 26876'-msse4.1' 26877'-msse4.2' 26878'-mavx' 26879'-mavx2' 26880'-mavx512f' 26881'-mavx512pf' 26882'-mavx512er' 26883'-mavx512cd' 26884'-mavx512vl' 26885'-mavx512bw' 26886'-mavx512dq' 26887'-mavx512ifma' 26888'-mavx512vbmi' 26889'-msha' 26890'-maes' 26891'-mpclmul' 26892'-mclflushopt' 26893'-mclwb' 26894'-mfsgsbase' 26895'-mptwrite' 26896'-mrdrnd' 26897'-mf16c' 26898'-mfma' 26899'-mpconfig' 26900'-mwbnoinvd' 26901'-mfma4' 26902'-mprfchw' 26903'-mrdpid' 26904'-mprefetchwt1' 26905'-mrdseed' 26906'-msgx' 26907'-mxop' 26908'-mlwp' 26909'-m3dnow' 26910'-m3dnowa' 26911'-mpopcnt' 26912'-mabm' 26913'-madx' 26914'-mbmi' 26915'-mbmi2' 26916'-mlzcnt' 26917'-mfxsr' 26918'-mxsave' 26919'-mxsaveopt' 26920'-mxsavec' 26921'-mxsaves' 26922'-mrtm' 26923'-mhle' 26924'-mtbm' 26925'-mmwaitx' 26926'-mclzero' 26927'-mpku' 26928'-mavx512vbmi2' 26929'-mavx512bf16' 26930'-mgfni' 26931'-mvaes' 26932'-mwaitpkg' 26933'-mvpclmulqdq' 26934'-mavx512bitalg' 26935'-mmovdiri' 26936'-mmovdir64b' 26937'-menqcmd' 26938'-mavx512vpopcntdq' 26939'-mavx512vp2intersect' 26940'-mavx5124fmaps' 26941'-mavx512vnni' 26942'-mavx5124vnniw' 26943'-mcldemote' 26944 These switches enable the use of instructions in the MMX, SSE, 26945 SSE2, SSE3, SSSE3, SSE4, SSE4A, SSE4.1, SSE4.2, AVX, AVX2, AVX512F, 26946 AVX512PF, AVX512ER, AVX512CD, AVX512VL, AVX512BW, AVX512DQ, 26947 AVX512IFMA, AVX512VBMI, SHA, AES, PCLMUL, CLFLUSHOPT, CLWB, 26948 FSGSBASE, PTWRITE, RDRND, F16C, FMA, PCONFIG, WBNOINVD, FMA4, 26949 PREFETCHW, RDPID, PREFETCHWT1, RDSEED, SGX, XOP, LWP, 3DNow!, 26950 enhanced 3DNow!, POPCNT, ABM, ADX, BMI, BMI2, LZCNT, FXSR, XSAVE, 26951 XSAVEOPT, XSAVEC, XSAVES, RTM, HLE, TBM, MWAITX, CLZERO, PKU, 26952 AVX512VBMI2, GFNI, VAES, WAITPKG, VPCLMULQDQ, AVX512BITALG, 26953 MOVDIRI, MOVDIR64B, AVX512BF16, ENQCMD, AVX512VPOPCNTDQ, 26954 AVX5124FMAPS, AVX512VNNI, AVX5124VNNIW, or CLDEMOTE extended 26955 instruction sets. Each has a corresponding '-mno-' option to 26956 disable use of these instructions. 26957 26958 These extensions are also available as built-in functions: see 26959 *note x86 Built-in Functions::, for details of the functions 26960 enabled and disabled by these switches. 26961 26962 To generate SSE/SSE2 instructions automatically from floating-point 26963 code (as opposed to 387 instructions), see '-mfpmath=sse'. 26964 26965 GCC depresses SSEx instructions when '-mavx' is used. Instead, it 26966 generates new AVX instructions or AVX equivalence for all SSEx 26967 instructions when needed. 26968 26969 These options enable GCC to use these extended instructions in 26970 generated code, even without '-mfpmath=sse'. Applications that 26971 perform run-time CPU detection must compile separate files for each 26972 supported architecture, using the appropriate flags. In 26973 particular, the file containing the CPU detection code should be 26974 compiled without these options. 26975 26976'-mdump-tune-features' 26977 This option instructs GCC to dump the names of the x86 performance 26978 tuning features and default settings. The names can be used in 26979 '-mtune-ctrl=FEATURE-LIST'. 26980 26981'-mtune-ctrl=FEATURE-LIST' 26982 This option is used to do fine grain control of x86 code generation 26983 features. FEATURE-LIST is a comma separated list of FEATURE names. 26984 See also '-mdump-tune-features'. When specified, the FEATURE is 26985 turned on if it is not preceded with '^', otherwise, it is turned 26986 off. '-mtune-ctrl=FEATURE-LIST' is intended to be used by GCC 26987 developers. Using it may lead to code paths not covered by testing 26988 and can potentially result in compiler ICEs or runtime errors. 26989 26990'-mno-default' 26991 This option instructs GCC to turn off all tunable features. See 26992 also '-mtune-ctrl=FEATURE-LIST' and '-mdump-tune-features'. 26993 26994'-mcld' 26995 This option instructs GCC to emit a 'cld' instruction in the 26996 prologue of functions that use string instructions. String 26997 instructions depend on the DF flag to select between autoincrement 26998 or autodecrement mode. While the ABI specifies the DF flag to be 26999 cleared on function entry, some operating systems violate this 27000 specification by not clearing the DF flag in their exception 27001 dispatchers. The exception handler can be invoked with the DF flag 27002 set, which leads to wrong direction mode when string instructions 27003 are used. This option can be enabled by default on 32-bit x86 27004 targets by configuring GCC with the '--enable-cld' configure 27005 option. Generation of 'cld' instructions can be suppressed with 27006 the '-mno-cld' compiler option in this case. 27007 27008'-mvzeroupper' 27009 This option instructs GCC to emit a 'vzeroupper' instruction before 27010 a transfer of control flow out of the function to minimize the AVX 27011 to SSE transition penalty as well as remove unnecessary 'zeroupper' 27012 intrinsics. 27013 27014'-mprefer-avx128' 27015 This option instructs GCC to use 128-bit AVX instructions instead 27016 of 256-bit AVX instructions in the auto-vectorizer. 27017 27018'-mprefer-vector-width=OPT' 27019 This option instructs GCC to use OPT-bit vector width in 27020 instructions instead of default on the selected platform. 27021 27022 'none' 27023 No extra limitations applied to GCC other than defined by the 27024 selected platform. 27025 27026 '128' 27027 Prefer 128-bit vector width for instructions. 27028 27029 '256' 27030 Prefer 256-bit vector width for instructions. 27031 27032 '512' 27033 Prefer 512-bit vector width for instructions. 27034 27035'-mcx16' 27036 This option enables GCC to generate 'CMPXCHG16B' instructions in 27037 64-bit code to implement compare-and-exchange operations on 16-byte 27038 aligned 128-bit objects. This is useful for atomic updates of data 27039 structures exceeding one machine word in size. The compiler uses 27040 this instruction to implement *note __sync Builtins::. However, 27041 for *note __atomic Builtins:: operating on 128-bit integers, a 27042 library call is always used. 27043 27044'-msahf' 27045 This option enables generation of 'SAHF' instructions in 64-bit 27046 code. Early Intel Pentium 4 CPUs with Intel 64 support, prior to 27047 the introduction of Pentium 4 G1 step in December 2005, lacked the 27048 'LAHF' and 'SAHF' instructions which are supported by AMD64. These 27049 are load and store instructions, respectively, for certain status 27050 flags. In 64-bit mode, the 'SAHF' instruction is used to optimize 27051 'fmod', 'drem', and 'remainder' built-in functions; see *note Other 27052 Builtins:: for details. 27053 27054'-mmovbe' 27055 This option enables use of the 'movbe' instruction to implement 27056 '__builtin_bswap32' and '__builtin_bswap64'. 27057 27058'-mshstk' 27059 The '-mshstk' option enables shadow stack built-in functions from 27060 x86 Control-flow Enforcement Technology (CET). 27061 27062'-mcrc32' 27063 This option enables built-in functions '__builtin_ia32_crc32qi', 27064 '__builtin_ia32_crc32hi', '__builtin_ia32_crc32si' and 27065 '__builtin_ia32_crc32di' to generate the 'crc32' machine 27066 instruction. 27067 27068'-mrecip' 27069 This option enables use of 'RCPSS' and 'RSQRTSS' instructions (and 27070 their vectorized variants 'RCPPS' and 'RSQRTPS') with an additional 27071 Newton-Raphson step to increase precision instead of 'DIVSS' and 27072 'SQRTSS' (and their vectorized variants) for single-precision 27073 floating-point arguments. These instructions are generated only 27074 when '-funsafe-math-optimizations' is enabled together with 27075 '-ffinite-math-only' and '-fno-trapping-math'. Note that while the 27076 throughput of the sequence is higher than the throughput of the 27077 non-reciprocal instruction, the precision of the sequence can be 27078 decreased by up to 2 ulp (i.e. the inverse of 1.0 equals 27079 0.99999994). 27080 27081 Note that GCC implements '1.0f/sqrtf(X)' in terms of 'RSQRTSS' (or 27082 'RSQRTPS') already with '-ffast-math' (or the above option 27083 combination), and doesn't need '-mrecip'. 27084 27085 Also note that GCC emits the above sequence with additional 27086 Newton-Raphson step for vectorized single-float division and 27087 vectorized 'sqrtf(X)' already with '-ffast-math' (or the above 27088 option combination), and doesn't need '-mrecip'. 27089 27090'-mrecip=OPT' 27091 This option controls which reciprocal estimate instructions may be 27092 used. OPT is a comma-separated list of options, which may be 27093 preceded by a '!' to invert the option: 27094 27095 'all' 27096 Enable all estimate instructions. 27097 27098 'default' 27099 Enable the default instructions, equivalent to '-mrecip'. 27100 27101 'none' 27102 Disable all estimate instructions, equivalent to '-mno-recip'. 27103 27104 'div' 27105 Enable the approximation for scalar division. 27106 27107 'vec-div' 27108 Enable the approximation for vectorized division. 27109 27110 'sqrt' 27111 Enable the approximation for scalar square root. 27112 27113 'vec-sqrt' 27114 Enable the approximation for vectorized square root. 27115 27116 So, for example, '-mrecip=all,!sqrt' enables all of the reciprocal 27117 approximations, except for square root. 27118 27119'-mveclibabi=TYPE' 27120 Specifies the ABI type to use for vectorizing intrinsics using an 27121 external library. Supported values for TYPE are 'svml' for the 27122 Intel short vector math library and 'acml' for the AMD math core 27123 library. To use this option, both '-ftree-vectorize' and 27124 '-funsafe-math-optimizations' have to be enabled, and an SVML or 27125 ACML ABI-compatible library must be specified at link time. 27126 27127 GCC currently emits calls to 'vmldExp2', 'vmldLn2', 'vmldLog102', 27128 'vmldPow2', 'vmldTanh2', 'vmldTan2', 'vmldAtan2', 'vmldAtanh2', 27129 'vmldCbrt2', 'vmldSinh2', 'vmldSin2', 'vmldAsinh2', 'vmldAsin2', 27130 'vmldCosh2', 'vmldCos2', 'vmldAcosh2', 'vmldAcos2', 'vmlsExp4', 27131 'vmlsLn4', 'vmlsLog104', 'vmlsPow4', 'vmlsTanh4', 'vmlsTan4', 27132 'vmlsAtan4', 'vmlsAtanh4', 'vmlsCbrt4', 'vmlsSinh4', 'vmlsSin4', 27133 'vmlsAsinh4', 'vmlsAsin4', 'vmlsCosh4', 'vmlsCos4', 'vmlsAcosh4' 27134 and 'vmlsAcos4' for corresponding function type when 27135 '-mveclibabi=svml' is used, and '__vrd2_sin', '__vrd2_cos', 27136 '__vrd2_exp', '__vrd2_log', '__vrd2_log2', '__vrd2_log10', 27137 '__vrs4_sinf', '__vrs4_cosf', '__vrs4_expf', '__vrs4_logf', 27138 '__vrs4_log2f', '__vrs4_log10f' and '__vrs4_powf' for the 27139 corresponding function type when '-mveclibabi=acml' is used. 27140 27141'-mabi=NAME' 27142 Generate code for the specified calling convention. Permissible 27143 values are 'sysv' for the ABI used on GNU/Linux and other systems, 27144 and 'ms' for the Microsoft ABI. The default is to use the Microsoft 27145 ABI when targeting Microsoft Windows and the SysV ABI on all other 27146 systems. You can control this behavior for specific functions by 27147 using the function attributes 'ms_abi' and 'sysv_abi'. *Note 27148 Function Attributes::. 27149 27150'-mforce-indirect-call' 27151 Force all calls to functions to be indirect. This is useful when 27152 using Intel Processor Trace where it generates more precise timing 27153 information for function calls. 27154 27155'-mmanual-endbr' 27156 Insert ENDBR instruction at function entry only via the 'cf_check' 27157 function attribute. This is useful when used with the option 27158 '-fcf-protection=branch' to control ENDBR insertion at the function 27159 entry. 27160 27161'-mcall-ms2sysv-xlogues' 27162 Due to differences in 64-bit ABIs, any Microsoft ABI function that 27163 calls a System V ABI function must consider RSI, RDI and XMM6-15 as 27164 clobbered. By default, the code for saving and restoring these 27165 registers is emitted inline, resulting in fairly lengthy prologues 27166 and epilogues. Using '-mcall-ms2sysv-xlogues' emits prologues and 27167 epilogues that use stubs in the static portion of libgcc to perform 27168 these saves and restores, thus reducing function size at the cost 27169 of a few extra instructions. 27170 27171'-mtls-dialect=TYPE' 27172 Generate code to access thread-local storage using the 'gnu' or 27173 'gnu2' conventions. 'gnu' is the conservative default; 'gnu2' is 27174 more efficient, but it may add compile- and run-time requirements 27175 that cannot be satisfied on all systems. 27176 27177'-mpush-args' 27178'-mno-push-args' 27179 Use PUSH operations to store outgoing parameters. This method is 27180 shorter and usually equally fast as method using SUB/MOV operations 27181 and is enabled by default. In some cases disabling it may improve 27182 performance because of improved scheduling and reduced 27183 dependencies. 27184 27185'-maccumulate-outgoing-args' 27186 If enabled, the maximum amount of space required for outgoing 27187 arguments is computed in the function prologue. This is faster on 27188 most modern CPUs because of reduced dependencies, improved 27189 scheduling and reduced stack usage when the preferred stack 27190 boundary is not equal to 2. The drawback is a notable increase in 27191 code size. This switch implies '-mno-push-args'. 27192 27193'-mthreads' 27194 Support thread-safe exception handling on MinGW. Programs that rely 27195 on thread-safe exception handling must compile and link all code 27196 with the '-mthreads' option. When compiling, '-mthreads' defines 27197 '-D_MT'; when linking, it links in a special thread helper library 27198 '-lmingwthrd' which cleans up per-thread exception-handling data. 27199 27200'-mms-bitfields' 27201'-mno-ms-bitfields' 27202 27203 Enable/disable bit-field layout compatible with the native 27204 Microsoft Windows compiler. 27205 27206 If 'packed' is used on a structure, or if bit-fields are used, it 27207 may be that the Microsoft ABI lays out the structure differently 27208 than the way GCC normally does. Particularly when moving packed 27209 data between functions compiled with GCC and the native Microsoft 27210 compiler (either via function call or as data in a file), it may be 27211 necessary to access either format. 27212 27213 This option is enabled by default for Microsoft Windows targets. 27214 This behavior can also be controlled locally by use of variable or 27215 type attributes. For more information, see *note x86 Variable 27216 Attributes:: and *note x86 Type Attributes::. 27217 27218 The Microsoft structure layout algorithm is fairly simple with the 27219 exception of the bit-field packing. The padding and alignment of 27220 members of structures and whether a bit-field can straddle a 27221 storage-unit boundary are determine by these rules: 27222 27223 1. Structure members are stored sequentially in the order in 27224 which they are declared: the first member has the lowest 27225 memory address and the last member the highest. 27226 27227 2. Every data object has an alignment requirement. The alignment 27228 requirement for all data except structures, unions, and arrays 27229 is either the size of the object or the current packing size 27230 (specified with either the 'aligned' attribute or the 'pack' 27231 pragma), whichever is less. For structures, unions, and 27232 arrays, the alignment requirement is the largest alignment 27233 requirement of its members. Every object is allocated an 27234 offset so that: 27235 27236 offset % alignment_requirement == 0 27237 27238 3. Adjacent bit-fields are packed into the same 1-, 2-, or 4-byte 27239 allocation unit if the integral types are the same size and if 27240 the next bit-field fits into the current allocation unit 27241 without crossing the boundary imposed by the common alignment 27242 requirements of the bit-fields. 27243 27244 MSVC interprets zero-length bit-fields in the following ways: 27245 27246 1. If a zero-length bit-field is inserted between two bit-fields 27247 that are normally coalesced, the bit-fields are not coalesced. 27248 27249 For example: 27250 27251 struct 27252 { 27253 unsigned long bf_1 : 12; 27254 unsigned long : 0; 27255 unsigned long bf_2 : 12; 27256 } t1; 27257 27258 The size of 't1' is 8 bytes with the zero-length bit-field. 27259 If the zero-length bit-field were removed, 't1''s size would 27260 be 4 bytes. 27261 27262 2. If a zero-length bit-field is inserted after a bit-field, 27263 'foo', and the alignment of the zero-length bit-field is 27264 greater than the member that follows it, 'bar', 'bar' is 27265 aligned as the type of the zero-length bit-field. 27266 27267 For example: 27268 27269 struct 27270 { 27271 char foo : 4; 27272 short : 0; 27273 char bar; 27274 } t2; 27275 27276 struct 27277 { 27278 char foo : 4; 27279 short : 0; 27280 double bar; 27281 } t3; 27282 27283 For 't2', 'bar' is placed at offset 2, rather than offset 1. 27284 Accordingly, the size of 't2' is 4. For 't3', the zero-length 27285 bit-field does not affect the alignment of 'bar' or, as a 27286 result, the size of the structure. 27287 27288 Taking this into account, it is important to note the 27289 following: 27290 27291 1. If a zero-length bit-field follows a normal bit-field, 27292 the type of the zero-length bit-field may affect the 27293 alignment of the structure as whole. For example, 't2' 27294 has a size of 4 bytes, since the zero-length bit-field 27295 follows a normal bit-field, and is of type short. 27296 27297 2. Even if a zero-length bit-field is not followed by a 27298 normal bit-field, it may still affect the alignment of 27299 the structure: 27300 27301 struct 27302 { 27303 char foo : 6; 27304 long : 0; 27305 } t4; 27306 27307 Here, 't4' takes up 4 bytes. 27308 27309 3. Zero-length bit-fields following non-bit-field members are 27310 ignored: 27311 27312 struct 27313 { 27314 char foo; 27315 long : 0; 27316 char bar; 27317 } t5; 27318 27319 Here, 't5' takes up 2 bytes. 27320 27321'-mno-align-stringops' 27322 Do not align the destination of inlined string operations. This 27323 switch reduces code size and improves performance in case the 27324 destination is already aligned, but GCC doesn't know about it. 27325 27326'-minline-all-stringops' 27327 By default GCC inlines string operations only when the destination 27328 is known to be aligned to least a 4-byte boundary. This enables 27329 more inlining and increases code size, but may improve performance 27330 of code that depends on fast 'memcpy' and 'memset' for short 27331 lengths. The option enables inline expansion of 'strlen' for all 27332 pointer alignments. 27333 27334'-minline-stringops-dynamically' 27335 For string operations of unknown size, use run-time checks with 27336 inline code for small blocks and a library call for large blocks. 27337 27338'-mstringop-strategy=ALG' 27339 Override the internal decision heuristic for the particular 27340 algorithm to use for inlining string operations. The allowed 27341 values for ALG are: 27342 27343 'rep_byte' 27344 'rep_4byte' 27345 'rep_8byte' 27346 Expand using i386 'rep' prefix of the specified size. 27347 27348 'byte_loop' 27349 'loop' 27350 'unrolled_loop' 27351 Expand into an inline loop. 27352 27353 'libcall' 27354 Always use a library call. 27355 27356'-mmemcpy-strategy=STRATEGY' 27357 Override the internal decision heuristic to decide if 27358 '__builtin_memcpy' should be inlined and what inline algorithm to 27359 use when the expected size of the copy operation is known. 27360 STRATEGY is a comma-separated list of ALG:MAX_SIZE:DEST_ALIGN 27361 triplets. ALG is specified in '-mstringop-strategy', MAX_SIZE 27362 specifies the max byte size with which inline algorithm ALG is 27363 allowed. For the last triplet, the MAX_SIZE must be '-1'. The 27364 MAX_SIZE of the triplets in the list must be specified in 27365 increasing order. The minimal byte size for ALG is '0' for the 27366 first triplet and 'MAX_SIZE + 1' of the preceding range. 27367 27368'-mmemset-strategy=STRATEGY' 27369 The option is similar to '-mmemcpy-strategy=' except that it is to 27370 control '__builtin_memset' expansion. 27371 27372'-momit-leaf-frame-pointer' 27373 Don't keep the frame pointer in a register for leaf functions. 27374 This avoids the instructions to save, set up, and restore frame 27375 pointers and makes an extra register available in leaf functions. 27376 The option '-fomit-leaf-frame-pointer' removes the frame pointer 27377 for leaf functions, which might make debugging harder. 27378 27379'-mtls-direct-seg-refs' 27380'-mno-tls-direct-seg-refs' 27381 Controls whether TLS variables may be accessed with offsets from 27382 the TLS segment register ('%gs' for 32-bit, '%fs' for 64-bit), or 27383 whether the thread base pointer must be added. Whether or not this 27384 is valid depends on the operating system, and whether it maps the 27385 segment to cover the entire TLS area. 27386 27387 For systems that use the GNU C Library, the default is on. 27388 27389'-msse2avx' 27390'-mno-sse2avx' 27391 Specify that the assembler should encode SSE instructions with VEX 27392 prefix. The option '-mavx' turns this on by default. 27393 27394'-mfentry' 27395'-mno-fentry' 27396 If profiling is active ('-pg'), put the profiling counter call 27397 before the prologue. Note: On x86 architectures the attribute 27398 'ms_hook_prologue' isn't possible at the moment for '-mfentry' and 27399 '-pg'. 27400 27401'-mrecord-mcount' 27402'-mno-record-mcount' 27403 If profiling is active ('-pg'), generate a __mcount_loc section 27404 that contains pointers to each profiling call. This is useful for 27405 automatically patching and out calls. 27406 27407'-mnop-mcount' 27408'-mno-nop-mcount' 27409 If profiling is active ('-pg'), generate the calls to the profiling 27410 functions as NOPs. This is useful when they should be patched in 27411 later dynamically. This is likely only useful together with 27412 '-mrecord-mcount'. 27413 27414'-minstrument-return=TYPE' 27415 Instrument function exit in -pg -mfentry instrumented functions 27416 with call to specified function. This only instruments true 27417 returns ending with ret, but not sibling calls ending with jump. 27418 Valid types are NONE to not instrument, CALL to generate a call to 27419 __return__, or NOP5 to generate a 5 byte nop. 27420 27421'-mrecord-return' 27422'-mno-record-return' 27423 Generate a __return_loc section pointing to all return 27424 instrumentation code. 27425 27426'-mfentry-name=NAME' 27427 Set name of __fentry__ symbol called at function entry for -pg 27428 -mfentry functions. 27429 27430'-mfentry-section=NAME' 27431 Set name of section to record -mrecord-mcount calls (default 27432 __mcount_loc). 27433 27434'-mskip-rax-setup' 27435'-mno-skip-rax-setup' 27436 When generating code for the x86-64 architecture with SSE 27437 extensions disabled, '-mskip-rax-setup' can be used to skip setting 27438 up RAX register when there are no variable arguments passed in 27439 vector registers. 27440 27441 *Warning:* Since RAX register is used to avoid unnecessarily saving 27442 vector registers on stack when passing variable arguments, the 27443 impacts of this option are callees may waste some stack space, 27444 misbehave or jump to a random location. GCC 4.4 or newer don't 27445 have those issues, regardless the RAX register value. 27446 27447'-m8bit-idiv' 27448'-mno-8bit-idiv' 27449 On some processors, like Intel Atom, 8-bit unsigned integer divide 27450 is much faster than 32-bit/64-bit integer divide. This option 27451 generates a run-time check. If both dividend and divisor are 27452 within range of 0 to 255, 8-bit unsigned integer divide is used 27453 instead of 32-bit/64-bit integer divide. 27454 27455'-mavx256-split-unaligned-load' 27456'-mavx256-split-unaligned-store' 27457 Split 32-byte AVX unaligned load and store. 27458 27459'-mstack-protector-guard=GUARD' 27460'-mstack-protector-guard-reg=REG' 27461'-mstack-protector-guard-offset=OFFSET' 27462 Generate stack protection code using canary at GUARD. Supported 27463 locations are 'global' for global canary or 'tls' for per-thread 27464 canary in the TLS block (the default). This option has effect only 27465 when '-fstack-protector' or '-fstack-protector-all' is specified. 27466 27467 With the latter choice the options 27468 '-mstack-protector-guard-reg=REG' and 27469 '-mstack-protector-guard-offset=OFFSET' furthermore specify which 27470 segment register ('%fs' or '%gs') to use as base register for 27471 reading the canary, and from what offset from that base register. 27472 The default for those is as specified in the relevant ABI. 27473 27474'-mgeneral-regs-only' 27475 Generate code that uses only the general-purpose registers. This 27476 prevents the compiler from using floating-point, vector, mask and 27477 bound registers. 27478 27479'-mindirect-branch=CHOICE' 27480 Convert indirect call and jump with CHOICE. The default is 'keep', 27481 which keeps indirect call and jump unmodified. 'thunk' converts 27482 indirect call and jump to call and return thunk. 'thunk-inline' 27483 converts indirect call and jump to inlined call and return thunk. 27484 'thunk-extern' converts indirect call and jump to external call and 27485 return thunk provided in a separate object file. You can control 27486 this behavior for a specific function by using the function 27487 attribute 'indirect_branch'. *Note Function Attributes::. 27488 27489 Note that '-mcmodel=large' is incompatible with 27490 '-mindirect-branch=thunk' and '-mindirect-branch=thunk-extern' 27491 since the thunk function may not be reachable in the large code 27492 model. 27493 27494 Note that '-mindirect-branch=thunk-extern' is compatible with 27495 '-fcf-protection=branch' since the external thunk can be made to 27496 enable control-flow check. 27497 27498'-mfunction-return=CHOICE' 27499 Convert function return with CHOICE. The default is 'keep', which 27500 keeps function return unmodified. 'thunk' converts function return 27501 to call and return thunk. 'thunk-inline' converts function return 27502 to inlined call and return thunk. 'thunk-extern' converts function 27503 return to external call and return thunk provided in a separate 27504 object file. You can control this behavior for a specific function 27505 by using the function attribute 'function_return'. *Note Function 27506 Attributes::. 27507 27508 Note that '-mindirect-return=thunk-extern' is compatible with 27509 '-fcf-protection=branch' since the external thunk can be made to 27510 enable control-flow check. 27511 27512 Note that '-mcmodel=large' is incompatible with 27513 '-mfunction-return=thunk' and '-mfunction-return=thunk-extern' 27514 since the thunk function may not be reachable in the large code 27515 model. 27516 27517'-mindirect-branch-register' 27518 Force indirect call and jump via register. 27519 27520 These '-m' switches are supported in addition to the above on x86-64 27521processors in 64-bit environments. 27522 27523'-m32' 27524'-m64' 27525'-mx32' 27526'-m16' 27527'-miamcu' 27528 Generate code for a 16-bit, 32-bit or 64-bit environment. The 27529 '-m32' option sets 'int', 'long', and pointer types to 32 bits, and 27530 generates code that runs in 32-bit mode. 27531 27532 The '-m64' option sets 'int' to 32 bits and 'long' and pointer 27533 types to 64 bits, and generates code for the x86-64 architecture. 27534 For Darwin only the '-m64' option also turns off the '-fno-pic' and 27535 '-mdynamic-no-pic' options. 27536 27537 The '-mx32' option sets 'int', 'long', and pointer types to 32 27538 bits, and generates code for the x86-64 architecture. 27539 27540 The '-m16' option is the same as '-m32', except for that it outputs 27541 the '.code16gcc' assembly directive at the beginning of the 27542 assembly output so that the binary can run in 16-bit mode. 27543 27544 The '-miamcu' option generates code which conforms to Intel MCU 27545 psABI. It requires the '-m32' option to be turned on. 27546 27547'-mno-red-zone' 27548 Do not use a so-called "red zone" for x86-64 code. The red zone is 27549 mandated by the x86-64 ABI; it is a 128-byte area beyond the 27550 location of the stack pointer that is not modified by signal or 27551 interrupt handlers and therefore can be used for temporary data 27552 without adjusting the stack pointer. The flag '-mno-red-zone' 27553 disables this red zone. 27554 27555'-mcmodel=small' 27556 Generate code for the small code model: the program and its symbols 27557 must be linked in the lower 2 GB of the address space. Pointers 27558 are 64 bits. Programs can be statically or dynamically linked. 27559 This is the default code model. 27560 27561'-mcmodel=kernel' 27562 Generate code for the kernel code model. The kernel runs in the 27563 negative 2 GB of the address space. This model has to be used for 27564 Linux kernel code. 27565 27566'-mcmodel=medium' 27567 Generate code for the medium model: the program is linked in the 27568 lower 2 GB of the address space. Small symbols are also placed 27569 there. Symbols with sizes larger than '-mlarge-data-threshold' are 27570 put into large data or BSS sections and can be located above 2GB. 27571 Programs can be statically or dynamically linked. 27572 27573'-mcmodel=large' 27574 Generate code for the large model. This model makes no assumptions 27575 about addresses and sizes of sections. 27576 27577'-maddress-mode=long' 27578 Generate code for long address mode. This is only supported for 27579 64-bit and x32 environments. It is the default address mode for 27580 64-bit environments. 27581 27582'-maddress-mode=short' 27583 Generate code for short address mode. This is only supported for 27584 32-bit and x32 environments. It is the default address mode for 27585 32-bit and x32 environments. 27586 27587 27588File: gcc.info, Node: x86 Windows Options, Next: Xstormy16 Options, Prev: x86 Options, Up: Submodel Options 27589 275903.19.60 x86 Windows Options 27591--------------------------- 27592 27593These additional options are available for Microsoft Windows targets: 27594 27595'-mconsole' 27596 This option specifies that a console application is to be 27597 generated, by instructing the linker to set the PE header subsystem 27598 type required for console applications. This option is available 27599 for Cygwin and MinGW targets and is enabled by default on those 27600 targets. 27601 27602'-mdll' 27603 This option is available for Cygwin and MinGW targets. It 27604 specifies that a DLL--a dynamic link library--is to be generated, 27605 enabling the selection of the required runtime startup object and 27606 entry point. 27607 27608'-mnop-fun-dllimport' 27609 This option is available for Cygwin and MinGW targets. It 27610 specifies that the 'dllimport' attribute should be ignored. 27611 27612'-mthread' 27613 This option is available for MinGW targets. It specifies that 27614 MinGW-specific thread support is to be used. 27615 27616'-municode' 27617 This option is available for MinGW-w64 targets. It causes the 27618 'UNICODE' preprocessor macro to be predefined, and chooses 27619 Unicode-capable runtime startup code. 27620 27621'-mwin32' 27622 This option is available for Cygwin and MinGW targets. It 27623 specifies that the typical Microsoft Windows predefined macros are 27624 to be set in the pre-processor, but does not influence the choice 27625 of runtime library/startup code. 27626 27627'-mwindows' 27628 This option is available for Cygwin and MinGW targets. It 27629 specifies that a GUI application is to be generated by instructing 27630 the linker to set the PE header subsystem type appropriately. 27631 27632'-fno-set-stack-executable' 27633 This option is available for MinGW targets. It specifies that the 27634 executable flag for the stack used by nested functions isn't set. 27635 This is necessary for binaries running in kernel mode of Microsoft 27636 Windows, as there the User32 API, which is used to set executable 27637 privileges, isn't available. 27638 27639'-fwritable-relocated-rdata' 27640 This option is available for MinGW and Cygwin targets. It 27641 specifies that relocated-data in read-only section is put into the 27642 '.data' section. This is a necessary for older runtimes not 27643 supporting modification of '.rdata' sections for pseudo-relocation. 27644 27645'-mpe-aligned-commons' 27646 This option is available for Cygwin and MinGW targets. It 27647 specifies that the GNU extension to the PE file format that permits 27648 the correct alignment of COMMON variables should be used when 27649 generating code. It is enabled by default if GCC detects that the 27650 target assembler found during configuration supports the feature. 27651 27652 See also under *note x86 Options:: for standard options. 27653 27654 27655File: gcc.info, Node: Xstormy16 Options, Next: Xtensa Options, Prev: x86 Windows Options, Up: Submodel Options 27656 276573.19.61 Xstormy16 Options 27658------------------------- 27659 27660These options are defined for Xstormy16: 27661 27662'-msim' 27663 Choose startup files and linker script suitable for the simulator. 27664 27665 27666File: gcc.info, Node: Xtensa Options, Next: zSeries Options, Prev: Xstormy16 Options, Up: Submodel Options 27667 276683.19.62 Xtensa Options 27669---------------------- 27670 27671These options are supported for Xtensa targets: 27672 27673'-mconst16' 27674'-mno-const16' 27675 Enable or disable use of 'CONST16' instructions for loading 27676 constant values. The 'CONST16' instruction is currently not a 27677 standard option from Tensilica. When enabled, 'CONST16' 27678 instructions are always used in place of the standard 'L32R' 27679 instructions. The use of 'CONST16' is enabled by default only if 27680 the 'L32R' instruction is not available. 27681 27682'-mfused-madd' 27683'-mno-fused-madd' 27684 Enable or disable use of fused multiply/add and multiply/subtract 27685 instructions in the floating-point option. This has no effect if 27686 the floating-point option is not also enabled. Disabling fused 27687 multiply/add and multiply/subtract instructions forces the compiler 27688 to use separate instructions for the multiply and add/subtract 27689 operations. This may be desirable in some cases where strict IEEE 27690 754-compliant results are required: the fused multiply add/subtract 27691 instructions do not round the intermediate result, thereby 27692 producing results with _more_ bits of precision than specified by 27693 the IEEE standard. Disabling fused multiply add/subtract 27694 instructions also ensures that the program output is not sensitive 27695 to the compiler's ability to combine multiply and add/subtract 27696 operations. 27697 27698'-mserialize-volatile' 27699'-mno-serialize-volatile' 27700 When this option is enabled, GCC inserts 'MEMW' instructions before 27701 'volatile' memory references to guarantee sequential consistency. 27702 The default is '-mserialize-volatile'. Use 27703 '-mno-serialize-volatile' to omit the 'MEMW' instructions. 27704 27705'-mforce-no-pic' 27706 For targets, like GNU/Linux, where all user-mode Xtensa code must 27707 be position-independent code (PIC), this option disables PIC for 27708 compiling kernel code. 27709 27710'-mtext-section-literals' 27711'-mno-text-section-literals' 27712 These options control the treatment of literal pools. The default 27713 is '-mno-text-section-literals', which places literals in a 27714 separate section in the output file. This allows the literal pool 27715 to be placed in a data RAM/ROM, and it also allows the linker to 27716 combine literal pools from separate object files to remove 27717 redundant literals and improve code size. With 27718 '-mtext-section-literals', the literals are interspersed in the 27719 text section in order to keep them as close as possible to their 27720 references. This may be necessary for large assembly files. 27721 Literals for each function are placed right before that function. 27722 27723'-mauto-litpools' 27724'-mno-auto-litpools' 27725 These options control the treatment of literal pools. The default 27726 is '-mno-auto-litpools', which places literals in a separate 27727 section in the output file unless '-mtext-section-literals' is 27728 used. With '-mauto-litpools' the literals are interspersed in the 27729 text section by the assembler. Compiler does not produce explicit 27730 '.literal' directives and loads literals into registers with 'MOVI' 27731 instructions instead of 'L32R' to let the assembler do relaxation 27732 and place literals as necessary. This option allows assembler to 27733 create several literal pools per function and assemble very big 27734 functions, which may not be possible with 27735 '-mtext-section-literals'. 27736 27737'-mtarget-align' 27738'-mno-target-align' 27739 When this option is enabled, GCC instructs the assembler to 27740 automatically align instructions to reduce branch penalties at the 27741 expense of some code density. The assembler attempts to widen 27742 density instructions to align branch targets and the instructions 27743 following call instructions. If there are not enough preceding 27744 safe density instructions to align a target, no widening is 27745 performed. The default is '-mtarget-align'. These options do not 27746 affect the treatment of auto-aligned instructions like 'LOOP', 27747 which the assembler always aligns, either by widening density 27748 instructions or by inserting NOP instructions. 27749 27750'-mlongcalls' 27751'-mno-longcalls' 27752 When this option is enabled, GCC instructs the assembler to 27753 translate direct calls to indirect calls unless it can determine 27754 that the target of a direct call is in the range allowed by the 27755 call instruction. This translation typically occurs for calls to 27756 functions in other source files. Specifically, the assembler 27757 translates a direct 'CALL' instruction into an 'L32R' followed by a 27758 'CALLX' instruction. The default is '-mno-longcalls'. This option 27759 should be used in programs where the call target can potentially be 27760 out of range. This option is implemented in the assembler, not the 27761 compiler, so the assembly code generated by GCC still shows direct 27762 call instructions--look at the disassembled object code to see the 27763 actual instructions. Note that the assembler uses an indirect call 27764 for every cross-file call, not just those that really are out of 27765 range. 27766 27767 27768File: gcc.info, Node: zSeries Options, Prev: Xtensa Options, Up: Submodel Options 27769 277703.19.63 zSeries Options 27771----------------------- 27772 27773These are listed under *Note S/390 and zSeries Options::. 27774 27775 27776File: gcc.info, Node: Spec Files, Next: Environment Variables, Prev: Submodel Options, Up: Invoking GCC 27777 277783.20 Specifying Subprocesses and the Switches to Pass to Them 27779============================================================= 27780 27781'gcc' is a driver program. It performs its job by invoking a sequence 27782of other programs to do the work of compiling, assembling and linking. 27783GCC interprets its command-line parameters and uses these to deduce 27784which programs it should invoke, and which command-line options it ought 27785to place on their command lines. This behavior is controlled by "spec 27786strings". In most cases there is one spec string for each program that 27787GCC can invoke, but a few programs have multiple spec strings to control 27788their behavior. The spec strings built into GCC can be overridden by 27789using the '-specs=' command-line switch to specify a spec file. 27790 27791 "Spec files" are plain-text files that are used to construct spec 27792strings. They consist of a sequence of directives separated by blank 27793lines. The type of directive is determined by the first non-whitespace 27794character on the line, which can be one of the following: 27795 27796'%COMMAND' 27797 Issues a COMMAND to the spec file processor. The commands that can 27798 appear here are: 27799 27800 '%include <FILE>' 27801 Search for FILE and insert its text at the current point in 27802 the specs file. 27803 27804 '%include_noerr <FILE>' 27805 Just like '%include', but do not generate an error message if 27806 the include file cannot be found. 27807 27808 '%rename OLD_NAME NEW_NAME' 27809 Rename the spec string OLD_NAME to NEW_NAME. 27810 27811'*[SPEC_NAME]:' 27812 This tells the compiler to create, override or delete the named 27813 spec string. All lines after this directive up to the next 27814 directive or blank line are considered to be the text for the spec 27815 string. If this results in an empty string then the spec is 27816 deleted. (Or, if the spec did not exist, then nothing happens.) 27817 Otherwise, if the spec does not currently exist a new spec is 27818 created. If the spec does exist then its contents are overridden 27819 by the text of this directive, unless the first character of that 27820 text is the '+' character, in which case the text is appended to 27821 the spec. 27822 27823'[SUFFIX]:' 27824 Creates a new '[SUFFIX] spec' pair. All lines after this directive 27825 and up to the next directive or blank line are considered to make 27826 up the spec string for the indicated suffix. When the compiler 27827 encounters an input file with the named suffix, it processes the 27828 spec string in order to work out how to compile that file. For 27829 example: 27830 27831 .ZZ: 27832 z-compile -input %i 27833 27834 This says that any input file whose name ends in '.ZZ' should be 27835 passed to the program 'z-compile', which should be invoked with the 27836 command-line switch '-input' and with the result of performing the 27837 '%i' substitution. (See below.) 27838 27839 As an alternative to providing a spec string, the text following a 27840 suffix directive can be one of the following: 27841 27842 '@LANGUAGE' 27843 This says that the suffix is an alias for a known LANGUAGE. 27844 This is similar to using the '-x' command-line switch to GCC 27845 to specify a language explicitly. For example: 27846 27847 .ZZ: 27848 @c++ 27849 27850 Says that .ZZ files are, in fact, C++ source files. 27851 27852 '#NAME' 27853 This causes an error messages saying: 27854 27855 NAME compiler not installed on this system. 27856 27857 GCC already has an extensive list of suffixes built into it. This 27858 directive adds an entry to the end of the list of suffixes, but 27859 since the list is searched from the end backwards, it is 27860 effectively possible to override earlier entries using this 27861 technique. 27862 27863 GCC has the following spec strings built into it. Spec files can 27864override these strings or create their own. Note that individual 27865targets can also add their own spec strings to this list. 27866 27867 asm Options to pass to the assembler 27868 asm_final Options to pass to the assembler post-processor 27869 cpp Options to pass to the C preprocessor 27870 cc1 Options to pass to the C compiler 27871 cc1plus Options to pass to the C++ compiler 27872 endfile Object files to include at the end of the link 27873 link Options to pass to the linker 27874 lib Libraries to include on the command line to the linker 27875 libgcc Decides which GCC support library to pass to the linker 27876 linker Sets the name of the linker 27877 predefines Defines to be passed to the C preprocessor 27878 signed_char Defines to pass to CPP to say whether char is signed 27879 by default 27880 startfile Object files to include at the start of the link 27881 27882 Here is a small example of a spec file: 27883 27884 %rename lib old_lib 27885 27886 *lib: 27887 --start-group -lgcc -lc -leval1 --end-group %(old_lib) 27888 27889 This example renames the spec called 'lib' to 'old_lib' and then 27890overrides the previous definition of 'lib' with a new one. The new 27891definition adds in some extra command-line options before including the 27892text of the old definition. 27893 27894 "Spec strings" are a list of command-line options to be passed to their 27895corresponding program. In addition, the spec strings can contain 27896'%'-prefixed sequences to substitute variable text or to conditionally 27897insert text into the command line. Using these constructs it is 27898possible to generate quite complex command lines. 27899 27900 Here is a table of all defined '%'-sequences for spec strings. Note 27901that spaces are not generated automatically around the results of 27902expanding these sequences. Therefore you can concatenate them together 27903or combine them with constant text in a single argument. 27904 27905'%%' 27906 Substitute one '%' into the program name or argument. 27907 27908'%i' 27909 Substitute the name of the input file being processed. 27910 27911'%b' 27912 Substitute the basename of the input file being processed. This is 27913 the substring up to (and not including) the last period and not 27914 including the directory. 27915 27916'%B' 27917 This is the same as '%b', but include the file suffix (text after 27918 the last period). 27919 27920'%d' 27921 Marks the argument containing or following the '%d' as a temporary 27922 file name, so that that file is deleted if GCC exits successfully. 27923 Unlike '%g', this contributes no text to the argument. 27924 27925'%gSUFFIX' 27926 Substitute a file name that has suffix SUFFIX and is chosen once 27927 per compilation, and mark the argument in the same way as '%d'. To 27928 reduce exposure to denial-of-service attacks, the file name is now 27929 chosen in a way that is hard to predict even when previously chosen 27930 file names are known. For example, '%g.s ... %g.o ... %g.s' might 27931 turn into 'ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s'. SUFFIX matches the 27932 regexp '[.A-Za-z]*' or the special string '%O', which is treated 27933 exactly as if '%O' had been preprocessed. Previously, '%g' was 27934 simply substituted with a file name chosen once per compilation, 27935 without regard to any appended suffix (which was therefore treated 27936 just like ordinary text), making such attacks more likely to 27937 succeed. 27938 27939'%uSUFFIX' 27940 Like '%g', but generates a new temporary file name each time it 27941 appears instead of once per compilation. 27942 27943'%USUFFIX' 27944 Substitutes the last file name generated with '%uSUFFIX', 27945 generating a new one if there is no such last file name. In the 27946 absence of any '%uSUFFIX', this is just like '%gSUFFIX', except 27947 they don't share the same suffix _space_, so '%g.s ... %U.s ... 27948 %g.s ... %U.s' involves the generation of two distinct file names, 27949 one for each '%g.s' and another for each '%U.s'. Previously, '%U' 27950 was simply substituted with a file name chosen for the previous 27951 '%u', without regard to any appended suffix. 27952 27953'%jSUFFIX' 27954 Substitutes the name of the 'HOST_BIT_BUCKET', if any, and if it is 27955 writable, and if '-save-temps' is not used; otherwise, substitute 27956 the name of a temporary file, just like '%u'. This temporary file 27957 is not meant for communication between processes, but rather as a 27958 junk disposal mechanism. 27959 27960'%|SUFFIX' 27961'%mSUFFIX' 27962 Like '%g', except if '-pipe' is in effect. In that case '%|' 27963 substitutes a single dash and '%m' substitutes nothing at all. 27964 These are the two most common ways to instruct a program that it 27965 should read from standard input or write to standard output. If 27966 you need something more elaborate you can use an '%{pipe:'X'}' 27967 construct: see for example 'gcc/fortran/lang-specs.h'. 27968 27969'%.SUFFIX' 27970 Substitutes .SUFFIX for the suffixes of a matched switch's args 27971 when it is subsequently output with '%*'. SUFFIX is terminated by 27972 the next space or %. 27973 27974'%w' 27975 Marks the argument containing or following the '%w' as the 27976 designated output file of this compilation. This puts the argument 27977 into the sequence of arguments that '%o' substitutes. 27978 27979'%o' 27980 Substitutes the names of all the output files, with spaces 27981 automatically placed around them. You should write spaces around 27982 the '%o' as well or the results are undefined. '%o' is for use in 27983 the specs for running the linker. Input files whose names have no 27984 recognized suffix are not compiled at all, but they are included 27985 among the output files, so they are linked. 27986 27987'%O' 27988 Substitutes the suffix for object files. Note that this is handled 27989 specially when it immediately follows '%g, %u, or %U', because of 27990 the need for those to form complete file names. The handling is 27991 such that '%O' is treated exactly as if it had already been 27992 substituted, except that '%g, %u, and %U' do not currently support 27993 additional SUFFIX characters following '%O' as they do following, 27994 for example, '.o'. 27995 27996'%p' 27997 Substitutes the standard macro predefinitions for the current 27998 target machine. Use this when running 'cpp'. 27999 28000'%P' 28001 Like '%p', but puts '__' before and after the name of each 28002 predefined macro, except for macros that start with '__' or with 28003 '_L', where L is an uppercase letter. This is for ISO C. 28004 28005'%I' 28006 Substitute any of '-iprefix' (made from 'GCC_EXEC_PREFIX'), 28007 '-isysroot' (made from 'TARGET_SYSTEM_ROOT'), '-isystem' (made from 28008 'COMPILER_PATH' and '-B' options) and '-imultilib' as necessary. 28009 28010'%s' 28011 Current argument is the name of a library or startup file of some 28012 sort. Search for that file in a standard list of directories and 28013 substitute the full name found. The current working directory is 28014 included in the list of directories scanned. 28015 28016'%T' 28017 Current argument is the name of a linker script. Search for that 28018 file in the current list of directories to scan for libraries. If 28019 the file is located insert a '--script' option into the command 28020 line followed by the full path name found. If the file is not 28021 found then generate an error message. Note: the current working 28022 directory is not searched. 28023 28024'%eSTR' 28025 Print STR as an error message. STR is terminated by a newline. 28026 Use this when inconsistent options are detected. 28027 28028'%(NAME)' 28029 Substitute the contents of spec string NAME at this point. 28030 28031'%x{OPTION}' 28032 Accumulate an option for '%X'. 28033 28034'%X' 28035 Output the accumulated linker options specified by a '%x' spec 28036 string. 28037 28038'%Y' 28039 Output the accumulated assembler options specified by '-Wa'. 28040 28041'%Z' 28042 Output the accumulated preprocessor options specified by '-Wp'. 28043 28044'%a' 28045 Process the 'asm' spec. This is used to compute the switches to be 28046 passed to the assembler. 28047 28048'%A' 28049 Process the 'asm_final' spec. This is a spec string for passing 28050 switches to an assembler post-processor, if such a program is 28051 needed. 28052 28053'%l' 28054 Process the 'link' spec. This is the spec for computing the 28055 command line passed to the linker. Typically it makes use of the 28056 '%L %G %S %D and %E' sequences. 28057 28058'%D' 28059 Dump out a '-L' option for each directory that GCC believes might 28060 contain startup files. If the target supports multilibs then the 28061 current multilib directory is prepended to each of these paths. 28062 28063'%L' 28064 Process the 'lib' spec. This is a spec string for deciding which 28065 libraries are included on the command line to the linker. 28066 28067'%G' 28068 Process the 'libgcc' spec. This is a spec string for deciding 28069 which GCC support library is included on the command line to the 28070 linker. 28071 28072'%S' 28073 Process the 'startfile' spec. This is a spec for deciding which 28074 object files are the first ones passed to the linker. Typically 28075 this might be a file named 'crt0.o'. 28076 28077'%E' 28078 Process the 'endfile' spec. This is a spec string that specifies 28079 the last object files that are passed to the linker. 28080 28081'%C' 28082 Process the 'cpp' spec. This is used to construct the arguments to 28083 be passed to the C preprocessor. 28084 28085'%1' 28086 Process the 'cc1' spec. This is used to construct the options to 28087 be passed to the actual C compiler ('cc1'). 28088 28089'%2' 28090 Process the 'cc1plus' spec. This is used to construct the options 28091 to be passed to the actual C++ compiler ('cc1plus'). 28092 28093'%*' 28094 Substitute the variable part of a matched option. See below. Note 28095 that each comma in the substituted string is replaced by a single 28096 space. 28097 28098'%<S' 28099 Remove all occurrences of '-S' from the command line. Note--this 28100 command is position dependent. '%' commands in the spec string 28101 before this one see '-S', '%' commands in the spec string after 28102 this one do not. 28103 28104'%:FUNCTION(ARGS)' 28105 Call the named function FUNCTION, passing it ARGS. ARGS is first 28106 processed as a nested spec string, then split into an argument 28107 vector in the usual fashion. The function returns a string which 28108 is processed as if it had appeared literally as part of the current 28109 spec. 28110 28111 The following built-in spec functions are provided: 28112 28113 'getenv' 28114 The 'getenv' spec function takes two arguments: an environment 28115 variable name and a string. If the environment variable is 28116 not defined, a fatal error is issued. Otherwise, the return 28117 value is the value of the environment variable concatenated 28118 with the string. For example, if 'TOPDIR' is defined as 28119 '/path/to/top', then: 28120 28121 %:getenv(TOPDIR /include) 28122 28123 expands to '/path/to/top/include'. 28124 28125 'if-exists' 28126 The 'if-exists' spec function takes one argument, an absolute 28127 pathname to a file. If the file exists, 'if-exists' returns 28128 the pathname. Here is a small example of its usage: 28129 28130 *startfile: 28131 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s 28132 28133 'if-exists-else' 28134 The 'if-exists-else' spec function is similar to the 28135 'if-exists' spec function, except that it takes two arguments. 28136 The first argument is an absolute pathname to a file. If the 28137 file exists, 'if-exists-else' returns the pathname. If it 28138 does not exist, it returns the second argument. This way, 28139 'if-exists-else' can be used to select one file or another, 28140 based on the existence of the first. Here is a small example 28141 of its usage: 28142 28143 *startfile: 28144 crt0%O%s %:if-exists(crti%O%s) \ 28145 %:if-exists-else(crtbeginT%O%s crtbegin%O%s) 28146 28147 'replace-outfile' 28148 The 'replace-outfile' spec function takes two arguments. It 28149 looks for the first argument in the outfiles array and 28150 replaces it with the second argument. Here is a small example 28151 of its usage: 28152 28153 %{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)} 28154 28155 'remove-outfile' 28156 The 'remove-outfile' spec function takes one argument. It 28157 looks for the first argument in the outfiles array and removes 28158 it. Here is a small example its usage: 28159 28160 %:remove-outfile(-lm) 28161 28162 'pass-through-libs' 28163 The 'pass-through-libs' spec function takes any number of 28164 arguments. It finds any '-l' options and any non-options 28165 ending in '.a' (which it assumes are the names of linker input 28166 library archive files) and returns a result containing all the 28167 found arguments each prepended by '-plugin-opt=-pass-through=' 28168 and joined by spaces. This list is intended to be passed to 28169 the LTO linker plugin. 28170 28171 %:pass-through-libs(%G %L %G) 28172 28173 'print-asm-header' 28174 The 'print-asm-header' function takes no arguments and simply 28175 prints a banner like: 28176 28177 Assembler options 28178 ================= 28179 28180 Use "-Wa,OPTION" to pass "OPTION" to the assembler. 28181 28182 It is used to separate compiler options from assembler options 28183 in the '--target-help' output. 28184 28185'%{S}' 28186 Substitutes the '-S' switch, if that switch is given to GCC. If 28187 that switch is not specified, this substitutes nothing. Note that 28188 the leading dash is omitted when specifying this option, and it is 28189 automatically inserted if the substitution is performed. Thus the 28190 spec string '%{foo}' matches the command-line option '-foo' and 28191 outputs the command-line option '-foo'. 28192 28193'%W{S}' 28194 Like %{'S'} but mark last argument supplied within as a file to be 28195 deleted on failure. 28196 28197'%{S*}' 28198 Substitutes all the switches specified to GCC whose names start 28199 with '-S', but which also take an argument. This is used for 28200 switches like '-o', '-D', '-I', etc. GCC considers '-o foo' as 28201 being one switch whose name starts with 'o'. %{o*} substitutes 28202 this text, including the space. Thus two arguments are generated. 28203 28204'%{S*&T*}' 28205 Like %{'S'*}, but preserve order of 'S' and 'T' options (the order 28206 of 'S' and 'T' in the spec is not significant). There can be any 28207 number of ampersand-separated variables; for each the wild card is 28208 optional. Useful for CPP as '%{D*&U*&A*}'. 28209 28210'%{S:X}' 28211 Substitutes 'X', if the '-S' switch is given to GCC. 28212 28213'%{!S:X}' 28214 Substitutes 'X', if the '-S' switch is _not_ given to GCC. 28215 28216'%{S*:X}' 28217 Substitutes 'X' if one or more switches whose names start with '-S' 28218 are specified to GCC. Normally 'X' is substituted only once, no 28219 matter how many such switches appeared. However, if '%*' appears 28220 somewhere in 'X', then 'X' is substituted once for each matching 28221 switch, with the '%*' replaced by the part of that switch matching 28222 the '*'. 28223 28224 If '%*' appears as the last part of a spec sequence then a space is 28225 added after the end of the last substitution. If there is more 28226 text in the sequence, however, then a space is not generated. This 28227 allows the '%*' substitution to be used as part of a larger string. 28228 For example, a spec string like this: 28229 28230 %{mcu=*:--script=%*/memory.ld} 28231 28232 when matching an option like '-mcu=newchip' produces: 28233 28234 --script=newchip/memory.ld 28235 28236'%{.S:X}' 28237 Substitutes 'X', if processing a file with suffix 'S'. 28238 28239'%{!.S:X}' 28240 Substitutes 'X', if _not_ processing a file with suffix 'S'. 28241 28242'%{,S:X}' 28243 Substitutes 'X', if processing a file for language 'S'. 28244 28245'%{!,S:X}' 28246 Substitutes 'X', if not processing a file for language 'S'. 28247 28248'%{S|P:X}' 28249 Substitutes 'X' if either '-S' or '-P' is given to GCC. This may 28250 be combined with '!', '.', ',', and '*' sequences as well, although 28251 they have a stronger binding than the '|'. If '%*' appears in 'X', 28252 all of the alternatives must be starred, and only the first 28253 matching alternative is substituted. 28254 28255 For example, a spec string like this: 28256 28257 %{.c:-foo} %{!.c:-bar} %{.c|d:-baz} %{!.c|d:-boggle} 28258 28259 outputs the following command-line options from the following input 28260 command-line options: 28261 28262 fred.c -foo -baz 28263 jim.d -bar -boggle 28264 -d fred.c -foo -baz -boggle 28265 -d jim.d -bar -baz -boggle 28266 28267'%{S:X; T:Y; :D}' 28268 28269 If 'S' is given to GCC, substitutes 'X'; else if 'T' is given to 28270 GCC, substitutes 'Y'; else substitutes 'D'. There can be as many 28271 clauses as you need. This may be combined with '.', ',', '!', '|', 28272 and '*' as needed. 28273 28274 The switch matching text 'S' in a '%{S}', '%{S:X}' or similar construct 28275can use a backslash to ignore the special meaning of the character 28276following it, thus allowing literal matching of a character that is 28277otherwise specially treated. For example, '%{std=iso9899\:1999:X}' 28278substitutes 'X' if the '-std=iso9899:1999' option is given. 28279 28280 The conditional text 'X' in a '%{S:X}' or similar construct may contain 28281other nested '%' constructs or spaces, or even newlines. They are 28282processed as usual, as described above. Trailing white space in 'X' is 28283ignored. White space may also appear anywhere on the left side of the 28284colon in these constructs, except between '.' or '*' and the 28285corresponding word. 28286 28287 The '-O', '-f', '-m', and '-W' switches are handled specifically in 28288these constructs. If another value of '-O' or the negated form of a 28289'-f', '-m', or '-W' switch is found later in the command line, the 28290earlier switch value is ignored, except with {'S'*} where 'S' is just 28291one letter, which passes all matching options. 28292 28293 The character '|' at the beginning of the predicate text is used to 28294indicate that a command should be piped to the following command, but 28295only if '-pipe' is specified. 28296 28297 It is built into GCC which switches take arguments and which do not. 28298(You might think it would be useful to generalize this to allow each 28299compiler's spec to say which switches take arguments. But this cannot 28300be done in a consistent fashion. GCC cannot even decide which input 28301files have been specified without knowing which switches take arguments, 28302and it must know which input files to compile in order to tell which 28303compilers to run). 28304 28305 GCC also knows implicitly that arguments starting in '-l' are to be 28306treated as compiler output files, and passed to the linker in their 28307proper position among the other output files. 28308 28309 28310File: gcc.info, Node: Environment Variables, Next: Precompiled Headers, Prev: Spec Files, Up: Invoking GCC 28311 283123.21 Environment Variables Affecting GCC 28313======================================== 28314 28315This section describes several environment variables that affect how GCC 28316operates. Some of them work by specifying directories or prefixes to 28317use when searching for various kinds of files. Some are used to specify 28318other aspects of the compilation environment. 28319 28320 Note that you can also specify places to search using options such as 28321'-B', '-I' and '-L' (*note Directory Options::). These take precedence 28322over places specified using environment variables, which in turn take 28323precedence over those specified by the configuration of GCC. *Note 28324Controlling the Compilation Driver 'gcc': (gccint)Driver. 28325 28326'LANG' 28327'LC_CTYPE' 28328'LC_MESSAGES' 28329'LC_ALL' 28330 These environment variables control the way that GCC uses 28331 localization information which allows GCC to work with different 28332 national conventions. GCC inspects the locale categories 28333 'LC_CTYPE' and 'LC_MESSAGES' if it has been configured to do so. 28334 These locale categories can be set to any value supported by your 28335 installation. A typical value is 'en_GB.UTF-8' for English in the 28336 United Kingdom encoded in UTF-8. 28337 28338 The 'LC_CTYPE' environment variable specifies character 28339 classification. GCC uses it to determine the character boundaries 28340 in a string; this is needed for some multibyte encodings that 28341 contain quote and escape characters that are otherwise interpreted 28342 as a string end or escape. 28343 28344 The 'LC_MESSAGES' environment variable specifies the language to 28345 use in diagnostic messages. 28346 28347 If the 'LC_ALL' environment variable is set, it overrides the value 28348 of 'LC_CTYPE' and 'LC_MESSAGES'; otherwise, 'LC_CTYPE' and 28349 'LC_MESSAGES' default to the value of the 'LANG' environment 28350 variable. If none of these variables are set, GCC defaults to 28351 traditional C English behavior. 28352 28353'TMPDIR' 28354 If 'TMPDIR' is set, it specifies the directory to use for temporary 28355 files. GCC uses temporary files to hold the output of one stage of 28356 compilation which is to be used as input to the next stage: for 28357 example, the output of the preprocessor, which is the input to the 28358 compiler proper. 28359 28360'GCC_COMPARE_DEBUG' 28361 Setting 'GCC_COMPARE_DEBUG' is nearly equivalent to passing 28362 '-fcompare-debug' to the compiler driver. See the documentation of 28363 this option for more details. 28364 28365'GCC_EXEC_PREFIX' 28366 If 'GCC_EXEC_PREFIX' is set, it specifies a prefix to use in the 28367 names of the subprograms executed by the compiler. No slash is 28368 added when this prefix is combined with the name of a subprogram, 28369 but you can specify a prefix that ends with a slash if you wish. 28370 28371 If 'GCC_EXEC_PREFIX' is not set, GCC attempts to figure out an 28372 appropriate prefix to use based on the pathname it is invoked with. 28373 28374 If GCC cannot find the subprogram using the specified prefix, it 28375 tries looking in the usual places for the subprogram. 28376 28377 The default value of 'GCC_EXEC_PREFIX' is 'PREFIX/lib/gcc/' where 28378 PREFIX is the prefix to the installed compiler. In many cases 28379 PREFIX is the value of 'prefix' when you ran the 'configure' 28380 script. 28381 28382 Other prefixes specified with '-B' take precedence over this 28383 prefix. 28384 28385 This prefix is also used for finding files such as 'crt0.o' that 28386 are used for linking. 28387 28388 In addition, the prefix is used in an unusual way in finding the 28389 directories to search for header files. For each of the standard 28390 directories whose name normally begins with '/usr/local/lib/gcc' 28391 (more precisely, with the value of 'GCC_INCLUDE_DIR'), GCC tries 28392 replacing that beginning with the specified prefix to produce an 28393 alternate directory name. Thus, with '-Bfoo/', GCC searches 28394 'foo/bar' just before it searches the standard directory 28395 '/usr/local/lib/bar'. If a standard directory begins with the 28396 configured PREFIX then the value of PREFIX is replaced by 28397 'GCC_EXEC_PREFIX' when looking for header files. 28398 28399'COMPILER_PATH' 28400 The value of 'COMPILER_PATH' is a colon-separated list of 28401 directories, much like 'PATH'. GCC tries the directories thus 28402 specified when searching for subprograms, if it cannot find the 28403 subprograms using 'GCC_EXEC_PREFIX'. 28404 28405'LIBRARY_PATH' 28406 The value of 'LIBRARY_PATH' is a colon-separated list of 28407 directories, much like 'PATH'. When configured as a native 28408 compiler, GCC tries the directories thus specified when searching 28409 for special linker files, if it cannot find them using 28410 'GCC_EXEC_PREFIX'. Linking using GCC also uses these directories 28411 when searching for ordinary libraries for the '-l' option (but 28412 directories specified with '-L' come first). 28413 28414'LANG' 28415 This variable is used to pass locale information to the compiler. 28416 One way in which this information is used is to determine the 28417 character set to be used when character literals, string literals 28418 and comments are parsed in C and C++. When the compiler is 28419 configured to allow multibyte characters, the following values for 28420 'LANG' are recognized: 28421 28422 'C-JIS' 28423 Recognize JIS characters. 28424 'C-SJIS' 28425 Recognize SJIS characters. 28426 'C-EUCJP' 28427 Recognize EUCJP characters. 28428 28429 If 'LANG' is not defined, or if it has some other value, then the 28430 compiler uses 'mblen' and 'mbtowc' as defined by the default locale 28431 to recognize and translate multibyte characters. 28432 28433Some additional environment variables affect the behavior of the 28434preprocessor. 28435 28436'CPATH' 28437'C_INCLUDE_PATH' 28438'CPLUS_INCLUDE_PATH' 28439'OBJC_INCLUDE_PATH' 28440 Each variable's value is a list of directories separated by a 28441 special character, much like 'PATH', in which to look for header 28442 files. The special character, 'PATH_SEPARATOR', is 28443 target-dependent and determined at GCC build time. For Microsoft 28444 Windows-based targets it is a semicolon, and for almost all other 28445 targets it is a colon. 28446 28447 'CPATH' specifies a list of directories to be searched as if 28448 specified with '-I', but after any paths given with '-I' options on 28449 the command line. This environment variable is used regardless of 28450 which language is being preprocessed. 28451 28452 The remaining environment variables apply only when preprocessing 28453 the particular language indicated. Each specifies a list of 28454 directories to be searched as if specified with '-isystem', but 28455 after any paths given with '-isystem' options on the command line. 28456 28457 In all these variables, an empty element instructs the compiler to 28458 search its current working directory. Empty elements can appear at 28459 the beginning or end of a path. For instance, if the value of 28460 'CPATH' is ':/special/include', that has the same effect as 28461 '-I. -I/special/include'. 28462 28463'DEPENDENCIES_OUTPUT' 28464 If this variable is set, its value specifies how to output 28465 dependencies for Make based on the non-system header files 28466 processed by the compiler. System header files are ignored in the 28467 dependency output. 28468 28469 The value of 'DEPENDENCIES_OUTPUT' can be just a file name, in 28470 which case the Make rules are written to that file, guessing the 28471 target name from the source file name. Or the value can have the 28472 form 'FILE TARGET', in which case the rules are written to file 28473 FILE using TARGET as the target name. 28474 28475 In other words, this environment variable is equivalent to 28476 combining the options '-MM' and '-MF' (*note Preprocessor 28477 Options::), with an optional '-MT' switch too. 28478 28479'SUNPRO_DEPENDENCIES' 28480 This variable is the same as 'DEPENDENCIES_OUTPUT' (see above), 28481 except that system header files are not ignored, so it implies '-M' 28482 rather than '-MM'. However, the dependence on the main input file 28483 is omitted. *Note Preprocessor Options::. 28484 28485'SOURCE_DATE_EPOCH' 28486 If this variable is set, its value specifies a UNIX timestamp to be 28487 used in replacement of the current date and time in the '__DATE__' 28488 and '__TIME__' macros, so that the embedded timestamps become 28489 reproducible. 28490 28491 The value of 'SOURCE_DATE_EPOCH' must be a UNIX timestamp, defined 28492 as the number of seconds (excluding leap seconds) since 01 Jan 1970 28493 00:00:00 represented in ASCII; identical to the output of ''date 28494 +%s'' on GNU/Linux and other systems that support the '%s' 28495 extension in the 'date' command. 28496 28497 The value should be a known timestamp such as the last modification 28498 time of the source or package and it should be set by the build 28499 process. 28500 28501 28502File: gcc.info, Node: Precompiled Headers, Prev: Environment Variables, Up: Invoking GCC 28503 285043.22 Using Precompiled Headers 28505============================== 28506 28507Often large projects have many header files that are included in every 28508source file. The time the compiler takes to process these header files 28509over and over again can account for nearly all of the time required to 28510build the project. To make builds faster, GCC allows you to 28511"precompile" a header file. 28512 28513 To create a precompiled header file, simply compile it as you would any 28514other file, if necessary using the '-x' option to make the driver treat 28515it as a C or C++ header file. You may want to use a tool like 'make' to 28516keep the precompiled header up-to-date when the headers it contains 28517change. 28518 28519 A precompiled header file is searched for when '#include' is seen in 28520the compilation. As it searches for the included file (*note Search 28521Path: (cpp)Search Path.) the compiler looks for a precompiled header in 28522each directory just before it looks for the include file in that 28523directory. The name searched for is the name specified in the 28524'#include' with '.gch' appended. If the precompiled header file cannot 28525be used, it is ignored. 28526 28527 For instance, if you have '#include "all.h"', and you have 'all.h.gch' 28528in the same directory as 'all.h', then the precompiled header file is 28529used if possible, and the original header is used otherwise. 28530 28531 Alternatively, you might decide to put the precompiled header file in a 28532directory and use '-I' to ensure that directory is searched before (or 28533instead of) the directory containing the original header. Then, if you 28534want to check that the precompiled header file is always used, you can 28535put a file of the same name as the original header in this directory 28536containing an '#error' command. 28537 28538 This also works with '-include'. So yet another way to use precompiled 28539headers, good for projects not designed with precompiled header files in 28540mind, is to simply take most of the header files used by a project, 28541include them from another header file, precompile that header file, and 28542'-include' the precompiled header. If the header files have guards 28543against multiple inclusion, they are skipped because they've already 28544been included (in the precompiled header). 28545 28546 If you need to precompile the same header file for different languages, 28547targets, or compiler options, you can instead make a _directory_ named 28548like 'all.h.gch', and put each precompiled header in the directory, 28549perhaps using '-o'. It doesn't matter what you call the files in the 28550directory; every precompiled header in the directory is considered. The 28551first precompiled header encountered in the directory that is valid for 28552this compilation is used; they're searched in no particular order. 28553 28554 There are many other possibilities, limited only by your imagination, 28555good sense, and the constraints of your build system. 28556 28557 A precompiled header file can be used only when these conditions apply: 28558 28559 * Only one precompiled header can be used in a particular 28560 compilation. 28561 28562 * A precompiled header cannot be used once the first C token is seen. 28563 You can have preprocessor directives before a precompiled header; 28564 you cannot include a precompiled header from inside another header. 28565 28566 * The precompiled header file must be produced for the same language 28567 as the current compilation. You cannot use a C precompiled header 28568 for a C++ compilation. 28569 28570 * The precompiled header file must have been produced by the same 28571 compiler binary as the current compilation is using. 28572 28573 * Any macros defined before the precompiled header is included must 28574 either be defined in the same way as when the precompiled header 28575 was generated, or must not affect the precompiled header, which 28576 usually means that they don't appear in the precompiled header at 28577 all. 28578 28579 The '-D' option is one way to define a macro before a precompiled 28580 header is included; using a '#define' can also do it. There are 28581 also some options that define macros implicitly, like '-O' and 28582 '-Wdeprecated'; the same rule applies to macros defined this way. 28583 28584 * If debugging information is output when using the precompiled 28585 header, using '-g' or similar, the same kind of debugging 28586 information must have been output when building the precompiled 28587 header. However, a precompiled header built using '-g' can be used 28588 in a compilation when no debugging information is being output. 28589 28590 * The same '-m' options must generally be used when building and 28591 using the precompiled header. *Note Submodel Options::, for any 28592 cases where this rule is relaxed. 28593 28594 * Each of the following options must be the same when building and 28595 using the precompiled header: 28596 28597 -fexceptions 28598 28599 * Some other command-line options starting with '-f', '-p', or '-O' 28600 must be defined in the same way as when the precompiled header was 28601 generated. At present, it's not clear which options are safe to 28602 change and which are not; the safest choice is to use exactly the 28603 same options when generating and using the precompiled header. The 28604 following are known to be safe: 28605 28606 -fmessage-length= -fpreprocessed -fsched-interblock 28607 -fsched-spec -fsched-spec-load -fsched-spec-load-dangerous 28608 -fsched-verbose=NUMBER -fschedule-insns -fvisibility= 28609 -pedantic-errors 28610 28611 * Address space layout randomization (ASLR) can lead to not binary 28612 identical PCH files. If you rely on stable PCH file contents 28613 disable ASLR when generating PCH files. 28614 28615 For all of these except the last, the compiler automatically ignores 28616the precompiled header if the conditions aren't met. If you find an 28617option combination that doesn't work and doesn't cause the precompiled 28618header to be ignored, please consider filing a bug report, see *note 28619Bugs::. 28620 28621 If you do use differing options when generating and using the 28622precompiled header, the actual behavior is a mixture of the behavior for 28623the options. For instance, if you use '-g' to generate the precompiled 28624header but not when using it, you may or may not get debugging 28625information for routines in the precompiled header. 28626 28627 28628File: gcc.info, Node: C Implementation, Next: C++ Implementation, Prev: Invoking GCC, Up: Top 28629 286304 C Implementation-Defined Behavior 28631*********************************** 28632 28633A conforming implementation of ISO C is required to document its choice 28634of behavior in each of the areas that are designated "implementation 28635defined". The following lists all such areas, along with the section 28636numbers from the ISO/IEC 9899:1990, ISO/IEC 9899:1999 and ISO/IEC 286379899:2011 standards. Some areas are only implementation-defined in one 28638version of the standard. 28639 28640 Some choices depend on the externally determined ABI for the platform 28641(including standard character encodings) which GCC follows; these are 28642listed as "determined by ABI" below. *Note Binary Compatibility: 28643Compatibility, and <http://gcc.gnu.org/readings.html>. Some choices are 28644documented in the preprocessor manual. *Note Implementation-defined 28645behavior: (cpp)Implementation-defined behavior. Some choices are made 28646by the library and operating system (or other environment when compiling 28647for a freestanding environment); refer to their documentation for 28648details. 28649 28650* Menu: 28651 28652* Translation implementation:: 28653* Environment implementation:: 28654* Identifiers implementation:: 28655* Characters implementation:: 28656* Integers implementation:: 28657* Floating point implementation:: 28658* Arrays and pointers implementation:: 28659* Hints implementation:: 28660* Structures unions enumerations and bit-fields implementation:: 28661* Qualifiers implementation:: 28662* Declarators implementation:: 28663* Statements implementation:: 28664* Preprocessing directives implementation:: 28665* Library functions implementation:: 28666* Architecture implementation:: 28667* Locale-specific behavior implementation:: 28668 28669 28670File: gcc.info, Node: Translation implementation, Next: Environment implementation, Up: C Implementation 28671 286724.1 Translation 28673=============== 28674 28675 * 'How a diagnostic is identified (C90 3.7, C99 and C11 3.10, C90, 28676 C99 and C11 5.1.1.3).' 28677 28678 Diagnostics consist of all the output sent to stderr by GCC. 28679 28680 * 'Whether each nonempty sequence of white-space characters other 28681 than new-line is retained or replaced by one space character in 28682 translation phase 3 (C90, C99 and C11 5.1.1.2).' 28683 28684 *Note Implementation-defined behavior: (cpp)Implementation-defined 28685 behavior. 28686 28687 28688File: gcc.info, Node: Environment implementation, Next: Identifiers implementation, Prev: Translation implementation, Up: C Implementation 28689 286904.2 Environment 28691=============== 28692 28693The behavior of most of these points are dependent on the implementation 28694of the C library, and are not defined by GCC itself. 28695 28696 * 'The mapping between physical source file multibyte characters and 28697 the source character set in translation phase 1 (C90, C99 and C11 28698 5.1.1.2).' 28699 28700 *Note Implementation-defined behavior: (cpp)Implementation-defined 28701 behavior. 28702 28703 28704File: gcc.info, Node: Identifiers implementation, Next: Characters implementation, Prev: Environment implementation, Up: C Implementation 28705 287064.3 Identifiers 28707=============== 28708 28709 * 'Which additional multibyte characters may appear in identifiers 28710 and their correspondence to universal character names (C99 and C11 28711 6.4.2).' 28712 28713 *Note Implementation-defined behavior: (cpp)Implementation-defined 28714 behavior. 28715 28716 * 'The number of significant initial characters in an identifier (C90 28717 6.1.2, C90, C99 and C11 5.2.4.1, C99 and C11 6.4.2).' 28718 28719 For internal names, all characters are significant. For external 28720 names, the number of significant characters are defined by the 28721 linker; for almost all targets, all characters are significant. 28722 28723 * 'Whether case distinctions are significant in an identifier with 28724 external linkage (C90 6.1.2).' 28725 28726 This is a property of the linker. C99 and C11 require that case 28727 distinctions are always significant in identifiers with external 28728 linkage and systems without this property are not supported by GCC. 28729 28730 28731File: gcc.info, Node: Characters implementation, Next: Integers implementation, Prev: Identifiers implementation, Up: C Implementation 28732 287334.4 Characters 28734============== 28735 28736 * 'The number of bits in a byte (C90 3.4, C99 and C11 3.6).' 28737 28738 Determined by ABI. 28739 28740 * 'The values of the members of the execution character set (C90, C99 28741 and C11 5.2.1).' 28742 28743 Determined by ABI. 28744 28745 * 'The unique value of the member of the execution character set 28746 produced for each of the standard alphabetic escape sequences (C90, 28747 C99 and C11 5.2.2).' 28748 28749 Determined by ABI. 28750 28751 * 'The value of a 'char' object into which has been stored any 28752 character other than a member of the basic execution character set 28753 (C90 6.1.2.5, C99 and C11 6.2.5).' 28754 28755 Determined by ABI. 28756 28757 * 'Which of 'signed char' or 'unsigned char' has the same range, 28758 representation, and behavior as "plain" 'char' (C90 6.1.2.5, C90 28759 6.2.1.1, C99 and C11 6.2.5, C99 and C11 6.3.1.1).' 28760 28761 Determined by ABI. The options '-funsigned-char' and 28762 '-fsigned-char' change the default. *Note Options Controlling C 28763 Dialect: C Dialect Options. 28764 28765 * 'The mapping of members of the source character set (in character 28766 constants and string literals) to members of the execution 28767 character set (C90 6.1.3.4, C99 and C11 6.4.4.4, C90, C99 and C11 28768 5.1.1.2).' 28769 28770 Determined by ABI. 28771 28772 * 'The value of an integer character constant containing more than 28773 one character or containing a character or escape sequence that 28774 does not map to a single-byte execution character (C90 6.1.3.4, C99 28775 and C11 6.4.4.4).' 28776 28777 *Note Implementation-defined behavior: (cpp)Implementation-defined 28778 behavior. 28779 28780 * 'The value of a wide character constant containing more than one 28781 multibyte character or a single multibyte character that maps to 28782 multiple members of the extended execution character set, or 28783 containing a multibyte character or escape sequence not represented 28784 in the extended execution character set (C90 6.1.3.4, C99 and C11 28785 6.4.4.4).' 28786 28787 *Note Implementation-defined behavior: (cpp)Implementation-defined 28788 behavior. 28789 28790 * 'The current locale used to convert a wide character constant 28791 consisting of a single multibyte character that maps to a member of 28792 the extended execution character set into a corresponding wide 28793 character code (C90 6.1.3.4, C99 and C11 6.4.4.4).' 28794 28795 *Note Implementation-defined behavior: (cpp)Implementation-defined 28796 behavior. 28797 28798 * 'Whether differently-prefixed wide string literal tokens can be 28799 concatenated and, if so, the treatment of the resulting multibyte 28800 character sequence (C11 6.4.5).' 28801 28802 Such tokens may not be concatenated. 28803 28804 * 'The current locale used to convert a wide string literal into 28805 corresponding wide character codes (C90 6.1.4, C99 and C11 6.4.5).' 28806 28807 *Note Implementation-defined behavior: (cpp)Implementation-defined 28808 behavior. 28809 28810 * 'The value of a string literal containing a multibyte character or 28811 escape sequence not represented in the execution character set (C90 28812 6.1.4, C99 and C11 6.4.5).' 28813 28814 *Note Implementation-defined behavior: (cpp)Implementation-defined 28815 behavior. 28816 28817 * 'The encoding of any of 'wchar_t', 'char16_t', and 'char32_t' where 28818 the corresponding standard encoding macro ('__STDC_ISO_10646__', 28819 '__STDC_UTF_16__', or '__STDC_UTF_32__') is not defined (C11 28820 6.10.8.2).' 28821 28822 *Note Implementation-defined behavior: (cpp)Implementation-defined 28823 behavior. 'char16_t' and 'char32_t' literals are always encoded in 28824 UTF-16 and UTF-32 respectively. 28825 28826 28827File: gcc.info, Node: Integers implementation, Next: Floating point implementation, Prev: Characters implementation, Up: C Implementation 28828 288294.5 Integers 28830============ 28831 28832 * 'Any extended integer types that exist in the implementation (C99 28833 and C11 6.2.5).' 28834 28835 GCC does not support any extended integer types. 28836 28837 * 'Whether signed integer types are represented using sign and 28838 magnitude, two's complement, or one's complement, and whether the 28839 extraordinary value is a trap representation or an ordinary value 28840 (C99 and C11 6.2.6.2).' 28841 28842 GCC supports only two's complement integer types, and all bit 28843 patterns are ordinary values. 28844 28845 * 'The rank of any extended integer type relative to another extended 28846 integer type with the same precision (C99 and C11 6.3.1.1).' 28847 28848 GCC does not support any extended integer types. 28849 28850 * 'The result of, or the signal raised by, converting an integer to a 28851 signed integer type when the value cannot be represented in an 28852 object of that type (C90 6.2.1.2, C99 and C11 6.3.1.3).' 28853 28854 For conversion to a type of width N, the value is reduced modulo 28855 2^N to be within range of the type; no signal is raised. 28856 28857 * 'The results of some bitwise operations on signed integers (C90 28858 6.3, C99 and C11 6.5).' 28859 28860 Bitwise operators act on the representation of the value including 28861 both the sign and value bits, where the sign bit is considered 28862 immediately above the highest-value value bit. Signed '>>' acts on 28863 negative numbers by sign extension. 28864 28865 As an extension to the C language, GCC does not use the latitude 28866 given in C99 and C11 only to treat certain aspects of signed '<<' 28867 as undefined. However, '-fsanitize=shift' (and 28868 '-fsanitize=undefined') will diagnose such cases. They are also 28869 diagnosed where constant expressions are required. 28870 28871 * 'The sign of the remainder on integer division (C90 6.3.5).' 28872 28873 GCC always follows the C99 and C11 requirement that the result of 28874 division is truncated towards zero. 28875 28876 28877File: gcc.info, Node: Floating point implementation, Next: Arrays and pointers implementation, Prev: Integers implementation, Up: C Implementation 28878 288794.6 Floating Point 28880================== 28881 28882 * 'The accuracy of the floating-point operations and of the library 28883 functions in '<math.h>' and '<complex.h>' that return 28884 floating-point results (C90, C99 and C11 5.2.4.2.2).' 28885 28886 The accuracy is unknown. 28887 28888 * 'The rounding behaviors characterized by non-standard values of 28889 'FLT_ROUNDS' (C90, C99 and C11 5.2.4.2.2).' 28890 28891 GCC does not use such values. 28892 28893 * 'The evaluation methods characterized by non-standard negative 28894 values of 'FLT_EVAL_METHOD' (C99 and C11 5.2.4.2.2).' 28895 28896 GCC does not use such values. 28897 28898 * 'The direction of rounding when an integer is converted to a 28899 floating-point number that cannot exactly represent the original 28900 value (C90 6.2.1.3, C99 and C11 6.3.1.4).' 28901 28902 C99 Annex F is followed. 28903 28904 * 'The direction of rounding when a floating-point number is 28905 converted to a narrower floating-point number (C90 6.2.1.4, C99 and 28906 C11 6.3.1.5).' 28907 28908 C99 Annex F is followed. 28909 28910 * 'How the nearest representable value or the larger or smaller 28911 representable value immediately adjacent to the nearest 28912 representable value is chosen for certain floating constants (C90 28913 6.1.3.1, C99 and C11 6.4.4.2).' 28914 28915 C99 Annex F is followed. 28916 28917 * 'Whether and how floating expressions are contracted when not 28918 disallowed by the 'FP_CONTRACT' pragma (C99 and C11 6.5).' 28919 28920 Expressions are currently only contracted if '-ffp-contract=fast', 28921 '-funsafe-math-optimizations' or '-ffast-math' are used. This is 28922 subject to change. 28923 28924 * 'The default state for the 'FENV_ACCESS' pragma (C99 and C11 28925 7.6.1).' 28926 28927 This pragma is not implemented, but the default is to "off" unless 28928 '-frounding-math' is used in which case it is "on". 28929 28930 * 'Additional floating-point exceptions, rounding modes, 28931 environments, and classifications, and their macro names (C99 and 28932 C11 7.6, C99 and C11 7.12).' 28933 28934 This is dependent on the implementation of the C library, and is 28935 not defined by GCC itself. 28936 28937 * 'The default state for the 'FP_CONTRACT' pragma (C99 and C11 28938 7.12.2).' 28939 28940 This pragma is not implemented. Expressions are currently only 28941 contracted if '-ffp-contract=fast', '-funsafe-math-optimizations' 28942 or '-ffast-math' are used. This is subject to change. 28943 28944 * 'Whether the "inexact" floating-point exception can be raised when 28945 the rounded result actually does equal the mathematical result in 28946 an IEC 60559 conformant implementation (C99 F.9).' 28947 28948 This is dependent on the implementation of the C library, and is 28949 not defined by GCC itself. 28950 28951 * 'Whether the "underflow" (and "inexact") floating-point exception 28952 can be raised when a result is tiny but not inexact in an IEC 60559 28953 conformant implementation (C99 F.9).' 28954 28955 This is dependent on the implementation of the C library, and is 28956 not defined by GCC itself. 28957 28958 28959File: gcc.info, Node: Arrays and pointers implementation, Next: Hints implementation, Prev: Floating point implementation, Up: C Implementation 28960 289614.7 Arrays and Pointers 28962======================= 28963 28964 * 'The result of converting a pointer to an integer or vice versa 28965 (C90 6.3.4, C99 and C11 6.3.2.3).' 28966 28967 A cast from pointer to integer discards most-significant bits if 28968 the pointer representation is larger than the integer type, 28969 sign-extends(1) if the pointer representation is smaller than the 28970 integer type, otherwise the bits are unchanged. 28971 28972 A cast from integer to pointer discards most-significant bits if 28973 the pointer representation is smaller than the integer type, 28974 extends according to the signedness of the integer type if the 28975 pointer representation is larger than the integer type, otherwise 28976 the bits are unchanged. 28977 28978 When casting from pointer to integer and back again, the resulting 28979 pointer must reference the same object as the original pointer, 28980 otherwise the behavior is undefined. That is, one may not use 28981 integer arithmetic to avoid the undefined behavior of pointer 28982 arithmetic as proscribed in C99 and C11 6.5.6/8. 28983 28984 * 'The size of the result of subtracting two pointers to elements of 28985 the same array (C90 6.3.6, C99 and C11 6.5.6).' 28986 28987 The value is as specified in the standard and the type is 28988 determined by the ABI. 28989 28990 ---------- Footnotes ---------- 28991 28992 (1) Future versions of GCC may zero-extend, or use a target-defined 28993'ptr_extend' pattern. Do not rely on sign extension. 28994 28995 28996File: gcc.info, Node: Hints implementation, Next: Structures unions enumerations and bit-fields implementation, Prev: Arrays and pointers implementation, Up: C Implementation 28997 289984.8 Hints 28999========= 29000 29001 * 'The extent to which suggestions made by using the 'register' 29002 storage-class specifier are effective (C90 6.5.1, C99 and C11 29003 6.7.1).' 29004 29005 The 'register' specifier affects code generation only in these 29006 ways: 29007 29008 * When used as part of the register variable extension, see 29009 *note Explicit Register Variables::. 29010 29011 * When '-O0' is in use, the compiler allocates distinct stack 29012 memory for all variables that do not have the 'register' 29013 storage-class specifier; if 'register' is specified, the 29014 variable may have a shorter lifespan than the code would 29015 indicate and may never be placed in memory. 29016 29017 * On some rare x86 targets, 'setjmp' doesn't save the registers 29018 in all circumstances. In those cases, GCC doesn't allocate 29019 any variables in registers unless they are marked 'register'. 29020 29021 * 'The extent to which suggestions made by using the inline function 29022 specifier are effective (C99 and C11 6.7.4).' 29023 29024 GCC will not inline any functions if the '-fno-inline' option is 29025 used or if '-O0' is used. Otherwise, GCC may still be unable to 29026 inline a function for many reasons; the '-Winline' option may be 29027 used to determine if a function has not been inlined and why not. 29028 29029 29030File: gcc.info, Node: Structures unions enumerations and bit-fields implementation, Next: Qualifiers implementation, Prev: Hints implementation, Up: C Implementation 29031 290324.9 Structures, Unions, Enumerations, and Bit-Fields 29033==================================================== 29034 29035 * 'A member of a union object is accessed using a member of a 29036 different type (C90 6.3.2.3).' 29037 29038 The relevant bytes of the representation of the object are treated 29039 as an object of the type used for the access. *Note 29040 Type-punning::. This may be a trap representation. 29041 29042 * 'Whether a "plain" 'int' bit-field is treated as a 'signed int' 29043 bit-field or as an 'unsigned int' bit-field (C90 6.5.2, C90 29044 6.5.2.1, C99 and C11 6.7.2, C99 and C11 6.7.2.1).' 29045 29046 By default it is treated as 'signed int' but this may be changed by 29047 the '-funsigned-bitfields' option. 29048 29049 * 'Allowable bit-field types other than '_Bool', 'signed int', and 29050 'unsigned int' (C99 and C11 6.7.2.1).' 29051 29052 Other integer types, such as 'long int', and enumerated types are 29053 permitted even in strictly conforming mode. 29054 29055 * 'Whether atomic types are permitted for bit-fields (C11 6.7.2.1).' 29056 29057 Atomic types are not permitted for bit-fields. 29058 29059 * 'Whether a bit-field can straddle a storage-unit boundary (C90 29060 6.5.2.1, C99 and C11 6.7.2.1).' 29061 29062 Determined by ABI. 29063 29064 * 'The order of allocation of bit-fields within a unit (C90 6.5.2.1, 29065 C99 and C11 6.7.2.1).' 29066 29067 Determined by ABI. 29068 29069 * 'The alignment of non-bit-field members of structures (C90 6.5.2.1, 29070 C99 and C11 6.7.2.1).' 29071 29072 Determined by ABI. 29073 29074 * 'The integer type compatible with each enumerated type (C90 29075 6.5.2.2, C99 and C11 6.7.2.2).' 29076 29077 Normally, the type is 'unsigned int' if there are no negative 29078 values in the enumeration, otherwise 'int'. If '-fshort-enums' is 29079 specified, then if there are negative values it is the first of 29080 'signed char', 'short' and 'int' that can represent all the values, 29081 otherwise it is the first of 'unsigned char', 'unsigned short' and 29082 'unsigned int' that can represent all the values. 29083 29084 On some targets, '-fshort-enums' is the default; this is determined 29085 by the ABI. 29086 29087 29088File: gcc.info, Node: Qualifiers implementation, Next: Declarators implementation, Prev: Structures unions enumerations and bit-fields implementation, Up: C Implementation 29089 290904.10 Qualifiers 29091=============== 29092 29093 * 'What constitutes an access to an object that has 29094 volatile-qualified type (C90 6.5.3, C99 and C11 6.7.3).' 29095 29096 Such an object is normally accessed by pointers and used for 29097 accessing hardware. In most expressions, it is intuitively obvious 29098 what is a read and what is a write. For example 29099 29100 volatile int *dst = SOMEVALUE; 29101 volatile int *src = SOMEOTHERVALUE; 29102 *dst = *src; 29103 29104 will cause a read of the volatile object pointed to by SRC and 29105 store the value into the volatile object pointed to by DST. There 29106 is no guarantee that these reads and writes are atomic, especially 29107 for objects larger than 'int'. 29108 29109 However, if the volatile storage is not being modified, and the 29110 value of the volatile storage is not used, then the situation is 29111 less obvious. For example 29112 29113 volatile int *src = SOMEVALUE; 29114 *src; 29115 29116 According to the C standard, such an expression is an rvalue whose 29117 type is the unqualified version of its original type, i.e. 'int'. 29118 Whether GCC interprets this as a read of the volatile object being 29119 pointed to or only as a request to evaluate the expression for its 29120 side effects depends on this type. 29121 29122 If it is a scalar type, or on most targets an aggregate type whose 29123 only member object is of a scalar type, or a union type whose 29124 member objects are of scalar types, the expression is interpreted 29125 by GCC as a read of the volatile object; in the other cases, the 29126 expression is only evaluated for its side effects. 29127 29128 When an object of an aggregate type, with the same size and 29129 alignment as a scalar type 'S', is the subject of a volatile access 29130 by an assignment expression or an atomic function, the access to it 29131 is performed as if the object's declared type were 'volatile S'. 29132 29133 29134File: gcc.info, Node: Declarators implementation, Next: Statements implementation, Prev: Qualifiers implementation, Up: C Implementation 29135 291364.11 Declarators 29137================ 29138 29139 * 'The maximum number of declarators that may modify an arithmetic, 29140 structure or union type (C90 6.5.4).' 29141 29142 GCC is only limited by available memory. 29143 29144 29145File: gcc.info, Node: Statements implementation, Next: Preprocessing directives implementation, Prev: Declarators implementation, Up: C Implementation 29146 291474.12 Statements 29148=============== 29149 29150 * 'The maximum number of 'case' values in a 'switch' statement (C90 29151 6.6.4.2).' 29152 29153 GCC is only limited by available memory. 29154 29155 29156File: gcc.info, Node: Preprocessing directives implementation, Next: Library functions implementation, Prev: Statements implementation, Up: C Implementation 29157 291584.13 Preprocessing Directives 29159============================= 29160 29161*Note Implementation-defined behavior: (cpp)Implementation-defined 29162behavior, for details of these aspects of implementation-defined 29163behavior. 29164 29165 * 'The locations within '#pragma' directives where header name 29166 preprocessing tokens are recognized (C11 6.4, C11 6.4.7).' 29167 29168 * 'How sequences in both forms of header names are mapped to headers 29169 or external source file names (C90 6.1.7, C99 and C11 6.4.7).' 29170 29171 * 'Whether the value of a character constant in a constant expression 29172 that controls conditional inclusion matches the value of the same 29173 character constant in the execution character set (C90 6.8.1, C99 29174 and C11 6.10.1).' 29175 29176 * 'Whether the value of a single-character character constant in a 29177 constant expression that controls conditional inclusion may have a 29178 negative value (C90 6.8.1, C99 and C11 6.10.1).' 29179 29180 * 'The places that are searched for an included '<>' delimited 29181 header, and how the places are specified or the header is 29182 identified (C90 6.8.2, C99 and C11 6.10.2).' 29183 29184 * 'How the named source file is searched for in an included '""' 29185 delimited header (C90 6.8.2, C99 and C11 6.10.2).' 29186 29187 * 'The method by which preprocessing tokens (possibly resulting from 29188 macro expansion) in a '#include' directive are combined into a 29189 header name (C90 6.8.2, C99 and C11 6.10.2).' 29190 29191 * 'The nesting limit for '#include' processing (C90 6.8.2, C99 and 29192 C11 6.10.2).' 29193 29194 * 'Whether the '#' operator inserts a '\' character before the '\' 29195 character that begins a universal character name in a character 29196 constant or string literal (C99 and C11 6.10.3.2).' 29197 29198 * 'The behavior on each recognized non-'STDC #pragma' directive (C90 29199 6.8.6, C99 and C11 6.10.6).' 29200 29201 *Note Pragmas: (cpp)Pragmas, for details of pragmas accepted by GCC 29202 on all targets. *Note Pragmas Accepted by GCC: Pragmas, for 29203 details of target-specific pragmas. 29204 29205 * 'The definitions for '__DATE__' and '__TIME__' when respectively, 29206 the date and time of translation are not available (C90 6.8.8, C99 29207 6.10.8, C11 6.10.8.1).' 29208 29209 29210File: gcc.info, Node: Library functions implementation, Next: Architecture implementation, Prev: Preprocessing directives implementation, Up: C Implementation 29211 292124.14 Library Functions 29213====================== 29214 29215The behavior of most of these points are dependent on the implementation 29216of the C library, and are not defined by GCC itself. 29217 29218 * 'The null pointer constant to which the macro 'NULL' expands (C90 29219 7.1.6, C99 7.17, C11 7.19).' 29220 29221 In '<stddef.h>', 'NULL' expands to '((void *)0)'. GCC does not 29222 provide the other headers which define 'NULL' and some library 29223 implementations may use other definitions in those headers. 29224 29225 29226File: gcc.info, Node: Architecture implementation, Next: Locale-specific behavior implementation, Prev: Library functions implementation, Up: C Implementation 29227 292284.15 Architecture 29229================= 29230 29231 * 'The values or expressions assigned to the macros specified in the 29232 headers '<float.h>', '<limits.h>', and '<stdint.h>' (C90, C99 and 29233 C11 5.2.4.2, C99 7.18.2, C99 7.18.3, C11 7.20.2, C11 7.20.3).' 29234 29235 Determined by ABI. 29236 29237 * 'The result of attempting to indirectly access an object with 29238 automatic or thread storage duration from a thread other than the 29239 one with which it is associated (C11 6.2.4).' 29240 29241 Such accesses are supported, subject to the same requirements for 29242 synchronization for concurrent accesses as for concurrent accesses 29243 to any object. 29244 29245 * 'The number, order, and encoding of bytes in any object (when not 29246 explicitly specified in this International Standard) (C99 and C11 29247 6.2.6.1).' 29248 29249 Determined by ABI. 29250 29251 * 'Whether any extended alignments are supported and the contexts in 29252 which they are supported (C11 6.2.8).' 29253 29254 Extended alignments up to 2^{28} (bytes) are supported for objects 29255 of automatic storage duration. Alignments supported for objects of 29256 static and thread storage duration are determined by the ABI. 29257 29258 * 'Valid alignment values other than those returned by an _Alignof 29259 expression for fundamental types, if any (C11 6.2.8).' 29260 29261 Valid alignments are powers of 2 up to and including 2^{28}. 29262 29263 * 'The value of the result of the 'sizeof' and '_Alignof' operators 29264 (C90 6.3.3.4, C99 and C11 6.5.3.4).' 29265 29266 Determined by ABI. 29267 29268 29269File: gcc.info, Node: Locale-specific behavior implementation, Prev: Architecture implementation, Up: C Implementation 29270 292714.16 Locale-Specific Behavior 29272============================= 29273 29274The behavior of these points are dependent on the implementation of the 29275C library, and are not defined by GCC itself. 29276 29277 29278File: gcc.info, Node: C++ Implementation, Next: C Extensions, Prev: C Implementation, Up: Top 29279 292805 C++ Implementation-Defined Behavior 29281************************************* 29282 29283A conforming implementation of ISO C++ is required to document its 29284choice of behavior in each of the areas that are designated 29285"implementation defined". The following lists all such areas, along 29286with the section numbers from the ISO/IEC 14882:1998 and ISO/IEC 2928714882:2003 standards. Some areas are only implementation-defined in one 29288version of the standard. 29289 29290 Some choices depend on the externally determined ABI for the platform 29291(including standard character encodings) which GCC follows; these are 29292listed as "determined by ABI" below. *Note Binary Compatibility: 29293Compatibility, and <http://gcc.gnu.org/readings.html>. Some choices are 29294documented in the preprocessor manual. *Note Implementation-defined 29295behavior: (cpp)Implementation-defined behavior. Some choices are 29296documented in the corresponding document for the C language. *Note C 29297Implementation::. Some choices are made by the library and operating 29298system (or other environment when compiling for a freestanding 29299environment); refer to their documentation for details. 29300 29301* Menu: 29302 29303* Conditionally-supported behavior:: 29304* Exception handling:: 29305 29306 29307File: gcc.info, Node: Conditionally-supported behavior, Next: Exception handling, Up: C++ Implementation 29308 293095.1 Conditionally-Supported Behavior 29310==================================== 29311 29312'Each implementation shall include documentation that identifies all 29313conditionally-supported constructs that it does not support (C++0x 293141.4).' 29315 29316 * 'Whether an argument of class type with a non-trivial copy 29317 constructor or destructor can be passed to ... (C++0x 5.2.2).' 29318 29319 Such argument passing is supported, using the same 29320 pass-by-invisible-reference approach used for normal function 29321 arguments of such types. 29322 29323 29324File: gcc.info, Node: Exception handling, Prev: Conditionally-supported behavior, Up: C++ Implementation 29325 293265.2 Exception Handling 29327====================== 29328 29329 * 'In the situation where no matching handler is found, it is 29330 implementation-defined whether or not the stack is unwound before 29331 std::terminate() is called (C++98 15.5.1).' 29332 29333 The stack is not unwound before std::terminate is called. 29334 29335 c Copyright (C) 1988-2020 Free Software Foundation, Inc. 29336 29337 29338File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: C++ Implementation, Up: Top 29339 293406 Extensions to the C Language Family 29341************************************* 29342 29343GNU C provides several language features not found in ISO standard C. 29344(The '-pedantic' option directs GCC to print a warning message if any of 29345these features is used.) To test for the availability of these features 29346in conditional compilation, check for a predefined macro '__GNUC__', 29347which is always defined under GCC. 29348 29349 These extensions are available in C and Objective-C. Most of them are 29350also available in C++. *Note Extensions to the C++ Language: C++ 29351Extensions, for extensions that apply _only_ to C++. 29352 29353 Some features that are in ISO C99 but not C90 or C++ are also, as 29354extensions, accepted by GCC in C90 mode and in C++. 29355 29356* Menu: 29357 29358* Statement Exprs:: Putting statements and declarations inside expressions. 29359* Local Labels:: Labels local to a block. 29360* Labels as Values:: Getting pointers to labels, and computed gotos. 29361* Nested Functions:: Nested function in GNU C. 29362* Nonlocal Gotos:: Nonlocal gotos. 29363* Constructing Calls:: Dispatching a call to another function. 29364* Typeof:: 'typeof': referring to the type of an expression. 29365* Conditionals:: Omitting the middle operand of a '?:' expression. 29366* __int128:: 128-bit integers--'__int128'. 29367* Long Long:: Double-word integers--'long long int'. 29368* Complex:: Data types for complex numbers. 29369* Floating Types:: Additional Floating Types. 29370* Half-Precision:: Half-Precision Floating Point. 29371* Decimal Float:: Decimal Floating Types. 29372* Hex Floats:: Hexadecimal floating-point constants. 29373* Fixed-Point:: Fixed-Point Types. 29374* Named Address Spaces::Named address spaces. 29375* Zero Length:: Zero-length arrays. 29376* Empty Structures:: Structures with no members. 29377* Variable Length:: Arrays whose length is computed at run time. 29378* Variadic Macros:: Macros with a variable number of arguments. 29379* Escaped Newlines:: Slightly looser rules for escaped newlines. 29380* Subscripting:: Any array can be subscripted, even if not an lvalue. 29381* Pointer Arith:: Arithmetic on 'void'-pointers and function pointers. 29382* Variadic Pointer Args:: Pointer arguments to variadic functions. 29383* Pointers to Arrays:: Pointers to arrays with qualifiers work as expected. 29384* Initializers:: Non-constant initializers. 29385* Compound Literals:: Compound literals give structures, unions 29386 or arrays as values. 29387* Designated Inits:: Labeling elements of initializers. 29388* Case Ranges:: 'case 1 ... 9' and such. 29389* Cast to Union:: Casting to union type from any member of the union. 29390* Mixed Declarations:: Mixing declarations and code. 29391* Function Attributes:: Declaring that functions have no side effects, 29392 or that they can never return. 29393* Variable Attributes:: Specifying attributes of variables. 29394* Type Attributes:: Specifying attributes of types. 29395* Label Attributes:: Specifying attributes on labels. 29396* Enumerator Attributes:: Specifying attributes on enumerators. 29397* Statement Attributes:: Specifying attributes on statements. 29398* Attribute Syntax:: Formal syntax for attributes. 29399* Function Prototypes:: Prototype declarations and old-style definitions. 29400* C++ Comments:: C++ comments are recognized. 29401* Dollar Signs:: Dollar sign is allowed in identifiers. 29402* Character Escapes:: '\e' stands for the character <ESC>. 29403* Alignment:: Determining the alignment of a function, type or variable. 29404* Inline:: Defining inline functions (as fast as macros). 29405* Volatiles:: What constitutes an access to a volatile object. 29406* Using Assembly Language with C:: Instructions and extensions for interfacing C with assembler. 29407* Alternate Keywords:: '__const__', '__asm__', etc., for header files. 29408* Incomplete Enums:: 'enum foo;', with details to follow. 29409* Function Names:: Printable strings which are the name of the current 29410 function. 29411* Return Address:: Getting the return or frame address of a function. 29412* Vector Extensions:: Using vector instructions through built-in functions. 29413* Offsetof:: Special syntax for implementing 'offsetof'. 29414* __sync Builtins:: Legacy built-in functions for atomic memory access. 29415* __atomic Builtins:: Atomic built-in functions with memory model. 29416* Integer Overflow Builtins:: Built-in functions to perform arithmetics and 29417 arithmetic overflow checking. 29418* x86 specific memory model extensions for transactional memory:: x86 memory models. 29419* Object Size Checking:: Built-in functions for limited buffer overflow 29420 checking. 29421* Other Builtins:: Other built-in functions. 29422* Target Builtins:: Built-in functions specific to particular targets. 29423* Target Format Checks:: Format checks specific to particular targets. 29424* Pragmas:: Pragmas accepted by GCC. 29425* Unnamed Fields:: Unnamed struct/union fields within structs/unions. 29426* Thread-Local:: Per-thread variables. 29427* Binary constants:: Binary constants using the '0b' prefix. 29428 29429 29430File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions 29431 294326.1 Statements and Declarations in Expressions 29433============================================== 29434 29435A compound statement enclosed in parentheses may appear as an expression 29436in GNU C. This allows you to use loops, switches, and local variables 29437within an expression. 29438 29439 Recall that a compound statement is a sequence of statements surrounded 29440by braces; in this construct, parentheses go around the braces. For 29441example: 29442 29443 ({ int y = foo (); int z; 29444 if (y > 0) z = y; 29445 else z = - y; 29446 z; }) 29447 29448is a valid (though slightly more complex than necessary) expression for 29449the absolute value of 'foo ()'. 29450 29451 The last thing in the compound statement should be an expression 29452followed by a semicolon; the value of this subexpression serves as the 29453value of the entire construct. (If you use some other kind of statement 29454last within the braces, the construct has type 'void', and thus 29455effectively no value.) 29456 29457 This feature is especially useful in making macro definitions "safe" 29458(so that they evaluate each operand exactly once). For example, the 29459"maximum" function is commonly defined as a macro in standard C as 29460follows: 29461 29462 #define max(a,b) ((a) > (b) ? (a) : (b)) 29463 29464But this definition computes either A or B twice, with bad results if 29465the operand has side effects. In GNU C, if you know the type of the 29466operands (here taken as 'int'), you can avoid this problem by defining 29467the macro as follows: 29468 29469 #define maxint(a,b) \ 29470 ({int _a = (a), _b = (b); _a > _b ? _a : _b; }) 29471 29472 Note that introducing variable declarations (as we do in 'maxint') can 29473cause variable shadowing, so while this example using the 'max' macro 29474produces correct results: 29475 int _a = 1, _b = 2, c; 29476 c = max (_a, _b); 29477this example using maxint will not: 29478 int _a = 1, _b = 2, c; 29479 c = maxint (_a, _b); 29480 29481 This problem may for instance occur when we use this pattern 29482recursively, like so: 29483 29484 #define maxint3(a, b, c) \ 29485 ({int _a = (a), _b = (b), _c = (c); maxint (maxint (_a, _b), _c); }) 29486 29487 Embedded statements are not allowed in constant expressions, such as 29488the value of an enumeration constant, the width of a bit-field, or the 29489initial value of a static variable. 29490 29491 If you don't know the type of the operand, you can still do this, but 29492you must use 'typeof' or '__auto_type' (*note Typeof::). 29493 29494 In G++, the result value of a statement expression undergoes array and 29495function pointer decay, and is returned by value to the enclosing 29496expression. For instance, if 'A' is a class, then 29497 29498 A a; 29499 29500 ({a;}).Foo () 29501 29502constructs a temporary 'A' object to hold the result of the statement 29503expression, and that is used to invoke 'Foo'. Therefore the 'this' 29504pointer observed by 'Foo' is not the address of 'a'. 29505 29506 In a statement expression, any temporaries created within a statement 29507are destroyed at that statement's end. This makes statement expressions 29508inside macros slightly different from function calls. In the latter 29509case temporaries introduced during argument evaluation are destroyed at 29510the end of the statement that includes the function call. In the 29511statement expression case they are destroyed during the statement 29512expression. For instance, 29513 29514 #define macro(a) ({__typeof__(a) b = (a); b + 3; }) 29515 template<typename T> T function(T a) { T b = a; return b + 3; } 29516 29517 void foo () 29518 { 29519 macro (X ()); 29520 function (X ()); 29521 } 29522 29523has different places where temporaries are destroyed. For the 'macro' 29524case, the temporary 'X' is destroyed just after the initialization of 29525'b'. In the 'function' case that temporary is destroyed when the 29526function returns. 29527 29528 These considerations mean that it is probably a bad idea to use 29529statement expressions of this form in header files that are designed to 29530work with C++. (Note that some versions of the GNU C Library contained 29531header files using statement expressions that lead to precisely this 29532bug.) 29533 29534 Jumping into a statement expression with 'goto' or using a 'switch' 29535statement outside the statement expression with a 'case' or 'default' 29536label inside the statement expression is not permitted. Jumping into a 29537statement expression with a computed 'goto' (*note Labels as Values::) 29538has undefined behavior. Jumping out of a statement expression is 29539permitted, but if the statement expression is part of a larger 29540expression then it is unspecified which other subexpressions of that 29541expression have been evaluated except where the language definition 29542requires certain subexpressions to be evaluated before or after the 29543statement expression. A 'break' or 'continue' statement inside of a 29544statement expression used in 'while', 'do' or 'for' loop or 'switch' 29545statement condition or 'for' statement init or increment expressions 29546jumps to an outer loop or 'switch' statement if any (otherwise it is an 29547error), rather than to the loop or 'switch' statement in whose condition 29548or init or increment expression it appears. In any case, as with a 29549function call, the evaluation of a statement expression is not 29550interleaved with the evaluation of other parts of the containing 29551expression. For example, 29552 29553 foo (), (({ bar1 (); goto a; 0; }) + bar2 ()), baz(); 29554 29555calls 'foo' and 'bar1' and does not call 'baz' but may or may not call 29556'bar2'. If 'bar2' is called, it is called after 'foo' and before 29557'bar1'. 29558 29559 29560File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions 29561 295626.2 Locally Declared Labels 29563=========================== 29564 29565GCC allows you to declare "local labels" in any nested block scope. A 29566local label is just like an ordinary label, but you can only reference 29567it (with a 'goto' statement, or by taking its address) within the block 29568in which it is declared. 29569 29570 A local label declaration looks like this: 29571 29572 __label__ LABEL; 29573 29574or 29575 29576 __label__ LABEL1, LABEL2, /* ... */; 29577 29578 Local label declarations must come at the beginning of the block, 29579before any ordinary declarations or statements. 29580 29581 The label declaration defines the label _name_, but does not define the 29582label itself. You must do this in the usual way, with 'LABEL:', within 29583the statements of the statement expression. 29584 29585 The local label feature is useful for complex macros. If a macro 29586contains nested loops, a 'goto' can be useful for breaking out of them. 29587However, an ordinary label whose scope is the whole function cannot be 29588used: if the macro can be expanded several times in one function, the 29589label is multiply defined in that function. A local label avoids this 29590problem. For example: 29591 29592 #define SEARCH(value, array, target) \ 29593 do { \ 29594 __label__ found; \ 29595 typeof (target) _SEARCH_target = (target); \ 29596 typeof (*(array)) *_SEARCH_array = (array); \ 29597 int i, j; \ 29598 int value; \ 29599 for (i = 0; i < max; i++) \ 29600 for (j = 0; j < max; j++) \ 29601 if (_SEARCH_array[i][j] == _SEARCH_target) \ 29602 { (value) = i; goto found; } \ 29603 (value) = -1; \ 29604 found:; \ 29605 } while (0) 29606 29607 This could also be written using a statement expression: 29608 29609 #define SEARCH(array, target) \ 29610 ({ \ 29611 __label__ found; \ 29612 typeof (target) _SEARCH_target = (target); \ 29613 typeof (*(array)) *_SEARCH_array = (array); \ 29614 int i, j; \ 29615 int value; \ 29616 for (i = 0; i < max; i++) \ 29617 for (j = 0; j < max; j++) \ 29618 if (_SEARCH_array[i][j] == _SEARCH_target) \ 29619 { value = i; goto found; } \ 29620 value = -1; \ 29621 found: \ 29622 value; \ 29623 }) 29624 29625 Local label declarations also make the labels they declare visible to 29626nested functions, if there are any. *Note Nested Functions::, for 29627details. 29628 29629 29630File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions 29631 296326.3 Labels as Values 29633==================== 29634 29635You can get the address of a label defined in the current function (or a 29636containing function) with the unary operator '&&'. The value has type 29637'void *'. This value is a constant and can be used wherever a constant 29638of that type is valid. For example: 29639 29640 void *ptr; 29641 /* ... */ 29642 ptr = &&foo; 29643 29644 To use these values, you need to be able to jump to one. This is done 29645with the computed goto statement(1), 'goto *EXP;'. For example, 29646 29647 goto *ptr; 29648 29649Any expression of type 'void *' is allowed. 29650 29651 One way of using these constants is in initializing a static array that 29652serves as a jump table: 29653 29654 static void *array[] = { &&foo, &&bar, &&hack }; 29655 29656Then you can select a label with indexing, like this: 29657 29658 goto *array[i]; 29659 29660Note that this does not check whether the subscript is in bounds--array 29661indexing in C never does that. 29662 29663 Such an array of label values serves a purpose much like that of the 29664'switch' statement. The 'switch' statement is cleaner, so use that 29665rather than an array unless the problem does not fit a 'switch' 29666statement very well. 29667 29668 Another use of label values is in an interpreter for threaded code. 29669The labels within the interpreter function can be stored in the threaded 29670code for super-fast dispatching. 29671 29672 You may not use this mechanism to jump to code in a different function. 29673If you do that, totally unpredictable things happen. The best way to 29674avoid this is to store the label address only in automatic variables and 29675never pass it as an argument. 29676 29677 An alternate way to write the above example is 29678 29679 static const int array[] = { &&foo - &&foo, &&bar - &&foo, 29680 &&hack - &&foo }; 29681 goto *(&&foo + array[i]); 29682 29683This is more friendly to code living in shared libraries, as it reduces 29684the number of dynamic relocations that are needed, and by consequence, 29685allows the data to be read-only. This alternative with label 29686differences is not supported for the AVR target, please use the first 29687approach for AVR programs. 29688 29689 The '&&foo' expressions for the same label might have different values 29690if the containing function is inlined or cloned. If a program relies on 29691them being always the same, '__attribute__((__noinline__,__noclone__))' 29692should be used to prevent inlining and cloning. If '&&foo' is used in a 29693static variable initializer, inlining and cloning is forbidden. 29694 29695 ---------- Footnotes ---------- 29696 29697 (1) The analogous feature in Fortran is called an assigned goto, but 29698that name seems inappropriate in C, where one can do more than simply 29699store label addresses in label variables. 29700 29701 29702File: gcc.info, Node: Nested Functions, Next: Nonlocal Gotos, Prev: Labels as Values, Up: C Extensions 29703 297046.4 Nested Functions 29705==================== 29706 29707A "nested function" is a function defined inside another function. 29708Nested functions are supported as an extension in GNU C, but are not 29709supported by GNU C++. 29710 29711 The nested function's name is local to the block where it is defined. 29712For example, here we define a nested function named 'square', and call 29713it twice: 29714 29715 foo (double a, double b) 29716 { 29717 double square (double z) { return z * z; } 29718 29719 return square (a) + square (b); 29720 } 29721 29722 The nested function can access all the variables of the containing 29723function that are visible at the point of its definition. This is 29724called "lexical scoping". For example, here we show a nested function 29725which uses an inherited variable named 'offset': 29726 29727 bar (int *array, int offset, int size) 29728 { 29729 int access (int *array, int index) 29730 { return array[index + offset]; } 29731 int i; 29732 /* ... */ 29733 for (i = 0; i < size; i++) 29734 /* ... */ access (array, i) /* ... */ 29735 } 29736 29737 Nested function definitions are permitted within functions in the 29738places where variable definitions are allowed; that is, in any block, 29739mixed with the other declarations and statements in the block. 29740 29741 It is possible to call the nested function from outside the scope of 29742its name by storing its address or passing the address to another 29743function: 29744 29745 hack (int *array, int size) 29746 { 29747 void store (int index, int value) 29748 { array[index] = value; } 29749 29750 intermediate (store, size); 29751 } 29752 29753 Here, the function 'intermediate' receives the address of 'store' as an 29754argument. If 'intermediate' calls 'store', the arguments given to 29755'store' are used to store into 'array'. But this technique works only 29756so long as the containing function ('hack', in this example) does not 29757exit. 29758 29759 If you try to call the nested function through its address after the 29760containing function exits, all hell breaks loose. If you try to call it 29761after a containing scope level exits, and if it refers to some of the 29762variables that are no longer in scope, you may be lucky, but it's not 29763wise to take the risk. If, however, the nested function does not refer 29764to anything that has gone out of scope, you should be safe. 29765 29766 GCC implements taking the address of a nested function using a 29767technique called "trampolines". This technique was described in 29768'Lexical Closures for C++' (Thomas M. Breuel, USENIX C++ Conference 29769Proceedings, October 17-21, 1988). 29770 29771 A nested function can jump to a label inherited from a containing 29772function, provided the label is explicitly declared in the containing 29773function (*note Local Labels::). Such a jump returns instantly to the 29774containing function, exiting the nested function that did the 'goto' and 29775any intermediate functions as well. Here is an example: 29776 29777 bar (int *array, int offset, int size) 29778 { 29779 __label__ failure; 29780 int access (int *array, int index) 29781 { 29782 if (index > size) 29783 goto failure; 29784 return array[index + offset]; 29785 } 29786 int i; 29787 /* ... */ 29788 for (i = 0; i < size; i++) 29789 /* ... */ access (array, i) /* ... */ 29790 /* ... */ 29791 return 0; 29792 29793 /* Control comes here from 'access' 29794 if it detects an error. */ 29795 failure: 29796 return -1; 29797 } 29798 29799 A nested function always has no linkage. Declaring one with 'extern' 29800or 'static' is erroneous. If you need to declare the nested function 29801before its definition, use 'auto' (which is otherwise meaningless for 29802function declarations). 29803 29804 bar (int *array, int offset, int size) 29805 { 29806 __label__ failure; 29807 auto int access (int *, int); 29808 /* ... */ 29809 int access (int *array, int index) 29810 { 29811 if (index > size) 29812 goto failure; 29813 return array[index + offset]; 29814 } 29815 /* ... */ 29816 } 29817 29818 29819File: gcc.info, Node: Nonlocal Gotos, Next: Constructing Calls, Prev: Nested Functions, Up: C Extensions 29820 298216.5 Nonlocal Gotos 29822================== 29823 29824GCC provides the built-in functions '__builtin_setjmp' and 29825'__builtin_longjmp' which are similar to, but not interchangeable with, 29826the C library functions 'setjmp' and 'longjmp'. The built-in versions 29827are used internally by GCC's libraries to implement exception handling 29828on some targets. You should use the standard C library functions 29829declared in '<setjmp.h>' in user code instead of the builtins. 29830 29831 The built-in versions of these functions use GCC's normal mechanisms to 29832save and restore registers using the stack on function entry and exit. 29833The jump buffer argument BUF holds only the information needed to 29834restore the stack frame, rather than the entire set of saved register 29835values. 29836 29837 An important caveat is that GCC arranges to save and restore only those 29838registers known to the specific architecture variant being compiled for. 29839This can make '__builtin_setjmp' and '__builtin_longjmp' more efficient 29840than their library counterparts in some cases, but it can also cause 29841incorrect and mysterious behavior when mixing with code that uses the 29842full register set. 29843 29844 You should declare the jump buffer argument BUF to the built-in 29845functions as: 29846 29847 #include <stdint.h> 29848 intptr_t BUF[5]; 29849 29850 -- Built-in Function: int __builtin_setjmp (intptr_t *BUF) 29851 This function saves the current stack context in BUF. 29852 '__builtin_setjmp' returns 0 when returning directly, and 1 when 29853 returning from '__builtin_longjmp' using the same BUF. 29854 29855 -- Built-in Function: void __builtin_longjmp (intptr_t *BUF, int VAL) 29856 This function restores the stack context in BUF, saved by a 29857 previous call to '__builtin_setjmp'. After '__builtin_longjmp' is 29858 finished, the program resumes execution as if the matching 29859 '__builtin_setjmp' returns the value VAL, which must be 1. 29860 29861 Because '__builtin_longjmp' depends on the function return 29862 mechanism to restore the stack context, it cannot be called from 29863 the same function calling '__builtin_setjmp' to initialize BUF. It 29864 can only be called from a function called (directly or indirectly) 29865 from the function calling '__builtin_setjmp'. 29866 29867 29868File: gcc.info, Node: Constructing Calls, Next: Typeof, Prev: Nonlocal Gotos, Up: C Extensions 29869 298706.6 Constructing Function Calls 29871=============================== 29872 29873Using the built-in functions described below, you can record the 29874arguments a function received, and call another function with the same 29875arguments, without knowing the number or types of the arguments. 29876 29877 You can also record the return value of that function call, and later 29878return that value, without knowing what data type the function tried to 29879return (as long as your caller expects that data type). 29880 29881 However, these built-in functions may interact badly with some 29882sophisticated features or other extensions of the language. It is, 29883therefore, not recommended to use them outside very simple functions 29884acting as mere forwarders for their arguments. 29885 29886 -- Built-in Function: void * __builtin_apply_args () 29887 This built-in function returns a pointer to data describing how to 29888 perform a call with the same arguments as are passed to the current 29889 function. 29890 29891 The function saves the arg pointer register, structure value 29892 address, and all registers that might be used to pass arguments to 29893 a function into a block of memory allocated on the stack. Then it 29894 returns the address of that block. 29895 29896 -- Built-in Function: void * __builtin_apply (void (*FUNCTION)(), void 29897 *ARGUMENTS, size_t SIZE) 29898 This built-in function invokes FUNCTION with a copy of the 29899 parameters described by ARGUMENTS and SIZE. 29900 29901 The value of ARGUMENTS should be the value returned by 29902 '__builtin_apply_args'. The argument SIZE specifies the size of 29903 the stack argument data, in bytes. 29904 29905 This function returns a pointer to data describing how to return 29906 whatever value is returned by FUNCTION. The data is saved in a 29907 block of memory allocated on the stack. 29908 29909 It is not always simple to compute the proper value for SIZE. The 29910 value is used by '__builtin_apply' to compute the amount of data 29911 that should be pushed on the stack and copied from the incoming 29912 argument area. 29913 29914 -- Built-in Function: void __builtin_return (void *RESULT) 29915 This built-in function returns the value described by RESULT from 29916 the containing function. You should specify, for RESULT, a value 29917 returned by '__builtin_apply'. 29918 29919 -- Built-in Function: __builtin_va_arg_pack () 29920 This built-in function represents all anonymous arguments of an 29921 inline function. It can be used only in inline functions that are 29922 always inlined, never compiled as a separate function, such as 29923 those using '__attribute__ ((__always_inline__))' or '__attribute__ 29924 ((__gnu_inline__))' extern inline functions. It must be only 29925 passed as last argument to some other function with variable 29926 arguments. This is useful for writing small wrapper inlines for 29927 variable argument functions, when using preprocessor macros is 29928 undesirable. For example: 29929 extern int myprintf (FILE *f, const char *format, ...); 29930 extern inline __attribute__ ((__gnu_inline__)) int 29931 myprintf (FILE *f, const char *format, ...) 29932 { 29933 int r = fprintf (f, "myprintf: "); 29934 if (r < 0) 29935 return r; 29936 int s = fprintf (f, format, __builtin_va_arg_pack ()); 29937 if (s < 0) 29938 return s; 29939 return r + s; 29940 } 29941 29942 -- Built-in Function: int __builtin_va_arg_pack_len () 29943 This built-in function returns the number of anonymous arguments of 29944 an inline function. It can be used only in inline functions that 29945 are always inlined, never compiled as a separate function, such as 29946 those using '__attribute__ ((__always_inline__))' or '__attribute__ 29947 ((__gnu_inline__))' extern inline functions. For example following 29948 does link- or run-time checking of open arguments for optimized 29949 code: 29950 #ifdef __OPTIMIZE__ 29951 extern inline __attribute__((__gnu_inline__)) int 29952 myopen (const char *path, int oflag, ...) 29953 { 29954 if (__builtin_va_arg_pack_len () > 1) 29955 warn_open_too_many_arguments (); 29956 29957 if (__builtin_constant_p (oflag)) 29958 { 29959 if ((oflag & O_CREAT) != 0 && __builtin_va_arg_pack_len () < 1) 29960 { 29961 warn_open_missing_mode (); 29962 return __open_2 (path, oflag); 29963 } 29964 return open (path, oflag, __builtin_va_arg_pack ()); 29965 } 29966 29967 if (__builtin_va_arg_pack_len () < 1) 29968 return __open_2 (path, oflag); 29969 29970 return open (path, oflag, __builtin_va_arg_pack ()); 29971 } 29972 #endif 29973 29974 29975File: gcc.info, Node: Typeof, Next: Conditionals, Prev: Constructing Calls, Up: C Extensions 29976 299776.7 Referring to a Type with 'typeof' 29978===================================== 29979 29980Another way to refer to the type of an expression is with 'typeof'. The 29981syntax of using of this keyword looks like 'sizeof', but the construct 29982acts semantically like a type name defined with 'typedef'. 29983 29984 There are two ways of writing the argument to 'typeof': with an 29985expression or with a type. Here is an example with an expression: 29986 29987 typeof (x[0](1)) 29988 29989This assumes that 'x' is an array of pointers to functions; the type 29990described is that of the values of the functions. 29991 29992 Here is an example with a typename as the argument: 29993 29994 typeof (int *) 29995 29996Here the type described is that of pointers to 'int'. 29997 29998 If you are writing a header file that must work when included in ISO C 29999programs, write '__typeof__' instead of 'typeof'. *Note Alternate 30000Keywords::. 30001 30002 A 'typeof' construct can be used anywhere a typedef name can be used. 30003For example, you can use it in a declaration, in a cast, or inside of 30004'sizeof' or 'typeof'. 30005 30006 The operand of 'typeof' is evaluated for its side effects if and only 30007if it is an expression of variably modified type or the name of such a 30008type. 30009 30010 'typeof' is often useful in conjunction with statement expressions 30011(*note Statement Exprs::). Here is how the two together can be used to 30012define a safe "maximum" macro which operates on any arithmetic type and 30013evaluates each of its arguments exactly once: 30014 30015 #define max(a,b) \ 30016 ({ typeof (a) _a = (a); \ 30017 typeof (b) _b = (b); \ 30018 _a > _b ? _a : _b; }) 30019 30020 The reason for using names that start with underscores for the local 30021variables is to avoid conflicts with variable names that occur within 30022the expressions that are substituted for 'a' and 'b'. Eventually we 30023hope to design a new form of declaration syntax that allows you to 30024declare variables whose scopes start only after their initializers; this 30025will be a more reliable way to prevent such conflicts. 30026 30027Some more examples of the use of 'typeof': 30028 30029 * This declares 'y' with the type of what 'x' points to. 30030 30031 typeof (*x) y; 30032 30033 * This declares 'y' as an array of such values. 30034 30035 typeof (*x) y[4]; 30036 30037 * This declares 'y' as an array of pointers to characters: 30038 30039 typeof (typeof (char *)[4]) y; 30040 30041 It is equivalent to the following traditional C declaration: 30042 30043 char *y[4]; 30044 30045 To see the meaning of the declaration using 'typeof', and why it 30046 might be a useful way to write, rewrite it with these macros: 30047 30048 #define pointer(T) typeof(T *) 30049 #define array(T, N) typeof(T [N]) 30050 30051 Now the declaration can be rewritten this way: 30052 30053 array (pointer (char), 4) y; 30054 30055 Thus, 'array (pointer (char), 4)' is the type of arrays of 4 30056 pointers to 'char'. 30057 30058 In GNU C, but not GNU C++, you may also declare the type of a variable 30059as '__auto_type'. In that case, the declaration must declare only one 30060variable, whose declarator must just be an identifier, the declaration 30061must be initialized, and the type of the variable is determined by the 30062initializer; the name of the variable is not in scope until after the 30063initializer. (In C++, you should use C++11 'auto' for this purpose.) 30064Using '__auto_type', the "maximum" macro above could be written as: 30065 30066 #define max(a,b) \ 30067 ({ __auto_type _a = (a); \ 30068 __auto_type _b = (b); \ 30069 _a > _b ? _a : _b; }) 30070 30071 Using '__auto_type' instead of 'typeof' has two advantages: 30072 30073 * Each argument to the macro appears only once in the expansion of 30074 the macro. This prevents the size of the macro expansion growing 30075 exponentially when calls to such macros are nested inside arguments 30076 of such macros. 30077 30078 * If the argument to the macro has variably modified type, it is 30079 evaluated only once when using '__auto_type', but twice if 'typeof' 30080 is used. 30081 30082 30083File: gcc.info, Node: Conditionals, Next: __int128, Prev: Typeof, Up: C Extensions 30084 300856.8 Conditionals with Omitted Operands 30086====================================== 30087 30088The middle operand in a conditional expression may be omitted. Then if 30089the first operand is nonzero, its value is the value of the conditional 30090expression. 30091 30092 Therefore, the expression 30093 30094 x ? : y 30095 30096has the value of 'x' if that is nonzero; otherwise, the value of 'y'. 30097 30098 This example is perfectly equivalent to 30099 30100 x ? x : y 30101 30102In this simple case, the ability to omit the middle operand is not 30103especially useful. When it becomes useful is when the first operand 30104does, or may (if it is a macro argument), contain a side effect. Then 30105repeating the operand in the middle would perform the side effect twice. 30106Omitting the middle operand uses the value already computed without the 30107undesirable effects of recomputing it. 30108 30109 30110File: gcc.info, Node: __int128, Next: Long Long, Prev: Conditionals, Up: C Extensions 30111 301126.9 128-bit Integers 30113==================== 30114 30115As an extension the integer scalar type '__int128' is supported for 30116targets which have an integer mode wide enough to hold 128 bits. Simply 30117write '__int128' for a signed 128-bit integer, or 'unsigned __int128' 30118for an unsigned 128-bit integer. There is no support in GCC for 30119expressing an integer constant of type '__int128' for targets with 'long 30120long' integer less than 128 bits wide. 30121 30122 30123File: gcc.info, Node: Long Long, Next: Complex, Prev: __int128, Up: C Extensions 30124 301256.10 Double-Word Integers 30126========================= 30127 30128ISO C99 and ISO C++11 support data types for integers that are at least 3012964 bits wide, and as an extension GCC supports them in C90 and C++98 30130modes. Simply write 'long long int' for a signed integer, or 'unsigned 30131long long int' for an unsigned integer. To make an integer constant of 30132type 'long long int', add the suffix 'LL' to the integer. To make an 30133integer constant of type 'unsigned long long int', add the suffix 'ULL' 30134to the integer. 30135 30136 You can use these types in arithmetic like any other integer types. 30137Addition, subtraction, and bitwise boolean operations on these types are 30138open-coded on all types of machines. Multiplication is open-coded if 30139the machine supports a fullword-to-doubleword widening multiply 30140instruction. Division and shifts are open-coded only on machines that 30141provide special support. The operations that are not open-coded use 30142special library routines that come with GCC. 30143 30144 There may be pitfalls when you use 'long long' types for function 30145arguments without function prototypes. If a function expects type 'int' 30146for its argument, and you pass a value of type 'long long int', 30147confusion results because the caller and the subroutine disagree about 30148the number of bytes for the argument. Likewise, if the function expects 30149'long long int' and you pass 'int'. The best way to avoid such problems 30150is to use prototypes. 30151 30152 30153File: gcc.info, Node: Complex, Next: Floating Types, Prev: Long Long, Up: C Extensions 30154 301556.11 Complex Numbers 30156==================== 30157 30158ISO C99 supports complex floating data types, and as an extension GCC 30159supports them in C90 mode and in C++. GCC also supports complex integer 30160data types which are not part of ISO C99. You can declare complex types 30161using the keyword '_Complex'. As an extension, the older GNU keyword 30162'__complex__' is also supported. 30163 30164 For example, '_Complex double x;' declares 'x' as a variable whose real 30165part and imaginary part are both of type 'double'. '_Complex short int 30166y;' declares 'y' to have real and imaginary parts of type 'short int'; 30167this is not likely to be useful, but it shows that the set of complex 30168types is complete. 30169 30170 To write a constant with a complex data type, use the suffix 'i' or 'j' 30171(either one; they are equivalent). For example, '2.5fi' has type 30172'_Complex float' and '3i' has type '_Complex int'. Such a constant 30173always has a pure imaginary value, but you can form any complex value 30174you like by adding one to a real constant. This is a GNU extension; if 30175you have an ISO C99 conforming C library (such as the GNU C Library), 30176and want to construct complex constants of floating type, you should 30177include '<complex.h>' and use the macros 'I' or '_Complex_I' instead. 30178 30179 The ISO C++14 library also defines the 'i' suffix, so C++14 code that 30180includes the '<complex>' header cannot use 'i' for the GNU extension. 30181The 'j' suffix still has the GNU meaning. 30182 30183 To extract the real part of a complex-valued expression EXP, write 30184'__real__ EXP'. Likewise, use '__imag__' to extract the imaginary part. 30185This is a GNU extension; for values of floating type, you should use the 30186ISO C99 functions 'crealf', 'creal', 'creall', 'cimagf', 'cimag' and 30187'cimagl', declared in '<complex.h>' and also provided as built-in 30188functions by GCC. 30189 30190 The operator '~' performs complex conjugation when used on a value with 30191a complex type. This is a GNU extension; for values of floating type, 30192you should use the ISO C99 functions 'conjf', 'conj' and 'conjl', 30193declared in '<complex.h>' and also provided as built-in functions by 30194GCC. 30195 30196 GCC can allocate complex automatic variables in a noncontiguous 30197fashion; it's even possible for the real part to be in a register while 30198the imaginary part is on the stack (or vice versa). Only the DWARF 30199debug info format can represent this, so use of DWARF is recommended. 30200If you are using the stabs debug info format, GCC describes a 30201noncontiguous complex variable as if it were two separate variables of 30202noncomplex type. If the variable's actual name is 'foo', the two 30203fictitious variables are named 'foo$real' and 'foo$imag'. You can 30204examine and set these two fictitious variables with your debugger. 30205 30206 30207File: gcc.info, Node: Floating Types, Next: Half-Precision, Prev: Complex, Up: C Extensions 30208 302096.12 Additional Floating Types 30210============================== 30211 30212ISO/IEC TS 18661-3:2015 defines C support for additional floating types 30213'_FloatN' and '_FloatNx', and GCC supports these type names; the set of 30214types supported depends on the target architecture. These types are not 30215supported when compiling C++. Constants with these types use suffixes 30216'fN' or 'FN' and 'fNx' or 'FNx'. These type names can be used together 30217with '_Complex' to declare complex types. 30218 30219 As an extension, GNU C and GNU C++ support additional floating types, 30220which are not supported by all targets. 30221 * '__float128' is available on i386, x86_64, IA-64, and hppa HP-UX, 30222 as well as on PowerPC GNU/Linux targets that enable the vector 30223 scalar (VSX) instruction set. '__float128' supports the 128-bit 30224 floating type. On i386, x86_64, PowerPC, and IA-64 other than 30225 HP-UX, '__float128' is an alias for '_Float128'. On hppa and IA-64 30226 HP-UX, '__float128' is an alias for 'long double'. 30227 30228 * '__float80' is available on the i386, x86_64, and IA-64 targets, 30229 and supports the 80-bit ('XFmode') floating type. It is an alias 30230 for the type name '_Float64x' on these targets. 30231 30232 * '__ibm128' is available on PowerPC targets, and provides access to 30233 the IBM extended double format which is the current format used for 30234 'long double'. When 'long double' transitions to '__float128' on 30235 PowerPC in the future, '__ibm128' will remain for use in 30236 conversions between the two types. 30237 30238 Support for these additional types includes the arithmetic operators: 30239add, subtract, multiply, divide; unary arithmetic operators; relational 30240operators; equality operators; and conversions to and from integer and 30241other floating types. Use a suffix 'w' or 'W' in a literal constant of 30242type '__float80' or type '__ibm128'. Use a suffix 'q' or 'Q' for 30243'_float128'. 30244 30245 In order to use '_Float128', '__float128', and '__ibm128' on PowerPC 30246Linux systems, you must use the '-mfloat128' option. It is expected in 30247future versions of GCC that '_Float128' and '__float128' will be enabled 30248automatically. 30249 30250 The '_Float128' type is supported on all systems where '__float128' is 30251supported or where 'long double' has the IEEE binary128 format. The 30252'_Float64x' type is supported on all systems where '__float128' is 30253supported. The '_Float32' type is supported on all systems supporting 30254IEEE binary32; the '_Float64' and '_Float32x' types are supported on all 30255systems supporting IEEE binary64. The '_Float16' type is supported on 30256AArch64 systems by default, and on ARM systems when the IEEE format for 3025716-bit floating-point types is selected with '-mfp16-format=ieee'. GCC 30258does not currently support '_Float128x' on any systems. 30259 30260 On the i386, x86_64, IA-64, and HP-UX targets, you can declare complex 30261types using the corresponding internal complex type, 'XCmode' for 30262'__float80' type and 'TCmode' for '__float128' type: 30263 30264 typedef _Complex float __attribute__((mode(TC))) _Complex128; 30265 typedef _Complex float __attribute__((mode(XC))) _Complex80; 30266 30267 On the PowerPC Linux VSX targets, you can declare complex types using 30268the corresponding internal complex type, 'KCmode' for '__float128' type 30269and 'ICmode' for '__ibm128' type: 30270 30271 typedef _Complex float __attribute__((mode(KC))) _Complex_float128; 30272 typedef _Complex float __attribute__((mode(IC))) _Complex_ibm128; 30273 30274 30275File: gcc.info, Node: Half-Precision, Next: Decimal Float, Prev: Floating Types, Up: C Extensions 30276 302776.13 Half-Precision Floating Point 30278================================== 30279 30280On ARM and AArch64 targets, GCC supports half-precision (16-bit) 30281floating point via the '__fp16' type defined in the ARM C Language 30282Extensions. On ARM systems, you must enable this type explicitly with 30283the '-mfp16-format' command-line option in order to use it. 30284 30285 ARM targets support two incompatible representations for half-precision 30286floating-point values. You must choose one of the representations and 30287use it consistently in your program. 30288 30289 Specifying '-mfp16-format=ieee' selects the IEEE 754-2008 format. This 30290format can represent normalized values in the range of 2^{-14} to 65504. 30291There are 11 bits of significand precision, approximately 3 decimal 30292digits. 30293 30294 Specifying '-mfp16-format=alternative' selects the ARM alternative 30295format. This representation is similar to the IEEE format, but does not 30296support infinities or NaNs. Instead, the range of exponents is 30297extended, so that this format can represent normalized values in the 30298range of 2^{-14} to 131008. 30299 30300 The GCC port for AArch64 only supports the IEEE 754-2008 format, and 30301does not require use of the '-mfp16-format' command-line option. 30302 30303 The '__fp16' type may only be used as an argument to intrinsics defined 30304in '<arm_fp16.h>', or as a storage format. For purposes of arithmetic 30305and other operations, '__fp16' values in C or C++ expressions are 30306automatically promoted to 'float'. 30307 30308 The ARM target provides hardware support for conversions between 30309'__fp16' and 'float' values as an extension to VFP and NEON (Advanced 30310SIMD), and from ARMv8-A provides hardware support for conversions 30311between '__fp16' and 'double' values. GCC generates code using these 30312hardware instructions if you compile with options to select an FPU that 30313provides them; for example, '-mfpu=neon-fp16 -mfloat-abi=softfp', in 30314addition to the '-mfp16-format' option to select a half-precision 30315format. 30316 30317 Language-level support for the '__fp16' data type is independent of 30318whether GCC generates code using hardware floating-point instructions. 30319In cases where hardware support is not specified, GCC implements 30320conversions between '__fp16' and other types as library calls. 30321 30322 It is recommended that portable code use the '_Float16' type defined by 30323ISO/IEC TS 18661-3:2015. *Note Floating Types::. 30324 30325 30326File: gcc.info, Node: Decimal Float, Next: Hex Floats, Prev: Half-Precision, Up: C Extensions 30327 303286.14 Decimal Floating Types 30329=========================== 30330 30331As an extension, GNU C supports decimal floating types as defined in the 30332N1312 draft of ISO/IEC WDTR24732. Support for decimal floating types in 30333GCC will evolve as the draft technical report changes. Calling 30334conventions for any target might also change. Not all targets support 30335decimal floating types. 30336 30337 The decimal floating types are '_Decimal32', '_Decimal64', and 30338'_Decimal128'. They use a radix of ten, unlike the floating types 30339'float', 'double', and 'long double' whose radix is not specified by the 30340C standard but is usually two. 30341 30342 Support for decimal floating types includes the arithmetic operators 30343add, subtract, multiply, divide; unary arithmetic operators; relational 30344operators; equality operators; and conversions to and from integer and 30345other floating types. Use a suffix 'df' or 'DF' in a literal constant 30346of type '_Decimal32', 'dd' or 'DD' for '_Decimal64', and 'dl' or 'DL' 30347for '_Decimal128'. 30348 30349 GCC support of decimal float as specified by the draft technical report 30350is incomplete: 30351 30352 * When the value of a decimal floating type cannot be represented in 30353 the integer type to which it is being converted, the result is 30354 undefined rather than the result value specified by the draft 30355 technical report. 30356 30357 * GCC does not provide the C library functionality associated with 30358 'math.h', 'fenv.h', 'stdio.h', 'stdlib.h', and 'wchar.h', which 30359 must come from a separate C library implementation. Because of 30360 this the GNU C compiler does not define macro '__STDC_DEC_FP__' to 30361 indicate that the implementation conforms to the technical report. 30362 30363 Types '_Decimal32', '_Decimal64', and '_Decimal128' are supported by 30364the DWARF debug information format. 30365 30366 30367File: gcc.info, Node: Hex Floats, Next: Fixed-Point, Prev: Decimal Float, Up: C Extensions 30368 303696.15 Hex Floats 30370=============== 30371 30372ISO C99 and ISO C++17 support floating-point numbers written not only in 30373the usual decimal notation, such as '1.55e1', but also numbers such as 30374'0x1.fp3' written in hexadecimal format. As a GNU extension, GCC 30375supports this in C90 mode (except in some cases when strictly 30376conforming) and in C++98, C++11 and C++14 modes. In that format the 30377'0x' hex introducer and the 'p' or 'P' exponent field are mandatory. 30378The exponent is a decimal number that indicates the power of 2 by which 30379the significant part is multiplied. Thus '0x1.f' is 1 15/16, 'p3' 30380multiplies it by 8, and the value of '0x1.fp3' is the same as '1.55e1'. 30381 30382 Unlike for floating-point numbers in the decimal notation the exponent 30383is always required in the hexadecimal notation. Otherwise the compiler 30384would not be able to resolve the ambiguity of, e.g., '0x1.f'. This 30385could mean '1.0f' or '1.9375' since 'f' is also the extension for 30386floating-point constants of type 'float'. 30387 30388 30389File: gcc.info, Node: Fixed-Point, Next: Named Address Spaces, Prev: Hex Floats, Up: C Extensions 30390 303916.16 Fixed-Point Types 30392====================== 30393 30394As an extension, GNU C supports fixed-point types as defined in the 30395N1169 draft of ISO/IEC DTR 18037. Support for fixed-point types in GCC 30396will evolve as the draft technical report changes. Calling conventions 30397for any target might also change. Not all targets support fixed-point 30398types. 30399 30400 The fixed-point types are 'short _Fract', '_Fract', 'long _Fract', 30401'long long _Fract', 'unsigned short _Fract', 'unsigned _Fract', 30402'unsigned long _Fract', 'unsigned long long _Fract', '_Sat short 30403_Fract', '_Sat _Fract', '_Sat long _Fract', '_Sat long long _Fract', 30404'_Sat unsigned short _Fract', '_Sat unsigned _Fract', '_Sat unsigned 30405long _Fract', '_Sat unsigned long long _Fract', 'short _Accum', 30406'_Accum', 'long _Accum', 'long long _Accum', 'unsigned short _Accum', 30407'unsigned _Accum', 'unsigned long _Accum', 'unsigned long long _Accum', 30408'_Sat short _Accum', '_Sat _Accum', '_Sat long _Accum', '_Sat long long 30409_Accum', '_Sat unsigned short _Accum', '_Sat unsigned _Accum', '_Sat 30410unsigned long _Accum', '_Sat unsigned long long _Accum'. 30411 30412 Fixed-point data values contain fractional and optional integral parts. 30413The format of fixed-point data varies and depends on the target machine. 30414 30415 Support for fixed-point types includes: 30416 * prefix and postfix increment and decrement operators ('++', '--') 30417 * unary arithmetic operators ('+', '-', '!') 30418 * binary arithmetic operators ('+', '-', '*', '/') 30419 * binary shift operators ('<<', '>>') 30420 * relational operators ('<', '<=', '>=', '>') 30421 * equality operators ('==', '!=') 30422 * assignment operators ('+=', '-=', '*=', '/=', '<<=', '>>=') 30423 * conversions to and from integer, floating-point, or fixed-point 30424 types 30425 30426 Use a suffix in a fixed-point literal constant: 30427 * 'hr' or 'HR' for 'short _Fract' and '_Sat short _Fract' 30428 * 'r' or 'R' for '_Fract' and '_Sat _Fract' 30429 * 'lr' or 'LR' for 'long _Fract' and '_Sat long _Fract' 30430 * 'llr' or 'LLR' for 'long long _Fract' and '_Sat long long _Fract' 30431 * 'uhr' or 'UHR' for 'unsigned short _Fract' and '_Sat unsigned short 30432 _Fract' 30433 * 'ur' or 'UR' for 'unsigned _Fract' and '_Sat unsigned _Fract' 30434 * 'ulr' or 'ULR' for 'unsigned long _Fract' and '_Sat unsigned long 30435 _Fract' 30436 * 'ullr' or 'ULLR' for 'unsigned long long _Fract' and '_Sat unsigned 30437 long long _Fract' 30438 * 'hk' or 'HK' for 'short _Accum' and '_Sat short _Accum' 30439 * 'k' or 'K' for '_Accum' and '_Sat _Accum' 30440 * 'lk' or 'LK' for 'long _Accum' and '_Sat long _Accum' 30441 * 'llk' or 'LLK' for 'long long _Accum' and '_Sat long long _Accum' 30442 * 'uhk' or 'UHK' for 'unsigned short _Accum' and '_Sat unsigned short 30443 _Accum' 30444 * 'uk' or 'UK' for 'unsigned _Accum' and '_Sat unsigned _Accum' 30445 * 'ulk' or 'ULK' for 'unsigned long _Accum' and '_Sat unsigned long 30446 _Accum' 30447 * 'ullk' or 'ULLK' for 'unsigned long long _Accum' and '_Sat unsigned 30448 long long _Accum' 30449 30450 GCC support of fixed-point types as specified by the draft technical 30451report is incomplete: 30452 30453 * Pragmas to control overflow and rounding behaviors are not 30454 implemented. 30455 30456 Fixed-point types are supported by the DWARF debug information format. 30457 30458 30459File: gcc.info, Node: Named Address Spaces, Next: Zero Length, Prev: Fixed-Point, Up: C Extensions 30460 304616.17 Named Address Spaces 30462========================= 30463 30464As an extension, GNU C supports named address spaces as defined in the 30465N1275 draft of ISO/IEC DTR 18037. Support for named address spaces in 30466GCC will evolve as the draft technical report changes. Calling 30467conventions for any target might also change. At present, only the AVR, 30468M32C, RL78, and x86 targets support address spaces other than the 30469generic address space. 30470 30471 Address space identifiers may be used exactly like any other C type 30472qualifier (e.g., 'const' or 'volatile'). See the N1275 document for 30473more details. 30474 304756.17.1 AVR Named Address Spaces 30476------------------------------- 30477 30478On the AVR target, there are several address spaces that can be used in 30479order to put read-only data into the flash memory and access that data 30480by means of the special instructions 'LPM' or 'ELPM' needed to read from 30481flash. 30482 30483 Devices belonging to 'avrtiny' and 'avrxmega3' can access flash memory 30484by means of 'LD*' instructions because the flash memory is mapped into 30485the RAM address space. There is _no need_ for language extensions like 30486'__flash' or attribute *note 'progmem': AVR Variable Attributes. The 30487default linker description files for these devices cater for that 30488feature and '.rodata' stays in flash: The compiler just generates 'LD*' 30489instructions, and the linker script adds core specific offsets to all 30490'.rodata' symbols: '0x4000' in the case of 'avrtiny' and '0x8000' in the 30491case of 'avrxmega3'. See *note AVR Options:: for a list of respective 30492devices. 30493 30494 For devices not in 'avrtiny' or 'avrxmega3', any data including 30495read-only data is located in RAM (the generic address space) because 30496flash memory is not visible in the RAM address space. In order to 30497locate read-only data in flash memory _and_ to generate the right 30498instructions to access this data without using (inline) assembler code, 30499special address spaces are needed. 30500 30501'__flash' 30502 The '__flash' qualifier locates data in the '.progmem.data' 30503 section. Data is read using the 'LPM' instruction. Pointers to 30504 this address space are 16 bits wide. 30505 30506'__flash1' 30507'__flash2' 30508'__flash3' 30509'__flash4' 30510'__flash5' 30511 These are 16-bit address spaces locating data in section 30512 '.progmemN.data' where N refers to address space '__flashN'. The 30513 compiler sets the 'RAMPZ' segment register appropriately before 30514 reading data by means of the 'ELPM' instruction. 30515 30516'__memx' 30517 This is a 24-bit address space that linearizes flash and RAM: If 30518 the high bit of the address is set, data is read from RAM using the 30519 lower two bytes as RAM address. If the high bit of the address is 30520 clear, data is read from flash with 'RAMPZ' set according to the 30521 high byte of the address. *Note '__builtin_avr_flash_segment': AVR 30522 Built-in Functions. 30523 30524 Objects in this address space are located in '.progmemx.data'. 30525 30526 Example 30527 30528 char my_read (const __flash char ** p) 30529 { 30530 /* p is a pointer to RAM that points to a pointer to flash. 30531 The first indirection of p reads that flash pointer 30532 from RAM and the second indirection reads a char from this 30533 flash address. */ 30534 30535 return **p; 30536 } 30537 30538 /* Locate array[] in flash memory */ 30539 const __flash int array[] = { 3, 5, 7, 11, 13, 17, 19 }; 30540 30541 int i = 1; 30542 30543 int main (void) 30544 { 30545 /* Return 17 by reading from flash memory */ 30546 return array[array[i]]; 30547 } 30548 30549For each named address space supported by avr-gcc there is an equally 30550named but uppercase built-in macro defined. The purpose is to 30551facilitate testing if respective address space support is available or 30552not: 30553 30554 #ifdef __FLASH 30555 const __flash int var = 1; 30556 30557 int read_var (void) 30558 { 30559 return var; 30560 } 30561 #else 30562 #include <avr/pgmspace.h> /* From AVR-LibC */ 30563 30564 const int var PROGMEM = 1; 30565 30566 int read_var (void) 30567 { 30568 return (int) pgm_read_word (&var); 30569 } 30570 #endif /* __FLASH */ 30571 30572Notice that attribute *note 'progmem': AVR Variable Attributes. locates 30573data in flash but accesses to these data read from generic address 30574space, i.e. from RAM, so that you need special accessors like 30575'pgm_read_byte' from AVR-LibC (http://nongnu.org/avr-libc/user-manual/) 30576together with attribute 'progmem'. 30577 30578Limitations and caveats 30579 30580 * Reading across the 64 KiB section boundary of the '__flash' or 30581 '__flashN' address spaces shows undefined behavior. The only 30582 address space that supports reading across the 64 KiB flash segment 30583 boundaries is '__memx'. 30584 30585 * If you use one of the '__flashN' address spaces you must arrange 30586 your linker script to locate the '.progmemN.data' sections 30587 according to your needs. 30588 30589 * Any data or pointers to the non-generic address spaces must be 30590 qualified as 'const', i.e. as read-only data. This still applies 30591 if the data in one of these address spaces like software version 30592 number or calibration lookup table are intended to be changed after 30593 load time by, say, a boot loader. In this case the right 30594 qualification is 'const' 'volatile' so that the compiler must not 30595 optimize away known values or insert them as immediates into 30596 operands of instructions. 30597 30598 * The following code initializes a variable 'pfoo' located in static 30599 storage with a 24-bit address: 30600 extern const __memx char foo; 30601 const __memx void *pfoo = &foo; 30602 30603 * On the reduced Tiny devices like ATtiny40, no address spaces are 30604 supported. Just use vanilla C / C++ code without overhead as 30605 outlined above. Attribute 'progmem' is supported but works 30606 differently, see *note AVR Variable Attributes::. 30607 306086.17.2 M32C Named Address Spaces 30609-------------------------------- 30610 30611On the M32C target, with the R8C and M16C CPU variants, variables 30612qualified with '__far' are accessed using 32-bit addresses in order to 30613access memory beyond the first 64 Ki bytes. If '__far' is used with the 30614M32CM or M32C CPU variants, it has no effect. 30615 306166.17.3 RL78 Named Address Spaces 30617-------------------------------- 30618 30619On the RL78 target, variables qualified with '__far' are accessed with 3062032-bit pointers (20-bit addresses) rather than the default 16-bit 30621addresses. Non-far variables are assumed to appear in the topmost 3062264 KiB of the address space. 30623 306246.17.4 x86 Named Address Spaces 30625------------------------------- 30626 30627On the x86 target, variables may be declared as being relative to the 30628'%fs' or '%gs' segments. 30629 30630'__seg_fs' 30631'__seg_gs' 30632 The object is accessed with the respective segment override prefix. 30633 30634 The respective segment base must be set via some method specific to 30635 the operating system. Rather than require an expensive system call 30636 to retrieve the segment base, these address spaces are not 30637 considered to be subspaces of the generic (flat) address space. 30638 This means that explicit casts are required to convert pointers 30639 between these address spaces and the generic address space. In 30640 practice the application should cast to 'uintptr_t' and apply the 30641 segment base offset that it installed previously. 30642 30643 The preprocessor symbols '__SEG_FS' and '__SEG_GS' are defined when 30644 these address spaces are supported. 30645 30646 30647File: gcc.info, Node: Zero Length, Next: Empty Structures, Prev: Named Address Spaces, Up: C Extensions 30648 306496.18 Arrays of Length Zero 30650========================== 30651 30652Declaring zero-length arrays is allowed in GNU C as an extension. A 30653zero-length array can be useful as the last element of a structure that 30654is really a header for a variable-length object: 30655 30656 struct line { 30657 int length; 30658 char contents[0]; 30659 }; 30660 30661 struct line *thisline = (struct line *) 30662 malloc (sizeof (struct line) + this_length); 30663 thisline->length = this_length; 30664 30665 Although the size of a zero-length array is zero, an array member of 30666this kind may increase the size of the enclosing type as a result of 30667tail padding. The offset of a zero-length array member from the 30668beginning of the enclosing structure is the same as the offset of an 30669array with one or more elements of the same type. The alignment of a 30670zero-length array is the same as the alignment of its elements. 30671 30672 Declaring zero-length arrays in other contexts, including as interior 30673members of structure objects or as non-member objects, is discouraged. 30674Accessing elements of zero-length arrays declared in such contexts is 30675undefined and may be diagnosed. 30676 30677 In the absence of the zero-length array extension, in ISO C90 the 30678'contents' array in the example above would typically be declared to 30679have a single element. Unlike a zero-length array which only 30680contributes to the size of the enclosing structure for the purposes of 30681alignment, a one-element array always occupies at least as much space as 30682a single object of the type. Although using one-element arrays this way 30683is discouraged, GCC handles accesses to trailing one-element array 30684members analogously to zero-length arrays. 30685 30686 The preferred mechanism to declare variable-length types like 'struct 30687line' above is the ISO C99 "flexible array member", with slightly 30688different syntax and semantics: 30689 30690 * Flexible array members are written as 'contents[]' without the '0'. 30691 30692 * Flexible array members have incomplete type, and so the 'sizeof' 30693 operator may not be applied. As a quirk of the original 30694 implementation of zero-length arrays, 'sizeof' evaluates to zero. 30695 30696 * Flexible array members may only appear as the last member of a 30697 'struct' that is otherwise non-empty. 30698 30699 * A structure containing a flexible array member, or a union 30700 containing such a structure (possibly recursively), may not be a 30701 member of a structure or an element of an array. (However, these 30702 uses are permitted by GCC as extensions.) 30703 30704 Non-empty initialization of zero-length arrays is treated like any case 30705where there are more initializer elements than the array holds, in that 30706a suitable warning about "excess elements in array" is given, and the 30707excess elements (all of them, in this case) are ignored. 30708 30709 GCC allows static initialization of flexible array members. This is 30710equivalent to defining a new structure containing the original structure 30711followed by an array of sufficient size to contain the data. E.g. in 30712the following, 'f1' is constructed as if it were declared like 'f2'. 30713 30714 struct f1 { 30715 int x; int y[]; 30716 } f1 = { 1, { 2, 3, 4 } }; 30717 30718 struct f2 { 30719 struct f1 f1; int data[3]; 30720 } f2 = { { 1 }, { 2, 3, 4 } }; 30721 30722The convenience of this extension is that 'f1' has the desired type, 30723eliminating the need to consistently refer to 'f2.f1'. 30724 30725 This has symmetry with normal static arrays, in that an array of 30726unknown size is also written with '[]'. 30727 30728 Of course, this extension only makes sense if the extra data comes at 30729the end of a top-level object, as otherwise we would be overwriting data 30730at subsequent offsets. To avoid undue complication and confusion with 30731initialization of deeply nested arrays, we simply disallow any non-empty 30732initialization except when the structure is the top-level object. For 30733example: 30734 30735 struct foo { int x; int y[]; }; 30736 struct bar { struct foo z; }; 30737 30738 struct foo a = { 1, { 2, 3, 4 } }; // Valid. 30739 struct bar b = { { 1, { 2, 3, 4 } } }; // Invalid. 30740 struct bar c = { { 1, { } } }; // Valid. 30741 struct foo d[1] = { { 1, { 2, 3, 4 } } }; // Invalid. 30742 30743 30744File: gcc.info, Node: Empty Structures, Next: Variable Length, Prev: Zero Length, Up: C Extensions 30745 307466.19 Structures with No Members 30747=============================== 30748 30749GCC permits a C structure to have no members: 30750 30751 struct empty { 30752 }; 30753 30754 The structure has size zero. In C++, empty structures are part of the 30755language. G++ treats empty structures as if they had a single member of 30756type 'char'. 30757 30758 30759File: gcc.info, Node: Variable Length, Next: Variadic Macros, Prev: Empty Structures, Up: C Extensions 30760 307616.20 Arrays of Variable Length 30762============================== 30763 30764Variable-length automatic arrays are allowed in ISO C99, and as an 30765extension GCC accepts them in C90 mode and in C++. These arrays are 30766declared like any other automatic arrays, but with a length that is not 30767a constant expression. The storage is allocated at the point of 30768declaration and deallocated when the block scope containing the 30769declaration exits. For example: 30770 30771 FILE * 30772 concat_fopen (char *s1, char *s2, char *mode) 30773 { 30774 char str[strlen (s1) + strlen (s2) + 1]; 30775 strcpy (str, s1); 30776 strcat (str, s2); 30777 return fopen (str, mode); 30778 } 30779 30780 Jumping or breaking out of the scope of the array name deallocates the 30781storage. Jumping into the scope is not allowed; you get an error 30782message for it. 30783 30784 As an extension, GCC accepts variable-length arrays as a member of a 30785structure or a union. For example: 30786 30787 void 30788 foo (int n) 30789 { 30790 struct S { int x[n]; }; 30791 } 30792 30793 You can use the function 'alloca' to get an effect much like 30794variable-length arrays. The function 'alloca' is available in many 30795other C implementations (but not in all). On the other hand, 30796variable-length arrays are more elegant. 30797 30798 There are other differences between these two methods. Space allocated 30799with 'alloca' exists until the containing _function_ returns. The space 30800for a variable-length array is deallocated as soon as the array name's 30801scope ends, unless you also use 'alloca' in this scope. 30802 30803 You can also use variable-length arrays as arguments to functions: 30804 30805 struct entry 30806 tester (int len, char data[len][len]) 30807 { 30808 /* ... */ 30809 } 30810 30811 The length of an array is computed once when the storage is allocated 30812and is remembered for the scope of the array in case you access it with 30813'sizeof'. 30814 30815 If you want to pass the array first and the length afterward, you can 30816use a forward declaration in the parameter list--another GNU extension. 30817 30818 struct entry 30819 tester (int len; char data[len][len], int len) 30820 { 30821 /* ... */ 30822 } 30823 30824 The 'int len' before the semicolon is a "parameter forward 30825declaration", and it serves the purpose of making the name 'len' known 30826when the declaration of 'data' is parsed. 30827 30828 You can write any number of such parameter forward declarations in the 30829parameter list. They can be separated by commas or semicolons, but the 30830last one must end with a semicolon, which is followed by the "real" 30831parameter declarations. Each forward declaration must match a "real" 30832declaration in parameter name and data type. ISO C99 does not support 30833parameter forward declarations. 30834 30835 30836File: gcc.info, Node: Variadic Macros, Next: Escaped Newlines, Prev: Variable Length, Up: C Extensions 30837 308386.21 Macros with a Variable Number of Arguments. 30839================================================ 30840 30841In the ISO C standard of 1999, a macro can be declared to accept a 30842variable number of arguments much as a function can. The syntax for 30843defining the macro is similar to that of a function. Here is an 30844example: 30845 30846 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__) 30847 30848Here '...' is a "variable argument". In the invocation of such a macro, 30849it represents the zero or more tokens until the closing parenthesis that 30850ends the invocation, including any commas. This set of tokens replaces 30851the identifier '__VA_ARGS__' in the macro body wherever it appears. See 30852the CPP manual for more information. 30853 30854 GCC has long supported variadic macros, and used a different syntax 30855that allowed you to give a name to the variable arguments just like any 30856other argument. Here is an example: 30857 30858 #define debug(format, args...) fprintf (stderr, format, args) 30859 30860This is in all ways equivalent to the ISO C example above, but arguably 30861more readable and descriptive. 30862 30863 GNU CPP has two further variadic macro extensions, and permits them to 30864be used with either of the above forms of macro definition. 30865 30866 In standard C, you are not allowed to leave the variable argument out 30867entirely; but you are allowed to pass an empty argument. For example, 30868this invocation is invalid in ISO C, because there is no comma after the 30869string: 30870 30871 debug ("A message") 30872 30873 GNU CPP permits you to completely omit the variable arguments in this 30874way. In the above examples, the compiler would complain, though since 30875the expansion of the macro still has the extra comma after the format 30876string. 30877 30878 To help solve this problem, CPP behaves specially for variable 30879arguments used with the token paste operator, '##'. If instead you 30880write 30881 30882 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__) 30883 30884and if the variable arguments are omitted or empty, the '##' operator 30885causes the preprocessor to remove the comma before it. If you do 30886provide some variable arguments in your macro invocation, GNU CPP does 30887not complain about the paste operation and instead places the variable 30888arguments after the comma. Just like any other pasted macro argument, 30889these arguments are not macro expanded. 30890 30891 30892File: gcc.info, Node: Escaped Newlines, Next: Subscripting, Prev: Variadic Macros, Up: C Extensions 30893 308946.22 Slightly Looser Rules for Escaped Newlines 30895=============================================== 30896 30897The preprocessor treatment of escaped newlines is more relaxed than that 30898specified by the C90 standard, which requires the newline to immediately 30899follow a backslash. GCC's implementation allows whitespace in the form 30900of spaces, horizontal and vertical tabs, and form feeds between the 30901backslash and the subsequent newline. The preprocessor issues a 30902warning, but treats it as a valid escaped newline and combines the two 30903lines to form a single logical line. This works within comments and 30904tokens, as well as between tokens. Comments are _not_ treated as 30905whitespace for the purposes of this relaxation, since they have not yet 30906been replaced with spaces. 30907 30908 30909File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Escaped Newlines, Up: C Extensions 30910 309116.23 Non-Lvalue Arrays May Have Subscripts 30912========================================== 30913 30914In ISO C99, arrays that are not lvalues still decay to pointers, and may 30915be subscripted, although they may not be modified or used after the next 30916sequence point and the unary '&' operator may not be applied to them. 30917As an extension, GNU C allows such arrays to be subscripted in C90 mode, 30918though otherwise they do not decay to pointers outside C99 mode. For 30919example, this is valid in GNU C though not valid in C90: 30920 30921 struct foo {int a[4];}; 30922 30923 struct foo f(); 30924 30925 bar (int index) 30926 { 30927 return f().a[index]; 30928 } 30929 30930 30931File: gcc.info, Node: Pointer Arith, Next: Variadic Pointer Args, Prev: Subscripting, Up: C Extensions 30932 309336.24 Arithmetic on 'void'- and Function-Pointers 30934================================================ 30935 30936In GNU C, addition and subtraction operations are supported on pointers 30937to 'void' and on pointers to functions. This is done by treating the 30938size of a 'void' or of a function as 1. 30939 30940 A consequence of this is that 'sizeof' is also allowed on 'void' and on 30941function types, and returns 1. 30942 30943 The option '-Wpointer-arith' requests a warning if these extensions are 30944used. 30945 30946 30947File: gcc.info, Node: Variadic Pointer Args, Next: Pointers to Arrays, Prev: Pointer Arith, Up: C Extensions 30948 309496.25 Pointer Arguments in Variadic Functions 30950============================================ 30951 30952Standard C requires that pointer types used with 'va_arg' in functions 30953with variable argument lists either must be compatible with that of the 30954actual argument, or that one type must be a pointer to 'void' and the 30955other a pointer to a character type. GNU C implements the POSIX XSI 30956extension that additionally permits the use of 'va_arg' with a pointer 30957type to receive arguments of any other pointer type. 30958 30959 In particular, in GNU C 'va_arg (ap, void *)' can safely be used to 30960consume an argument of any pointer type. 30961 30962 30963File: gcc.info, Node: Pointers to Arrays, Next: Initializers, Prev: Variadic Pointer Args, Up: C Extensions 30964 309656.26 Pointers to Arrays with Qualifiers Work as Expected 30966======================================================== 30967 30968In GNU C, pointers to arrays with qualifiers work similar to pointers to 30969other qualified types. For example, a value of type 'int (*)[5]' can be 30970used to initialize a variable of type 'const int (*)[5]'. These types 30971are incompatible in ISO C because the 'const' qualifier is formally 30972attached to the element type of the array and not the array itself. 30973 30974 extern void 30975 transpose (int N, int M, double out[M][N], const double in[N][M]); 30976 double x[3][2]; 30977 double y[2][3]; 30978 ... 30979 transpose(3, 2, y, x); 30980 30981 30982File: gcc.info, Node: Initializers, Next: Compound Literals, Prev: Pointers to Arrays, Up: C Extensions 30983 309846.27 Non-Constant Initializers 30985============================== 30986 30987As in standard C++ and ISO C99, the elements of an aggregate initializer 30988for an automatic variable are not required to be constant expressions in 30989GNU C. Here is an example of an initializer with run-time varying 30990elements: 30991 30992 foo (float f, float g) 30993 { 30994 float beat_freqs[2] = { f-g, f+g }; 30995 /* ... */ 30996 } 30997 30998 30999File: gcc.info, Node: Compound Literals, Next: Designated Inits, Prev: Initializers, Up: C Extensions 31000 310016.28 Compound Literals 31002====================== 31003 31004A compound literal looks like a cast of a brace-enclosed aggregate 31005initializer list. Its value is an object of the type specified in the 31006cast, containing the elements specified in the initializer. Unlike the 31007result of a cast, a compound literal is an lvalue. ISO C99 and later 31008support compound literals. As an extension, GCC supports compound 31009literals also in C90 mode and in C++, although as explained below, the 31010C++ semantics are somewhat different. 31011 31012 Usually, the specified type of a compound literal is a structure. 31013Assume that 'struct foo' and 'structure' are declared as shown: 31014 31015 struct foo {int a; char b[2];} structure; 31016 31017Here is an example of constructing a 'struct foo' with a compound 31018literal: 31019 31020 structure = ((struct foo) {x + y, 'a', 0}); 31021 31022This is equivalent to writing the following: 31023 31024 { 31025 struct foo temp = {x + y, 'a', 0}; 31026 structure = temp; 31027 } 31028 31029 You can also construct an array, though this is dangerous in C++, as 31030explained below. If all the elements of the compound literal are (made 31031up of) simple constant expressions suitable for use in initializers of 31032objects of static storage duration, then the compound literal can be 31033coerced to a pointer to its first element and used in such an 31034initializer, as shown here: 31035 31036 char **foo = (char *[]) { "x", "y", "z" }; 31037 31038 Compound literals for scalar types and union types are also allowed. 31039In the following example the variable 'i' is initialized to the value 31040'2', the result of incrementing the unnamed object created by the 31041compound literal. 31042 31043 int i = ++(int) { 1 }; 31044 31045 As a GNU extension, GCC allows initialization of objects with static 31046storage duration by compound literals (which is not possible in ISO C99 31047because the initializer is not a constant). It is handled as if the 31048object were initialized only with the brace-enclosed list if the types 31049of the compound literal and the object match. The elements of the 31050compound literal must be constant. If the object being initialized has 31051array type of unknown size, the size is determined by the size of the 31052compound literal. 31053 31054 static struct foo x = (struct foo) {1, 'a', 'b'}; 31055 static int y[] = (int []) {1, 2, 3}; 31056 static int z[] = (int [3]) {1}; 31057 31058The above lines are equivalent to the following: 31059 static struct foo x = {1, 'a', 'b'}; 31060 static int y[] = {1, 2, 3}; 31061 static int z[] = {1, 0, 0}; 31062 31063 In C, a compound literal designates an unnamed object with static or 31064automatic storage duration. In C++, a compound literal designates a 31065temporary object that only lives until the end of its full-expression. 31066As a result, well-defined C code that takes the address of a subobject 31067of a compound literal can be undefined in C++, so G++ rejects the 31068conversion of a temporary array to a pointer. For instance, if the 31069array compound literal example above appeared inside a function, any 31070subsequent use of 'foo' in C++ would have undefined behavior because the 31071lifetime of the array ends after the declaration of 'foo'. 31072 31073 As an optimization, G++ sometimes gives array compound literals longer 31074lifetimes: when the array either appears outside a function or has a 31075'const'-qualified type. If 'foo' and its initializer had elements of 31076type 'char *const' rather than 'char *', or if 'foo' were a global 31077variable, the array would have static storage duration. But it is 31078probably safest just to avoid the use of array compound literals in C++ 31079code. 31080 31081 31082File: gcc.info, Node: Designated Inits, Next: Case Ranges, Prev: Compound Literals, Up: C Extensions 31083 310846.29 Designated Initializers 31085============================ 31086 31087Standard C90 requires the elements of an initializer to appear in a 31088fixed order, the same as the order of the elements in the array or 31089structure being initialized. 31090 31091 In ISO C99 you can give the elements in any order, specifying the array 31092indices or structure field names they apply to, and GNU C allows this as 31093an extension in C90 mode as well. This extension is not implemented in 31094GNU C++. 31095 31096 To specify an array index, write '[INDEX] =' before the element value. 31097For example, 31098 31099 int a[6] = { [4] = 29, [2] = 15 }; 31100 31101is equivalent to 31102 31103 int a[6] = { 0, 0, 15, 0, 29, 0 }; 31104 31105The index values must be constant expressions, even if the array being 31106initialized is automatic. 31107 31108 An alternative syntax for this that has been obsolete since GCC 2.5 but 31109GCC still accepts is to write '[INDEX]' before the element value, with 31110no '='. 31111 31112 To initialize a range of elements to the same value, write '[FIRST ... 31113LAST] = VALUE'. This is a GNU extension. For example, 31114 31115 int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 }; 31116 31117If the value in it has side effects, the side effects happen only once, 31118not for each initialized field by the range initializer. 31119 31120Note that the length of the array is the highest value specified plus 31121one. 31122 31123 In a structure initializer, specify the name of a field to initialize 31124with '.FIELDNAME =' before the element value. For example, given the 31125following structure, 31126 31127 struct point { int x, y; }; 31128 31129the following initialization 31130 31131 struct point p = { .y = yvalue, .x = xvalue }; 31132 31133is equivalent to 31134 31135 struct point p = { xvalue, yvalue }; 31136 31137 Another syntax that has the same meaning, obsolete since GCC 2.5, is 31138'FIELDNAME:', as shown here: 31139 31140 struct point p = { y: yvalue, x: xvalue }; 31141 31142 Omitted fields are implicitly initialized the same as for objects that 31143have static storage duration. 31144 31145 The '[INDEX]' or '.FIELDNAME' is known as a "designator". You can also 31146use a designator (or the obsolete colon syntax) when initializing a 31147union, to specify which element of the union should be used. For 31148example, 31149 31150 union foo { int i; double d; }; 31151 31152 union foo f = { .d = 4 }; 31153 31154converts 4 to a 'double' to store it in the union using the second 31155element. By contrast, casting 4 to type 'union foo' stores it into the 31156union as the integer 'i', since it is an integer. *Note Cast to 31157Union::. 31158 31159 You can combine this technique of naming elements with ordinary C 31160initialization of successive elements. Each initializer element that 31161does not have a designator applies to the next consecutive element of 31162the array or structure. For example, 31163 31164 int a[6] = { [1] = v1, v2, [4] = v4 }; 31165 31166is equivalent to 31167 31168 int a[6] = { 0, v1, v2, 0, v4, 0 }; 31169 31170 Labeling the elements of an array initializer is especially useful when 31171the indices are characters or belong to an 'enum' type. For example: 31172 31173 int whitespace[256] 31174 = { [' '] = 1, ['\t'] = 1, ['\h'] = 1, 31175 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 }; 31176 31177 You can also write a series of '.FIELDNAME' and '[INDEX]' designators 31178before an '=' to specify a nested subobject to initialize; the list is 31179taken relative to the subobject corresponding to the closest surrounding 31180brace pair. For example, with the 'struct point' declaration above: 31181 31182 struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 }; 31183 31184 If the same field is initialized multiple times, or overlapping fields 31185of a union are initialized, the value from the last initialization is 31186used. When a field of a union is itself a structure, the entire 31187structure from the last field initialized is used. If any previous 31188initializer has side effect, it is unspecified whether the side effect 31189happens or not. Currently, GCC discards the side-effecting initializer 31190expressions and issues a warning. 31191 31192 31193File: gcc.info, Node: Case Ranges, Next: Cast to Union, Prev: Designated Inits, Up: C Extensions 31194 311956.30 Case Ranges 31196================ 31197 31198You can specify a range of consecutive values in a single 'case' label, 31199like this: 31200 31201 case LOW ... HIGH: 31202 31203This has the same effect as the proper number of individual 'case' 31204labels, one for each integer value from LOW to HIGH, inclusive. 31205 31206 This feature is especially useful for ranges of ASCII character codes: 31207 31208 case 'A' ... 'Z': 31209 31210 *Be careful:* Write spaces around the '...', for otherwise it may be 31211parsed wrong when you use it with integer values. For example, write 31212this: 31213 31214 case 1 ... 5: 31215 31216rather than this: 31217 31218 case 1...5: 31219 31220 31221File: gcc.info, Node: Cast to Union, Next: Mixed Declarations, Prev: Case Ranges, Up: C Extensions 31222 312236.31 Cast to a Union Type 31224========================= 31225 31226A cast to a union type is a C extension not available in C++. It looks 31227just like ordinary casts with the constraint that the type specified is 31228a union type. You can specify the type either with the 'union' keyword 31229or with a 'typedef' name that refers to a union. The result of a cast 31230to a union is a temporary rvalue of the union type with a member whose 31231type matches that of the operand initialized to the value of the 31232operand. The effect of a cast to a union is similar to a compound 31233literal except that it yields an rvalue like standard casts do. *Note 31234Compound Literals::. 31235 31236 Expressions that may be cast to the union type are those whose type 31237matches at least one of the members of the union. Thus, given the 31238following union and variables: 31239 31240 union foo { int i; double d; }; 31241 int x; 31242 double y; 31243 union foo z; 31244 31245both 'x' and 'y' can be cast to type 'union foo' and the following 31246assignments 31247 z = (union foo) x; 31248 z = (union foo) y; 31249 are shorthand equivalents of these 31250 z = (union foo) { .i = x }; 31251 z = (union foo) { .d = y }; 31252 31253 However, '(union foo) FLT_MAX;' is not a valid cast because the union 31254has no member of type 'float'. 31255 31256 Using the cast as the right-hand side of an assignment to a variable of 31257union type is equivalent to storing in a member of the union with the 31258same type 31259 31260 union foo u; 31261 /* ... */ 31262 u = (union foo) x == u.i = x 31263 u = (union foo) y == u.d = y 31264 31265 You can also use the union cast as a function argument: 31266 31267 void hack (union foo); 31268 /* ... */ 31269 hack ((union foo) x); 31270 31271 31272File: gcc.info, Node: Mixed Declarations, Next: Function Attributes, Prev: Cast to Union, Up: C Extensions 31273 312746.32 Mixed Declarations and Code 31275================================ 31276 31277ISO C99 and ISO C++ allow declarations and code to be freely mixed 31278within compound statements. As an extension, GNU C also allows this in 31279C90 mode. For example, you could do: 31280 31281 int i; 31282 /* ... */ 31283 i++; 31284 int j = i + 2; 31285 31286 Each identifier is visible from where it is declared until the end of 31287the enclosing block. 31288 31289 31290File: gcc.info, Node: Function Attributes, Next: Variable Attributes, Prev: Mixed Declarations, Up: C Extensions 31291 312926.33 Declaring Attributes of Functions 31293====================================== 31294 31295In GNU C and C++, you can use function attributes to specify certain 31296function properties that may help the compiler optimize calls or check 31297code more carefully for correctness. For example, you can use 31298attributes to specify that a function never returns ('noreturn'), 31299returns a value depending only on the values of its arguments ('const'), 31300or has 'printf'-style arguments ('format'). 31301 31302 You can also use attributes to control memory placement, code 31303generation options or call/return conventions within the function being 31304annotated. Many of these attributes are target-specific. For example, 31305many targets support attributes for defining interrupt handler 31306functions, which typically must follow special register usage and return 31307conventions. Such attributes are described in the subsection for each 31308target. However, a considerable number of attributes are supported by 31309most, if not all targets. Those are described in the *note Common 31310Function Attributes:: section. 31311 31312 Function attributes are introduced by the '__attribute__' keyword in 31313the declaration of a function, followed by an attribute specification 31314enclosed in double parentheses. You can specify multiple attributes in 31315a declaration by separating them by commas within the double parentheses 31316or by immediately following one attribute specification with another. 31317*Note Attribute Syntax::, for the exact rules on attribute syntax and 31318placement. Compatible attribute specifications on distinct declarations 31319of the same function are merged. An attribute specification that is not 31320compatible with attributes already applied to a declaration of the same 31321function is ignored with a warning. 31322 31323 Some function attributes take one or more arguments that refer to the 31324function's parameters by their positions within the function parameter 31325list. Such attribute arguments are referred to as "positional 31326arguments". Unless specified otherwise, positional arguments that 31327specify properties of parameters with pointer types can also specify the 31328same properties of the implicit C++ 'this' argument in non-static member 31329functions, and of parameters of reference to a pointer type. For 31330ordinary functions, position one refers to the first parameter on the 31331list. In C++ non-static member functions, position one refers to the 31332implicit 'this' pointer. The same restrictions and effects apply to 31333function attributes used with ordinary functions or C++ member 31334functions. 31335 31336 GCC also supports attributes on variable declarations (*note Variable 31337Attributes::), labels (*note Label Attributes::), enumerators (*note 31338Enumerator Attributes::), statements (*note Statement Attributes::), and 31339types (*note Type Attributes::). 31340 31341 There is some overlap between the purposes of attributes and pragmas 31342(*note Pragmas Accepted by GCC: Pragmas.). It has been found convenient 31343to use '__attribute__' to achieve a natural attachment of attributes to 31344their corresponding declarations, whereas '#pragma' is of use for 31345compatibility with other compilers or constructs that do not naturally 31346form part of the grammar. 31347 31348 In addition to the attributes documented here, GCC plugins may provide 31349their own attributes. 31350 31351* Menu: 31352 31353* Common Function Attributes:: 31354* AArch64 Function Attributes:: 31355* AMD GCN Function Attributes:: 31356* ARC Function Attributes:: 31357* ARM Function Attributes:: 31358* AVR Function Attributes:: 31359* Blackfin Function Attributes:: 31360* BPF Function Attributes:: 31361* CR16 Function Attributes:: 31362* C-SKY Function Attributes:: 31363* Epiphany Function Attributes:: 31364* H8/300 Function Attributes:: 31365* IA-64 Function Attributes:: 31366* M32C Function Attributes:: 31367* M32R/D Function Attributes:: 31368* m68k Function Attributes:: 31369* MCORE Function Attributes:: 31370* MeP Function Attributes:: 31371* MicroBlaze Function Attributes:: 31372* Microsoft Windows Function Attributes:: 31373* MIPS Function Attributes:: 31374* MSP430 Function Attributes:: 31375* NDS32 Function Attributes:: 31376* Nios II Function Attributes:: 31377* Nvidia PTX Function Attributes:: 31378* PowerPC Function Attributes:: 31379* RISC-V Function Attributes:: 31380* RL78 Function Attributes:: 31381* RX Function Attributes:: 31382* S/390 Function Attributes:: 31383* SH Function Attributes:: 31384* Symbian OS Function Attributes:: 31385* V850 Function Attributes:: 31386* Visium Function Attributes:: 31387* x86 Function Attributes:: 31388* Xstormy16 Function Attributes:: 31389 31390 31391File: gcc.info, Node: Common Function Attributes, Next: AArch64 Function Attributes, Up: Function Attributes 31392 313936.33.1 Common Function Attributes 31394--------------------------------- 31395 31396The following attributes are supported on most targets. 31397 31398'access' 31399'access (ACCESS-MODE, REF-INDEX)' 31400'access (ACCESS-MODE, REF-INDEX, SIZE-INDEX)' 31401 31402 The 'access' attribute enables the detection of invalid or unsafe 31403 accesses by functions to which they apply or their callers, as well 31404 as write-only accesses to objects that are never read from. Such 31405 accesses may be diagnosed by warnings such as 31406 '-Wstringop-overflow', '-Wuninitialized', '-Wunused', and others. 31407 31408 The 'access' attribute specifies that a function to whose 31409 by-reference arguments the attribute applies accesses the 31410 referenced object according to ACCESS-MODE. The ACCESS-MODE 31411 argument is required and must be one of three names: 'read_only', 31412 'read_write', or 'write_only'. The remaining two are positional 31413 arguments. 31414 31415 The required REF-INDEX positional argument denotes a function 31416 argument of pointer (or in C++, reference) type that is subject to 31417 the access. The same pointer argument can be referenced by at most 31418 one distinct 'access' attribute. 31419 31420 The optional SIZE-INDEX positional argument denotes a function 31421 argument of integer type that specifies the maximum size of the 31422 access. The size is the number of elements of the type referenced 31423 by REF-INDEX, or the number of bytes when the pointer type is 31424 'void*'. When no SIZE-INDEX argument is specified, the pointer 31425 argument must be either null or point to a space that is suitably 31426 aligned and large for at least one object of the referenced type 31427 (this implies that a past-the-end pointer is not a valid argument). 31428 The actual size of the access may be less but it must not be more. 31429 31430 The 'read_only' access mode specifies that the pointer to which it 31431 applies is used to read the referenced object but not write to it. 31432 Unless the argument specifying the size of the access denoted by 31433 SIZE-INDEX is zero, the referenced object must be initialized. The 31434 mode implies a stronger guarantee than the 'const' qualifier which, 31435 when cast away from a pointer, does not prevent the pointed-to 31436 object from being modified. Examples of the use of the 'read_only' 31437 access mode is the argument to the 'puts' function, or the second 31438 and third arguments to the 'memcpy' function. 31439 31440 __attribute__ ((access (read_only, 1))) int puts (const char*); 31441 __attribute__ ((access (read_only, 1, 2))) void* memcpy (void*, const void*, size_t); 31442 31443 The 'read_write' access mode applies to arguments of pointer types 31444 without the 'const' qualifier. It specifies that the pointer to 31445 which it applies is used to both read and write the referenced 31446 object. Unless the argument specifying the size of the access 31447 denoted by SIZE-INDEX is zero, the object referenced by the pointer 31448 must be initialized. An example of the use of the 'read_write' 31449 access mode is the first argument to the 'strcat' function. 31450 31451 __attribute__ ((access (read_write, 1), access (read_only, 2))) char* strcat (char*, const char*); 31452 31453 The 'write_only' access mode applies to arguments of pointer types 31454 without the 'const' qualifier. It specifies that the pointer to 31455 which it applies is used to write to the referenced object but not 31456 read from it. The object referenced by the pointer need not be 31457 initialized. An example of the use of the 'write_only' access mode 31458 is the first argument to the 'strcpy' function, or the first two 31459 arguments to the 'fgets' function. 31460 31461 __attribute__ ((access (write_only, 1), access (read_only, 2))) char* strcpy (char*, const char*); 31462 __attribute__ ((access (write_only, 1, 2), access (read_write, 3))) int fgets (char*, int, FILE*); 31463 31464'alias ("TARGET")' 31465 The 'alias' attribute causes the declaration to be emitted as an 31466 alias for another symbol, which must have been previously declared 31467 with the same type, and for variables, also the same size and 31468 alignment. Declaring an alias with a different type than the 31469 target is undefined and may be diagnosed. As an example, the 31470 following declarations: 31471 31472 void __f () { /* Do something. */; } 31473 void f () __attribute__ ((weak, alias ("__f"))); 31474 31475 define 'f' to be a weak alias for '__f'. In C++, the mangled name 31476 for the target must be used. It is an error if '__f' is not 31477 defined in the same translation unit. 31478 31479 This attribute requires assembler and object file support, and may 31480 not be available on all targets. 31481 31482'aligned' 31483'aligned (ALIGNMENT)' 31484 The 'aligned' attribute specifies a minimum alignment for the first 31485 instruction of the function, measured in bytes. When specified, 31486 ALIGNMENT must be an integer constant power of 2. Specifying no 31487 ALIGNMENT argument implies the ideal alignment for the target. The 31488 '__alignof__' operator can be used to determine what that is (*note 31489 Alignment::). The attribute has no effect when a definition for 31490 the function is not provided in the same translation unit. 31491 31492 The attribute cannot be used to decrease the alignment of a 31493 function previously declared with a more restrictive alignment; 31494 only to increase it. Attempts to do otherwise are diagnosed. Some 31495 targets specify a minimum default alignment for functions that is 31496 greater than 1. On such targets, specifying a less restrictive 31497 alignment is silently ignored. Using the attribute overrides the 31498 effect of the '-falign-functions' (*note Optimize Options::) option 31499 for this function. 31500 31501 Note that the effectiveness of 'aligned' attributes may be limited 31502 by inherent limitations in the system linker and/or object file 31503 format. On some systems, the linker is only able to arrange for 31504 functions to be aligned up to a certain maximum alignment. (For 31505 some linkers, the maximum supported alignment may be very very 31506 small.) See your linker documentation for further information. 31507 31508 The 'aligned' attribute can also be used for variables and fields 31509 (*note Variable Attributes::.) 31510 31511'alloc_align (POSITION)' 31512 The 'alloc_align' attribute may be applied to a function that 31513 returns a pointer and takes at least one argument of an integer or 31514 enumerated type. It indicates that the returned pointer is aligned 31515 on a boundary given by the function argument at POSITION. 31516 Meaningful alignments are powers of 2 greater than one. GCC uses 31517 this information to improve pointer alignment analysis. 31518 31519 The function parameter denoting the allocated alignment is 31520 specified by one constant integer argument whose number is the 31521 argument of the attribute. Argument numbering starts at one. 31522 31523 For instance, 31524 31525 void* my_memalign (size_t, size_t) __attribute__ ((alloc_align (1))); 31526 31527 declares that 'my_memalign' returns memory with minimum alignment 31528 given by parameter 1. 31529 31530'alloc_size (POSITION)' 31531'alloc_size (POSITION-1, POSITION-2)' 31532 The 'alloc_size' attribute may be applied to a function that 31533 returns a pointer and takes at least one argument of an integer or 31534 enumerated type. It indicates that the returned pointer points to 31535 memory whose size is given by the function argument at POSITION-1, 31536 or by the product of the arguments at POSITION-1 and POSITION-2. 31537 Meaningful sizes are positive values less than 'PTRDIFF_MAX'. GCC 31538 uses this information to improve the results of 31539 '__builtin_object_size'. 31540 31541 The function parameter(s) denoting the allocated size are specified 31542 by one or two integer arguments supplied to the attribute. The 31543 allocated size is either the value of the single function argument 31544 specified or the product of the two function arguments specified. 31545 Argument numbering starts at one for ordinary functions, and at two 31546 for C++ non-static member functions. 31547 31548 For instance, 31549 31550 void* my_calloc (size_t, size_t) __attribute__ ((alloc_size (1, 2))); 31551 void* my_realloc (void*, size_t) __attribute__ ((alloc_size (2))); 31552 31553 declares that 'my_calloc' returns memory of the size given by the 31554 product of parameter 1 and 2 and that 'my_realloc' returns memory 31555 of the size given by parameter 2. 31556 31557'always_inline' 31558 Generally, functions are not inlined unless optimization is 31559 specified. For functions declared inline, this attribute inlines 31560 the function independent of any restrictions that otherwise apply 31561 to inlining. Failure to inline such a function is diagnosed as an 31562 error. Note that if such a function is called indirectly the 31563 compiler may or may not inline it depending on optimization level 31564 and a failure to inline an indirect call may or may not be 31565 diagnosed. 31566 31567'artificial' 31568 This attribute is useful for small inline wrappers that if possible 31569 should appear during debugging as a unit. Depending on the debug 31570 info format it either means marking the function as artificial or 31571 using the caller location for all instructions within the inlined 31572 body. 31573 31574'assume_aligned (ALIGNMENT)' 31575'assume_aligned (ALIGNMENT, OFFSET)' 31576 The 'assume_aligned' attribute may be applied to a function that 31577 returns a pointer. It indicates that the returned pointer is 31578 aligned on a boundary given by ALIGNMENT. If the attribute has two 31579 arguments, the second argument is misalignment OFFSET. Meaningful 31580 values of ALIGNMENT are powers of 2 greater than one. Meaningful 31581 values of OFFSET are greater than zero and less than ALIGNMENT. 31582 31583 For instance 31584 31585 void* my_alloc1 (size_t) __attribute__((assume_aligned (16))); 31586 void* my_alloc2 (size_t) __attribute__((assume_aligned (32, 8))); 31587 31588 declares that 'my_alloc1' returns 16-byte aligned pointers and that 31589 'my_alloc2' returns a pointer whose value modulo 32 is equal to 8. 31590 31591'cold' 31592 The 'cold' attribute on functions is used to inform the compiler 31593 that the function is unlikely to be executed. The function is 31594 optimized for size rather than speed and on many targets it is 31595 placed into a special subsection of the text section so all cold 31596 functions appear close together, improving code locality of 31597 non-cold parts of program. The paths leading to calls of cold 31598 functions within code are marked as unlikely by the branch 31599 prediction mechanism. It is thus useful to mark functions used to 31600 handle unlikely conditions, such as 'perror', as cold to improve 31601 optimization of hot functions that do call marked functions in rare 31602 occasions. 31603 31604 When profile feedback is available, via '-fprofile-use', cold 31605 functions are automatically detected and this attribute is ignored. 31606 31607'const' 31608 Calls to functions whose return value is not affected by changes to 31609 the observable state of the program and that have no observable 31610 effects on such state other than to return a value may lend 31611 themselves to optimizations such as common subexpression 31612 elimination. Declaring such functions with the 'const' attribute 31613 allows GCC to avoid emitting some calls in repeated invocations of 31614 the function with the same argument values. 31615 31616 For example, 31617 31618 int square (int) __attribute__ ((const)); 31619 31620 tells GCC that subsequent calls to function 'square' with the same 31621 argument value can be replaced by the result of the first call 31622 regardless of the statements in between. 31623 31624 The 'const' attribute prohibits a function from reading objects 31625 that affect its return value between successive invocations. 31626 However, functions declared with the attribute can safely read 31627 objects that do not change their return value, such as non-volatile 31628 constants. 31629 31630 The 'const' attribute imposes greater restrictions on a function's 31631 definition than the similar 'pure' attribute. Declaring the same 31632 function with both the 'const' and the 'pure' attribute is 31633 diagnosed. Because a const function cannot have any observable 31634 side effects it does not make sense for it to return 'void'. 31635 Declaring such a function is diagnosed. 31636 31637 Note that a function that has pointer arguments and examines the 31638 data pointed to must _not_ be declared 'const' if the pointed-to 31639 data might change between successive invocations of the function. 31640 In general, since a function cannot distinguish data that might 31641 change from data that cannot, const functions should never take 31642 pointer or, in C++, reference arguments. Likewise, a function that 31643 calls a non-const function usually must not be const itself. 31644 31645'constructor' 31646'destructor' 31647'constructor (PRIORITY)' 31648'destructor (PRIORITY)' 31649 The 'constructor' attribute causes the function to be called 31650 automatically before execution enters 'main ()'. Similarly, the 31651 'destructor' attribute causes the function to be called 31652 automatically after 'main ()' completes or 'exit ()' is called. 31653 Functions with these attributes are useful for initializing data 31654 that is used implicitly during the execution of the program. 31655 31656 On some targets the attributes also accept an integer argument to 31657 specify a priority to control the order in which constructor and 31658 destructor functions are run. A constructor with a smaller 31659 priority number runs before a constructor with a larger priority 31660 number; the opposite relationship holds for destructors. So, if 31661 you have a constructor that allocates a resource and a destructor 31662 that deallocates the same resource, both functions typically have 31663 the same priority. The priorities for constructor and destructor 31664 functions are the same as those specified for namespace-scope C++ 31665 objects (*note C++ Attributes::). However, at present, the order 31666 in which constructors for C++ objects with static storage duration 31667 and functions decorated with attribute 'constructor' are invoked is 31668 unspecified. In mixed declarations, attribute 'init_priority' can 31669 be used to impose a specific ordering. 31670 31671 Using the argument forms of the 'constructor' and 'destructor' 31672 attributes on targets where the feature is not supported is 31673 rejected with an error. 31674 31675'copy' 31676'copy (FUNCTION)' 31677 The 'copy' attribute applies the set of attributes with which 31678 FUNCTION has been declared to the declaration of the function to 31679 which the attribute is applied. The attribute is designed for 31680 libraries that define aliases or function resolvers that are 31681 expected to specify the same set of attributes as their targets. 31682 The 'copy' attribute can be used with functions, variables, or 31683 types. However, the kind of symbol to which the attribute is 31684 applied (either function or variable) must match the kind of symbol 31685 to which the argument refers. The 'copy' attribute copies only 31686 syntactic and semantic attributes but not attributes that affect a 31687 symbol's linkage or visibility such as 'alias', 'visibility', or 31688 'weak'. The 'deprecated' and 'target_clones' attribute are also 31689 not copied. *Note Common Type Attributes::. *Note Common Variable 31690 Attributes::. 31691 31692 For example, the STRONGALIAS macro below makes use of the 'alias' 31693 and 'copy' attributes to define an alias named ALLOC for function 31694 ALLOCATE declared with attributes ALLOC_SIZE, MALLOC, and NOTHROW. 31695 Thanks to the '__typeof__' operator the alias has the same type as 31696 the target function. As a result of the 'copy' attribute the alias 31697 also shares the same attributes as the target. 31698 31699 #define StrongAlias(TargetFunc, AliasDecl) \ 31700 extern __typeof__ (TargetFunc) AliasDecl \ 31701 __attribute__ ((alias (#TargetFunc), copy (TargetFunc))); 31702 31703 extern __attribute__ ((alloc_size (1), malloc, nothrow)) 31704 void* allocate (size_t); 31705 StrongAlias (allocate, alloc); 31706 31707'deprecated' 31708'deprecated (MSG)' 31709 The 'deprecated' attribute results in a warning if the function is 31710 used anywhere in the source file. This is useful when identifying 31711 functions that are expected to be removed in a future version of a 31712 program. The warning also includes the location of the declaration 31713 of the deprecated function, to enable users to easily find further 31714 information about why the function is deprecated, or what they 31715 should do instead. Note that the warnings only occurs for uses: 31716 31717 int old_fn () __attribute__ ((deprecated)); 31718 int old_fn (); 31719 int (*fn_ptr)() = old_fn; 31720 31721 results in a warning on line 3 but not line 2. The optional MSG 31722 argument, which must be a string, is printed in the warning if 31723 present. 31724 31725 The 'deprecated' attribute can also be used for variables and types 31726 (*note Variable Attributes::, *note Type Attributes::.) 31727 31728 The message attached to the attribute is affected by the setting of 31729 the '-fmessage-length' option. 31730 31731'error ("MESSAGE")' 31732'warning ("MESSAGE")' 31733 If the 'error' or 'warning' attribute is used on a function 31734 declaration and a call to such a function is not eliminated through 31735 dead code elimination or other optimizations, an error or warning 31736 (respectively) that includes MESSAGE is diagnosed. This is useful 31737 for compile-time checking, especially together with 31738 '__builtin_constant_p' and inline functions where checking the 31739 inline function arguments is not possible through 'extern char 31740 [(condition) ? 1 : -1];' tricks. 31741 31742 While it is possible to leave the function undefined and thus 31743 invoke a link failure (to define the function with a message in 31744 '.gnu.warning*' section), when using these attributes the problem 31745 is diagnosed earlier and with exact location of the call even in 31746 presence of inline functions or when not emitting debugging 31747 information. 31748 31749'externally_visible' 31750 This attribute, attached to a global variable or function, 31751 nullifies the effect of the '-fwhole-program' command-line option, 31752 so the object remains visible outside the current compilation unit. 31753 31754 If '-fwhole-program' is used together with '-flto' and 'gold' is 31755 used as the linker plugin, 'externally_visible' attributes are 31756 automatically added to functions (not variable yet due to a current 31757 'gold' issue) that are accessed outside of LTO objects according to 31758 resolution file produced by 'gold'. For other linkers that cannot 31759 generate resolution file, explicit 'externally_visible' attributes 31760 are still necessary. 31761 31762'flatten' 31763 Generally, inlining into a function is limited. For a function 31764 marked with this attribute, every call inside this function is 31765 inlined, if possible. Functions declared with attribute 'noinline' 31766 and similar are not inlined. Whether the function itself is 31767 considered for inlining depends on its size and the current 31768 inlining parameters. 31769 31770'format (ARCHETYPE, STRING-INDEX, FIRST-TO-CHECK)' 31771 The 'format' attribute specifies that a function takes 'printf', 31772 'scanf', 'strftime' or 'strfmon' style arguments that should be 31773 type-checked against a format string. For example, the 31774 declaration: 31775 31776 extern int 31777 my_printf (void *my_object, const char *my_format, ...) 31778 __attribute__ ((format (printf, 2, 3))); 31779 31780 causes the compiler to check the arguments in calls to 'my_printf' 31781 for consistency with the 'printf' style format string argument 31782 'my_format'. 31783 31784 The parameter ARCHETYPE determines how the format string is 31785 interpreted, and should be 'printf', 'scanf', 'strftime', 31786 'gnu_printf', 'gnu_scanf', 'gnu_strftime' or 'strfmon'. (You can 31787 also use '__printf__', '__scanf__', '__strftime__' or 31788 '__strfmon__'.) On MinGW targets, 'ms_printf', 'ms_scanf', and 31789 'ms_strftime' are also present. ARCHETYPE values such as 'printf' 31790 refer to the formats accepted by the system's C runtime library, 31791 while values prefixed with 'gnu_' always refer to the formats 31792 accepted by the GNU C Library. On Microsoft Windows targets, 31793 values prefixed with 'ms_' refer to the formats accepted by the 31794 'msvcrt.dll' library. The parameter STRING-INDEX specifies which 31795 argument is the format string argument (starting from 1), while 31796 FIRST-TO-CHECK is the number of the first argument to check against 31797 the format string. For functions where the arguments are not 31798 available to be checked (such as 'vprintf'), specify the third 31799 parameter as zero. In this case the compiler only checks the 31800 format string for consistency. For 'strftime' formats, the third 31801 parameter is required to be zero. Since non-static C++ methods 31802 have an implicit 'this' argument, the arguments of such methods 31803 should be counted from two, not one, when giving values for 31804 STRING-INDEX and FIRST-TO-CHECK. 31805 31806 In the example above, the format string ('my_format') is the second 31807 argument of the function 'my_print', and the arguments to check 31808 start with the third argument, so the correct parameters for the 31809 format attribute are 2 and 3. 31810 31811 The 'format' attribute allows you to identify your own functions 31812 that take format strings as arguments, so that GCC can check the 31813 calls to these functions for errors. The compiler always (unless 31814 '-ffreestanding' or '-fno-builtin' is used) checks formats for the 31815 standard library functions 'printf', 'fprintf', 'sprintf', 'scanf', 31816 'fscanf', 'sscanf', 'strftime', 'vprintf', 'vfprintf' and 31817 'vsprintf' whenever such warnings are requested (using '-Wformat'), 31818 so there is no need to modify the header file 'stdio.h'. In C99 31819 mode, the functions 'snprintf', 'vsnprintf', 'vscanf', 'vfscanf' 31820 and 'vsscanf' are also checked. Except in strictly conforming C 31821 standard modes, the X/Open function 'strfmon' is also checked as 31822 are 'printf_unlocked' and 'fprintf_unlocked'. *Note Options 31823 Controlling C Dialect: C Dialect Options. 31824 31825 For Objective-C dialects, 'NSString' (or '__NSString__') is 31826 recognized in the same context. Declarations including these 31827 format attributes are parsed for correct syntax, however the result 31828 of checking of such format strings is not yet defined, and is not 31829 carried out by this version of the compiler. 31830 31831 The target may also provide additional types of format checks. 31832 *Note Format Checks Specific to Particular Target Machines: Target 31833 Format Checks. 31834 31835'format_arg (STRING-INDEX)' 31836 The 'format_arg' attribute specifies that a function takes one or 31837 more format strings for a 'printf', 'scanf', 'strftime' or 31838 'strfmon' style function and modifies it (for example, to translate 31839 it into another language), so the result can be passed to a 31840 'printf', 'scanf', 'strftime' or 'strfmon' style function (with the 31841 remaining arguments to the format function the same as they would 31842 have been for the unmodified string). Multiple 'format_arg' 31843 attributes may be applied to the same function, each designating a 31844 distinct parameter as a format string. For example, the 31845 declaration: 31846 31847 extern char * 31848 my_dgettext (char *my_domain, const char *my_format) 31849 __attribute__ ((format_arg (2))); 31850 31851 causes the compiler to check the arguments in calls to a 'printf', 31852 'scanf', 'strftime' or 'strfmon' type function, whose format string 31853 argument is a call to the 'my_dgettext' function, for consistency 31854 with the format string argument 'my_format'. If the 'format_arg' 31855 attribute had not been specified, all the compiler could tell in 31856 such calls to format functions would be that the format string 31857 argument is not constant; this would generate a warning when 31858 '-Wformat-nonliteral' is used, but the calls could not be checked 31859 without the attribute. 31860 31861 In calls to a function declared with more than one 'format_arg' 31862 attribute, each with a distinct argument value, the corresponding 31863 actual function arguments are checked against all format strings 31864 designated by the attributes. This capability is designed to 31865 support the GNU 'ngettext' family of functions. 31866 31867 The parameter STRING-INDEX specifies which argument is the format 31868 string argument (starting from one). Since non-static C++ methods 31869 have an implicit 'this' argument, the arguments of such methods 31870 should be counted from two. 31871 31872 The 'format_arg' attribute allows you to identify your own 31873 functions that modify format strings, so that GCC can check the 31874 calls to 'printf', 'scanf', 'strftime' or 'strfmon' type function 31875 whose operands are a call to one of your own function. The 31876 compiler always treats 'gettext', 'dgettext', and 'dcgettext' in 31877 this manner except when strict ISO C support is requested by 31878 '-ansi' or an appropriate '-std' option, or '-ffreestanding' or 31879 '-fno-builtin' is used. *Note Options Controlling C Dialect: C 31880 Dialect Options. 31881 31882 For Objective-C dialects, the 'format-arg' attribute may refer to 31883 an 'NSString' reference for compatibility with the 'format' 31884 attribute above. 31885 31886 The target may also allow additional types in 'format-arg' 31887 attributes. *Note Format Checks Specific to Particular Target 31888 Machines: Target Format Checks. 31889 31890'gnu_inline' 31891 This attribute should be used with a function that is also declared 31892 with the 'inline' keyword. It directs GCC to treat the function as 31893 if it were defined in gnu90 mode even when compiling in C99 or 31894 gnu99 mode. 31895 31896 If the function is declared 'extern', then this definition of the 31897 function is used only for inlining. In no case is the function 31898 compiled as a standalone function, not even if you take its address 31899 explicitly. Such an address becomes an external reference, as if 31900 you had only declared the function, and had not defined it. This 31901 has almost the effect of a macro. The way to use this is to put a 31902 function definition in a header file with this attribute, and put 31903 another copy of the function, without 'extern', in a library file. 31904 The definition in the header file causes most calls to the function 31905 to be inlined. If any uses of the function remain, they refer to 31906 the single copy in the library. Note that the two definitions of 31907 the functions need not be precisely the same, although if they do 31908 not have the same effect your program may behave oddly. 31909 31910 In C, if the function is neither 'extern' nor 'static', then the 31911 function is compiled as a standalone function, as well as being 31912 inlined where possible. 31913 31914 This is how GCC traditionally handled functions declared 'inline'. 31915 Since ISO C99 specifies a different semantics for 'inline', this 31916 function attribute is provided as a transition measure and as a 31917 useful feature in its own right. This attribute is available in 31918 GCC 4.1.3 and later. It is available if either of the preprocessor 31919 macros '__GNUC_GNU_INLINE__' or '__GNUC_STDC_INLINE__' are defined. 31920 *Note An Inline Function is As Fast As a Macro: Inline. 31921 31922 In C++, this attribute does not depend on 'extern' in any way, but 31923 it still requires the 'inline' keyword to enable its special 31924 behavior. 31925 31926'hot' 31927 The 'hot' attribute on a function is used to inform the compiler 31928 that the function is a hot spot of the compiled program. The 31929 function is optimized more aggressively and on many targets it is 31930 placed into a special subsection of the text section so all hot 31931 functions appear close together, improving locality. 31932 31933 When profile feedback is available, via '-fprofile-use', hot 31934 functions are automatically detected and this attribute is ignored. 31935 31936'ifunc ("RESOLVER")' 31937 The 'ifunc' attribute is used to mark a function as an indirect 31938 function using the STT_GNU_IFUNC symbol type extension to the ELF 31939 standard. This allows the resolution of the symbol value to be 31940 determined dynamically at load time, and an optimized version of 31941 the routine to be selected for the particular processor or other 31942 system characteristics determined then. To use this attribute, 31943 first define the implementation functions available, and a resolver 31944 function that returns a pointer to the selected implementation 31945 function. The implementation functions' declarations must match 31946 the API of the function being implemented. The resolver should be 31947 declared to be a function taking no arguments and returning a 31948 pointer to a function of the same type as the implementation. For 31949 example: 31950 31951 void *my_memcpy (void *dst, const void *src, size_t len) 31952 { 31953 ... 31954 return dst; 31955 } 31956 31957 static void * (*resolve_memcpy (void))(void *, const void *, size_t) 31958 { 31959 return my_memcpy; // we will just always select this routine 31960 } 31961 31962 The exported header file declaring the function the user calls 31963 would contain: 31964 31965 extern void *memcpy (void *, const void *, size_t); 31966 31967 allowing the user to call 'memcpy' as a regular function, unaware 31968 of the actual implementation. Finally, the indirect function needs 31969 to be defined in the same translation unit as the resolver 31970 function: 31971 31972 void *memcpy (void *, const void *, size_t) 31973 __attribute__ ((ifunc ("resolve_memcpy"))); 31974 31975 In C++, the 'ifunc' attribute takes a string that is the mangled 31976 name of the resolver function. A C++ resolver for a non-static 31977 member function of class 'C' should be declared to return a pointer 31978 to a non-member function taking pointer to 'C' as the first 31979 argument, followed by the same arguments as of the implementation 31980 function. G++ checks the signatures of the two functions and 31981 issues a '-Wattribute-alias' warning for mismatches. To suppress a 31982 warning for the necessary cast from a pointer to the implementation 31983 member function to the type of the corresponding non-member 31984 function use the '-Wno-pmf-conversions' option. For example: 31985 31986 class S 31987 { 31988 private: 31989 int debug_impl (int); 31990 int optimized_impl (int); 31991 31992 typedef int Func (S*, int); 31993 31994 static Func* resolver (); 31995 public: 31996 31997 int interface (int); 31998 }; 31999 32000 int S::debug_impl (int) { /* ... */ } 32001 int S::optimized_impl (int) { /* ... */ } 32002 32003 S::Func* S::resolver () 32004 { 32005 int (S::*pimpl) (int) 32006 = getenv ("DEBUG") ? &S::debug_impl : &S::optimized_impl; 32007 32008 // Cast triggers -Wno-pmf-conversions. 32009 return reinterpret_cast<Func*>(pimpl); 32010 } 32011 32012 int S::interface (int) __attribute__ ((ifunc ("_ZN1S8resolverEv"))); 32013 32014 Indirect functions cannot be weak. Binutils version 2.20.1 or 32015 higher and GNU C Library version 2.11.1 are required to use this 32016 feature. 32017 32018'interrupt' 32019'interrupt_handler' 32020 Many GCC back ends support attributes to indicate that a function 32021 is an interrupt handler, which tells the compiler to generate 32022 function entry and exit sequences that differ from those from 32023 regular functions. The exact syntax and behavior are 32024 target-specific; refer to the following subsections for details. 32025 32026'leaf' 32027 Calls to external functions with this attribute must return to the 32028 current compilation unit only by return or by exception handling. 32029 In particular, a leaf function is not allowed to invoke callback 32030 functions passed to it from the current compilation unit, directly 32031 call functions exported by the unit, or 'longjmp' into the unit. 32032 Leaf functions might still call functions from other compilation 32033 units and thus they are not necessarily leaf in the sense that they 32034 contain no function calls at all. 32035 32036 The attribute is intended for library functions to improve dataflow 32037 analysis. The compiler takes the hint that any data not escaping 32038 the current compilation unit cannot be used or modified by the leaf 32039 function. For example, the 'sin' function is a leaf function, but 32040 'qsort' is not. 32041 32042 Note that leaf functions might indirectly run a signal handler 32043 defined in the current compilation unit that uses static variables. 32044 Similarly, when lazy symbol resolution is in effect, leaf functions 32045 might invoke indirect functions whose resolver function or 32046 implementation function is defined in the current compilation unit 32047 and uses static variables. There is no standard-compliant way to 32048 write such a signal handler, resolver function, or implementation 32049 function, and the best that you can do is to remove the 'leaf' 32050 attribute or mark all such static variables 'volatile'. Lastly, 32051 for ELF-based systems that support symbol interposition, care 32052 should be taken that functions defined in the current compilation 32053 unit do not unexpectedly interpose other symbols based on the 32054 defined standards mode and defined feature test macros; otherwise 32055 an inadvertent callback would be added. 32056 32057 The attribute has no effect on functions defined within the current 32058 compilation unit. This is to allow easy merging of multiple 32059 compilation units into one, for example, by using the link-time 32060 optimization. For this reason the attribute is not allowed on 32061 types to annotate indirect calls. 32062 32063'malloc' 32064 This tells the compiler that a function is 'malloc'-like, i.e., 32065 that the pointer P returned by the function cannot alias any other 32066 pointer valid when the function returns, and moreover no pointers 32067 to valid objects occur in any storage addressed by P. In addition, 32068 GCC predicts that a function with the attribute returns non-null in 32069 most cases. 32070 32071 Using the attribute is designed to improve optimization by relying 32072 on the aliasing property it implies. Functions like 'malloc' and 32073 'calloc' have this property because they return a pointer to 32074 uninitialized or zeroed-out, newly obtained storage. However, 32075 functions like 'realloc' do not have this property, as they may 32076 return pointers to storage containing pointers to existing objects. 32077 Additionally, since all such functions are assumed to return null 32078 only infrequently, callers can be optimized based on that 32079 assumption. 32080 32081'no_icf' 32082 This function attribute prevents a functions from being merged with 32083 another semantically equivalent function. 32084 32085'no_instrument_function' 32086 If any of '-finstrument-functions', '-p', or '-pg' are given, 32087 profiling function calls are generated at entry and exit of most 32088 user-compiled functions. Functions with this attribute are not so 32089 instrumented. 32090 32091'no_profile_instrument_function' 32092 The 'no_profile_instrument_function' attribute on functions is used 32093 to inform the compiler that it should not process any profile 32094 feedback based optimization code instrumentation. 32095 32096'no_reorder' 32097 Do not reorder functions or variables marked 'no_reorder' against 32098 each other or top level assembler statements the executable. The 32099 actual order in the program will depend on the linker command line. 32100 Static variables marked like this are also not removed. This has a 32101 similar effect as the '-fno-toplevel-reorder' option, but only 32102 applies to the marked symbols. 32103 32104'no_sanitize ("SANITIZE_OPTION")' 32105 The 'no_sanitize' attribute on functions is used to inform the 32106 compiler that it should not do sanitization of any option mentioned 32107 in SANITIZE_OPTION. A list of values acceptable by the 32108 '-fsanitize' option can be provided. 32109 32110 void __attribute__ ((no_sanitize ("alignment", "object-size"))) 32111 f () { /* Do something. */; } 32112 void __attribute__ ((no_sanitize ("alignment,object-size"))) 32113 g () { /* Do something. */; } 32114 32115'no_sanitize_address' 32116'no_address_safety_analysis' 32117 The 'no_sanitize_address' attribute on functions is used to inform 32118 the compiler that it should not instrument memory accesses in the 32119 function when compiling with the '-fsanitize=address' option. The 32120 'no_address_safety_analysis' is a deprecated alias of the 32121 'no_sanitize_address' attribute, new code should use 32122 'no_sanitize_address'. 32123 32124'no_sanitize_thread' 32125 The 'no_sanitize_thread' attribute on functions is used to inform 32126 the compiler that it should not instrument memory accesses in the 32127 function when compiling with the '-fsanitize=thread' option. 32128 32129'no_sanitize_undefined' 32130 The 'no_sanitize_undefined' attribute on functions is used to 32131 inform the compiler that it should not check for undefined behavior 32132 in the function when compiling with the '-fsanitize=undefined' 32133 option. 32134 32135'no_split_stack' 32136 If '-fsplit-stack' is given, functions have a small prologue which 32137 decides whether to split the stack. Functions with the 32138 'no_split_stack' attribute do not have that prologue, and thus may 32139 run with only a small amount of stack space available. 32140 32141'no_stack_limit' 32142 This attribute locally overrides the '-fstack-limit-register' and 32143 '-fstack-limit-symbol' command-line options; it has the effect of 32144 disabling stack limit checking in the function it applies to. 32145 32146'noclone' 32147 This function attribute prevents a function from being considered 32148 for cloning--a mechanism that produces specialized copies of 32149 functions and which is (currently) performed by interprocedural 32150 constant propagation. 32151 32152'noinline' 32153 This function attribute prevents a function from being considered 32154 for inlining. If the function does not have side effects, there 32155 are optimizations other than inlining that cause function calls to 32156 be optimized away, although the function call is live. To keep 32157 such calls from being optimized away, put 32158 asm (""); 32159 32160 (*note Extended Asm::) in the called function, to serve as a 32161 special side effect. 32162 32163'noipa' 32164 Disable interprocedural optimizations between the function with 32165 this attribute and its callers, as if the body of the function is 32166 not available when optimizing callers and the callers are 32167 unavailable when optimizing the body. This attribute implies 32168 'noinline', 'noclone' and 'no_icf' attributes. However, this 32169 attribute is not equivalent to a combination of other attributes, 32170 because its purpose is to suppress existing and future 32171 optimizations employing interprocedural analysis, including those 32172 that do not have an attribute suitable for disabling them 32173 individually. This attribute is supported mainly for the purpose 32174 of testing the compiler. 32175 32176'nonnull' 32177'nonnull (ARG-INDEX, ...)' 32178 The 'nonnull' attribute may be applied to a function that takes at 32179 least one argument of a pointer type. It indicates that the 32180 referenced arguments must be non-null pointers. For instance, the 32181 declaration: 32182 32183 extern void * 32184 my_memcpy (void *dest, const void *src, size_t len) 32185 __attribute__((nonnull (1, 2))); 32186 32187 causes the compiler to check that, in calls to 'my_memcpy', 32188 arguments DEST and SRC are non-null. If the compiler determines 32189 that a null pointer is passed in an argument slot marked as 32190 non-null, and the '-Wnonnull' option is enabled, a warning is 32191 issued. *Note Warning Options::. Unless disabled by the 32192 '-fno-delete-null-pointer-checks' option the compiler may also 32193 perform optimizations based on the knowledge that certain function 32194 arguments cannot be null. In addition, the 32195 '-fisolate-erroneous-paths-attribute' option can be specified to 32196 have GCC transform calls with null arguments to non-null functions 32197 into traps. *Note Optimize Options::. 32198 32199 If no ARG-INDEX is given to the 'nonnull' attribute, all pointer 32200 arguments are marked as non-null. To illustrate, the following 32201 declaration is equivalent to the previous example: 32202 32203 extern void * 32204 my_memcpy (void *dest, const void *src, size_t len) 32205 __attribute__((nonnull)); 32206 32207'noplt' 32208 The 'noplt' attribute is the counterpart to option '-fno-plt'. 32209 Calls to functions marked with this attribute in 32210 position-independent code do not use the PLT. 32211 32212 /* Externally defined function foo. */ 32213 int foo () __attribute__ ((noplt)); 32214 32215 int 32216 main (/* ... */) 32217 { 32218 /* ... */ 32219 foo (); 32220 /* ... */ 32221 } 32222 32223 The 'noplt' attribute on function 'foo' tells the compiler to 32224 assume that the function 'foo' is externally defined and that the 32225 call to 'foo' must avoid the PLT in position-independent code. 32226 32227 In position-dependent code, a few targets also convert calls to 32228 functions that are marked to not use the PLT to use the GOT 32229 instead. 32230 32231'noreturn' 32232 A few standard library functions, such as 'abort' and 'exit', 32233 cannot return. GCC knows this automatically. Some programs define 32234 their own functions that never return. You can declare them 32235 'noreturn' to tell the compiler this fact. For example, 32236 32237 void fatal () __attribute__ ((noreturn)); 32238 32239 void 32240 fatal (/* ... */) 32241 { 32242 /* ... */ /* Print error message. */ /* ... */ 32243 exit (1); 32244 } 32245 32246 The 'noreturn' keyword tells the compiler to assume that 'fatal' 32247 cannot return. It can then optimize without regard to what would 32248 happen if 'fatal' ever did return. This makes slightly better 32249 code. More importantly, it helps avoid spurious warnings of 32250 uninitialized variables. 32251 32252 The 'noreturn' keyword does not affect the exceptional path when 32253 that applies: a 'noreturn'-marked function may still return to the 32254 caller by throwing an exception or calling 'longjmp'. 32255 32256 In order to preserve backtraces, GCC will never turn calls to 32257 'noreturn' functions into tail calls. 32258 32259 Do not assume that registers saved by the calling function are 32260 restored before calling the 'noreturn' function. 32261 32262 It does not make sense for a 'noreturn' function to have a return 32263 type other than 'void'. 32264 32265'nothrow' 32266 The 'nothrow' attribute is used to inform the compiler that a 32267 function cannot throw an exception. For example, most functions in 32268 the standard C library can be guaranteed not to throw an exception 32269 with the notable exceptions of 'qsort' and 'bsearch' that take 32270 function pointer arguments. 32271 32272'optimize (LEVEL, ...)' 32273'optimize (STRING, ...)' 32274 The 'optimize' attribute is used to specify that a function is to 32275 be compiled with different optimization options than specified on 32276 the command line. Valid arguments are constant non-negative 32277 integers and strings. Each numeric argument specifies an 32278 optimization LEVEL. Each STRING argument consists of one or more 32279 comma-separated substrings. Each substring that begins with the 32280 letter 'O' refers to an optimization option such as '-O0' or '-Os'. 32281 Other substrings are taken as suffixes to the '-f' prefix jointly 32282 forming the name of an optimization option. *Note Optimize 32283 Options::. 32284 32285 '#pragma GCC optimize' can be used to set optimization options for 32286 more than one function. *Note Function Specific Option Pragmas::, 32287 for details about the pragma. 32288 32289 Providing multiple strings as arguments separated by commas to 32290 specify multiple options is equivalent to separating the option 32291 suffixes with a comma (',') within a single string. Spaces are not 32292 permitted within the strings. 32293 32294 Not every optimization option that starts with the -F prefix 32295 specified by the attribute necessarily has an effect on the 32296 function. The 'optimize' attribute should be used for debugging 32297 purposes only. It is not suitable in production code. 32298 32299'patchable_function_entry' 32300 In case the target's text segment can be made writable at run time 32301 by any means, padding the function entry with a number of NOPs can 32302 be used to provide a universal tool for instrumentation. 32303 32304 The 'patchable_function_entry' function attribute can be used to 32305 change the number of NOPs to any desired value. The two-value 32306 syntax is the same as for the command-line switch 32307 '-fpatchable-function-entry=N,M', generating N NOPs, with the 32308 function entry point before the Mth NOP instruction. M defaults to 32309 0 if omitted e.g. function entry point is before the first NOP. 32310 32311 If patchable function entries are enabled globally using the 32312 command-line option '-fpatchable-function-entry=N,M', then you must 32313 disable instrumentation on all functions that are part of the 32314 instrumentation framework with the attribute 32315 'patchable_function_entry (0)' to prevent recursion. 32316 32317'pure' 32318 32319 Calls to functions that have no observable effects on the state of 32320 the program other than to return a value may lend themselves to 32321 optimizations such as common subexpression elimination. Declaring 32322 such functions with the 'pure' attribute allows GCC to avoid 32323 emitting some calls in repeated invocations of the function with 32324 the same argument values. 32325 32326 The 'pure' attribute prohibits a function from modifying the state 32327 of the program that is observable by means other than inspecting 32328 the function's return value. However, functions declared with the 32329 'pure' attribute can safely read any non-volatile objects, and 32330 modify the value of objects in a way that does not affect their 32331 return value or the observable state of the program. 32332 32333 For example, 32334 32335 int hash (char *) __attribute__ ((pure)); 32336 32337 tells GCC that subsequent calls to the function 'hash' with the 32338 same string can be replaced by the result of the first call 32339 provided the state of the program observable by 'hash', including 32340 the contents of the array itself, does not change in between. Even 32341 though 'hash' takes a non-const pointer argument it must not modify 32342 the array it points to, or any other object whose value the rest of 32343 the program may depend on. However, the caller may safely change 32344 the contents of the array between successive calls to the function 32345 (doing so disables the optimization). The restriction also applies 32346 to member objects referenced by the 'this' pointer in C++ 32347 non-static member functions. 32348 32349 Some common examples of pure functions are 'strlen' or 'memcmp'. 32350 Interesting non-pure functions are functions with infinite loops or 32351 those depending on volatile memory or other system resource, that 32352 may change between consecutive calls (such as the standard C 'feof' 32353 function in a multithreading environment). 32354 32355 The 'pure' attribute imposes similar but looser restrictions on a 32356 function's definition than the 'const' attribute: 'pure' allows the 32357 function to read any non-volatile memory, even if it changes in 32358 between successive invocations of the function. Declaring the same 32359 function with both the 'pure' and the 'const' attribute is 32360 diagnosed. Because a pure function cannot have any observable side 32361 effects it does not make sense for such a function to return 32362 'void'. Declaring such a function is diagnosed. 32363 32364'returns_nonnull' 32365 The 'returns_nonnull' attribute specifies that the function return 32366 value should be a non-null pointer. For instance, the declaration: 32367 32368 extern void * 32369 mymalloc (size_t len) __attribute__((returns_nonnull)); 32370 32371 lets the compiler optimize callers based on the knowledge that the 32372 return value will never be null. 32373 32374'returns_twice' 32375 The 'returns_twice' attribute tells the compiler that a function 32376 may return more than one time. The compiler ensures that all 32377 registers are dead before calling such a function and emits a 32378 warning about the variables that may be clobbered after the second 32379 return from the function. Examples of such functions are 'setjmp' 32380 and 'vfork'. The 'longjmp'-like counterpart of such function, if 32381 any, might need to be marked with the 'noreturn' attribute. 32382 32383'section ("SECTION-NAME")' 32384 Normally, the compiler places the code it generates in the 'text' 32385 section. Sometimes, however, you need additional sections, or you 32386 need certain particular functions to appear in special sections. 32387 The 'section' attribute specifies that a function lives in a 32388 particular section. For example, the declaration: 32389 32390 extern void foobar (void) __attribute__ ((section ("bar"))); 32391 32392 puts the function 'foobar' in the 'bar' section. 32393 32394 Some file formats do not support arbitrary sections so the 32395 'section' attribute is not available on all platforms. If you need 32396 to map the entire contents of a module to a particular section, 32397 consider using the facilities of the linker instead. 32398 32399'sentinel' 32400'sentinel (POSITION)' 32401 This function attribute indicates that an argument in a call to the 32402 function is expected to be an explicit 'NULL'. The attribute is 32403 only valid on variadic functions. By default, the sentinel is 32404 expected to be the last argument of the function call. If the 32405 optional POSITION argument is specified to the attribute, the 32406 sentinel must be located at POSITION counting backwards from the 32407 end of the argument list. 32408 32409 __attribute__ ((sentinel)) 32410 is equivalent to 32411 __attribute__ ((sentinel(0))) 32412 32413 The attribute is automatically set with a position of 0 for the 32414 built-in functions 'execl' and 'execlp'. The built-in function 32415 'execle' has the attribute set with a position of 1. 32416 32417 A valid 'NULL' in this context is defined as zero with any object 32418 pointer type. If your system defines the 'NULL' macro with an 32419 integer type then you need to add an explicit cast. During 32420 installation GCC replaces the system '<stddef.h>' header with a 32421 copy that redefines NULL appropriately. 32422 32423 The warnings for missing or incorrect sentinels are enabled with 32424 '-Wformat'. 32425 32426'simd' 32427'simd("MASK")' 32428 This attribute enables creation of one or more function versions 32429 that can process multiple arguments using SIMD instructions from a 32430 single invocation. Specifying this attribute allows compiler to 32431 assume that such versions are available at link time (provided in 32432 the same or another translation unit). Generated versions are 32433 target-dependent and described in the corresponding Vector ABI 32434 document. For x86_64 target this document can be found 32435 here (https://sourceware.org/glibc/wiki/libmvec?action=AttachFile&do=view&target=VectorABI.txt). 32436 32437 The optional argument MASK may have the value 'notinbranch' or 32438 'inbranch', and instructs the compiler to generate non-masked or 32439 masked clones correspondingly. By default, all clones are 32440 generated. 32441 32442 If the attribute is specified and '#pragma omp declare simd' is 32443 present on a declaration and the '-fopenmp' or '-fopenmp-simd' 32444 switch is specified, then the attribute is ignored. 32445 32446'stack_protect' 32447 This attribute adds stack protection code to the function if flags 32448 '-fstack-protector', '-fstack-protector-strong' or 32449 '-fstack-protector-explicit' are set. 32450 32451'target (STRING, ...)' 32452 Multiple target back ends implement the 'target' attribute to 32453 specify that a function is to be compiled with different target 32454 options than specified on the command line. One or more strings 32455 can be provided as arguments. Each string consists of one or more 32456 comma-separated suffixes to the '-m' prefix jointly forming the 32457 name of a machine-dependent option. *Note Machine-Dependent 32458 Options: Submodel Options. 32459 32460 The 'target' attribute can be used for instance to have a function 32461 compiled with a different ISA (instruction set architecture) than 32462 the default. '#pragma GCC target' can be used to specify 32463 target-specific options for more than one function. *Note Function 32464 Specific Option Pragmas::, for details about the pragma. 32465 32466 For instance, on an x86, you could declare one function with the 32467 'target("sse4.1,arch=core2")' attribute and another with 32468 'target("sse4a,arch=amdfam10")'. This is equivalent to compiling 32469 the first function with '-msse4.1' and '-march=core2' options, and 32470 the second function with '-msse4a' and '-march=amdfam10' options. 32471 It is up to you to make sure that a function is only invoked on a 32472 machine that supports the particular ISA it is compiled for (for 32473 example by using 'cpuid' on x86 to determine what feature bits and 32474 architecture family are used). 32475 32476 int core2_func (void) __attribute__ ((__target__ ("arch=core2"))); 32477 int sse3_func (void) __attribute__ ((__target__ ("sse3"))); 32478 32479 Providing multiple strings as arguments separated by commas to 32480 specify multiple options is equivalent to separating the option 32481 suffixes with a comma (',') within a single string. Spaces are not 32482 permitted within the strings. 32483 32484 The options supported are specific to each target; refer to *note 32485 x86 Function Attributes::, *note PowerPC Function Attributes::, 32486 *note ARM Function Attributes::, *note AArch64 Function 32487 Attributes::, *note Nios II Function Attributes::, and *note S/390 32488 Function Attributes:: for details. 32489 32490'symver ("NAME2@NODENAME")' 32491 On ELF targets this attribute creates a symbol version. The NAME2 32492 part of the parameter is the actual name of the symbol by which it 32493 will be externally referenced. The 'nodename' portion should be 32494 the name of a node specified in the version script supplied to the 32495 linker when building a shared library. Versioned symbol must be 32496 defined and must be exported with default visibility. 32497 32498 __attribute__ ((__symver__ ("foo@VERS_1"))) int 32499 foo_v1 (void) 32500 { 32501 } 32502 32503 Will produce a '.symver foo_v1, foo@VERS_1' directive in the 32504 assembler output. 32505 32506 It's an error to define multiple version of a given symbol. In 32507 such case an alias can be used. 32508 32509 __attribute__ ((__symver__ ("foo@VERS_2"))) 32510 __attribute__ ((alias ("foo_v1"))) 32511 int symver_foo_v1 (void); 32512 32513 This example creates an alias of 'foo_v1' with symbol name 32514 'symver_foo_v1' which will be version 'VERS_2' of 'foo'. 32515 32516 Finally if the parameter is '"NAME2@@NODENAME"' then in addition to 32517 creating a symbol version (as if '"NAME2@NODENAME"' was used) the 32518 version will be also used to resolve NAME2 by the linker. 32519 32520'target_clones (OPTIONS)' 32521 The 'target_clones' attribute is used to specify that a function be 32522 cloned into multiple versions compiled with different target 32523 options than specified on the command line. The supported options 32524 and restrictions are the same as for 'target' attribute. 32525 32526 For instance, on an x86, you could compile a function with 32527 'target_clones("sse4.1,avx")'. GCC creates two function clones, 32528 one compiled with '-msse4.1' and another with '-mavx'. 32529 32530 On a PowerPC, you can compile a function with 32531 'target_clones("cpu=power9,default")'. GCC will create two 32532 function clones, one compiled with '-mcpu=power9' and another with 32533 the default options. GCC must be configured to use GLIBC 2.23 or 32534 newer in order to use the 'target_clones' attribute. 32535 32536 It also creates a resolver function (see the 'ifunc' attribute 32537 above) that dynamically selects a clone suitable for current 32538 architecture. The resolver is created only if there is a usage of 32539 a function with 'target_clones' attribute. 32540 32541 Note that any subsequent call of a function without 'target_clone' 32542 from a 'target_clone' caller will not lead to copying (target 32543 clone) of the called function. If you want to enforce such 32544 behaviour, we recommend declaring the calling function with the 32545 'flatten' attribute? 32546 32547'unused' 32548 This attribute, attached to a function, means that the function is 32549 meant to be possibly unused. GCC does not produce a warning for 32550 this function. 32551 32552'used' 32553 This attribute, attached to a function, means that code must be 32554 emitted for the function even if it appears that the function is 32555 not referenced. This is useful, for example, when the function is 32556 referenced only in inline assembly. 32557 32558 When applied to a member function of a C++ class template, the 32559 attribute also means that the function is instantiated if the class 32560 itself is instantiated. 32561 32562'visibility ("VISIBILITY_TYPE")' 32563 This attribute affects the linkage of the declaration to which it 32564 is attached. It can be applied to variables (*note Common Variable 32565 Attributes::) and types (*note Common Type Attributes::) as well as 32566 functions. 32567 32568 There are four supported VISIBILITY_TYPE values: default, hidden, 32569 protected or internal visibility. 32570 32571 void __attribute__ ((visibility ("protected"))) 32572 f () { /* Do something. */; } 32573 int i __attribute__ ((visibility ("hidden"))); 32574 32575 The possible values of VISIBILITY_TYPE correspond to the visibility 32576 settings in the ELF gABI. 32577 32578 'default' 32579 Default visibility is the normal case for the object file 32580 format. This value is available for the visibility attribute 32581 to override other options that may change the assumed 32582 visibility of entities. 32583 32584 On ELF, default visibility means that the declaration is 32585 visible to other modules and, in shared libraries, means that 32586 the declared entity may be overridden. 32587 32588 On Darwin, default visibility means that the declaration is 32589 visible to other modules. 32590 32591 Default visibility corresponds to "external linkage" in the 32592 language. 32593 32594 'hidden' 32595 Hidden visibility indicates that the entity declared has a new 32596 form of linkage, which we call "hidden linkage". Two 32597 declarations of an object with hidden linkage refer to the 32598 same object if they are in the same shared object. 32599 32600 'internal' 32601 Internal visibility is like hidden visibility, but with 32602 additional processor specific semantics. Unless otherwise 32603 specified by the psABI, GCC defines internal visibility to 32604 mean that a function is _never_ called from another module. 32605 Compare this with hidden functions which, while they cannot be 32606 referenced directly by other modules, can be referenced 32607 indirectly via function pointers. By indicating that a 32608 function cannot be called from outside the module, GCC may for 32609 instance omit the load of a PIC register since it is known 32610 that the calling function loaded the correct value. 32611 32612 'protected' 32613 Protected visibility is like default visibility except that it 32614 indicates that references within the defining module bind to 32615 the definition in that module. That is, the declared entity 32616 cannot be overridden by another module. 32617 32618 All visibilities are supported on many, but not all, ELF targets 32619 (supported when the assembler supports the '.visibility' 32620 pseudo-op). Default visibility is supported everywhere. Hidden 32621 visibility is supported on Darwin targets. 32622 32623 The visibility attribute should be applied only to declarations 32624 that would otherwise have external linkage. The attribute should 32625 be applied consistently, so that the same entity should not be 32626 declared with different settings of the attribute. 32627 32628 In C++, the visibility attribute applies to types as well as 32629 functions and objects, because in C++ types have linkage. A class 32630 must not have greater visibility than its non-static data member 32631 types and bases, and class members default to the visibility of 32632 their class. Also, a declaration without explicit visibility is 32633 limited to the visibility of its type. 32634 32635 In C++, you can mark member functions and static member variables 32636 of a class with the visibility attribute. This is useful if you 32637 know a particular method or static member variable should only be 32638 used from one shared object; then you can mark it hidden while the 32639 rest of the class has default visibility. Care must be taken to 32640 avoid breaking the One Definition Rule; for example, it is usually 32641 not useful to mark an inline method as hidden without marking the 32642 whole class as hidden. 32643 32644 A C++ namespace declaration can also have the visibility attribute. 32645 32646 namespace nspace1 __attribute__ ((visibility ("protected"))) 32647 { /* Do something. */; } 32648 32649 This attribute applies only to the particular namespace body, not 32650 to other definitions of the same namespace; it is equivalent to 32651 using '#pragma GCC visibility' before and after the namespace 32652 definition (*note Visibility Pragmas::). 32653 32654 In C++, if a template argument has limited visibility, this 32655 restriction is implicitly propagated to the template instantiation. 32656 Otherwise, template instantiations and specializations default to 32657 the visibility of their template. 32658 32659 If both the template and enclosing class have explicit visibility, 32660 the visibility from the template is used. 32661 32662'warn_unused_result' 32663 The 'warn_unused_result' attribute causes a warning to be emitted 32664 if a caller of the function with this attribute does not use its 32665 return value. This is useful for functions where not checking the 32666 result is either a security problem or always a bug, such as 32667 'realloc'. 32668 32669 int fn () __attribute__ ((warn_unused_result)); 32670 int foo () 32671 { 32672 if (fn () < 0) return -1; 32673 fn (); 32674 return 0; 32675 } 32676 32677 results in warning on line 5. 32678 32679'weak' 32680 The 'weak' attribute causes a declaration of an external symbol to 32681 be emitted as a weak symbol rather than a global. This is 32682 primarily useful in defining library functions that can be 32683 overridden in user code, though it can also be used with 32684 non-function declarations. The overriding symbol must have the 32685 same type as the weak symbol. In addition, if it designates a 32686 variable it must also have the same size and alignment as the weak 32687 symbol. Weak symbols are supported for ELF targets, and also for 32688 a.out targets when using the GNU assembler and linker. 32689 32690'weakref' 32691'weakref ("TARGET")' 32692 The 'weakref' attribute marks a declaration as a weak reference. 32693 Without arguments, it should be accompanied by an 'alias' attribute 32694 naming the target symbol. Alternatively, TARGET may be given as an 32695 argument to 'weakref' itself, naming the target definition of the 32696 alias. The TARGET must have the same type as the declaration. In 32697 addition, if it designates a variable it must also have the same 32698 size and alignment as the declaration. In either form of the 32699 declaration 'weakref' implicitly marks the declared symbol as 32700 'weak'. Without a TARGET given as an argument to 'weakref' or to 32701 'alias', 'weakref' is equivalent to 'weak' (in that case the 32702 declaration may be 'extern'). 32703 32704 /* Given the declaration: */ 32705 extern int y (void); 32706 32707 /* the following... */ 32708 static int x (void) __attribute__ ((weakref ("y"))); 32709 32710 /* is equivalent to... */ 32711 static int x (void) __attribute__ ((weakref, alias ("y"))); 32712 32713 /* or, alternatively, to... */ 32714 static int x (void) __attribute__ ((weakref)); 32715 static int x (void) __attribute__ ((alias ("y"))); 32716 32717 A weak reference is an alias that does not by itself require a 32718 definition to be given for the target symbol. If the target symbol 32719 is only referenced through weak references, then it becomes a 32720 'weak' undefined symbol. If it is directly referenced, however, 32721 then such strong references prevail, and a definition is required 32722 for the symbol, not necessarily in the same translation unit. 32723 32724 The effect is equivalent to moving all references to the alias to a 32725 separate translation unit, renaming the alias to the aliased 32726 symbol, declaring it as weak, compiling the two separate 32727 translation units and performing a link with relocatable output 32728 (i.e. 'ld -r') on them. 32729 32730 A declaration to which 'weakref' is attached and that is associated 32731 with a named 'target' must be 'static'. 32732 32733 32734File: gcc.info, Node: AArch64 Function Attributes, Next: AMD GCN Function Attributes, Prev: Common Function Attributes, Up: Function Attributes 32735 327366.33.2 AArch64 Function Attributes 32737---------------------------------- 32738 32739The following target-specific function attributes are available for the 32740AArch64 target. For the most part, these options mirror the behavior of 32741similar command-line options (*note AArch64 Options::), but on a 32742per-function basis. 32743 32744'general-regs-only' 32745 Indicates that no floating-point or Advanced SIMD registers should 32746 be used when generating code for this function. If the function 32747 explicitly uses floating-point code, then the compiler gives an 32748 error. This is the same behavior as that of the command-line 32749 option '-mgeneral-regs-only'. 32750 32751'fix-cortex-a53-835769' 32752 Indicates that the workaround for the Cortex-A53 erratum 835769 32753 should be applied to this function. To explicitly disable the 32754 workaround for this function specify the negated form: 32755 'no-fix-cortex-a53-835769'. This corresponds to the behavior of 32756 the command line options '-mfix-cortex-a53-835769' and 32757 '-mno-fix-cortex-a53-835769'. 32758 32759'cmodel=' 32760 Indicates that code should be generated for a particular code model 32761 for this function. The behavior and permissible arguments are the 32762 same as for the command line option '-mcmodel='. 32763 32764'strict-align' 32765'no-strict-align' 32766 'strict-align' indicates that the compiler should not assume that 32767 unaligned memory references are handled by the system. To allow 32768 the compiler to assume that aligned memory references are handled 32769 by the system, the inverse attribute 'no-strict-align' can be 32770 specified. The behavior is same as for the command-line option 32771 '-mstrict-align' and '-mno-strict-align'. 32772 32773'omit-leaf-frame-pointer' 32774 Indicates that the frame pointer should be omitted for a leaf 32775 function call. To keep the frame pointer, the inverse attribute 32776 'no-omit-leaf-frame-pointer' can be specified. These attributes 32777 have the same behavior as the command-line options 32778 '-momit-leaf-frame-pointer' and '-mno-omit-leaf-frame-pointer'. 32779 32780'tls-dialect=' 32781 Specifies the TLS dialect to use for this function. The behavior 32782 and permissible arguments are the same as for the command-line 32783 option '-mtls-dialect='. 32784 32785'arch=' 32786 Specifies the architecture version and architectural extensions to 32787 use for this function. The behavior and permissible arguments are 32788 the same as for the '-march=' command-line option. 32789 32790'tune=' 32791 Specifies the core for which to tune the performance of this 32792 function. The behavior and permissible arguments are the same as 32793 for the '-mtune=' command-line option. 32794 32795'cpu=' 32796 Specifies the core for which to tune the performance of this 32797 function and also whose architectural features to use. The 32798 behavior and valid arguments are the same as for the '-mcpu=' 32799 command-line option. 32800 32801'sign-return-address' 32802 Select the function scope on which return address signing will be 32803 applied. The behavior and permissible arguments are the same as 32804 for the command-line option '-msign-return-address='. The default 32805 value is 'none'. This attribute is deprecated. The 32806 'branch-protection' attribute should be used instead. 32807 32808'branch-protection' 32809 Select the function scope on which branch protection will be 32810 applied. The behavior and permissible arguments are the same as 32811 for the command-line option '-mbranch-protection='. The default 32812 value is 'none'. 32813 32814'outline-atomics' 32815 Enable or disable calls to out-of-line helpers to implement atomic 32816 operations. This corresponds to the behavior of the command line 32817 options '-moutline-atomics' and '-mno-outline-atomics'. 32818 32819 The above target attributes can be specified as follows: 32820 32821 __attribute__((target("ATTR-STRING"))) 32822 int 32823 f (int a) 32824 { 32825 return a + 5; 32826 } 32827 32828 where 'ATTR-STRING' is one of the attribute strings specified above. 32829 32830 Additionally, the architectural extension string may be specified on 32831its own. This can be used to turn on and off particular architectural 32832extensions without having to specify a particular architecture version 32833or core. Example: 32834 32835 __attribute__((target("+crc+nocrypto"))) 32836 int 32837 foo (int a) 32838 { 32839 return a + 5; 32840 } 32841 32842 In this example 'target("+crc+nocrypto")' enables the 'crc' extension 32843and disables the 'crypto' extension for the function 'foo' without 32844modifying an existing '-march=' or '-mcpu' option. 32845 32846 Multiple target function attributes can be specified by separating them 32847with a comma. For example: 32848 __attribute__((target("arch=armv8-a+crc+crypto,tune=cortex-a53"))) 32849 int 32850 foo (int a) 32851 { 32852 return a + 5; 32853 } 32854 32855 is valid and compiles function 'foo' for ARMv8-A with 'crc' and 32856'crypto' extensions and tunes it for 'cortex-a53'. 32857 328586.33.2.1 Inlining rules 32859....................... 32860 32861Specifying target attributes on individual functions or performing 32862link-time optimization across translation units compiled with different 32863target options can affect function inlining rules: 32864 32865 In particular, a caller function can inline a callee function only if 32866the architectural features available to the callee are a subset of the 32867features available to the caller. For example: A function 'foo' 32868compiled with '-march=armv8-a+crc', or tagged with the equivalent 32869'arch=armv8-a+crc' attribute, can inline a function 'bar' compiled with 32870'-march=armv8-a+nocrc' because the all the architectural features that 32871function 'bar' requires are available to function 'foo'. Conversely, 32872function 'bar' cannot inline function 'foo'. 32873 32874 Additionally inlining a function compiled with '-mstrict-align' into a 32875function compiled without '-mstrict-align' is not allowed. However, 32876inlining a function compiled without '-mstrict-align' into a function 32877compiled with '-mstrict-align' is allowed. 32878 32879 Note that CPU tuning options and attributes such as the '-mcpu=', 32880'-mtune=' do not inhibit inlining unless the CPU specified by the 32881'-mcpu=' option or the 'cpu=' attribute conflicts with the architectural 32882feature rules specified above. 32883 32884 32885File: gcc.info, Node: AMD GCN Function Attributes, Next: ARC Function Attributes, Prev: AArch64 Function Attributes, Up: Function Attributes 32886 328876.33.3 AMD GCN Function Attributes 32888---------------------------------- 32889 32890These function attributes are supported by the AMD GCN back end: 32891 32892'amdgpu_hsa_kernel' 32893 This attribute indicates that the corresponding function should be 32894 compiled as a kernel function, that is an entry point that can be 32895 invoked from the host via the HSA runtime library. By default 32896 functions are only callable only from other GCN functions. 32897 32898 This attribute is implicitly applied to any function named 'main', 32899 using default parameters. 32900 32901 Kernel functions may return an integer value, which will be written 32902 to a conventional place within the HSA "kernargs" region. 32903 32904 The attribute parameters configure what values are passed into the 32905 kernel function by the GPU drivers, via the initial register state. 32906 Some values are used by the compiler, and therefore forced on. 32907 Enabling other options may break assumptions in the compiler and/or 32908 run-time libraries. 32909 32910 'private_segment_buffer' 32911 Set 'enable_sgpr_private_segment_buffer' flag. Always on 32912 (required to locate the stack). 32913 32914 'dispatch_ptr' 32915 Set 'enable_sgpr_dispatch_ptr' flag. Always on (required to 32916 locate the launch dimensions). 32917 32918 'queue_ptr' 32919 Set 'enable_sgpr_queue_ptr' flag. Always on (required to 32920 convert address spaces). 32921 32922 'kernarg_segment_ptr' 32923 Set 'enable_sgpr_kernarg_segment_ptr' flag. Always on 32924 (required to locate the kernel arguments, "kernargs"). 32925 32926 'dispatch_id' 32927 Set 'enable_sgpr_dispatch_id' flag. 32928 32929 'flat_scratch_init' 32930 Set 'enable_sgpr_flat_scratch_init' flag. 32931 32932 'private_segment_size' 32933 Set 'enable_sgpr_private_segment_size' flag. 32934 32935 'grid_workgroup_count_X' 32936 Set 'enable_sgpr_grid_workgroup_count_x' flag. Always on 32937 (required to use OpenACC/OpenMP). 32938 32939 'grid_workgroup_count_Y' 32940 Set 'enable_sgpr_grid_workgroup_count_y' flag. 32941 32942 'grid_workgroup_count_Z' 32943 Set 'enable_sgpr_grid_workgroup_count_z' flag. 32944 32945 'workgroup_id_X' 32946 Set 'enable_sgpr_workgroup_id_x' flag. 32947 32948 'workgroup_id_Y' 32949 Set 'enable_sgpr_workgroup_id_y' flag. 32950 32951 'workgroup_id_Z' 32952 Set 'enable_sgpr_workgroup_id_z' flag. 32953 32954 'workgroup_info' 32955 Set 'enable_sgpr_workgroup_info' flag. 32956 32957 'private_segment_wave_offset' 32958 Set 'enable_sgpr_private_segment_wave_byte_offset' flag. 32959 Always on (required to locate the stack). 32960 32961 'work_item_id_X' 32962 Set 'enable_vgpr_workitem_id' parameter. Always on (can't be 32963 disabled). 32964 32965 'work_item_id_Y' 32966 Set 'enable_vgpr_workitem_id' parameter. Always on (required 32967 to enable vectorization.) 32968 32969 'work_item_id_Z' 32970 Set 'enable_vgpr_workitem_id' parameter. Always on (required 32971 to use OpenACC/OpenMP). 32972 32973 32974File: gcc.info, Node: ARC Function Attributes, Next: ARM Function Attributes, Prev: AMD GCN Function Attributes, Up: Function Attributes 32975 329766.33.4 ARC Function Attributes 32977------------------------------ 32978 32979These function attributes are supported by the ARC back end: 32980 32981'interrupt' 32982 Use this attribute to indicate that the specified function is an 32983 interrupt handler. The compiler generates function entry and exit 32984 sequences suitable for use in an interrupt handler when this 32985 attribute is present. 32986 32987 On the ARC, you must specify the kind of interrupt to be handled in 32988 a parameter to the interrupt attribute like this: 32989 32990 void f () __attribute__ ((interrupt ("ilink1"))); 32991 32992 Permissible values for this parameter are: 'ilink1' and 'ilink2' 32993 for ARCv1 architecture, and 'ilink' and 'firq' for ARCv2 32994 architecture. 32995 32996'long_call' 32997'medium_call' 32998'short_call' 32999 These attributes specify how a particular function is called. 33000 These attributes override the '-mlong-calls' and '-mmedium-calls' 33001 (*note ARC Options::) command-line switches and '#pragma 33002 long_calls' settings. 33003 33004 For ARC, a function marked with the 'long_call' attribute is always 33005 called using register-indirect jump-and-link instructions, thereby 33006 enabling the called function to be placed anywhere within the 33007 32-bit address space. A function marked with the 'medium_call' 33008 attribute will always be close enough to be called with an 33009 unconditional branch-and-link instruction, which has a 25-bit 33010 offset from the call site. A function marked with the 'short_call' 33011 attribute will always be close enough to be called with a 33012 conditional branch-and-link instruction, which has a 21-bit offset 33013 from the call site. 33014 33015'jli_always' 33016 Forces a particular function to be called using 'jli' instruction. 33017 The 'jli' instruction makes use of a table stored into '.jlitab' 33018 section, which holds the location of the functions which are 33019 addressed using this instruction. 33020 33021'jli_fixed' 33022 Identical like the above one, but the location of the function in 33023 the 'jli' table is known and given as an attribute parameter. 33024 33025'secure_call' 33026 This attribute allows one to mark secure-code functions that are 33027 callable from normal mode. The location of the secure call 33028 function into the 'sjli' table needs to be passed as argument. 33029 33030'naked' 33031 This attribute allows the compiler to construct the requisite 33032 function declaration, while allowing the body of the function to be 33033 assembly code. The specified function will not have 33034 prologue/epilogue sequences generated by the compiler. Only basic 33035 'asm' statements can safely be included in naked functions (*note 33036 Basic Asm::). While using extended 'asm' or a mixture of basic 33037 'asm' and C code may appear to work, they cannot be depended upon 33038 to work reliably and are not supported. 33039 33040 33041File: gcc.info, Node: ARM Function Attributes, Next: AVR Function Attributes, Prev: ARC Function Attributes, Up: Function Attributes 33042 330436.33.5 ARM Function Attributes 33044------------------------------ 33045 33046These function attributes are supported for ARM targets: 33047 33048'general-regs-only' 33049 Indicates that no floating-point or Advanced SIMD registers should 33050 be used when generating code for this function. If the function 33051 explicitly uses floating-point code, then the compiler gives an 33052 error. This is the same behavior as that of the command-line 33053 option '-mgeneral-regs-only'. 33054 33055'interrupt' 33056 Use this attribute to indicate that the specified function is an 33057 interrupt handler. The compiler generates function entry and exit 33058 sequences suitable for use in an interrupt handler when this 33059 attribute is present. 33060 33061 You can specify the kind of interrupt to be handled by adding an 33062 optional parameter to the interrupt attribute like this: 33063 33064 void f () __attribute__ ((interrupt ("IRQ"))); 33065 33066 Permissible values for this parameter are: 'IRQ', 'FIQ', 'SWI', 33067 'ABORT' and 'UNDEF'. 33068 33069 On ARMv7-M the interrupt type is ignored, and the attribute means 33070 the function may be called with a word-aligned stack pointer. 33071 33072'isr' 33073 Use this attribute on ARM to write Interrupt Service Routines. 33074 This is an alias to the 'interrupt' attribute above. 33075 33076'long_call' 33077'short_call' 33078 These attributes specify how a particular function is called. 33079 These attributes override the '-mlong-calls' (*note ARM Options::) 33080 command-line switch and '#pragma long_calls' settings. For ARM, 33081 the 'long_call' attribute indicates that the function might be far 33082 away from the call site and require a different (more expensive) 33083 calling sequence. The 'short_call' attribute always places the 33084 offset to the function from the call site into the 'BL' instruction 33085 directly. 33086 33087'naked' 33088 This attribute allows the compiler to construct the requisite 33089 function declaration, while allowing the body of the function to be 33090 assembly code. The specified function will not have 33091 prologue/epilogue sequences generated by the compiler. Only basic 33092 'asm' statements can safely be included in naked functions (*note 33093 Basic Asm::). While using extended 'asm' or a mixture of basic 33094 'asm' and C code may appear to work, they cannot be depended upon 33095 to work reliably and are not supported. 33096 33097'pcs' 33098 33099 The 'pcs' attribute can be used to control the calling convention 33100 used for a function on ARM. The attribute takes an argument that 33101 specifies the calling convention to use. 33102 33103 When compiling using the AAPCS ABI (or a variant of it) then valid 33104 values for the argument are '"aapcs"' and '"aapcs-vfp"'. In order 33105 to use a variant other than '"aapcs"' then the compiler must be 33106 permitted to use the appropriate co-processor registers (i.e., the 33107 VFP registers must be available in order to use '"aapcs-vfp"'). 33108 For example, 33109 33110 /* Argument passed in r0, and result returned in r0+r1. */ 33111 double f2d (float) __attribute__((pcs("aapcs"))); 33112 33113 Variadic functions always use the '"aapcs"' calling convention and 33114 the compiler rejects attempts to specify an alternative. 33115 33116'target (OPTIONS)' 33117 As discussed in *note Common Function Attributes::, this attribute 33118 allows specification of target-specific compilation options. 33119 33120 On ARM, the following options are allowed: 33121 33122 'thumb' 33123 Force code generation in the Thumb (T16/T32) ISA, depending on 33124 the architecture level. 33125 33126 'arm' 33127 Force code generation in the ARM (A32) ISA. 33128 33129 Functions from different modes can be inlined in the caller's 33130 mode. 33131 33132 'fpu=' 33133 Specifies the fpu for which to tune the performance of this 33134 function. The behavior and permissible arguments are the same 33135 as for the '-mfpu=' command-line option. 33136 33137 'arch=' 33138 Specifies the architecture version and architectural 33139 extensions to use for this function. The behavior and 33140 permissible arguments are the same as for the '-march=' 33141 command-line option. 33142 33143 The above target attributes can be specified as follows: 33144 33145 __attribute__((target("arch=armv8-a+crc"))) 33146 int 33147 f (int a) 33148 { 33149 return a + 5; 33150 } 33151 33152 Additionally, the architectural extension string may be 33153 specified on its own. This can be used to turn on and off 33154 particular architectural extensions without having to specify 33155 a particular architecture version or core. Example: 33156 33157 __attribute__((target("+crc+nocrypto"))) 33158 int 33159 foo (int a) 33160 { 33161 return a + 5; 33162 } 33163 33164 In this example 'target("+crc+nocrypto")' enables the 'crc' 33165 extension and disables the 'crypto' extension for the function 33166 'foo' without modifying an existing '-march=' or '-mcpu' 33167 option. 33168 33169 33170File: gcc.info, Node: AVR Function Attributes, Next: Blackfin Function Attributes, Prev: ARM Function Attributes, Up: Function Attributes 33171 331726.33.6 AVR Function Attributes 33173------------------------------ 33174 33175These function attributes are supported by the AVR back end: 33176 33177'interrupt' 33178 Use this attribute to indicate that the specified function is an 33179 interrupt handler. The compiler generates function entry and exit 33180 sequences suitable for use in an interrupt handler when this 33181 attribute is present. 33182 33183 On the AVR, the hardware globally disables interrupts when an 33184 interrupt is executed. The first instruction of an interrupt 33185 handler declared with this attribute is a 'SEI' instruction to 33186 re-enable interrupts. See also the 'signal' function attribute 33187 that does not insert a 'SEI' instruction. If both 'signal' and 33188 'interrupt' are specified for the same function, 'signal' is 33189 silently ignored. 33190 33191'naked' 33192 This attribute allows the compiler to construct the requisite 33193 function declaration, while allowing the body of the function to be 33194 assembly code. The specified function will not have 33195 prologue/epilogue sequences generated by the compiler. Only basic 33196 'asm' statements can safely be included in naked functions (*note 33197 Basic Asm::). While using extended 'asm' or a mixture of basic 33198 'asm' and C code may appear to work, they cannot be depended upon 33199 to work reliably and are not supported. 33200 33201'no_gccisr' 33202 Do not use '__gcc_isr' pseudo instructions in a function with the 33203 'interrupt' or 'signal' attribute aka. interrupt service routine 33204 (ISR). Use this attribute if the preamble of the ISR prologue 33205 should always read 33206 push __zero_reg__ 33207 push __tmp_reg__ 33208 in __tmp_reg__, __SREG__ 33209 push __tmp_reg__ 33210 clr __zero_reg__ 33211 and accordingly for the postamble of the epilogue -- no matter 33212 whether the mentioned registers are actually used in the ISR or 33213 not. Situations where you might want to use this attribute 33214 include: 33215 * Code that (effectively) clobbers bits of 'SREG' other than the 33216 'I'-flag by writing to the memory location of 'SREG'. 33217 * Code that uses inline assembler to jump to a different 33218 function which expects (parts of) the prologue code as 33219 outlined above to be present. 33220 To disable '__gcc_isr' generation for the whole compilation unit, 33221 there is option '-mno-gas-isr-prologues', *note AVR Options::. 33222 33223'OS_main' 33224'OS_task' 33225 On AVR, functions with the 'OS_main' or 'OS_task' attribute do not 33226 save/restore any call-saved register in their prologue/epilogue. 33227 33228 The 'OS_main' attribute can be used when there _is guarantee_ that 33229 interrupts are disabled at the time when the function is entered. 33230 This saves resources when the stack pointer has to be changed to 33231 set up a frame for local variables. 33232 33233 The 'OS_task' attribute can be used when there is _no guarantee_ 33234 that interrupts are disabled at that time when the function is 33235 entered like for, e.g. task functions in a multi-threading 33236 operating system. In that case, changing the stack pointer 33237 register is guarded by save/clear/restore of the global interrupt 33238 enable flag. 33239 33240 The differences to the 'naked' function attribute are: 33241 * 'naked' functions do not have a return instruction whereas 33242 'OS_main' and 'OS_task' functions have a 'RET' or 'RETI' 33243 return instruction. 33244 * 'naked' functions do not set up a frame for local variables or 33245 a frame pointer whereas 'OS_main' and 'OS_task' do this as 33246 needed. 33247 33248'signal' 33249 Use this attribute on the AVR to indicate that the specified 33250 function is an interrupt handler. The compiler generates function 33251 entry and exit sequences suitable for use in an interrupt handler 33252 when this attribute is present. 33253 33254 See also the 'interrupt' function attribute. 33255 33256 The AVR hardware globally disables interrupts when an interrupt is 33257 executed. Interrupt handler functions defined with the 'signal' 33258 attribute do not re-enable interrupts. It is save to enable 33259 interrupts in a 'signal' handler. This "save" only applies to the 33260 code generated by the compiler and not to the IRQ layout of the 33261 application which is responsibility of the application. 33262 33263 If both 'signal' and 'interrupt' are specified for the same 33264 function, 'signal' is silently ignored. 33265 33266 33267File: gcc.info, Node: Blackfin Function Attributes, Next: BPF Function Attributes, Prev: AVR Function Attributes, Up: Function Attributes 33268 332696.33.7 Blackfin Function Attributes 33270----------------------------------- 33271 33272These function attributes are supported by the Blackfin back end: 33273 33274'exception_handler' 33275 Use this attribute on the Blackfin to indicate that the specified 33276 function is an exception handler. The compiler generates function 33277 entry and exit sequences suitable for use in an exception handler 33278 when this attribute is present. 33279 33280'interrupt_handler' 33281 Use this attribute to indicate that the specified function is an 33282 interrupt handler. The compiler generates function entry and exit 33283 sequences suitable for use in an interrupt handler when this 33284 attribute is present. 33285 33286'kspisusp' 33287 When used together with 'interrupt_handler', 'exception_handler' or 33288 'nmi_handler', code is generated to load the stack pointer from the 33289 USP register in the function prologue. 33290 33291'l1_text' 33292 This attribute specifies a function to be placed into L1 33293 Instruction SRAM. The function is put into a specific section 33294 named '.l1.text'. With '-mfdpic', function calls with a such 33295 function as the callee or caller uses inlined PLT. 33296 33297'l2' 33298 This attribute specifies a function to be placed into L2 SRAM. The 33299 function is put into a specific section named '.l2.text'. With 33300 '-mfdpic', callers of such functions use an inlined PLT. 33301 33302'longcall' 33303'shortcall' 33304 The 'longcall' attribute indicates that the function might be far 33305 away from the call site and require a different (more expensive) 33306 calling sequence. The 'shortcall' attribute indicates that the 33307 function is always close enough for the shorter calling sequence to 33308 be used. These attributes override the '-mlongcall' switch. 33309 33310'nesting' 33311 Use this attribute together with 'interrupt_handler', 33312 'exception_handler' or 'nmi_handler' to indicate that the function 33313 entry code should enable nested interrupts or exceptions. 33314 33315'nmi_handler' 33316 Use this attribute on the Blackfin to indicate that the specified 33317 function is an NMI handler. The compiler generates function entry 33318 and exit sequences suitable for use in an NMI handler when this 33319 attribute is present. 33320 33321'saveall' 33322 Use this attribute to indicate that all registers except the stack 33323 pointer should be saved in the prologue regardless of whether they 33324 are used or not. 33325 33326 33327File: gcc.info, Node: BPF Function Attributes, Next: CR16 Function Attributes, Prev: Blackfin Function Attributes, Up: Function Attributes 33328 333296.33.8 BPF Function Attributes 33330------------------------------ 33331 33332These function attributes are supported by the BPF back end: 33333 33334'kernel_helper' 33335 use this attribute to indicate the specified function declaration 33336 is a kernel helper. The helper function is passed as an argument 33337 to the attribute. Example: 33338 33339 int bpf_probe_read (void *dst, int size, const void *unsafe_ptr) 33340 __attribute__ ((kernel_helper (4))); 33341 33342 33343File: gcc.info, Node: CR16 Function Attributes, Next: C-SKY Function Attributes, Prev: BPF Function Attributes, Up: Function Attributes 33344 333456.33.9 CR16 Function Attributes 33346------------------------------- 33347 33348These function attributes are supported by the CR16 back end: 33349 33350'interrupt' 33351 Use this attribute to indicate that the specified function is an 33352 interrupt handler. The compiler generates function entry and exit 33353 sequences suitable for use in an interrupt handler when this 33354 attribute is present. 33355 33356 33357File: gcc.info, Node: C-SKY Function Attributes, Next: Epiphany Function Attributes, Prev: CR16 Function Attributes, Up: Function Attributes 33358 333596.33.10 C-SKY Function Attributes 33360--------------------------------- 33361 33362These function attributes are supported by the C-SKY back end: 33363 33364'interrupt' 33365'isr' 33366 Use these attributes to indicate that the specified function is an 33367 interrupt handler. The compiler generates function entry and exit 33368 sequences suitable for use in an interrupt handler when either of 33369 these attributes are present. 33370 33371 Use of these options requires the '-mistack' command-line option to 33372 enable support for the necessary interrupt stack instructions. 33373 They are ignored with a warning otherwise. *Note C-SKY Options::. 33374 33375'naked' 33376 This attribute allows the compiler to construct the requisite 33377 function declaration, while allowing the body of the function to be 33378 assembly code. The specified function will not have 33379 prologue/epilogue sequences generated by the compiler. Only basic 33380 'asm' statements can safely be included in naked functions (*note 33381 Basic Asm::). While using extended 'asm' or a mixture of basic 33382 'asm' and C code may appear to work, they cannot be depended upon 33383 to work reliably and are not supported. 33384 33385 33386File: gcc.info, Node: Epiphany Function Attributes, Next: H8/300 Function Attributes, Prev: C-SKY Function Attributes, Up: Function Attributes 33387 333886.33.11 Epiphany Function Attributes 33389------------------------------------ 33390 33391These function attributes are supported by the Epiphany back end: 33392 33393'disinterrupt' 33394 This attribute causes the compiler to emit instructions to disable 33395 interrupts for the duration of the given function. 33396 33397'forwarder_section' 33398 This attribute modifies the behavior of an interrupt handler. The 33399 interrupt handler may be in external memory which cannot be reached 33400 by a branch instruction, so generate a local memory trampoline to 33401 transfer control. The single parameter identifies the section 33402 where the trampoline is placed. 33403 33404'interrupt' 33405 Use this attribute to indicate that the specified function is an 33406 interrupt handler. The compiler generates function entry and exit 33407 sequences suitable for use in an interrupt handler when this 33408 attribute is present. It may also generate a special section with 33409 code to initialize the interrupt vector table. 33410 33411 On Epiphany targets one or more optional parameters can be added 33412 like this: 33413 33414 void __attribute__ ((interrupt ("dma0, dma1"))) universal_dma_handler (); 33415 33416 Permissible values for these parameters are: 'reset', 33417 'software_exception', 'page_miss', 'timer0', 'timer1', 'message', 33418 'dma0', 'dma1', 'wand' and 'swi'. Multiple parameters indicate 33419 that multiple entries in the interrupt vector table should be 33420 initialized for this function, i.e. for each parameter NAME, a jump 33421 to the function is emitted in the section ivt_entry_NAME. The 33422 parameter(s) may be omitted entirely, in which case no interrupt 33423 vector table entry is provided. 33424 33425 Note that interrupts are enabled inside the function unless the 33426 'disinterrupt' attribute is also specified. 33427 33428 The following examples are all valid uses of these attributes on 33429 Epiphany targets: 33430 void __attribute__ ((interrupt)) universal_handler (); 33431 void __attribute__ ((interrupt ("dma1"))) dma1_handler (); 33432 void __attribute__ ((interrupt ("dma0, dma1"))) 33433 universal_dma_handler (); 33434 void __attribute__ ((interrupt ("timer0"), disinterrupt)) 33435 fast_timer_handler (); 33436 void __attribute__ ((interrupt ("dma0, dma1"), 33437 forwarder_section ("tramp"))) 33438 external_dma_handler (); 33439 33440'long_call' 33441'short_call' 33442 These attributes specify how a particular function is called. 33443 These attributes override the '-mlong-calls' (*note Adapteva 33444 Epiphany Options::) command-line switch and '#pragma long_calls' 33445 settings. 33446 33447 33448File: gcc.info, Node: H8/300 Function Attributes, Next: IA-64 Function Attributes, Prev: Epiphany Function Attributes, Up: Function Attributes 33449 334506.33.12 H8/300 Function Attributes 33451---------------------------------- 33452 33453These function attributes are available for H8/300 targets: 33454 33455'function_vector' 33456 Use this attribute on the H8/300, H8/300H, and H8S to indicate that 33457 the specified function should be called through the function 33458 vector. Calling a function through the function vector reduces 33459 code size; however, the function vector has a limited size (maximum 33460 128 entries on the H8/300 and 64 entries on the H8/300H and H8S) 33461 and shares space with the interrupt vector. 33462 33463'interrupt_handler' 33464 Use this attribute on the H8/300, H8/300H, and H8S to indicate that 33465 the specified function is an interrupt handler. The compiler 33466 generates function entry and exit sequences suitable for use in an 33467 interrupt handler when this attribute is present. 33468 33469'saveall' 33470 Use this attribute on the H8/300, H8/300H, and H8S to indicate that 33471 all registers except the stack pointer should be saved in the 33472 prologue regardless of whether they are used or not. 33473 33474 33475File: gcc.info, Node: IA-64 Function Attributes, Next: M32C Function Attributes, Prev: H8/300 Function Attributes, Up: Function Attributes 33476 334776.33.13 IA-64 Function Attributes 33478--------------------------------- 33479 33480These function attributes are supported on IA-64 targets: 33481 33482'syscall_linkage' 33483 This attribute is used to modify the IA-64 calling convention by 33484 marking all input registers as live at all function exits. This 33485 makes it possible to restart a system call after an interrupt 33486 without having to save/restore the input registers. This also 33487 prevents kernel data from leaking into application code. 33488 33489'version_id' 33490 This IA-64 HP-UX attribute, attached to a global variable or 33491 function, renames a symbol to contain a version string, thus 33492 allowing for function level versioning. HP-UX system header files 33493 may use function level versioning for some system calls. 33494 33495 extern int foo () __attribute__((version_id ("20040821"))); 33496 33497 Calls to 'foo' are mapped to calls to 'foo{20040821}'. 33498 33499 33500File: gcc.info, Node: M32C Function Attributes, Next: M32R/D Function Attributes, Prev: IA-64 Function Attributes, Up: Function Attributes 33501 335026.33.14 M32C Function Attributes 33503-------------------------------- 33504 33505These function attributes are supported by the M32C back end: 33506 33507'bank_switch' 33508 When added to an interrupt handler with the M32C port, causes the 33509 prologue and epilogue to use bank switching to preserve the 33510 registers rather than saving them on the stack. 33511 33512'fast_interrupt' 33513 Use this attribute on the M32C port to indicate that the specified 33514 function is a fast interrupt handler. This is just like the 33515 'interrupt' attribute, except that 'freit' is used to return 33516 instead of 'reit'. 33517 33518'function_vector' 33519 On M16C/M32C targets, the 'function_vector' attribute declares a 33520 special page subroutine call function. Use of this attribute 33521 reduces the code size by 2 bytes for each call generated to the 33522 subroutine. The argument to the attribute is the vector number 33523 entry from the special page vector table which contains the 16 33524 low-order bits of the subroutine's entry address. Each vector 33525 table has special page number (18 to 255) that is used in 'jsrs' 33526 instructions. Jump addresses of the routines are generated by 33527 adding 0x0F0000 (in case of M16C targets) or 0xFF0000 (in case of 33528 M32C targets), to the 2-byte addresses set in the vector table. 33529 Therefore you need to ensure that all the special page vector 33530 routines should get mapped within the address range 0x0F0000 to 33531 0x0FFFFF (for M16C) and 0xFF0000 to 0xFFFFFF (for M32C). 33532 33533 In the following example 2 bytes are saved for each call to 33534 function 'foo'. 33535 33536 void foo (void) __attribute__((function_vector(0x18))); 33537 void foo (void) 33538 { 33539 } 33540 33541 void bar (void) 33542 { 33543 foo(); 33544 } 33545 33546 If functions are defined in one file and are called in another 33547 file, then be sure to write this declaration in both files. 33548 33549 This attribute is ignored for R8C target. 33550 33551'interrupt' 33552 Use this attribute to indicate that the specified function is an 33553 interrupt handler. The compiler generates function entry and exit 33554 sequences suitable for use in an interrupt handler when this 33555 attribute is present. 33556 33557 33558File: gcc.info, Node: M32R/D Function Attributes, Next: m68k Function Attributes, Prev: M32C Function Attributes, Up: Function Attributes 33559 335606.33.15 M32R/D Function Attributes 33561---------------------------------- 33562 33563These function attributes are supported by the M32R/D back end: 33564 33565'interrupt' 33566 Use this attribute to indicate that the specified function is an 33567 interrupt handler. The compiler generates function entry and exit 33568 sequences suitable for use in an interrupt handler when this 33569 attribute is present. 33570 33571'model (MODEL-NAME)' 33572 33573 On the M32R/D, use this attribute to set the addressability of an 33574 object, and of the code generated for a function. The identifier 33575 MODEL-NAME is one of 'small', 'medium', or 'large', representing 33576 each of the code models. 33577 33578 Small model objects live in the lower 16MB of memory (so that their 33579 addresses can be loaded with the 'ld24' instruction), and are 33580 callable with the 'bl' instruction. 33581 33582 Medium model objects may live anywhere in the 32-bit address space 33583 (the compiler generates 'seth/add3' instructions to load their 33584 addresses), and are callable with the 'bl' instruction. 33585 33586 Large model objects may live anywhere in the 32-bit address space 33587 (the compiler generates 'seth/add3' instructions to load their 33588 addresses), and may not be reachable with the 'bl' instruction (the 33589 compiler generates the much slower 'seth/add3/jl' instruction 33590 sequence). 33591 33592 33593File: gcc.info, Node: m68k Function Attributes, Next: MCORE Function Attributes, Prev: M32R/D Function Attributes, Up: Function Attributes 33594 335956.33.16 m68k Function Attributes 33596-------------------------------- 33597 33598These function attributes are supported by the m68k back end: 33599 33600'interrupt' 33601'interrupt_handler' 33602 Use this attribute to indicate that the specified function is an 33603 interrupt handler. The compiler generates function entry and exit 33604 sequences suitable for use in an interrupt handler when this 33605 attribute is present. Either name may be used. 33606 33607'interrupt_thread' 33608 Use this attribute on fido, a subarchitecture of the m68k, to 33609 indicate that the specified function is an interrupt handler that 33610 is designed to run as a thread. The compiler omits generate 33611 prologue/epilogue sequences and replaces the return instruction 33612 with a 'sleep' instruction. This attribute is available only on 33613 fido. 33614 33615 33616File: gcc.info, Node: MCORE Function Attributes, Next: MeP Function Attributes, Prev: m68k Function Attributes, Up: Function Attributes 33617 336186.33.17 MCORE Function Attributes 33619--------------------------------- 33620 33621These function attributes are supported by the MCORE back end: 33622 33623'naked' 33624 This attribute allows the compiler to construct the requisite 33625 function declaration, while allowing the body of the function to be 33626 assembly code. The specified function will not have 33627 prologue/epilogue sequences generated by the compiler. Only basic 33628 'asm' statements can safely be included in naked functions (*note 33629 Basic Asm::). While using extended 'asm' or a mixture of basic 33630 'asm' and C code may appear to work, they cannot be depended upon 33631 to work reliably and are not supported. 33632 33633 33634File: gcc.info, Node: MeP Function Attributes, Next: MicroBlaze Function Attributes, Prev: MCORE Function Attributes, Up: Function Attributes 33635 336366.33.18 MeP Function Attributes 33637------------------------------- 33638 33639These function attributes are supported by the MeP back end: 33640 33641'disinterrupt' 33642 On MeP targets, this attribute causes the compiler to emit 33643 instructions to disable interrupts for the duration of the given 33644 function. 33645 33646'interrupt' 33647 Use this attribute to indicate that the specified function is an 33648 interrupt handler. The compiler generates function entry and exit 33649 sequences suitable for use in an interrupt handler when this 33650 attribute is present. 33651 33652'near' 33653 This attribute causes the compiler to assume the called function is 33654 close enough to use the normal calling convention, overriding the 33655 '-mtf' command-line option. 33656 33657'far' 33658 On MeP targets this causes the compiler to use a calling convention 33659 that assumes the called function is too far away for the built-in 33660 addressing modes. 33661 33662'vliw' 33663 The 'vliw' attribute tells the compiler to emit instructions in 33664 VLIW mode instead of core mode. Note that this attribute is not 33665 allowed unless a VLIW coprocessor has been configured and enabled 33666 through command-line options. 33667 33668 33669File: gcc.info, Node: MicroBlaze Function Attributes, Next: Microsoft Windows Function Attributes, Prev: MeP Function Attributes, Up: Function Attributes 33670 336716.33.19 MicroBlaze Function Attributes 33672-------------------------------------- 33673 33674These function attributes are supported on MicroBlaze targets: 33675 33676'save_volatiles' 33677 Use this attribute to indicate that the function is an interrupt 33678 handler. All volatile registers (in addition to non-volatile 33679 registers) are saved in the function prologue. If the function is 33680 a leaf function, only volatiles used by the function are saved. A 33681 normal function return is generated instead of a return from 33682 interrupt. 33683 33684'break_handler' 33685 Use this attribute to indicate that the specified function is a 33686 break handler. The compiler generates function entry and exit 33687 sequences suitable for use in an break handler when this attribute 33688 is present. The return from 'break_handler' is done through the 33689 'rtbd' instead of 'rtsd'. 33690 33691 void f () __attribute__ ((break_handler)); 33692 33693'interrupt_handler' 33694'fast_interrupt' 33695 These attributes indicate that the specified function is an 33696 interrupt handler. Use the 'fast_interrupt' attribute to indicate 33697 handlers used in low-latency interrupt mode, and 33698 'interrupt_handler' for interrupts that do not use low-latency 33699 handlers. In both cases, GCC emits appropriate prologue code and 33700 generates a return from the handler using 'rtid' instead of 'rtsd'. 33701 33702 33703File: gcc.info, Node: Microsoft Windows Function Attributes, Next: MIPS Function Attributes, Prev: MicroBlaze Function Attributes, Up: Function Attributes 33704 337056.33.20 Microsoft Windows Function Attributes 33706--------------------------------------------- 33707 33708The following attributes are available on Microsoft Windows and Symbian 33709OS targets. 33710 33711'dllexport' 33712 On Microsoft Windows targets and Symbian OS targets the 'dllexport' 33713 attribute causes the compiler to provide a global pointer to a 33714 pointer in a DLL, so that it can be referenced with the 'dllimport' 33715 attribute. On Microsoft Windows targets, the pointer name is 33716 formed by combining '_imp__' and the function or variable name. 33717 33718 You can use '__declspec(dllexport)' as a synonym for '__attribute__ 33719 ((dllexport))' for compatibility with other compilers. 33720 33721 On systems that support the 'visibility' attribute, this attribute 33722 also implies "default" visibility. It is an error to explicitly 33723 specify any other visibility. 33724 33725 GCC's default behavior is to emit all inline functions with the 33726 'dllexport' attribute. Since this can cause object file-size 33727 bloat, you can use '-fno-keep-inline-dllexport', which tells GCC to 33728 ignore the attribute for inlined functions unless the 33729 '-fkeep-inline-functions' flag is used instead. 33730 33731 The attribute is ignored for undefined symbols. 33732 33733 When applied to C++ classes, the attribute marks defined 33734 non-inlined member functions and static data members as exports. 33735 Static consts initialized in-class are not marked unless they are 33736 also defined out-of-class. 33737 33738 For Microsoft Windows targets there are alternative methods for 33739 including the symbol in the DLL's export table such as using a 33740 '.def' file with an 'EXPORTS' section or, with GNU ld, using the 33741 '--export-all' linker flag. 33742 33743'dllimport' 33744 On Microsoft Windows and Symbian OS targets, the 'dllimport' 33745 attribute causes the compiler to reference a function or variable 33746 via a global pointer to a pointer that is set up by the DLL 33747 exporting the symbol. The attribute implies 'extern'. On 33748 Microsoft Windows targets, the pointer name is formed by combining 33749 '_imp__' and the function or variable name. 33750 33751 You can use '__declspec(dllimport)' as a synonym for '__attribute__ 33752 ((dllimport))' for compatibility with other compilers. 33753 33754 On systems that support the 'visibility' attribute, this attribute 33755 also implies "default" visibility. It is an error to explicitly 33756 specify any other visibility. 33757 33758 Currently, the attribute is ignored for inlined functions. If the 33759 attribute is applied to a symbol _definition_, an error is 33760 reported. If a symbol previously declared 'dllimport' is later 33761 defined, the attribute is ignored in subsequent references, and a 33762 warning is emitted. The attribute is also overridden by a 33763 subsequent declaration as 'dllexport'. 33764 33765 When applied to C++ classes, the attribute marks non-inlined member 33766 functions and static data members as imports. However, the 33767 attribute is ignored for virtual methods to allow creation of 33768 vtables using thunks. 33769 33770 On the SH Symbian OS target the 'dllimport' attribute also has 33771 another affect--it can cause the vtable and run-time type 33772 information for a class to be exported. This happens when the 33773 class has a dllimported constructor or a non-inline, non-pure 33774 virtual function and, for either of those two conditions, the class 33775 also has an inline constructor or destructor and has a key function 33776 that is defined in the current translation unit. 33777 33778 For Microsoft Windows targets the use of the 'dllimport' attribute 33779 on functions is not necessary, but provides a small performance 33780 benefit by eliminating a thunk in the DLL. The use of the 33781 'dllimport' attribute on imported variables can be avoided by 33782 passing the '--enable-auto-import' switch to the GNU linker. As 33783 with functions, using the attribute for a variable eliminates a 33784 thunk in the DLL. 33785 33786 One drawback to using this attribute is that a pointer to a 33787 _variable_ marked as 'dllimport' cannot be used as a constant 33788 address. However, a pointer to a _function_ with the 'dllimport' 33789 attribute can be used as a constant initializer; in this case, the 33790 address of a stub function in the import lib is referenced. On 33791 Microsoft Windows targets, the attribute can be disabled for 33792 functions by setting the '-mnop-fun-dllimport' flag. 33793 33794 33795File: gcc.info, Node: MIPS Function Attributes, Next: MSP430 Function Attributes, Prev: Microsoft Windows Function Attributes, Up: Function Attributes 33796 337976.33.21 MIPS Function Attributes 33798-------------------------------- 33799 33800These function attributes are supported by the MIPS back end: 33801 33802'interrupt' 33803 Use this attribute to indicate that the specified function is an 33804 interrupt handler. The compiler generates function entry and exit 33805 sequences suitable for use in an interrupt handler when this 33806 attribute is present. An optional argument is supported for the 33807 interrupt attribute which allows the interrupt mode to be 33808 described. By default GCC assumes the external interrupt 33809 controller (EIC) mode is in use, this can be explicitly set using 33810 'eic'. When interrupts are non-masked then the requested Interrupt 33811 Priority Level (IPL) is copied to the current IPL which has the 33812 effect of only enabling higher priority interrupts. To use 33813 vectored interrupt mode use the argument 33814 'vector=[sw0|sw1|hw0|hw1|hw2|hw3|hw4|hw5]', this will change the 33815 behavior of the non-masked interrupt support and GCC will arrange 33816 to mask all interrupts from sw0 up to and including the specified 33817 interrupt vector. 33818 33819 You can use the following attributes to modify the behavior of an 33820 interrupt handler: 33821 'use_shadow_register_set' 33822 Assume that the handler uses a shadow register set, instead of 33823 the main general-purpose registers. An optional argument 33824 'intstack' is supported to indicate that the shadow register 33825 set contains a valid stack pointer. 33826 33827 'keep_interrupts_masked' 33828 Keep interrupts masked for the whole function. Without this 33829 attribute, GCC tries to reenable interrupts for as much of the 33830 function as it can. 33831 33832 'use_debug_exception_return' 33833 Return using the 'deret' instruction. Interrupt handlers that 33834 don't have this attribute return using 'eret' instead. 33835 33836 You can use any combination of these attributes, as shown below: 33837 void __attribute__ ((interrupt)) v0 (); 33838 void __attribute__ ((interrupt, use_shadow_register_set)) v1 (); 33839 void __attribute__ ((interrupt, keep_interrupts_masked)) v2 (); 33840 void __attribute__ ((interrupt, use_debug_exception_return)) v3 (); 33841 void __attribute__ ((interrupt, use_shadow_register_set, 33842 keep_interrupts_masked)) v4 (); 33843 void __attribute__ ((interrupt, use_shadow_register_set, 33844 use_debug_exception_return)) v5 (); 33845 void __attribute__ ((interrupt, keep_interrupts_masked, 33846 use_debug_exception_return)) v6 (); 33847 void __attribute__ ((interrupt, use_shadow_register_set, 33848 keep_interrupts_masked, 33849 use_debug_exception_return)) v7 (); 33850 void __attribute__ ((interrupt("eic"))) v8 (); 33851 void __attribute__ ((interrupt("vector=hw3"))) v9 (); 33852 33853'long_call' 33854'short_call' 33855'near' 33856'far' 33857 These attributes specify how a particular function is called on 33858 MIPS. The attributes override the '-mlong-calls' (*note MIPS 33859 Options::) command-line switch. The 'long_call' and 'far' 33860 attributes are synonyms, and cause the compiler to always call the 33861 function by first loading its address into a register, and then 33862 using the contents of that register. The 'short_call' and 'near' 33863 attributes are synonyms, and have the opposite effect; they specify 33864 that non-PIC calls should be made using the more efficient 'jal' 33865 instruction. 33866 33867'mips16' 33868'nomips16' 33869 33870 On MIPS targets, you can use the 'mips16' and 'nomips16' function 33871 attributes to locally select or turn off MIPS16 code generation. A 33872 function with the 'mips16' attribute is emitted as MIPS16 code, 33873 while MIPS16 code generation is disabled for functions with the 33874 'nomips16' attribute. These attributes override the '-mips16' and 33875 '-mno-mips16' options on the command line (*note MIPS Options::). 33876 33877 When compiling files containing mixed MIPS16 and non-MIPS16 code, 33878 the preprocessor symbol '__mips16' reflects the setting on the 33879 command line, not that within individual functions. Mixed MIPS16 33880 and non-MIPS16 code may interact badly with some GCC extensions 33881 such as '__builtin_apply' (*note Constructing Calls::). 33882 33883'micromips, MIPS' 33884'nomicromips, MIPS' 33885 33886 On MIPS targets, you can use the 'micromips' and 'nomicromips' 33887 function attributes to locally select or turn off microMIPS code 33888 generation. A function with the 'micromips' attribute is emitted 33889 as microMIPS code, while microMIPS code generation is disabled for 33890 functions with the 'nomicromips' attribute. These attributes 33891 override the '-mmicromips' and '-mno-micromips' options on the 33892 command line (*note MIPS Options::). 33893 33894 When compiling files containing mixed microMIPS and non-microMIPS 33895 code, the preprocessor symbol '__mips_micromips' reflects the 33896 setting on the command line, not that within individual functions. 33897 Mixed microMIPS and non-microMIPS code may interact badly with some 33898 GCC extensions such as '__builtin_apply' (*note Constructing 33899 Calls::). 33900 33901'nocompression' 33902 On MIPS targets, you can use the 'nocompression' function attribute 33903 to locally turn off MIPS16 and microMIPS code generation. This 33904 attribute overrides the '-mips16' and '-mmicromips' options on the 33905 command line (*note MIPS Options::). 33906 33907 33908File: gcc.info, Node: MSP430 Function Attributes, Next: NDS32 Function Attributes, Prev: MIPS Function Attributes, Up: Function Attributes 33909 339106.33.22 MSP430 Function Attributes 33911---------------------------------- 33912 33913These function attributes are supported by the MSP430 back end: 33914 33915'critical' 33916 Critical functions disable interrupts upon entry and restore the 33917 previous interrupt state upon exit. Critical functions cannot also 33918 have the 'naked', 'reentrant' or 'interrupt' attributes. 33919 33920 The MSP430 hardware ensures that interrupts are disabled on entry 33921 to 'interrupt' functions, and restores the previous interrupt state 33922 on exit. The 'critical' attribute is therefore redundant on 33923 'interrupt' functions. 33924 33925'interrupt' 33926 Use this attribute to indicate that the specified function is an 33927 interrupt handler. The compiler generates function entry and exit 33928 sequences suitable for use in an interrupt handler when this 33929 attribute is present. 33930 33931 You can provide an argument to the interrupt attribute which 33932 specifies a name or number. If the argument is a number it 33933 indicates the slot in the interrupt vector table (0 - 31) to which 33934 this handler should be assigned. If the argument is a name it is 33935 treated as a symbolic name for the vector slot. These names should 33936 match up with appropriate entries in the linker script. By default 33937 the names 'watchdog' for vector 26, 'nmi' for vector 30 and 'reset' 33938 for vector 31 are recognized. 33939 33940'naked' 33941 This attribute allows the compiler to construct the requisite 33942 function declaration, while allowing the body of the function to be 33943 assembly code. The specified function will not have 33944 prologue/epilogue sequences generated by the compiler. Only basic 33945 'asm' statements can safely be included in naked functions (*note 33946 Basic Asm::). While using extended 'asm' or a mixture of basic 33947 'asm' and C code may appear to work, they cannot be depended upon 33948 to work reliably and are not supported. 33949 33950'reentrant' 33951 Reentrant functions disable interrupts upon entry and enable them 33952 upon exit. Reentrant functions cannot also have the 'naked' or 33953 'critical' attributes. They can have the 'interrupt' attribute. 33954 33955'wakeup' 33956 This attribute only applies to interrupt functions. It is silently 33957 ignored if applied to a non-interrupt function. A wakeup interrupt 33958 function will rouse the processor from any low-power state that it 33959 might be in when the function exits. 33960 33961'lower' 33962'upper' 33963'either' 33964 On the MSP430 target these attributes can be used to specify 33965 whether the function or variable should be placed into low memory, 33966 high memory, or the placement should be left to the linker to 33967 decide. The attributes are only significant if compiling for the 33968 MSP430X architecture in the large memory model. 33969 33970 The attributes work in conjunction with a linker script that has 33971 been augmented to specify where to place sections with a '.lower' 33972 and a '.upper' prefix. So, for example, as well as placing the 33973 '.data' section, the script also specifies the placement of a 33974 '.lower.data' and a '.upper.data' section. The intention is that 33975 'lower' sections are placed into a small but easier to access 33976 memory region and the upper sections are placed into a larger, but 33977 slower to access, region. 33978 33979 The 'either' attribute is special. It tells the linker to place 33980 the object into the corresponding 'lower' section if there is room 33981 for it. If there is insufficient room then the object is placed 33982 into the corresponding 'upper' section instead. Note that the 33983 placement algorithm is not very sophisticated. It does not attempt 33984 to find an optimal packing of the 'lower' sections. It just makes 33985 one pass over the objects and does the best that it can. Using the 33986 '-ffunction-sections' and '-fdata-sections' command-line options 33987 can help the packing, however, since they produce smaller, easier 33988 to pack regions. 33989 33990 33991File: gcc.info, Node: NDS32 Function Attributes, Next: Nios II Function Attributes, Prev: MSP430 Function Attributes, Up: Function Attributes 33992 339936.33.23 NDS32 Function Attributes 33994--------------------------------- 33995 33996These function attributes are supported by the NDS32 back end: 33997 33998'exception' 33999 Use this attribute on the NDS32 target to indicate that the 34000 specified function is an exception handler. The compiler will 34001 generate corresponding sections for use in an exception handler. 34002 34003'interrupt' 34004 On NDS32 target, this attribute indicates that the specified 34005 function is an interrupt handler. The compiler generates 34006 corresponding sections for use in an interrupt handler. You can 34007 use the following attributes to modify the behavior: 34008 'nested' 34009 This interrupt service routine is interruptible. 34010 'not_nested' 34011 This interrupt service routine is not interruptible. 34012 'nested_ready' 34013 This interrupt service routine is interruptible after 34014 'PSW.GIE' (global interrupt enable) is set. This allows 34015 interrupt service routine to finish some short critical code 34016 before enabling interrupts. 34017 'save_all' 34018 The system will help save all registers into stack before 34019 entering interrupt handler. 34020 'partial_save' 34021 The system will help save caller registers into stack before 34022 entering interrupt handler. 34023 34024'naked' 34025 This attribute allows the compiler to construct the requisite 34026 function declaration, while allowing the body of the function to be 34027 assembly code. The specified function will not have 34028 prologue/epilogue sequences generated by the compiler. Only basic 34029 'asm' statements can safely be included in naked functions (*note 34030 Basic Asm::). While using extended 'asm' or a mixture of basic 34031 'asm' and C code may appear to work, they cannot be depended upon 34032 to work reliably and are not supported. 34033 34034'reset' 34035 Use this attribute on the NDS32 target to indicate that the 34036 specified function is a reset handler. The compiler will generate 34037 corresponding sections for use in a reset handler. You can use the 34038 following attributes to provide extra exception handling: 34039 'nmi' 34040 Provide a user-defined function to handle NMI exception. 34041 'warm' 34042 Provide a user-defined function to handle warm reset 34043 exception. 34044 34045 34046File: gcc.info, Node: Nios II Function Attributes, Next: Nvidia PTX Function Attributes, Prev: NDS32 Function Attributes, Up: Function Attributes 34047 340486.33.24 Nios II Function Attributes 34049----------------------------------- 34050 34051These function attributes are supported by the Nios II back end: 34052 34053'target (OPTIONS)' 34054 As discussed in *note Common Function Attributes::, this attribute 34055 allows specification of target-specific compilation options. 34056 34057 When compiling for Nios II, the following options are allowed: 34058 34059 'custom-INSN=N' 34060 'no-custom-INSN' 34061 Each 'custom-INSN=N' attribute locally enables use of a custom 34062 instruction with encoding N when generating code that uses 34063 INSN. Similarly, 'no-custom-INSN' locally inhibits use of the 34064 custom instruction INSN. These target attributes correspond 34065 to the '-mcustom-INSN=N' and '-mno-custom-INSN' command-line 34066 options, and support the same set of INSN keywords. *Note 34067 Nios II Options::, for more information. 34068 34069 'custom-fpu-cfg=NAME' 34070 This attribute corresponds to the '-mcustom-fpu-cfg=NAME' 34071 command-line option, to select a predefined set of custom 34072 instructions named NAME. *Note Nios II Options::, for more 34073 information. 34074 34075 34076File: gcc.info, Node: Nvidia PTX Function Attributes, Next: PowerPC Function Attributes, Prev: Nios II Function Attributes, Up: Function Attributes 34077 340786.33.25 Nvidia PTX Function Attributes 34079-------------------------------------- 34080 34081These function attributes are supported by the Nvidia PTX back end: 34082 34083'kernel' 34084 This attribute indicates that the corresponding function should be 34085 compiled as a kernel function, which can be invoked from the host 34086 via the CUDA RT library. By default functions are only callable 34087 only from other PTX functions. 34088 34089 Kernel functions must have 'void' return type. 34090 34091 34092File: gcc.info, Node: PowerPC Function Attributes, Next: RISC-V Function Attributes, Prev: Nvidia PTX Function Attributes, Up: Function Attributes 34093 340946.33.26 PowerPC Function Attributes 34095----------------------------------- 34096 34097These function attributes are supported by the PowerPC back end: 34098 34099'longcall' 34100'shortcall' 34101 The 'longcall' attribute indicates that the function might be far 34102 away from the call site and require a different (more expensive) 34103 calling sequence. The 'shortcall' attribute indicates that the 34104 function is always close enough for the shorter calling sequence to 34105 be used. These attributes override both the '-mlongcall' switch 34106 and the '#pragma longcall' setting. 34107 34108 *Note RS/6000 and PowerPC Options::, for more information on 34109 whether long calls are necessary. 34110 34111'target (OPTIONS)' 34112 As discussed in *note Common Function Attributes::, this attribute 34113 allows specification of target-specific compilation options. 34114 34115 On the PowerPC, the following options are allowed: 34116 34117 'altivec' 34118 'no-altivec' 34119 Generate code that uses (does not use) AltiVec instructions. 34120 In 32-bit code, you cannot enable AltiVec instructions unless 34121 '-mabi=altivec' is used on the command line. 34122 34123 'cmpb' 34124 'no-cmpb' 34125 Generate code that uses (does not use) the compare bytes 34126 instruction implemented on the POWER6 processor and other 34127 processors that support the PowerPC V2.05 architecture. 34128 34129 'dlmzb' 34130 'no-dlmzb' 34131 Generate code that uses (does not use) the string-search 34132 'dlmzb' instruction on the IBM 405, 440, 464 and 476 34133 processors. This instruction is generated by default when 34134 targeting those processors. 34135 34136 'fprnd' 34137 'no-fprnd' 34138 Generate code that uses (does not use) the FP round to integer 34139 instructions implemented on the POWER5+ processor and other 34140 processors that support the PowerPC V2.03 architecture. 34141 34142 'hard-dfp' 34143 'no-hard-dfp' 34144 Generate code that uses (does not use) the decimal 34145 floating-point instructions implemented on some POWER 34146 processors. 34147 34148 'isel' 34149 'no-isel' 34150 Generate code that uses (does not use) ISEL instruction. 34151 34152 'mfcrf' 34153 'no-mfcrf' 34154 Generate code that uses (does not use) the move from condition 34155 register field instruction implemented on the POWER4 processor 34156 and other processors that support the PowerPC V2.01 34157 architecture. 34158 34159 'mulhw' 34160 'no-mulhw' 34161 Generate code that uses (does not use) the half-word multiply 34162 and multiply-accumulate instructions on the IBM 405, 440, 464 34163 and 476 processors. These instructions are generated by 34164 default when targeting those processors. 34165 34166 'multiple' 34167 'no-multiple' 34168 Generate code that uses (does not use) the load multiple word 34169 instructions and the store multiple word instructions. 34170 34171 'update' 34172 'no-update' 34173 Generate code that uses (does not use) the load or store 34174 instructions that update the base register to the address of 34175 the calculated memory location. 34176 34177 'popcntb' 34178 'no-popcntb' 34179 Generate code that uses (does not use) the popcount and 34180 double-precision FP reciprocal estimate instruction 34181 implemented on the POWER5 processor and other processors that 34182 support the PowerPC V2.02 architecture. 34183 34184 'popcntd' 34185 'no-popcntd' 34186 Generate code that uses (does not use) the popcount 34187 instruction implemented on the POWER7 processor and other 34188 processors that support the PowerPC V2.06 architecture. 34189 34190 'powerpc-gfxopt' 34191 'no-powerpc-gfxopt' 34192 Generate code that uses (does not use) the optional PowerPC 34193 architecture instructions in the Graphics group, including 34194 floating-point select. 34195 34196 'powerpc-gpopt' 34197 'no-powerpc-gpopt' 34198 Generate code that uses (does not use) the optional PowerPC 34199 architecture instructions in the General Purpose group, 34200 including floating-point square root. 34201 34202 'recip-precision' 34203 'no-recip-precision' 34204 Assume (do not assume) that the reciprocal estimate 34205 instructions provide higher-precision estimates than is 34206 mandated by the PowerPC ABI. 34207 34208 'string' 34209 'no-string' 34210 Generate code that uses (does not use) the load string 34211 instructions and the store string word instructions to save 34212 multiple registers and do small block moves. 34213 34214 'vsx' 34215 'no-vsx' 34216 Generate code that uses (does not use) vector/scalar (VSX) 34217 instructions, and also enable the use of built-in functions 34218 that allow more direct access to the VSX instruction set. In 34219 32-bit code, you cannot enable VSX or AltiVec instructions 34220 unless '-mabi=altivec' is used on the command line. 34221 34222 'friz' 34223 'no-friz' 34224 Generate (do not generate) the 'friz' instruction when the 34225 '-funsafe-math-optimizations' option is used to optimize 34226 rounding a floating-point value to 64-bit integer and back to 34227 floating point. The 'friz' instruction does not return the 34228 same value if the floating-point number is too large to fit in 34229 an integer. 34230 34231 'avoid-indexed-addresses' 34232 'no-avoid-indexed-addresses' 34233 Generate code that tries to avoid (not avoid) the use of 34234 indexed load or store instructions. 34235 34236 'paired' 34237 'no-paired' 34238 Generate code that uses (does not use) the generation of 34239 PAIRED simd instructions. 34240 34241 'longcall' 34242 'no-longcall' 34243 Generate code that assumes (does not assume) that all calls 34244 are far away so that a longer more expensive calling sequence 34245 is required. 34246 34247 'cpu=CPU' 34248 Specify the architecture to generate code for when compiling 34249 the function. If you select the 'target("cpu=power7")' 34250 attribute when generating 32-bit code, VSX and AltiVec 34251 instructions are not generated unless you use the 34252 '-mabi=altivec' option on the command line. 34253 34254 'tune=TUNE' 34255 Specify the architecture to tune for when compiling the 34256 function. If you do not specify the 'target("tune=TUNE")' 34257 attribute and you do specify the 'target("cpu=CPU")' 34258 attribute, compilation tunes for the CPU architecture, and not 34259 the default tuning specified on the command line. 34260 34261 On the PowerPC, the inliner does not inline a function that has 34262 different target options than the caller, unless the callee has a 34263 subset of the target options of the caller. 34264 34265 34266File: gcc.info, Node: RISC-V Function Attributes, Next: RL78 Function Attributes, Prev: PowerPC Function Attributes, Up: Function Attributes 34267 342686.33.27 RISC-V Function Attributes 34269---------------------------------- 34270 34271These function attributes are supported by the RISC-V back end: 34272 34273'naked' 34274 This attribute allows the compiler to construct the requisite 34275 function declaration, while allowing the body of the function to be 34276 assembly code. The specified function will not have 34277 prologue/epilogue sequences generated by the compiler. Only basic 34278 'asm' statements can safely be included in naked functions (*note 34279 Basic Asm::). While using extended 'asm' or a mixture of basic 34280 'asm' and C code may appear to work, they cannot be depended upon 34281 to work reliably and are not supported. 34282 34283'interrupt' 34284 Use this attribute to indicate that the specified function is an 34285 interrupt handler. The compiler generates function entry and exit 34286 sequences suitable for use in an interrupt handler when this 34287 attribute is present. 34288 34289 You can specify the kind of interrupt to be handled by adding an 34290 optional parameter to the interrupt attribute like this: 34291 34292 void f (void) __attribute__ ((interrupt ("user"))); 34293 34294 Permissible values for this parameter are 'user', 'supervisor', and 34295 'machine'. If there is no parameter, then it defaults to 34296 'machine'. 34297 34298 34299File: gcc.info, Node: RL78 Function Attributes, Next: RX Function Attributes, Prev: RISC-V Function Attributes, Up: Function Attributes 34300 343016.33.28 RL78 Function Attributes 34302-------------------------------- 34303 34304These function attributes are supported by the RL78 back end: 34305 34306'interrupt' 34307'brk_interrupt' 34308 These attributes indicate that the specified function is an 34309 interrupt handler. The compiler generates function entry and exit 34310 sequences suitable for use in an interrupt handler when this 34311 attribute is present. 34312 34313 Use 'brk_interrupt' instead of 'interrupt' for handlers intended to 34314 be used with the 'BRK' opcode (i.e. those that must end with 'RETB' 34315 instead of 'RETI'). 34316 34317'naked' 34318 This attribute allows the compiler to construct the requisite 34319 function declaration, while allowing the body of the function to be 34320 assembly code. The specified function will not have 34321 prologue/epilogue sequences generated by the compiler. Only basic 34322 'asm' statements can safely be included in naked functions (*note 34323 Basic Asm::). While using extended 'asm' or a mixture of basic 34324 'asm' and C code may appear to work, they cannot be depended upon 34325 to work reliably and are not supported. 34326 34327 34328File: gcc.info, Node: RX Function Attributes, Next: S/390 Function Attributes, Prev: RL78 Function Attributes, Up: Function Attributes 34329 343306.33.29 RX Function Attributes 34331------------------------------ 34332 34333These function attributes are supported by the RX back end: 34334 34335'fast_interrupt' 34336 Use this attribute on the RX port to indicate that the specified 34337 function is a fast interrupt handler. This is just like the 34338 'interrupt' attribute, except that 'freit' is used to return 34339 instead of 'reit'. 34340 34341'interrupt' 34342 Use this attribute to indicate that the specified function is an 34343 interrupt handler. The compiler generates function entry and exit 34344 sequences suitable for use in an interrupt handler when this 34345 attribute is present. 34346 34347 On RX and RL78 targets, you may specify one or more vector numbers 34348 as arguments to the attribute, as well as naming an alternate table 34349 name. Parameters are handled sequentially, so one handler can be 34350 assigned to multiple entries in multiple tables. One may also pass 34351 the magic string '"$default"' which causes the function to be used 34352 for any unfilled slots in the current table. 34353 34354 This example shows a simple assignment of a function to one vector 34355 in the default table (note that preprocessor macros may be used for 34356 chip-specific symbolic vector names): 34357 void __attribute__ ((interrupt (5))) txd1_handler (); 34358 34359 This example assigns a function to two slots in the default table 34360 (using preprocessor macros defined elsewhere) and makes it the 34361 default for the 'dct' table: 34362 void __attribute__ ((interrupt (RXD1_VECT,RXD2_VECT,"dct","$default"))) 34363 txd1_handler (); 34364 34365'naked' 34366 This attribute allows the compiler to construct the requisite 34367 function declaration, while allowing the body of the function to be 34368 assembly code. The specified function will not have 34369 prologue/epilogue sequences generated by the compiler. Only basic 34370 'asm' statements can safely be included in naked functions (*note 34371 Basic Asm::). While using extended 'asm' or a mixture of basic 34372 'asm' and C code may appear to work, they cannot be depended upon 34373 to work reliably and are not supported. 34374 34375'vector' 34376 This RX attribute is similar to the 'interrupt' attribute, 34377 including its parameters, but does not make the function an 34378 interrupt-handler type function (i.e. it retains the normal C 34379 function calling ABI). See the 'interrupt' attribute for a 34380 description of its arguments. 34381 34382 34383File: gcc.info, Node: S/390 Function Attributes, Next: SH Function Attributes, Prev: RX Function Attributes, Up: Function Attributes 34384 343856.33.30 S/390 Function Attributes 34386--------------------------------- 34387 34388These function attributes are supported on the S/390: 34389 34390'hotpatch (HALFWORDS-BEFORE-FUNCTION-LABEL,HALFWORDS-AFTER-FUNCTION-LABEL)' 34391 34392 On S/390 System z targets, you can use this function attribute to 34393 make GCC generate a "hot-patching" function prologue. If the 34394 '-mhotpatch=' command-line option is used at the same time, the 34395 'hotpatch' attribute takes precedence. The first of the two 34396 arguments specifies the number of halfwords to be added before the 34397 function label. A second argument can be used to specify the 34398 number of halfwords to be added after the function label. For both 34399 arguments the maximum allowed value is 1000000. 34400 34401 If both arguments are zero, hotpatching is disabled. 34402 34403'target (OPTIONS)' 34404 As discussed in *note Common Function Attributes::, this attribute 34405 allows specification of target-specific compilation options. 34406 34407 On S/390, the following options are supported: 34408 34409 'arch=' 34410 'tune=' 34411 'stack-guard=' 34412 'stack-size=' 34413 'branch-cost=' 34414 'warn-framesize=' 34415 'backchain' 34416 'no-backchain' 34417 'hard-dfp' 34418 'no-hard-dfp' 34419 'hard-float' 34420 'soft-float' 34421 'htm' 34422 'no-htm' 34423 'vx' 34424 'no-vx' 34425 'packed-stack' 34426 'no-packed-stack' 34427 'small-exec' 34428 'no-small-exec' 34429 'mvcle' 34430 'no-mvcle' 34431 'warn-dynamicstack' 34432 'no-warn-dynamicstack' 34433 34434 The options work exactly like the S/390 specific command line 34435 options (without the prefix '-m') except that they do not change 34436 any feature macros. For example, 34437 34438 target("no-vx") 34439 34440 does not undefine the '__VEC__' macro. 34441 34442 34443File: gcc.info, Node: SH Function Attributes, Next: Symbian OS Function Attributes, Prev: S/390 Function Attributes, Up: Function Attributes 34444 344456.33.31 SH Function Attributes 34446------------------------------ 34447 34448These function attributes are supported on the SH family of processors: 34449 34450'function_vector' 34451 On SH2A targets, this attribute declares a function to be called 34452 using the TBR relative addressing mode. The argument to this 34453 attribute is the entry number of the same function in a vector 34454 table containing all the TBR relative addressable functions. For 34455 correct operation the TBR must be setup accordingly to point to the 34456 start of the vector table before any functions with this attribute 34457 are invoked. Usually a good place to do the initialization is the 34458 startup routine. The TBR relative vector table can have at max 256 34459 function entries. The jumps to these functions are generated using 34460 a SH2A specific, non delayed branch instruction JSR/N @(disp8,TBR). 34461 You must use GAS and GLD from GNU binutils version 2.7 or later for 34462 this attribute to work correctly. 34463 34464 In an application, for a function being called once, this attribute 34465 saves at least 8 bytes of code; and if other successive calls are 34466 being made to the same function, it saves 2 bytes of code per each 34467 of these calls. 34468 34469'interrupt_handler' 34470 Use this attribute to indicate that the specified function is an 34471 interrupt handler. The compiler generates function entry and exit 34472 sequences suitable for use in an interrupt handler when this 34473 attribute is present. 34474 34475'nosave_low_regs' 34476 Use this attribute on SH targets to indicate that an 34477 'interrupt_handler' function should not save and restore registers 34478 R0..R7. This can be used on SH3* and SH4* targets that have a 34479 second R0..R7 register bank for non-reentrant interrupt handlers. 34480 34481'renesas' 34482 On SH targets this attribute specifies that the function or struct 34483 follows the Renesas ABI. 34484 34485'resbank' 34486 On the SH2A target, this attribute enables the high-speed register 34487 saving and restoration using a register bank for 34488 'interrupt_handler' routines. Saving to the bank is performed 34489 automatically after the CPU accepts an interrupt that uses a 34490 register bank. 34491 34492 The nineteen 32-bit registers comprising general register R0 to 34493 R14, control register GBR, and system registers MACH, MACL, and PR 34494 and the vector table address offset are saved into a register bank. 34495 Register banks are stacked in first-in last-out (FILO) sequence. 34496 Restoration from the bank is executed by issuing a RESBANK 34497 instruction. 34498 34499'sp_switch' 34500 Use this attribute on the SH to indicate an 'interrupt_handler' 34501 function should switch to an alternate stack. It expects a string 34502 argument that names a global variable holding the address of the 34503 alternate stack. 34504 34505 void *alt_stack; 34506 void f () __attribute__ ((interrupt_handler, 34507 sp_switch ("alt_stack"))); 34508 34509'trap_exit' 34510 Use this attribute on the SH for an 'interrupt_handler' to return 34511 using 'trapa' instead of 'rte'. This attribute expects an integer 34512 argument specifying the trap number to be used. 34513 34514'trapa_handler' 34515 On SH targets this function attribute is similar to 34516 'interrupt_handler' but it does not save and restore all registers. 34517 34518 34519File: gcc.info, Node: Symbian OS Function Attributes, Next: V850 Function Attributes, Prev: SH Function Attributes, Up: Function Attributes 34520 345216.33.32 Symbian OS Function Attributes 34522-------------------------------------- 34523 34524*Note Microsoft Windows Function Attributes::, for discussion of the 34525'dllexport' and 'dllimport' attributes. 34526 34527 34528File: gcc.info, Node: V850 Function Attributes, Next: Visium Function Attributes, Prev: Symbian OS Function Attributes, Up: Function Attributes 34529 345306.33.33 V850 Function Attributes 34531-------------------------------- 34532 34533The V850 back end supports these function attributes: 34534 34535'interrupt' 34536'interrupt_handler' 34537 Use these attributes to indicate that the specified function is an 34538 interrupt handler. The compiler generates function entry and exit 34539 sequences suitable for use in an interrupt handler when either 34540 attribute is present. 34541 34542 34543File: gcc.info, Node: Visium Function Attributes, Next: x86 Function Attributes, Prev: V850 Function Attributes, Up: Function Attributes 34544 345456.33.34 Visium Function Attributes 34546---------------------------------- 34547 34548These function attributes are supported by the Visium back end: 34549 34550'interrupt' 34551 Use this attribute to indicate that the specified function is an 34552 interrupt handler. The compiler generates function entry and exit 34553 sequences suitable for use in an interrupt handler when this 34554 attribute is present. 34555 34556 34557File: gcc.info, Node: x86 Function Attributes, Next: Xstormy16 Function Attributes, Prev: Visium Function Attributes, Up: Function Attributes 34558 345596.33.35 x86 Function Attributes 34560------------------------------- 34561 34562These function attributes are supported by the x86 back end: 34563 34564'cdecl' 34565 On the x86-32 targets, the 'cdecl' attribute causes the compiler to 34566 assume that the calling function pops off the stack space used to 34567 pass arguments. This is useful to override the effects of the 34568 '-mrtd' switch. 34569 34570'fastcall' 34571 On x86-32 targets, the 'fastcall' attribute causes the compiler to 34572 pass the first argument (if of integral type) in the register ECX 34573 and the second argument (if of integral type) in the register EDX. 34574 Subsequent and other typed arguments are passed on the stack. The 34575 called function pops the arguments off the stack. If the number of 34576 arguments is variable all arguments are pushed on the stack. 34577 34578'thiscall' 34579 On x86-32 targets, the 'thiscall' attribute causes the compiler to 34580 pass the first argument (if of integral type) in the register ECX. 34581 Subsequent and other typed arguments are passed on the stack. The 34582 called function pops the arguments off the stack. If the number of 34583 arguments is variable all arguments are pushed on the stack. The 34584 'thiscall' attribute is intended for C++ non-static member 34585 functions. As a GCC extension, this calling convention can be used 34586 for C functions and for static member methods. 34587 34588'ms_abi' 34589'sysv_abi' 34590 34591 On 32-bit and 64-bit x86 targets, you can use an ABI attribute to 34592 indicate which calling convention should be used for a function. 34593 The 'ms_abi' attribute tells the compiler to use the Microsoft ABI, 34594 while the 'sysv_abi' attribute tells the compiler to use the ABI 34595 used on GNU/Linux and other systems. The default is to use the 34596 Microsoft ABI when targeting Windows. On all other systems, the 34597 default is the x86/AMD ABI. 34598 34599 Note, the 'ms_abi' attribute for Microsoft Windows 64-bit targets 34600 currently requires the '-maccumulate-outgoing-args' option. 34601 34602'callee_pop_aggregate_return (NUMBER)' 34603 34604 On x86-32 targets, you can use this attribute to control how 34605 aggregates are returned in memory. If the caller is responsible 34606 for popping the hidden pointer together with the rest of the 34607 arguments, specify NUMBER equal to zero. If callee is responsible 34608 for popping the hidden pointer, specify NUMBER equal to one. 34609 34610 The default x86-32 ABI assumes that the callee pops the stack for 34611 hidden pointer. However, on x86-32 Microsoft Windows targets, the 34612 compiler assumes that the caller pops the stack for hidden pointer. 34613 34614'ms_hook_prologue' 34615 34616 On 32-bit and 64-bit x86 targets, you can use this function 34617 attribute to make GCC generate the "hot-patching" function prologue 34618 used in Win32 API functions in Microsoft Windows XP Service Pack 2 34619 and newer. 34620 34621'naked' 34622 This attribute allows the compiler to construct the requisite 34623 function declaration, while allowing the body of the function to be 34624 assembly code. The specified function will not have 34625 prologue/epilogue sequences generated by the compiler. Only basic 34626 'asm' statements can safely be included in naked functions (*note 34627 Basic Asm::). While using extended 'asm' or a mixture of basic 34628 'asm' and C code may appear to work, they cannot be depended upon 34629 to work reliably and are not supported. 34630 34631'regparm (NUMBER)' 34632 On x86-32 targets, the 'regparm' attribute causes the compiler to 34633 pass arguments number one to NUMBER if they are of integral type in 34634 registers EAX, EDX, and ECX instead of on the stack. Functions 34635 that take a variable number of arguments continue to be passed all 34636 of their arguments on the stack. 34637 34638 Beware that on some ELF systems this attribute is unsuitable for 34639 global functions in shared libraries with lazy binding (which is 34640 the default). Lazy binding sends the first call via resolving code 34641 in the loader, which might assume EAX, EDX and ECX can be 34642 clobbered, as per the standard calling conventions. Solaris 8 is 34643 affected by this. Systems with the GNU C Library version 2.1 or 34644 higher and FreeBSD are believed to be safe since the loaders there 34645 save EAX, EDX and ECX. (Lazy binding can be disabled with the 34646 linker or the loader if desired, to avoid the problem.) 34647 34648'sseregparm' 34649 On x86-32 targets with SSE support, the 'sseregparm' attribute 34650 causes the compiler to pass up to 3 floating-point arguments in SSE 34651 registers instead of on the stack. Functions that take a variable 34652 number of arguments continue to pass all of their floating-point 34653 arguments on the stack. 34654 34655'force_align_arg_pointer' 34656 On x86 targets, the 'force_align_arg_pointer' attribute may be 34657 applied to individual function definitions, generating an alternate 34658 prologue and epilogue that realigns the run-time stack if 34659 necessary. This supports mixing legacy codes that run with a 34660 4-byte aligned stack with modern codes that keep a 16-byte stack 34661 for SSE compatibility. 34662 34663'stdcall' 34664 On x86-32 targets, the 'stdcall' attribute causes the compiler to 34665 assume that the called function pops off the stack space used to 34666 pass arguments, unless it takes a variable number of arguments. 34667 34668'no_caller_saved_registers' 34669 Use this attribute to indicate that the specified function has no 34670 caller-saved registers. That is, all registers are callee-saved. 34671 For example, this attribute can be used for a function called from 34672 an interrupt handler. The compiler generates proper function entry 34673 and exit sequences to save and restore any modified registers, 34674 except for the EFLAGS register. Since GCC doesn't preserve SSE, 34675 MMX nor x87 states, the GCC option '-mgeneral-regs-only' should be 34676 used to compile functions with 'no_caller_saved_registers' 34677 attribute. 34678 34679'interrupt' 34680 Use this attribute to indicate that the specified function is an 34681 interrupt handler or an exception handler (depending on parameters 34682 passed to the function, explained further). The compiler generates 34683 function entry and exit sequences suitable for use in an interrupt 34684 handler when this attribute is present. The 'IRET' instruction, 34685 instead of the 'RET' instruction, is used to return from interrupt 34686 handlers. All registers, except for the EFLAGS register which is 34687 restored by the 'IRET' instruction, are preserved by the compiler. 34688 Since GCC doesn't preserve SSE, MMX nor x87 states, the GCC option 34689 '-mgeneral-regs-only' should be used to compile interrupt and 34690 exception handlers. 34691 34692 Any interruptible-without-stack-switch code must be compiled with 34693 '-mno-red-zone' since interrupt handlers can and will, because of 34694 the hardware design, touch the red zone. 34695 34696 An interrupt handler must be declared with a mandatory pointer 34697 argument: 34698 34699 struct interrupt_frame; 34700 34701 __attribute__ ((interrupt)) 34702 void 34703 f (struct interrupt_frame *frame) 34704 { 34705 } 34706 34707 and you must define 'struct interrupt_frame' as described in the 34708 processor's manual. 34709 34710 Exception handlers differ from interrupt handlers because the 34711 system pushes an error code on the stack. An exception handler 34712 declaration is similar to that for an interrupt handler, but with a 34713 different mandatory function signature. The compiler arranges to 34714 pop the error code off the stack before the 'IRET' instruction. 34715 34716 #ifdef __x86_64__ 34717 typedef unsigned long long int uword_t; 34718 #else 34719 typedef unsigned int uword_t; 34720 #endif 34721 34722 struct interrupt_frame; 34723 34724 __attribute__ ((interrupt)) 34725 void 34726 f (struct interrupt_frame *frame, uword_t error_code) 34727 { 34728 ... 34729 } 34730 34731 Exception handlers should only be used for exceptions that push an 34732 error code; you should use an interrupt handler in other cases. 34733 The system will crash if the wrong kind of handler is used. 34734 34735'target (OPTIONS)' 34736 As discussed in *note Common Function Attributes::, this attribute 34737 allows specification of target-specific compilation options. 34738 34739 On the x86, the following options are allowed: 34740 '3dnow' 34741 'no-3dnow' 34742 Enable/disable the generation of the 3DNow! instructions. 34743 34744 '3dnowa' 34745 'no-3dnowa' 34746 Enable/disable the generation of the enhanced 3DNow! 34747 instructions. 34748 34749 'abm' 34750 'no-abm' 34751 Enable/disable the generation of the advanced bit 34752 instructions. 34753 34754 'adx' 34755 'no-adx' 34756 Enable/disable the generation of the ADX instructions. 34757 34758 'aes' 34759 'no-aes' 34760 Enable/disable the generation of the AES instructions. 34761 34762 'avx' 34763 'no-avx' 34764 Enable/disable the generation of the AVX instructions. 34765 34766 'avx2' 34767 'no-avx2' 34768 Enable/disable the generation of the AVX2 instructions. 34769 34770 'avx5124fmaps' 34771 'no-avx5124fmaps' 34772 Enable/disable the generation of the AVX5124FMAPS 34773 instructions. 34774 34775 'avx5124vnniw' 34776 'no-avx5124vnniw' 34777 Enable/disable the generation of the AVX5124VNNIW 34778 instructions. 34779 34780 'avx512bitalg' 34781 'no-avx512bitalg' 34782 Enable/disable the generation of the AVX512BITALG 34783 instructions. 34784 34785 'avx512bw' 34786 'no-avx512bw' 34787 Enable/disable the generation of the AVX512BW instructions. 34788 34789 'avx512cd' 34790 'no-avx512cd' 34791 Enable/disable the generation of the AVX512CD instructions. 34792 34793 'avx512dq' 34794 'no-avx512dq' 34795 Enable/disable the generation of the AVX512DQ instructions. 34796 34797 'avx512er' 34798 'no-avx512er' 34799 Enable/disable the generation of the AVX512ER instructions. 34800 34801 'avx512f' 34802 'no-avx512f' 34803 Enable/disable the generation of the AVX512F instructions. 34804 34805 'avx512ifma' 34806 'no-avx512ifma' 34807 Enable/disable the generation of the AVX512IFMA instructions. 34808 34809 'avx512pf' 34810 'no-avx512pf' 34811 Enable/disable the generation of the AVX512PF instructions. 34812 34813 'avx512vbmi' 34814 'no-avx512vbmi' 34815 Enable/disable the generation of the AVX512VBMI instructions. 34816 34817 'avx512vbmi2' 34818 'no-avx512vbmi2' 34819 Enable/disable the generation of the AVX512VBMI2 instructions. 34820 34821 'avx512vl' 34822 'no-avx512vl' 34823 Enable/disable the generation of the AVX512VL instructions. 34824 34825 'avx512vnni' 34826 'no-avx512vnni' 34827 Enable/disable the generation of the AVX512VNNI instructions. 34828 34829 'avx512vpopcntdq' 34830 'no-avx512vpopcntdq' 34831 Enable/disable the generation of the AVX512VPOPCNTDQ 34832 instructions. 34833 34834 'bmi' 34835 'no-bmi' 34836 Enable/disable the generation of the BMI instructions. 34837 34838 'bmi2' 34839 'no-bmi2' 34840 Enable/disable the generation of the BMI2 instructions. 34841 34842 'cldemote' 34843 'no-cldemote' 34844 Enable/disable the generation of the CLDEMOTE instructions. 34845 34846 'clflushopt' 34847 'no-clflushopt' 34848 Enable/disable the generation of the CLFLUSHOPT instructions. 34849 34850 'clwb' 34851 'no-clwb' 34852 Enable/disable the generation of the CLWB instructions. 34853 34854 'clzero' 34855 'no-clzero' 34856 Enable/disable the generation of the CLZERO instructions. 34857 34858 'crc32' 34859 'no-crc32' 34860 Enable/disable the generation of the CRC32 instructions. 34861 34862 'cx16' 34863 'no-cx16' 34864 Enable/disable the generation of the CMPXCHG16B instructions. 34865 34866 'default' 34867 *Note Function Multiversioning::, where it is used to specify 34868 the default function version. 34869 34870 'f16c' 34871 'no-f16c' 34872 Enable/disable the generation of the F16C instructions. 34873 34874 'fma' 34875 'no-fma' 34876 Enable/disable the generation of the FMA instructions. 34877 34878 'fma4' 34879 'no-fma4' 34880 Enable/disable the generation of the FMA4 instructions. 34881 34882 'fsgsbase' 34883 'no-fsgsbase' 34884 Enable/disable the generation of the FSGSBASE instructions. 34885 34886 'fxsr' 34887 'no-fxsr' 34888 Enable/disable the generation of the FXSR instructions. 34889 34890 'gfni' 34891 'no-gfni' 34892 Enable/disable the generation of the GFNI instructions. 34893 34894 'hle' 34895 'no-hle' 34896 Enable/disable the generation of the HLE instruction prefixes. 34897 34898 'lwp' 34899 'no-lwp' 34900 Enable/disable the generation of the LWP instructions. 34901 34902 'lzcnt' 34903 'no-lzcnt' 34904 Enable/disable the generation of the LZCNT instructions. 34905 34906 'mmx' 34907 'no-mmx' 34908 Enable/disable the generation of the MMX instructions. 34909 34910 'movbe' 34911 'no-movbe' 34912 Enable/disable the generation of the MOVBE instructions. 34913 34914 'movdir64b' 34915 'no-movdir64b' 34916 Enable/disable the generation of the MOVDIR64B instructions. 34917 34918 'movdiri' 34919 'no-movdiri' 34920 Enable/disable the generation of the MOVDIRI instructions. 34921 34922 'mwaitx' 34923 'no-mwaitx' 34924 Enable/disable the generation of the MWAITX instructions. 34925 34926 'pclmul' 34927 'no-pclmul' 34928 Enable/disable the generation of the PCLMUL instructions. 34929 34930 'pconfig' 34931 'no-pconfig' 34932 Enable/disable the generation of the PCONFIG instructions. 34933 34934 'pku' 34935 'no-pku' 34936 Enable/disable the generation of the PKU instructions. 34937 34938 'popcnt' 34939 'no-popcnt' 34940 Enable/disable the generation of the POPCNT instruction. 34941 34942 'prefetchwt1' 34943 'no-prefetchwt1' 34944 Enable/disable the generation of the PREFETCHWT1 instructions. 34945 34946 'prfchw' 34947 'no-prfchw' 34948 Enable/disable the generation of the PREFETCHW instruction. 34949 34950 'ptwrite' 34951 'no-ptwrite' 34952 Enable/disable the generation of the PTWRITE instructions. 34953 34954 'rdpid' 34955 'no-rdpid' 34956 Enable/disable the generation of the RDPID instructions. 34957 34958 'rdrnd' 34959 'no-rdrnd' 34960 Enable/disable the generation of the RDRND instructions. 34961 34962 'rdseed' 34963 'no-rdseed' 34964 Enable/disable the generation of the RDSEED instructions. 34965 34966 'rtm' 34967 'no-rtm' 34968 Enable/disable the generation of the RTM instructions. 34969 34970 'sahf' 34971 'no-sahf' 34972 Enable/disable the generation of the SAHF instructions. 34973 34974 'sgx' 34975 'no-sgx' 34976 Enable/disable the generation of the SGX instructions. 34977 34978 'sha' 34979 'no-sha' 34980 Enable/disable the generation of the SHA instructions. 34981 34982 'shstk' 34983 'no-shstk' 34984 Enable/disable the shadow stack built-in functions from CET. 34985 34986 'sse' 34987 'no-sse' 34988 Enable/disable the generation of the SSE instructions. 34989 34990 'sse2' 34991 'no-sse2' 34992 Enable/disable the generation of the SSE2 instructions. 34993 34994 'sse3' 34995 'no-sse3' 34996 Enable/disable the generation of the SSE3 instructions. 34997 34998 'sse4' 34999 'no-sse4' 35000 Enable/disable the generation of the SSE4 instructions (both 35001 SSE4.1 and SSE4.2). 35002 35003 'sse4.1' 35004 'no-sse4.1' 35005 Enable/disable the generation of the sse4.1 instructions. 35006 35007 'sse4.2' 35008 'no-sse4.2' 35009 Enable/disable the generation of the sse4.2 instructions. 35010 35011 'sse4a' 35012 'no-sse4a' 35013 Enable/disable the generation of the SSE4A instructions. 35014 35015 'ssse3' 35016 'no-ssse3' 35017 Enable/disable the generation of the SSSE3 instructions. 35018 35019 'tbm' 35020 'no-tbm' 35021 Enable/disable the generation of the TBM instructions. 35022 35023 'vaes' 35024 'no-vaes' 35025 Enable/disable the generation of the VAES instructions. 35026 35027 'vpclmulqdq' 35028 'no-vpclmulqdq' 35029 Enable/disable the generation of the VPCLMULQDQ instructions. 35030 35031 'waitpkg' 35032 'no-waitpkg' 35033 Enable/disable the generation of the WAITPKG instructions. 35034 35035 'wbnoinvd' 35036 'no-wbnoinvd' 35037 Enable/disable the generation of the WBNOINVD instructions. 35038 35039 'xop' 35040 'no-xop' 35041 Enable/disable the generation of the XOP instructions. 35042 35043 'xsave' 35044 'no-xsave' 35045 Enable/disable the generation of the XSAVE instructions. 35046 35047 'xsavec' 35048 'no-xsavec' 35049 Enable/disable the generation of the XSAVEC instructions. 35050 35051 'xsaveopt' 35052 'no-xsaveopt' 35053 Enable/disable the generation of the XSAVEOPT instructions. 35054 35055 'xsaves' 35056 'no-xsaves' 35057 Enable/disable the generation of the XSAVES instructions. 35058 35059 'cld' 35060 'no-cld' 35061 Enable/disable the generation of the CLD before string moves. 35062 35063 'fancy-math-387' 35064 'no-fancy-math-387' 35065 Enable/disable the generation of the 'sin', 'cos', and 'sqrt' 35066 instructions on the 387 floating-point unit. 35067 35068 'ieee-fp' 35069 'no-ieee-fp' 35070 Enable/disable the generation of floating point that depends 35071 on IEEE arithmetic. 35072 35073 'inline-all-stringops' 35074 'no-inline-all-stringops' 35075 Enable/disable inlining of string operations. 35076 35077 'inline-stringops-dynamically' 35078 'no-inline-stringops-dynamically' 35079 Enable/disable the generation of the inline code to do small 35080 string operations and calling the library routines for large 35081 operations. 35082 35083 'align-stringops' 35084 'no-align-stringops' 35085 Do/do not align destination of inlined string operations. 35086 35087 'recip' 35088 'no-recip' 35089 Enable/disable the generation of RCPSS, RCPPS, RSQRTSS and 35090 RSQRTPS instructions followed an additional Newton-Raphson 35091 step instead of doing a floating-point division. 35092 35093 'arch=ARCH' 35094 Specify the architecture to generate code for in compiling the 35095 function. 35096 35097 'tune=TUNE' 35098 Specify the architecture to tune for in compiling the 35099 function. 35100 35101 'fpmath=FPMATH' 35102 Specify which floating-point unit to use. You must specify 35103 the 'target("fpmath=sse,387")' option as 35104 'target("fpmath=sse+387")' because the comma would separate 35105 different options. 35106 35107 'indirect_branch("CHOICE")' 35108 On x86 targets, the 'indirect_branch' attribute causes the 35109 compiler to convert indirect call and jump with CHOICE. 35110 'keep' keeps indirect call and jump unmodified. 'thunk' 35111 converts indirect call and jump to call and return thunk. 35112 'thunk-inline' converts indirect call and jump to inlined call 35113 and return thunk. 'thunk-extern' converts indirect call and 35114 jump to external call and return thunk provided in a separate 35115 object file. 35116 35117 'function_return("CHOICE")' 35118 On x86 targets, the 'function_return' attribute causes the 35119 compiler to convert function return with CHOICE. 'keep' keeps 35120 function return unmodified. 'thunk' converts function return 35121 to call and return thunk. 'thunk-inline' converts function 35122 return to inlined call and return thunk. 'thunk-extern' 35123 converts function return to external call and return thunk 35124 provided in a separate object file. 35125 35126 'nocf_check' 35127 The 'nocf_check' attribute on a function is used to inform the 35128 compiler that the function's prologue should not be 35129 instrumented when compiled with the '-fcf-protection=branch' 35130 option. The compiler assumes that the function's address is a 35131 valid target for a control-flow transfer. 35132 35133 The 'nocf_check' attribute on a type of pointer to function is 35134 used to inform the compiler that a call through the pointer 35135 should not be instrumented when compiled with the 35136 '-fcf-protection=branch' option. The compiler assumes that 35137 the function's address from the pointer is a valid target for 35138 a control-flow transfer. A direct function call through a 35139 function name is assumed to be a safe call thus direct calls 35140 are not instrumented by the compiler. 35141 35142 The 'nocf_check' attribute is applied to an object's type. In 35143 case of assignment of a function address or a function pointer 35144 to another pointer, the attribute is not carried over from the 35145 right-hand object's type; the type of left-hand object stays 35146 unchanged. The compiler checks for 'nocf_check' attribute 35147 mismatch and reports a warning in case of mismatch. 35148 35149 { 35150 int foo (void) __attribute__(nocf_check); 35151 void (*foo1)(void) __attribute__(nocf_check); 35152 void (*foo2)(void); 35153 35154 /* foo's address is assumed to be valid. */ 35155 int 35156 foo (void) 35157 35158 /* This call site is not checked for control-flow 35159 validity. */ 35160 (*foo1)(); 35161 35162 /* A warning is issued about attribute mismatch. */ 35163 foo1 = foo2; 35164 35165 /* This call site is still not checked. */ 35166 (*foo1)(); 35167 35168 /* This call site is checked. */ 35169 (*foo2)(); 35170 35171 /* A warning is issued about attribute mismatch. */ 35172 foo2 = foo1; 35173 35174 /* This call site is still checked. */ 35175 (*foo2)(); 35176 35177 return 0; 35178 } 35179 35180 'cf_check' 35181 35182 The 'cf_check' attribute on a function is used to inform the 35183 compiler that ENDBR instruction should be placed at the 35184 function entry when '-fcf-protection=branch' is enabled. 35185 35186 'indirect_return' 35187 35188 The 'indirect_return' attribute can be applied to a function, 35189 as well as variable or type of function pointer to inform the 35190 compiler that the function may return via indirect branch. 35191 35192 'fentry_name("NAME")' 35193 On x86 targets, the 'fentry_name' attribute sets the function 35194 to call on function entry when function instrumentation is 35195 enabled with '-pg -mfentry'. When NAME is nop then a 5 byte 35196 nop sequence is generated. 35197 35198 'fentry_section("NAME")' 35199 On x86 targets, the 'fentry_section' attribute sets the name 35200 of the section to record function entry instrumentation calls 35201 in when enabled with '-pg -mrecord-mcount' 35202 35203 On the x86, the inliner does not inline a function that has 35204 different target options than the caller, unless the callee has a 35205 subset of the target options of the caller. For example a function 35206 declared with 'target("sse3")' can inline a function with 35207 'target("sse2")', since '-msse3' implies '-msse2'. 35208 35209 35210File: gcc.info, Node: Xstormy16 Function Attributes, Prev: x86 Function Attributes, Up: Function Attributes 35211 352126.33.36 Xstormy16 Function Attributes 35213------------------------------------- 35214 35215These function attributes are supported by the Xstormy16 back end: 35216 35217'interrupt' 35218 Use this attribute to indicate that the specified function is an 35219 interrupt handler. The compiler generates function entry and exit 35220 sequences suitable for use in an interrupt handler when this 35221 attribute is present. 35222 35223 35224File: gcc.info, Node: Variable Attributes, Next: Type Attributes, Prev: Function Attributes, Up: C Extensions 35225 352266.34 Specifying Attributes of Variables 35227======================================= 35228 35229The keyword '__attribute__' allows you to specify special properties of 35230variables, function parameters, or structure, union, and, in C++, class 35231members. This '__attribute__' keyword is followed by an attribute 35232specification enclosed in double parentheses. Some attributes are 35233currently defined generically for variables. Other attributes are 35234defined for variables on particular target systems. Other attributes 35235are available for functions (*note Function Attributes::), labels (*note 35236Label Attributes::), enumerators (*note Enumerator Attributes::), 35237statements (*note Statement Attributes::), and for types (*note Type 35238Attributes::). Other front ends might define more attributes (*note 35239Extensions to the C++ Language: C++ Extensions.). 35240 35241 *Note Attribute Syntax::, for details of the exact syntax for using 35242attributes. 35243 35244* Menu: 35245 35246* Common Variable Attributes:: 35247* ARC Variable Attributes:: 35248* AVR Variable Attributes:: 35249* Blackfin Variable Attributes:: 35250* H8/300 Variable Attributes:: 35251* IA-64 Variable Attributes:: 35252* M32R/D Variable Attributes:: 35253* MeP Variable Attributes:: 35254* Microsoft Windows Variable Attributes:: 35255* MSP430 Variable Attributes:: 35256* Nvidia PTX Variable Attributes:: 35257* PowerPC Variable Attributes:: 35258* RL78 Variable Attributes:: 35259* V850 Variable Attributes:: 35260* x86 Variable Attributes:: 35261* Xstormy16 Variable Attributes:: 35262 35263 35264File: gcc.info, Node: Common Variable Attributes, Next: ARC Variable Attributes, Up: Variable Attributes 35265 352666.34.1 Common Variable Attributes 35267--------------------------------- 35268 35269The following attributes are supported on most targets. 35270 35271'alias ("TARGET")' 35272 The 'alias' variable attribute causes the declaration to be emitted 35273 as an alias for another symbol known as an "alias target". Except 35274 for top-level qualifiers the alias target must have the same type 35275 as the alias. For instance, the following 35276 35277 int var_target; 35278 extern int __attribute__ ((alias ("var_target"))) var_alias; 35279 35280 defines 'var_alias' to be an alias for the 'var_target' variable. 35281 35282 It is an error if the alias target is not defined in the same 35283 translation unit as the alias. 35284 35285 Note that in the absence of the attribute GCC assumes that distinct 35286 declarations with external linkage denote distinct objects. Using 35287 both the alias and the alias target to access the same object is 35288 undefined in a translation unit without a declaration of the alias 35289 with the attribute. 35290 35291 This attribute requires assembler and object file support, and may 35292 not be available on all targets. 35293 35294'aligned' 35295'aligned (ALIGNMENT)' 35296 The 'aligned' attribute specifies a minimum alignment for the 35297 variable or structure field, measured in bytes. When specified, 35298 ALIGNMENT must be an integer constant power of 2. Specifying no 35299 ALIGNMENT argument implies the maximum alignment for the target, 35300 which is often, but by no means always, 8 or 16 bytes. 35301 35302 For example, the declaration: 35303 35304 int x __attribute__ ((aligned (16))) = 0; 35305 35306 causes the compiler to allocate the global variable 'x' on a 35307 16-byte boundary. On a 68040, this could be used in conjunction 35308 with an 'asm' expression to access the 'move16' instruction which 35309 requires 16-byte aligned operands. 35310 35311 You can also specify the alignment of structure fields. For 35312 example, to create a double-word aligned 'int' pair, you could 35313 write: 35314 35315 struct foo { int x[2] __attribute__ ((aligned (8))); }; 35316 35317 This is an alternative to creating a union with a 'double' member, 35318 which forces the union to be double-word aligned. 35319 35320 As in the preceding examples, you can explicitly specify the 35321 alignment (in bytes) that you wish the compiler to use for a given 35322 variable or structure field. Alternatively, you can leave out the 35323 alignment factor and just ask the compiler to align a variable or 35324 field to the default alignment for the target architecture you are 35325 compiling for. The default alignment is sufficient for all scalar 35326 types, but may not be enough for all vector types on a target that 35327 supports vector operations. The default alignment is fixed for a 35328 particular target ABI. 35329 35330 GCC also provides a target specific macro '__BIGGEST_ALIGNMENT__', 35331 which is the largest alignment ever used for any data type on the 35332 target machine you are compiling for. For example, you could 35333 write: 35334 35335 short array[3] __attribute__ ((aligned (__BIGGEST_ALIGNMENT__))); 35336 35337 The compiler automatically sets the alignment for the declared 35338 variable or field to '__BIGGEST_ALIGNMENT__'. Doing this can often 35339 make copy operations more efficient, because the compiler can use 35340 whatever instructions copy the biggest chunks of memory when 35341 performing copies to or from the variables or fields that you have 35342 aligned this way. Note that the value of '__BIGGEST_ALIGNMENT__' 35343 may change depending on command-line options. 35344 35345 When used on a struct, or struct member, the 'aligned' attribute 35346 can only increase the alignment; in order to decrease it, the 35347 'packed' attribute must be specified as well. When used as part of 35348 a typedef, the 'aligned' attribute can both increase and decrease 35349 alignment, and specifying the 'packed' attribute generates a 35350 warning. 35351 35352 Note that the effectiveness of 'aligned' attributes for static 35353 variables may be limited by inherent limitations in the system 35354 linker and/or object file format. On some systems, the linker is 35355 only able to arrange for variables to be aligned up to a certain 35356 maximum alignment. (For some linkers, the maximum supported 35357 alignment may be very very small.) If your linker is only able to 35358 align variables up to a maximum of 8-byte alignment, then 35359 specifying 'aligned(16)' in an '__attribute__' still only provides 35360 you with 8-byte alignment. See your linker documentation for 35361 further information. 35362 35363 Stack variables are not affected by linker restrictions; GCC can 35364 properly align them on any target. 35365 35366 The 'aligned' attribute can also be used for functions (*note 35367 Common Function Attributes::.) 35368 35369'warn_if_not_aligned (ALIGNMENT)' 35370 This attribute specifies a threshold for the structure field, 35371 measured in bytes. If the structure field is aligned below the 35372 threshold, a warning will be issued. For example, the declaration: 35373 35374 struct foo 35375 { 35376 int i1; 35377 int i2; 35378 unsigned long long x __attribute__ ((warn_if_not_aligned (16))); 35379 }; 35380 35381 causes the compiler to issue an warning on 'struct foo', like 35382 'warning: alignment 8 of 'struct foo' is less than 16'. The 35383 compiler also issues a warning, like 'warning: 'x' offset 8 in 35384 'struct foo' isn't aligned to 16', when the structure field has the 35385 misaligned offset: 35386 35387 struct __attribute__ ((aligned (16))) foo 35388 { 35389 int i1; 35390 int i2; 35391 unsigned long long x __attribute__ ((warn_if_not_aligned (16))); 35392 }; 35393 35394 This warning can be disabled by '-Wno-if-not-aligned'. The 35395 'warn_if_not_aligned' attribute can also be used for types (*note 35396 Common Type Attributes::.) 35397 35398'alloc_size (POSITION)' 35399'alloc_size (POSITION-1, POSITION-2)' 35400 The 'alloc_size' variable attribute may be applied to the 35401 declaration of a pointer to a function that returns a pointer and 35402 takes at least one argument of an integer type. It indicates that 35403 the returned pointer points to an object whose size is given by the 35404 function argument at POSITION-1, or by the product of the arguments 35405 at POSITION-1 and POSITION-2. Meaningful sizes are positive values 35406 less than 'PTRDIFF_MAX'. Other sizes are disagnosed when detected. 35407 GCC uses this information to improve the results of 35408 '__builtin_object_size'. 35409 35410 For instance, the following declarations 35411 35412 typedef __attribute__ ((alloc_size (1, 2))) void* 35413 (*calloc_ptr) (size_t, size_t); 35414 typedef __attribute__ ((alloc_size (1))) void* 35415 (*malloc_ptr) (size_t); 35416 35417 specify that 'calloc_ptr' is a pointer of a function that, like the 35418 standard C function 'calloc', returns an object whose size is given 35419 by the product of arguments 1 and 2, and similarly, that 35420 'malloc_ptr', like the standard C function 'malloc', returns an 35421 object whose size is given by argument 1 to the function. 35422 35423'cleanup (CLEANUP_FUNCTION)' 35424 The 'cleanup' attribute runs a function when the variable goes out 35425 of scope. This attribute can only be applied to auto function 35426 scope variables; it may not be applied to parameters or variables 35427 with static storage duration. The function must take one 35428 parameter, a pointer to a type compatible with the variable. The 35429 return value of the function (if any) is ignored. 35430 35431 If '-fexceptions' is enabled, then CLEANUP_FUNCTION is run during 35432 the stack unwinding that happens during the processing of the 35433 exception. Note that the 'cleanup' attribute does not allow the 35434 exception to be caught, only to perform an action. It is undefined 35435 what happens if CLEANUP_FUNCTION does not return normally. 35436 35437'common' 35438'nocommon' 35439 The 'common' attribute requests GCC to place a variable in "common" 35440 storage. The 'nocommon' attribute requests the opposite--to 35441 allocate space for it directly. 35442 35443 These attributes override the default chosen by the '-fno-common' 35444 and '-fcommon' flags respectively. 35445 35446'copy' 35447'copy (VARIABLE)' 35448 The 'copy' attribute applies the set of attributes with which 35449 VARIABLE has been declared to the declaration of the variable to 35450 which the attribute is applied. The attribute is designed for 35451 libraries that define aliases that are expected to specify the same 35452 set of attributes as the aliased symbols. The 'copy' attribute can 35453 be used with variables, functions or types. However, the kind of 35454 symbol to which the attribute is applied (either varible or 35455 function) must match the kind of symbol to which the argument 35456 refers. The 'copy' attribute copies only syntactic and semantic 35457 attributes but not attributes that affect a symbol's linkage or 35458 visibility such as 'alias', 'visibility', or 'weak'. The 35459 'deprecated' attribute is also not copied. *Note Common Function 35460 Attributes::. *Note Common Type Attributes::. 35461 35462'deprecated' 35463'deprecated (MSG)' 35464 The 'deprecated' attribute results in a warning if the variable is 35465 used anywhere in the source file. This is useful when identifying 35466 variables that are expected to be removed in a future version of a 35467 program. The warning also includes the location of the declaration 35468 of the deprecated variable, to enable users to easily find further 35469 information about why the variable is deprecated, or what they 35470 should do instead. Note that the warning only occurs for uses: 35471 35472 extern int old_var __attribute__ ((deprecated)); 35473 extern int old_var; 35474 int new_fn () { return old_var; } 35475 35476 results in a warning on line 3 but not line 2. The optional MSG 35477 argument, which must be a string, is printed in the warning if 35478 present. 35479 35480 The 'deprecated' attribute can also be used for functions and types 35481 (*note Common Function Attributes::, *note Common Type 35482 Attributes::). 35483 35484 The message attached to the attribute is affected by the setting of 35485 the '-fmessage-length' option. 35486 35487'mode (MODE)' 35488 This attribute specifies the data type for the 35489 declaration--whichever type corresponds to the mode MODE. This in 35490 effect lets you request an integer or floating-point type according 35491 to its width. 35492 35493 *Note (gccint)Machine Modes::, for a list of the possible keywords 35494 for MODE. You may also specify a mode of 'byte' or '__byte__' to 35495 indicate the mode corresponding to a one-byte integer, 'word' or 35496 '__word__' for the mode of a one-word integer, and 'pointer' or 35497 '__pointer__' for the mode used to represent pointers. 35498 35499'nonstring' 35500 The 'nonstring' variable attribute specifies that an object or 35501 member declaration with type array of 'char', 'signed char', or 35502 'unsigned char', or pointer to such a type is intended to store 35503 character arrays that do not necessarily contain a terminating 35504 'NUL'. This is useful in detecting uses of such arrays or pointers 35505 with functions that expect 'NUL'-terminated strings, and to avoid 35506 warnings when such an array or pointer is used as an argument to a 35507 bounded string manipulation function such as 'strncpy'. For 35508 example, without the attribute, GCC will issue a warning for the 35509 'strncpy' call below because it may truncate the copy without 35510 appending the terminating 'NUL' character. Using the attribute 35511 makes it possible to suppress the warning. However, when the array 35512 is declared with the attribute the call to 'strlen' is diagnosed 35513 because when the array doesn't contain a 'NUL'-terminated string 35514 the call is undefined. To copy, compare, of search non-string 35515 character arrays use the 'memcpy', 'memcmp', 'memchr', and other 35516 functions that operate on arrays of bytes. In addition, calling 35517 'strnlen' and 'strndup' with such arrays is safe provided a 35518 suitable bound is specified, and not diagnosed. 35519 35520 struct Data 35521 { 35522 char name [32] __attribute__ ((nonstring)); 35523 }; 35524 35525 int f (struct Data *pd, const char *s) 35526 { 35527 strncpy (pd->name, s, sizeof pd->name); 35528 ... 35529 return strlen (pd->name); // unsafe, gets a warning 35530 } 35531 35532'packed' 35533 The 'packed' attribute specifies that a structure member should 35534 have the smallest possible alignment--one bit for a bit-field and 35535 one byte otherwise, unless a larger value is specified with the 35536 'aligned' attribute. The attribute does not apply to non-member 35537 objects. 35538 35539 For example in the structure below, the member array 'x' is packed 35540 so that it immediately follows 'a' with no intervening padding: 35541 35542 struct foo 35543 { 35544 char a; 35545 int x[2] __attribute__ ((packed)); 35546 }; 35547 35548 _Note:_ The 4.1, 4.2 and 4.3 series of GCC ignore the 'packed' 35549 attribute on bit-fields of type 'char'. This has been fixed in GCC 35550 4.4 but the change can lead to differences in the structure layout. 35551 See the documentation of '-Wpacked-bitfield-compat' for more 35552 information. 35553 35554'section ("SECTION-NAME")' 35555 Normally, the compiler places the objects it generates in sections 35556 like 'data' and 'bss'. Sometimes, however, you need additional 35557 sections, or you need certain particular variables to appear in 35558 special sections, for example to map to special hardware. The 35559 'section' attribute specifies that a variable (or function) lives 35560 in a particular section. For example, this small program uses 35561 several specific section names: 35562 35563 struct duart a __attribute__ ((section ("DUART_A"))) = { 0 }; 35564 struct duart b __attribute__ ((section ("DUART_B"))) = { 0 }; 35565 char stack[10000] __attribute__ ((section ("STACK"))) = { 0 }; 35566 int init_data __attribute__ ((section ("INITDATA"))); 35567 35568 main() 35569 { 35570 /* Initialize stack pointer */ 35571 init_sp (stack + sizeof (stack)); 35572 35573 /* Initialize initialized data */ 35574 memcpy (&init_data, &data, &edata - &data); 35575 35576 /* Turn on the serial ports */ 35577 init_duart (&a); 35578 init_duart (&b); 35579 } 35580 35581 Use the 'section' attribute with _global_ variables and not _local_ 35582 variables, as shown in the example. 35583 35584 You may use the 'section' attribute with initialized or 35585 uninitialized global variables but the linker requires each object 35586 be defined once, with the exception that uninitialized variables 35587 tentatively go in the 'common' (or 'bss') section and can be 35588 multiply "defined". Using the 'section' attribute changes what 35589 section the variable goes into and may cause the linker to issue an 35590 error if an uninitialized variable has multiple definitions. You 35591 can force a variable to be initialized with the '-fno-common' flag 35592 or the 'nocommon' attribute. 35593 35594 Some file formats do not support arbitrary sections so the 35595 'section' attribute is not available on all platforms. If you need 35596 to map the entire contents of a module to a particular section, 35597 consider using the facilities of the linker instead. 35598 35599'tls_model ("TLS_MODEL")' 35600 The 'tls_model' attribute sets thread-local storage model (*note 35601 Thread-Local::) of a particular '__thread' variable, overriding 35602 '-ftls-model=' command-line switch on a per-variable basis. The 35603 TLS_MODEL argument should be one of 'global-dynamic', 35604 'local-dynamic', 'initial-exec' or 'local-exec'. 35605 35606 Not all targets support this attribute. 35607 35608'unused' 35609 This attribute, attached to a variable, means that the variable is 35610 meant to be possibly unused. GCC does not produce a warning for 35611 this variable. 35612 35613'used' 35614 This attribute, attached to a variable with static storage, means 35615 that the variable must be emitted even if it appears that the 35616 variable is not referenced. 35617 35618 When applied to a static data member of a C++ class template, the 35619 attribute also means that the member is instantiated if the class 35620 itself is instantiated. 35621 35622'vector_size (BYTES)' 35623 This attribute specifies the vector size for the type of the 35624 declared variable, measured in bytes. The type to which it applies 35625 is known as the "base type". The BYTES argument must be a positive 35626 power-of-two multiple of the base type size. For example, the 35627 declaration: 35628 35629 int foo __attribute__ ((vector_size (16))); 35630 35631 causes the compiler to set the mode for 'foo', to be 16 bytes, 35632 divided into 'int' sized units. Assuming a 32-bit 'int', 'foo''s 35633 type is a vector of four units of four bytes each, and the 35634 corresponding mode of 'foo' is 'V4SI'. *Note Vector Extensions::, 35635 for details of manipulating vector variables. 35636 35637 This attribute is only applicable to integral and floating scalars, 35638 although arrays, pointers, and function return values are allowed 35639 in conjunction with this construct. 35640 35641 Aggregates with this attribute are invalid, even if they are of the 35642 same size as a corresponding scalar. For example, the declaration: 35643 35644 struct S { int a; }; 35645 struct S __attribute__ ((vector_size (16))) foo; 35646 35647 is invalid even if the size of the structure is the same as the 35648 size of the 'int'. 35649 35650'visibility ("VISIBILITY_TYPE")' 35651 This attribute affects the linkage of the declaration to which it 35652 is attached. The 'visibility' attribute is described in *note 35653 Common Function Attributes::. 35654 35655'weak' 35656 The 'weak' attribute is described in *note Common Function 35657 Attributes::. 35658 35659'noinit' 35660 Any data with the 'noinit' attribute will not be initialized by the 35661 C runtime startup code, or the program loader. Not initializing 35662 data in this way can reduce program startup times. This attribute 35663 is specific to ELF targets and relies on the linker to place such 35664 data in the right location 35665 35666 35667File: gcc.info, Node: ARC Variable Attributes, Next: AVR Variable Attributes, Prev: Common Variable Attributes, Up: Variable Attributes 35668 356696.34.2 ARC Variable Attributes 35670------------------------------ 35671 35672'aux' 35673 The 'aux' attribute is used to directly access the ARC's auxiliary 35674 register space from C. The auxilirary register number is given via 35675 attribute argument. 35676 35677 35678File: gcc.info, Node: AVR Variable Attributes, Next: Blackfin Variable Attributes, Prev: ARC Variable Attributes, Up: Variable Attributes 35679 356806.34.3 AVR Variable Attributes 35681------------------------------ 35682 35683'progmem' 35684 The 'progmem' attribute is used on the AVR to place read-only data 35685 in the non-volatile program memory (flash). The 'progmem' 35686 attribute accomplishes this by putting respective variables into a 35687 section whose name starts with '.progmem'. 35688 35689 This attribute works similar to the 'section' attribute but adds 35690 additional checking. 35691 35692 * Ordinary AVR cores with 32 general purpose registers: 35693 'progmem' affects the location of the data but not how this 35694 data is accessed. In order to read data located with the 35695 'progmem' attribute (inline) assembler must be used. 35696 /* Use custom macros from AVR-LibC (http://nongnu.org/avr-libc/user-manual/) */ 35697 #include <avr/pgmspace.h> 35698 35699 /* Locate var in flash memory */ 35700 const int var[2] PROGMEM = { 1, 2 }; 35701 35702 int read_var (int i) 35703 { 35704 /* Access var[] by accessor macro from avr/pgmspace.h */ 35705 return (int) pgm_read_word (& var[i]); 35706 } 35707 35708 AVR is a Harvard architecture processor and data and read-only 35709 data normally resides in the data memory (RAM). 35710 35711 See also the *note AVR Named Address Spaces:: section for an 35712 alternate way to locate and access data in flash memory. 35713 35714 * AVR cores with flash memory visible in the RAM address range: 35715 On such devices, there is no need for attribute 'progmem' or 35716 *note '__flash': AVR Named Address Spaces. qualifier at all. 35717 Just use standard C / C++. The compiler will generate 'LD*' 35718 instructions. As flash memory is visible in the RAM address 35719 range, and the default linker script does _not_ locate 35720 '.rodata' in RAM, no special features are needed in order not 35721 to waste RAM for read-only data or to read from flash. You 35722 might even get slightly better performance by avoiding 35723 'progmem' and '__flash'. This applies to devices from 35724 families 'avrtiny' and 'avrxmega3', see *note AVR Options:: 35725 for an overview. 35726 35727 * Reduced AVR Tiny cores like ATtiny40: 35728 The compiler adds '0x4000' to the addresses of objects and 35729 declarations in 'progmem' and locates the objects in flash 35730 memory, namely in section '.progmem.data'. The offset is 35731 needed because the flash memory is visible in the RAM address 35732 space starting at address '0x4000'. 35733 35734 Data in 'progmem' can be accessed by means of ordinary C code, 35735 no special functions or macros are needed. 35736 35737 /* var is located in flash memory */ 35738 extern const int var[2] __attribute__((progmem)); 35739 35740 int read_var (int i) 35741 { 35742 return var[i]; 35743 } 35744 35745 Please notice that on these devices, there is no need for 35746 'progmem' at all. 35747 35748'io' 35749'io (ADDR)' 35750 Variables with the 'io' attribute are used to address memory-mapped 35751 peripherals in the io address range. If an address is specified, 35752 the variable is assigned that address, and the value is interpreted 35753 as an address in the data address space. Example: 35754 35755 volatile int porta __attribute__((io (0x22))); 35756 35757 The address specified in the address in the data address range. 35758 35759 Otherwise, the variable it is not assigned an address, but the 35760 compiler will still use in/out instructions where applicable, 35761 assuming some other module assigns an address in the io address 35762 range. Example: 35763 35764 extern volatile int porta __attribute__((io)); 35765 35766'io_low' 35767'io_low (ADDR)' 35768 This is like the 'io' attribute, but additionally it informs the 35769 compiler that the object lies in the lower half of the I/O area, 35770 allowing the use of 'cbi', 'sbi', 'sbic' and 'sbis' instructions. 35771 35772'address' 35773'address (ADDR)' 35774 Variables with the 'address' attribute are used to address 35775 memory-mapped peripherals that may lie outside the io address 35776 range. 35777 35778 volatile int porta __attribute__((address (0x600))); 35779 35780'absdata' 35781 Variables in static storage and with the 'absdata' attribute can be 35782 accessed by the 'LDS' and 'STS' instructions which take absolute 35783 addresses. 35784 35785 * This attribute is only supported for the reduced AVR Tiny core 35786 like ATtiny40. 35787 35788 * You must make sure that respective data is located in the 35789 address range '0x40'...'0xbf' accessible by 'LDS' and 'STS'. 35790 One way to achieve this as an appropriate linker description 35791 file. 35792 35793 * If the location does not fit the address range of 'LDS' and 35794 'STS', there is currently (Binutils 2.26) just an unspecific 35795 warning like 35796 'module.c:(.text+0x1c): warning: internal error: out of 35797 range error' 35798 35799 See also the '-mabsdata' *note command-line option: AVR Options. 35800 35801 35802File: gcc.info, Node: Blackfin Variable Attributes, Next: H8/300 Variable Attributes, Prev: AVR Variable Attributes, Up: Variable Attributes 35803 358046.34.4 Blackfin Variable Attributes 35805----------------------------------- 35806 35807Three attributes are currently defined for the Blackfin. 35808 35809'l1_data' 35810'l1_data_A' 35811'l1_data_B' 35812 Use these attributes on the Blackfin to place the variable into L1 35813 Data SRAM. Variables with 'l1_data' attribute are put into the 35814 specific section named '.l1.data'. Those with 'l1_data_A' 35815 attribute are put into the specific section named '.l1.data.A'. 35816 Those with 'l1_data_B' attribute are put into the specific section 35817 named '.l1.data.B'. 35818 35819'l2' 35820 Use this attribute on the Blackfin to place the variable into L2 35821 SRAM. Variables with 'l2' attribute are put into the specific 35822 section named '.l2.data'. 35823 35824 35825File: gcc.info, Node: H8/300 Variable Attributes, Next: IA-64 Variable Attributes, Prev: Blackfin Variable Attributes, Up: Variable Attributes 35826 358276.34.5 H8/300 Variable Attributes 35828--------------------------------- 35829 35830These variable attributes are available for H8/300 targets: 35831 35832'eightbit_data' 35833 Use this attribute on the H8/300, H8/300H, and H8S to indicate that 35834 the specified variable should be placed into the eight-bit data 35835 section. The compiler generates more efficient code for certain 35836 operations on data in the eight-bit data area. Note the eight-bit 35837 data area is limited to 256 bytes of data. 35838 35839 You must use GAS and GLD from GNU binutils version 2.7 or later for 35840 this attribute to work correctly. 35841 35842'tiny_data' 35843 Use this attribute on the H8/300H and H8S to indicate that the 35844 specified variable should be placed into the tiny data section. 35845 The compiler generates more efficient code for loads and stores on 35846 data in the tiny data section. Note the tiny data area is limited 35847 to slightly under 32KB of data. 35848 35849 35850File: gcc.info, Node: IA-64 Variable Attributes, Next: M32R/D Variable Attributes, Prev: H8/300 Variable Attributes, Up: Variable Attributes 35851 358526.34.6 IA-64 Variable Attributes 35853-------------------------------- 35854 35855The IA-64 back end supports the following variable attribute: 35856 35857'model (MODEL-NAME)' 35858 35859 On IA-64, use this attribute to set the addressability of an 35860 object. At present, the only supported identifier for MODEL-NAME 35861 is 'small', indicating addressability via "small" (22-bit) 35862 addresses (so that their addresses can be loaded with the 'addl' 35863 instruction). Caveat: such addressing is by definition not 35864 position independent and hence this attribute must not be used for 35865 objects defined by shared libraries. 35866 35867 35868File: gcc.info, Node: M32R/D Variable Attributes, Next: MeP Variable Attributes, Prev: IA-64 Variable Attributes, Up: Variable Attributes 35869 358706.34.7 M32R/D Variable Attributes 35871--------------------------------- 35872 35873One attribute is currently defined for the M32R/D. 35874 35875'model (MODEL-NAME)' 35876 Use this attribute on the M32R/D to set the addressability of an 35877 object. The identifier MODEL-NAME is one of 'small', 'medium', or 35878 'large', representing each of the code models. 35879 35880 Small model objects live in the lower 16MB of memory (so that their 35881 addresses can be loaded with the 'ld24' instruction). 35882 35883 Medium and large model objects may live anywhere in the 32-bit 35884 address space (the compiler generates 'seth/add3' instructions to 35885 load their addresses). 35886 35887 35888File: gcc.info, Node: MeP Variable Attributes, Next: Microsoft Windows Variable Attributes, Prev: M32R/D Variable Attributes, Up: Variable Attributes 35889 358906.34.8 MeP Variable Attributes 35891------------------------------ 35892 35893The MeP target has a number of addressing modes and busses. The 'near' 35894space spans the standard memory space's first 16 megabytes (24 bits). 35895The 'far' space spans the entire 32-bit memory space. The 'based' space 35896is a 128-byte region in the memory space that is addressed relative to 35897the '$tp' register. The 'tiny' space is a 65536-byte region relative to 35898the '$gp' register. In addition to these memory regions, the MeP target 35899has a separate 16-bit control bus which is specified with 'cb' 35900attributes. 35901 35902'based' 35903 Any variable with the 'based' attribute is assigned to the '.based' 35904 section, and is accessed with relative to the '$tp' register. 35905 35906'tiny' 35907 Likewise, the 'tiny' attribute assigned variables to the '.tiny' 35908 section, relative to the '$gp' register. 35909 35910'near' 35911 Variables with the 'near' attribute are assumed to have addresses 35912 that fit in a 24-bit addressing mode. This is the default for 35913 large variables ('-mtiny=4' is the default) but this attribute can 35914 override '-mtiny=' for small variables, or override '-ml'. 35915 35916'far' 35917 Variables with the 'far' attribute are addressed using a full 35918 32-bit address. Since this covers the entire memory space, this 35919 allows modules to make no assumptions about where variables might 35920 be stored. 35921 35922'io' 35923'io (ADDR)' 35924 Variables with the 'io' attribute are used to address memory-mapped 35925 peripherals. If an address is specified, the variable is assigned 35926 that address, else it is not assigned an address (it is assumed 35927 some other module assigns an address). Example: 35928 35929 int timer_count __attribute__((io(0x123))); 35930 35931'cb' 35932'cb (ADDR)' 35933 Variables with the 'cb' attribute are used to access the control 35934 bus, using special instructions. 'addr' indicates the control bus 35935 address. Example: 35936 35937 int cpu_clock __attribute__((cb(0x123))); 35938 35939 35940File: gcc.info, Node: Microsoft Windows Variable Attributes, Next: MSP430 Variable Attributes, Prev: MeP Variable Attributes, Up: Variable Attributes 35941 359426.34.9 Microsoft Windows Variable Attributes 35943-------------------------------------------- 35944 35945You can use these attributes on Microsoft Windows targets. *note x86 35946Variable Attributes:: for additional Windows compatibility attributes 35947available on all x86 targets. 35948 35949'dllimport' 35950'dllexport' 35951 The 'dllimport' and 'dllexport' attributes are described in *note 35952 Microsoft Windows Function Attributes::. 35953 35954'selectany' 35955 The 'selectany' attribute causes an initialized global variable to 35956 have link-once semantics. When multiple definitions of the 35957 variable are encountered by the linker, the first is selected and 35958 the remainder are discarded. Following usage by the Microsoft 35959 compiler, the linker is told _not_ to warn about size or content 35960 differences of the multiple definitions. 35961 35962 Although the primary usage of this attribute is for POD types, the 35963 attribute can also be applied to global C++ objects that are 35964 initialized by a constructor. In this case, the static 35965 initialization and destruction code for the object is emitted in 35966 each translation defining the object, but the calls to the 35967 constructor and destructor are protected by a link-once guard 35968 variable. 35969 35970 The 'selectany' attribute is only available on Microsoft Windows 35971 targets. You can use '__declspec (selectany)' as a synonym for 35972 '__attribute__ ((selectany))' for compatibility with other 35973 compilers. 35974 35975'shared' 35976 On Microsoft Windows, in addition to putting variable definitions 35977 in a named section, the section can also be shared among all 35978 running copies of an executable or DLL. For example, this small 35979 program defines shared data by putting it in a named section 35980 'shared' and marking the section shareable: 35981 35982 int foo __attribute__((section ("shared"), shared)) = 0; 35983 35984 int 35985 main() 35986 { 35987 /* Read and write foo. All running 35988 copies see the same value. */ 35989 return 0; 35990 } 35991 35992 You may only use the 'shared' attribute along with 'section' 35993 attribute with a fully-initialized global definition because of the 35994 way linkers work. See 'section' attribute for more information. 35995 35996 The 'shared' attribute is only available on Microsoft Windows. 35997 35998 35999File: gcc.info, Node: MSP430 Variable Attributes, Next: Nvidia PTX Variable Attributes, Prev: Microsoft Windows Variable Attributes, Up: Variable Attributes 36000 360016.34.10 MSP430 Variable Attributes 36002---------------------------------- 36003 36004'noinit' 36005 Any data with the 'noinit' attribute will not be initialised by the 36006 C runtime startup code, or the program loader. Not initialising 36007 data in this way can reduce program startup times. 36008 36009'persistent' 36010 Any variable with the 'persistent' attribute will not be 36011 initialised by the C runtime startup code. Instead its value will 36012 be set once, when the application is loaded, and then never 36013 initialised again, even if the processor is reset or the program 36014 restarts. Persistent data is intended to be placed into FLASH RAM, 36015 where its value will be retained across resets. The linker script 36016 being used to create the application should ensure that persistent 36017 data is correctly placed. 36018 36019'upper' 36020'either' 36021 These attributes are the same as the MSP430 function attributes of 36022 the same name (*note MSP430 Function Attributes::). 36023 36024'lower' 36025 This option behaves mostly the same as the MSP430 function 36026 attribute of the same name (*note MSP430 Function Attributes::), 36027 but it has some additional functionality. 36028 36029 If '-mdata-region='{'upper,either,none'} has been passed, or the 36030 'section' attribute is applied to a variable, the compiler will 36031 generate 430X instructions to handle it. This is because the 36032 compiler has to assume that the variable could get placed in the 36033 upper memory region (above address 0xFFFF). Marking the variable 36034 with the 'lower' attribute informs the compiler that the variable 36035 will be placed in lower memory so it is safe to use 430 36036 instructions to handle it. 36037 36038 In the case of the 'section' attribute, the section name given will 36039 be used, and the '.lower' prefix will not be added. 36040 36041 36042File: gcc.info, Node: Nvidia PTX Variable Attributes, Next: PowerPC Variable Attributes, Prev: MSP430 Variable Attributes, Up: Variable Attributes 36043 360446.34.11 Nvidia PTX Variable Attributes 36045-------------------------------------- 36046 36047These variable attributes are supported by the Nvidia PTX back end: 36048 36049'shared' 36050 Use this attribute to place a variable in the '.shared' memory 36051 space. This memory space is private to each cooperative thread 36052 array; only threads within one thread block refer to the same 36053 instance of the variable. The runtime does not initialize 36054 variables in this memory space. 36055 36056 36057File: gcc.info, Node: PowerPC Variable Attributes, Next: RL78 Variable Attributes, Prev: Nvidia PTX Variable Attributes, Up: Variable Attributes 36058 360596.34.12 PowerPC Variable Attributes 36060----------------------------------- 36061 36062Three attributes currently are defined for PowerPC configurations: 36063'altivec', 'ms_struct' and 'gcc_struct'. 36064 36065 For full documentation of the struct attributes please see the 36066documentation in *note x86 Variable Attributes::. 36067 36068 For documentation of 'altivec' attribute please see the documentation 36069in *note PowerPC Type Attributes::. 36070 36071 36072File: gcc.info, Node: RL78 Variable Attributes, Next: V850 Variable Attributes, Prev: PowerPC Variable Attributes, Up: Variable Attributes 36073 360746.34.13 RL78 Variable Attributes 36075-------------------------------- 36076 36077The RL78 back end supports the 'saddr' variable attribute. This 36078specifies placement of the corresponding variable in the SADDR area, 36079which can be accessed more efficiently than the default memory region. 36080 36081 36082File: gcc.info, Node: V850 Variable Attributes, Next: x86 Variable Attributes, Prev: RL78 Variable Attributes, Up: Variable Attributes 36083 360846.34.14 V850 Variable Attributes 36085-------------------------------- 36086 36087These variable attributes are supported by the V850 back end: 36088 36089'sda' 36090 Use this attribute to explicitly place a variable in the small data 36091 area, which can hold up to 64 kilobytes. 36092 36093'tda' 36094 Use this attribute to explicitly place a variable in the tiny data 36095 area, which can hold up to 256 bytes in total. 36096 36097'zda' 36098 Use this attribute to explicitly place a variable in the first 32 36099 kilobytes of memory. 36100 36101 36102File: gcc.info, Node: x86 Variable Attributes, Next: Xstormy16 Variable Attributes, Prev: V850 Variable Attributes, Up: Variable Attributes 36103 361046.34.15 x86 Variable Attributes 36105------------------------------- 36106 36107Two attributes are currently defined for x86 configurations: 'ms_struct' 36108and 'gcc_struct'. 36109 36110'ms_struct' 36111'gcc_struct' 36112 36113 If 'packed' is used on a structure, or if bit-fields are used, it 36114 may be that the Microsoft ABI lays out the structure differently 36115 than the way GCC normally does. Particularly when moving packed 36116 data between functions compiled with GCC and the native Microsoft 36117 compiler (either via function call or as data in a file), it may be 36118 necessary to access either format. 36119 36120 The 'ms_struct' and 'gcc_struct' attributes correspond to the 36121 '-mms-bitfields' and '-mno-ms-bitfields' command-line options, 36122 respectively; see *note x86 Options::, for details of how structure 36123 layout is affected. *Note x86 Type Attributes::, for information 36124 about the corresponding attributes on types. 36125 36126 36127File: gcc.info, Node: Xstormy16 Variable Attributes, Prev: x86 Variable Attributes, Up: Variable Attributes 36128 361296.34.16 Xstormy16 Variable Attributes 36130------------------------------------- 36131 36132One attribute is currently defined for xstormy16 configurations: 36133'below100'. 36134 36135'below100' 36136 36137 If a variable has the 'below100' attribute ('BELOW100' is allowed 36138 also), GCC places the variable in the first 0x100 bytes of memory 36139 and use special opcodes to access it. Such variables are placed in 36140 either the '.bss_below100' section or the '.data_below100' section. 36141 36142 36143File: gcc.info, Node: Type Attributes, Next: Label Attributes, Prev: Variable Attributes, Up: C Extensions 36144 361456.35 Specifying Attributes of Types 36146=================================== 36147 36148The keyword '__attribute__' allows you to specify various special 36149properties of types. Some type attributes apply only to structure and 36150union types, and in C++, also class types, while others can apply to any 36151type defined via a 'typedef' declaration. Unless otherwise specified, 36152the same restrictions and effects apply to attributes regardless of 36153whether a type is a trivial structure or a C++ class with user-defined 36154constructors, destructors, or a copy assignment. 36155 36156 Other attributes are defined for functions (*note Function 36157Attributes::), labels (*note Label Attributes::), enumerators (*note 36158Enumerator Attributes::), statements (*note Statement Attributes::), and 36159for variables (*note Variable Attributes::). 36160 36161 The '__attribute__' keyword is followed by an attribute specification 36162enclosed in double parentheses. 36163 36164 You may specify type attributes in an enum, struct or union type 36165declaration or definition by placing them immediately after the 36166'struct', 'union' or 'enum' keyword. You can also place them just past 36167the closing curly brace of the definition, but this is less preferred 36168because logically the type should be fully defined at the closing brace. 36169 36170 You can also include type attributes in a 'typedef' declaration. *Note 36171Attribute Syntax::, for details of the exact syntax for using 36172attributes. 36173 36174* Menu: 36175 36176* Common Type Attributes:: 36177* ARC Type Attributes:: 36178* ARM Type Attributes:: 36179* MeP Type Attributes:: 36180* PowerPC Type Attributes:: 36181* x86 Type Attributes:: 36182 36183 36184File: gcc.info, Node: Common Type Attributes, Next: ARC Type Attributes, Up: Type Attributes 36185 361866.35.1 Common Type Attributes 36187----------------------------- 36188 36189The following type attributes are supported on most targets. 36190 36191'aligned' 36192'aligned (ALIGNMENT)' 36193 The 'aligned' attribute specifies a minimum alignment (in bytes) 36194 for variables of the specified type. When specified, ALIGNMENT 36195 must be a power of 2. Specifying no ALIGNMENT argument implies the 36196 maximum alignment for the target, which is often, but by no means 36197 always, 8 or 16 bytes. For example, the declarations: 36198 36199 struct __attribute__ ((aligned (8))) S { short f[3]; }; 36200 typedef int more_aligned_int __attribute__ ((aligned (8))); 36201 36202 force the compiler to ensure (as far as it can) that each variable 36203 whose type is 'struct S' or 'more_aligned_int' is allocated and 36204 aligned _at least_ on a 8-byte boundary. On a SPARC, having all 36205 variables of type 'struct S' aligned to 8-byte boundaries allows 36206 the compiler to use the 'ldd' and 'std' (doubleword load and store) 36207 instructions when copying one variable of type 'struct S' to 36208 another, thus improving run-time efficiency. 36209 36210 Note that the alignment of any given 'struct' or 'union' type is 36211 required by the ISO C standard to be at least a perfect multiple of 36212 the lowest common multiple of the alignments of all of the members 36213 of the 'struct' or 'union' in question. This means that you _can_ 36214 effectively adjust the alignment of a 'struct' or 'union' type by 36215 attaching an 'aligned' attribute to any one of the members of such 36216 a type, but the notation illustrated in the example above is a more 36217 obvious, intuitive, and readable way to request the compiler to 36218 adjust the alignment of an entire 'struct' or 'union' type. 36219 36220 As in the preceding example, you can explicitly specify the 36221 alignment (in bytes) that you wish the compiler to use for a given 36222 'struct' or 'union' type. Alternatively, you can leave out the 36223 alignment factor and just ask the compiler to align a type to the 36224 maximum useful alignment for the target machine you are compiling 36225 for. For example, you could write: 36226 36227 struct __attribute__ ((aligned)) S { short f[3]; }; 36228 36229 Whenever you leave out the alignment factor in an 'aligned' 36230 attribute specification, the compiler automatically sets the 36231 alignment for the type to the largest alignment that is ever used 36232 for any data type on the target machine you are compiling for. 36233 Doing this can often make copy operations more efficient, because 36234 the compiler can use whatever instructions copy the biggest chunks 36235 of memory when performing copies to or from the variables that have 36236 types that you have aligned this way. 36237 36238 In the example above, if the size of each 'short' is 2 bytes, then 36239 the size of the entire 'struct S' type is 6 bytes. The smallest 36240 power of two that is greater than or equal to that is 8, so the 36241 compiler sets the alignment for the entire 'struct S' type to 8 36242 bytes. 36243 36244 Note that although you can ask the compiler to select a 36245 time-efficient alignment for a given type and then declare only 36246 individual stand-alone objects of that type, the compiler's ability 36247 to select a time-efficient alignment is primarily useful only when 36248 you plan to create arrays of variables having the relevant 36249 (efficiently aligned) type. If you declare or use arrays of 36250 variables of an efficiently-aligned type, then it is likely that 36251 your program also does pointer arithmetic (or subscripting, which 36252 amounts to the same thing) on pointers to the relevant type, and 36253 the code that the compiler generates for these pointer arithmetic 36254 operations is often more efficient for efficiently-aligned types 36255 than for other types. 36256 36257 Note that the effectiveness of 'aligned' attributes may be limited 36258 by inherent limitations in your linker. On many systems, the 36259 linker is only able to arrange for variables to be aligned up to a 36260 certain maximum alignment. (For some linkers, the maximum 36261 supported alignment may be very very small.) If your linker is 36262 only able to align variables up to a maximum of 8-byte alignment, 36263 then specifying 'aligned (16)' in an '__attribute__' still only 36264 provides you with 8-byte alignment. See your linker documentation 36265 for further information. 36266 36267 When used on a struct, or struct member, the 'aligned' attribute 36268 can only increase the alignment; in order to decrease it, the 36269 'packed' attribute must be specified as well. When used as part of 36270 a typedef, the 'aligned' attribute can both increase and decrease 36271 alignment, and specifying the 'packed' attribute generates a 36272 warning. 36273 36274'warn_if_not_aligned (ALIGNMENT)' 36275 This attribute specifies a threshold for the structure field, 36276 measured in bytes. If the structure field is aligned below the 36277 threshold, a warning will be issued. For example, the declaration: 36278 36279 typedef unsigned long long __u64 36280 __attribute__((aligned (4), warn_if_not_aligned (8))); 36281 36282 struct foo 36283 { 36284 int i1; 36285 int i2; 36286 __u64 x; 36287 }; 36288 36289 causes the compiler to issue an warning on 'struct foo', like 36290 'warning: alignment 4 of 'struct foo' is less than 8'. It is used 36291 to define 'struct foo' in such a way that 'struct foo' has the same 36292 layout and the structure field 'x' has the same alignment when 36293 '__u64' is aligned at either 4 or 8 bytes. Align 'struct foo' to 8 36294 bytes: 36295 36296 struct __attribute__ ((aligned (8))) foo 36297 { 36298 int i1; 36299 int i2; 36300 __u64 x; 36301 }; 36302 36303 silences the warning. The compiler also issues a warning, like 36304 'warning: 'x' offset 12 in 'struct foo' isn't aligned to 8', when 36305 the structure field has the misaligned offset: 36306 36307 struct __attribute__ ((aligned (8))) foo 36308 { 36309 int i1; 36310 int i2; 36311 int i3; 36312 __u64 x; 36313 }; 36314 36315 This warning can be disabled by '-Wno-if-not-aligned'. 36316 36317'alloc_size (POSITION)' 36318'alloc_size (POSITION-1, POSITION-2)' 36319 The 'alloc_size' type attribute may be applied to the definition of 36320 a type of a function that returns a pointer and takes at least one 36321 argument of an integer type. It indicates that the returned 36322 pointer points to an object whose size is given by the function 36323 argument at POSITION-1, or by the product of the arguments at 36324 POSITION-1 and POSITION-2. Meaningful sizes are positive values 36325 less than 'PTRDIFF_MAX'. Other sizes are disagnosed when detected. 36326 GCC uses this information to improve the results of 36327 '__builtin_object_size'. 36328 36329 For instance, the following declarations 36330 36331 typedef __attribute__ ((alloc_size (1, 2))) void* 36332 calloc_type (size_t, size_t); 36333 typedef __attribute__ ((alloc_size (1))) void* 36334 malloc_type (size_t); 36335 36336 specify that 'calloc_type' is a type of a function that, like the 36337 standard C function 'calloc', returns an object whose size is given 36338 by the product of arguments 1 and 2, and that 'malloc_type', like 36339 the standard C function 'malloc', returns an object whose size is 36340 given by argument 1 to the function. 36341 36342'copy' 36343'copy (EXPRESSION)' 36344 The 'copy' attribute applies the set of attributes with which the 36345 type of the EXPRESSION has been declared to the declaration of the 36346 type to which the attribute is applied. The attribute is designed 36347 for libraries that define aliases that are expected to specify the 36348 same set of attributes as the aliased symbols. The 'copy' 36349 attribute can be used with types, variables, or functions. 36350 However, the kind of symbol to which the attribute is applied 36351 (either varible or function) must match the kind of symbol to which 36352 the argument refers. The 'copy' attribute copies only syntactic 36353 and semantic attributes but not attributes that affect a symbol's 36354 linkage or visibility such as 'alias', 'visibility', or 'weak'. 36355 The 'deprecated' attribute is also not copied. *Note Common 36356 Function Attributes::. *Note Common Variable Attributes::. 36357 36358 For example, suppose 'struct A' below is defined in some third 36359 party library header to have the alignment requirement 'N' and to 36360 force a warning whenever a variable of the type is not so aligned 36361 due to attribute 'packed'. Specifying the 'copy' attribute on the 36362 definition on the unrelated 'struct B' has the effect of copying 36363 all relevant attributes from the type referenced by the pointer 36364 expression to 'struct B'. 36365 36366 struct __attribute__ ((aligned (N), warn_if_not_aligned (N))) 36367 A { /* ... */ }; 36368 struct __attribute__ ((copy ( (struct A *)0)) B { /* ... */ }; 36369 36370'deprecated' 36371'deprecated (MSG)' 36372 The 'deprecated' attribute results in a warning if the type is used 36373 anywhere in the source file. This is useful when identifying types 36374 that are expected to be removed in a future version of a program. 36375 If possible, the warning also includes the location of the 36376 declaration of the deprecated type, to enable users to easily find 36377 further information about why the type is deprecated, or what they 36378 should do instead. Note that the warnings only occur for uses and 36379 then only if the type is being applied to an identifier that itself 36380 is not being declared as deprecated. 36381 36382 typedef int T1 __attribute__ ((deprecated)); 36383 T1 x; 36384 typedef T1 T2; 36385 T2 y; 36386 typedef T1 T3 __attribute__ ((deprecated)); 36387 T3 z __attribute__ ((deprecated)); 36388 36389 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No 36390 warning is issued for line 4 because T2 is not explicitly 36391 deprecated. Line 5 has no warning because T3 is explicitly 36392 deprecated. Similarly for line 6. The optional MSG argument, 36393 which must be a string, is printed in the warning if present. 36394 Control characters in the string will be replaced with escape 36395 sequences, and if the '-fmessage-length' option is set to 0 (its 36396 default value) then any newline characters will be ignored. 36397 36398 The 'deprecated' attribute can also be used for functions and 36399 variables (*note Function Attributes::, *note Variable 36400 Attributes::.) 36401 36402 The message attached to the attribute is affected by the setting of 36403 the '-fmessage-length' option. 36404 36405'designated_init' 36406 This attribute may only be applied to structure types. It 36407 indicates that any initialization of an object of this type must 36408 use designated initializers rather than positional initializers. 36409 The intent of this attribute is to allow the programmer to indicate 36410 that a structure's layout may change, and that therefore relying on 36411 positional initialization will result in future breakage. 36412 36413 GCC emits warnings based on this attribute by default; use 36414 '-Wno-designated-init' to suppress them. 36415 36416'may_alias' 36417 Accesses through pointers to types with this attribute are not 36418 subject to type-based alias analysis, but are instead assumed to be 36419 able to alias any other type of objects. In the context of section 36420 6.5 paragraph 7 of the C99 standard, an lvalue expression 36421 dereferencing such a pointer is treated like having a character 36422 type. See '-fstrict-aliasing' for more information on aliasing 36423 issues. This extension exists to support some vector APIs, in 36424 which pointers to one vector type are permitted to alias pointers 36425 to a different vector type. 36426 36427 Note that an object of a type with this attribute does not have any 36428 special semantics. 36429 36430 Example of use: 36431 36432 typedef short __attribute__ ((__may_alias__)) short_a; 36433 36434 int 36435 main (void) 36436 { 36437 int a = 0x12345678; 36438 short_a *b = (short_a *) &a; 36439 36440 b[1] = 0; 36441 36442 if (a == 0x12345678) 36443 abort(); 36444 36445 exit(0); 36446 } 36447 36448 If you replaced 'short_a' with 'short' in the variable declaration, 36449 the above program would abort when compiled with 36450 '-fstrict-aliasing', which is on by default at '-O2' or above. 36451 36452'mode (MODE)' 36453 This attribute specifies the data type for the 36454 declaration--whichever type corresponds to the mode MODE. This in 36455 effect lets you request an integer or floating-point type according 36456 to its width. 36457 36458 *Note (gccint)Machine Modes::, for a list of the possible keywords 36459 for MODE. You may also specify a mode of 'byte' or '__byte__' to 36460 indicate the mode corresponding to a one-byte integer, 'word' or 36461 '__word__' for the mode of a one-word integer, and 'pointer' or 36462 '__pointer__' for the mode used to represent pointers. 36463 36464'packed' 36465 This attribute, attached to a 'struct', 'union', or C++ 'class' 36466 type definition, specifies that each of its members (other than 36467 zero-width bit-fields) is placed to minimize the memory required. 36468 This is equivalent to specifying the 'packed' attribute on each of 36469 the members. 36470 36471 When attached to an 'enum' definition, the 'packed' attribute 36472 indicates that the smallest integral type should be used. 36473 Specifying the '-fshort-enums' flag on the command line is 36474 equivalent to specifying the 'packed' attribute on all 'enum' 36475 definitions. 36476 36477 In the following example 'struct my_packed_struct''s members are 36478 packed closely together, but the internal layout of its 's' member 36479 is not packed--to do that, 'struct my_unpacked_struct' needs to be 36480 packed too. 36481 36482 struct my_unpacked_struct 36483 { 36484 char c; 36485 int i; 36486 }; 36487 36488 struct __attribute__ ((__packed__)) my_packed_struct 36489 { 36490 char c; 36491 int i; 36492 struct my_unpacked_struct s; 36493 }; 36494 36495 You may only specify the 'packed' attribute on the definition of an 36496 'enum', 'struct', 'union', or 'class', not on a 'typedef' that does 36497 not also define the enumerated type, structure, union, or class. 36498 36499'scalar_storage_order ("ENDIANNESS")' 36500 When attached to a 'union' or a 'struct', this attribute sets the 36501 storage order, aka endianness, of the scalar fields of the type, as 36502 well as the array fields whose component is scalar. The supported 36503 endiannesses are 'big-endian' and 'little-endian'. The attribute 36504 has no effects on fields which are themselves a 'union', a 'struct' 36505 or an array whose component is a 'union' or a 'struct', and it is 36506 possible for these fields to have a different scalar storage order 36507 than the enclosing type. 36508 36509 This attribute is supported only for targets that use a uniform 36510 default scalar storage order (fortunately, most of them), i.e. 36511 targets that store the scalars either all in big-endian or all in 36512 little-endian. 36513 36514 Additional restrictions are enforced for types with the reverse 36515 scalar storage order with regard to the scalar storage order of the 36516 target: 36517 36518 * Taking the address of a scalar field of a 'union' or a 36519 'struct' with reverse scalar storage order is not permitted 36520 and yields an error. 36521 * Taking the address of an array field, whose component is 36522 scalar, of a 'union' or a 'struct' with reverse scalar storage 36523 order is permitted but yields a warning, unless 36524 '-Wno-scalar-storage-order' is specified. 36525 * Taking the address of a 'union' or a 'struct' with reverse 36526 scalar storage order is permitted. 36527 36528 These restrictions exist because the storage order attribute is 36529 lost when the address of a scalar or the address of an array with 36530 scalar component is taken, so storing indirectly through this 36531 address generally does not work. The second case is nevertheless 36532 allowed to be able to perform a block copy from or to the array. 36533 36534 Moreover, the use of type punning or aliasing to toggle the storage 36535 order is not supported; that is to say, a given scalar object 36536 cannot be accessed through distinct types that assign a different 36537 storage order to it. 36538 36539'transparent_union' 36540 36541 This attribute, attached to a 'union' type definition, indicates 36542 that any function parameter having that union type causes calls to 36543 that function to be treated in a special way. 36544 36545 First, the argument corresponding to a transparent union type can 36546 be of any type in the union; no cast is required. Also, if the 36547 union contains a pointer type, the corresponding argument can be a 36548 null pointer constant or a void pointer expression; and if the 36549 union contains a void pointer type, the corresponding argument can 36550 be any pointer expression. If the union member type is a pointer, 36551 qualifiers like 'const' on the referenced type must be respected, 36552 just as with normal pointer conversions. 36553 36554 Second, the argument is passed to the function using the calling 36555 conventions of the first member of the transparent union, not the 36556 calling conventions of the union itself. All members of the union 36557 must have the same machine representation; this is necessary for 36558 this argument passing to work properly. 36559 36560 Transparent unions are designed for library functions that have 36561 multiple interfaces for compatibility reasons. For example, 36562 suppose the 'wait' function must accept either a value of type 'int 36563 *' to comply with POSIX, or a value of type 'union wait *' to 36564 comply with the 4.1BSD interface. If 'wait''s parameter were 'void 36565 *', 'wait' would accept both kinds of arguments, but it would also 36566 accept any other pointer type and this would make argument type 36567 checking less useful. Instead, '<sys/wait.h>' might define the 36568 interface as follows: 36569 36570 typedef union __attribute__ ((__transparent_union__)) 36571 { 36572 int *__ip; 36573 union wait *__up; 36574 } wait_status_ptr_t; 36575 36576 pid_t wait (wait_status_ptr_t); 36577 36578 This interface allows either 'int *' or 'union wait *' arguments to 36579 be passed, using the 'int *' calling convention. The program can 36580 call 'wait' with arguments of either type: 36581 36582 int w1 () { int w; return wait (&w); } 36583 int w2 () { union wait w; return wait (&w); } 36584 36585 With this interface, 'wait''s implementation might look like this: 36586 36587 pid_t wait (wait_status_ptr_t p) 36588 { 36589 return waitpid (-1, p.__ip, 0); 36590 } 36591 36592'unused' 36593 When attached to a type (including a 'union' or a 'struct'), this 36594 attribute means that variables of that type are meant to appear 36595 possibly unused. GCC does not produce a warning for any variables 36596 of that type, even if the variable appears to do nothing. This is 36597 often the case with lock or thread classes, which are usually 36598 defined and then not referenced, but contain constructors and 36599 destructors that have nontrivial bookkeeping functions. 36600 36601'vector_size (BYTES)' 36602 This attribute specifies the vector size for the type, measured in 36603 bytes. The type to which it applies is known as the "base type". 36604 The BYTES argument must be a positive power-of-two multiple of the 36605 base type size. For example, the following declarations: 36606 36607 typedef __attribute__ ((vector_size (32))) int int_vec32_t ; 36608 typedef __attribute__ ((vector_size (32))) int* int_vec32_ptr_t; 36609 typedef __attribute__ ((vector_size (32))) int int_vec32_arr3_t[3]; 36610 36611 define 'int_vec32_t' to be a 32-byte vector type composed of 'int' 36612 sized units. With 'int' having a size of 4 bytes, the type defines 36613 a vector of eight units, four bytes each. The mode of variables of 36614 type 'int_vec32_t' is 'V8SI'. 'int_vec32_ptr_t' is then defined to 36615 be a pointer to such a vector type, and 'int_vec32_arr3_t' to be an 36616 array of three such vectors. *Note Vector Extensions::, for 36617 details of manipulating objects of vector types. 36618 36619 This attribute is only applicable to integral and floating scalar 36620 types. In function declarations the attribute applies to the 36621 function return type. 36622 36623 For example, the following: 36624 __attribute__ ((vector_size (16))) float get_flt_vec16 (void); 36625 declares 'get_flt_vec16' to be a function returning a 16-byte 36626 vector with the base type 'float'. 36627 36628'visibility' 36629 In C++, attribute visibility (*note Function Attributes::) can also 36630 be applied to class, struct, union and enum types. Unlike other 36631 type attributes, the attribute must appear between the initial 36632 keyword and the name of the type; it cannot appear after the body 36633 of the type. 36634 36635 Note that the type visibility is applied to vague linkage entities 36636 associated with the class (vtable, typeinfo node, etc.). In 36637 particular, if a class is thrown as an exception in one shared 36638 object and caught in another, the class must have default 36639 visibility. Otherwise the two shared objects are unable to use the 36640 same typeinfo node and exception handling will break. 36641 36642 To specify multiple attributes, separate them by commas within the 36643double parentheses: for example, '__attribute__ ((aligned (16), 36644packed))'. 36645 36646 36647File: gcc.info, Node: ARC Type Attributes, Next: ARM Type Attributes, Prev: Common Type Attributes, Up: Type Attributes 36648 366496.35.2 ARC Type Attributes 36650-------------------------- 36651 36652Declaring objects with 'uncached' allows you to exclude data-cache 36653participation in load and store operations on those objects without 36654involving the additional semantic implications of 'volatile'. The '.di' 36655instruction suffix is used for all loads and stores of data declared 36656'uncached'. 36657 36658 36659File: gcc.info, Node: ARM Type Attributes, Next: MeP Type Attributes, Prev: ARC Type Attributes, Up: Type Attributes 36660 366616.35.3 ARM Type Attributes 36662-------------------------- 36663 36664On those ARM targets that support 'dllimport' (such as Symbian OS), you 36665can use the 'notshared' attribute to indicate that the virtual table and 36666other similar data for a class should not be exported from a DLL. For 36667example: 36668 36669 class __declspec(notshared) C { 36670 public: 36671 __declspec(dllimport) C(); 36672 virtual void f(); 36673 } 36674 36675 __declspec(dllexport) 36676 C::C() {} 36677 36678In this code, 'C::C' is exported from the current DLL, but the virtual 36679table for 'C' is not exported. (You can use '__attribute__' instead of 36680'__declspec' if you prefer, but most Symbian OS code uses '__declspec'.) 36681 36682 36683File: gcc.info, Node: MeP Type Attributes, Next: PowerPC Type Attributes, Prev: ARM Type Attributes, Up: Type Attributes 36684 366856.35.4 MeP Type Attributes 36686-------------------------- 36687 36688Many of the MeP variable attributes may be applied to types as well. 36689Specifically, the 'based', 'tiny', 'near', and 'far' attributes may be 36690applied to either. The 'io' and 'cb' attributes may not be applied to 36691types. 36692 36693 36694File: gcc.info, Node: PowerPC Type Attributes, Next: x86 Type Attributes, Prev: MeP Type Attributes, Up: Type Attributes 36695 366966.35.5 PowerPC Type Attributes 36697------------------------------ 36698 36699Three attributes currently are defined for PowerPC configurations: 36700'altivec', 'ms_struct' and 'gcc_struct'. 36701 36702 For full documentation of the 'ms_struct' and 'gcc_struct' attributes 36703please see the documentation in *note x86 Type Attributes::. 36704 36705 The 'altivec' attribute allows one to declare AltiVec vector data types 36706supported by the AltiVec Programming Interface Manual. The attribute 36707requires an argument to specify one of three vector types: 'vector__', 36708'pixel__' (always followed by unsigned short), and 'bool__' (always 36709followed by unsigned). 36710 36711 __attribute__((altivec(vector__))) 36712 __attribute__((altivec(pixel__))) unsigned short 36713 __attribute__((altivec(bool__))) unsigned 36714 36715 These attributes mainly are intended to support the '__vector', 36716'__pixel', and '__bool' AltiVec keywords. 36717 36718 36719File: gcc.info, Node: x86 Type Attributes, Prev: PowerPC Type Attributes, Up: Type Attributes 36720 367216.35.6 x86 Type Attributes 36722-------------------------- 36723 36724Two attributes are currently defined for x86 configurations: 'ms_struct' 36725and 'gcc_struct'. 36726 36727'ms_struct' 36728'gcc_struct' 36729 36730 If 'packed' is used on a structure, or if bit-fields are used it 36731 may be that the Microsoft ABI packs them differently than GCC 36732 normally packs them. Particularly when moving packed data between 36733 functions compiled with GCC and the native Microsoft compiler 36734 (either via function call or as data in a file), it may be 36735 necessary to access either format. 36736 36737 The 'ms_struct' and 'gcc_struct' attributes correspond to the 36738 '-mms-bitfields' and '-mno-ms-bitfields' command-line options, 36739 respectively; see *note x86 Options::, for details of how structure 36740 layout is affected. *Note x86 Variable Attributes::, for 36741 information about the corresponding attributes on variables. 36742 36743 36744File: gcc.info, Node: Label Attributes, Next: Enumerator Attributes, Prev: Type Attributes, Up: C Extensions 36745 367466.36 Label Attributes 36747===================== 36748 36749GCC allows attributes to be set on C labels. *Note Attribute Syntax::, 36750for details of the exact syntax for using attributes. Other attributes 36751are available for functions (*note Function Attributes::), variables 36752(*note Variable Attributes::), enumerators (*note Enumerator 36753Attributes::), statements (*note Statement Attributes::), and for types 36754(*note Type Attributes::). 36755 36756 This example uses the 'cold' label attribute to indicate the 36757'ErrorHandling' branch is unlikely to be taken and that the 36758'ErrorHandling' label is unused: 36759 36760 36761 asm goto ("some asm" : : : : NoError); 36762 36763 /* This branch (the fall-through from the asm) is less commonly used */ 36764 ErrorHandling: 36765 __attribute__((cold, unused)); /* Semi-colon is required here */ 36766 printf("error\n"); 36767 return 0; 36768 36769 NoError: 36770 printf("no error\n"); 36771 return 1; 36772 36773'unused' 36774 This feature is intended for program-generated code that may 36775 contain unused labels, but which is compiled with '-Wall'. It is 36776 not normally appropriate to use in it human-written code, though it 36777 could be useful in cases where the code that jumps to the label is 36778 contained within an '#ifdef' conditional. 36779 36780'hot' 36781 The 'hot' attribute on a label is used to inform the compiler that 36782 the path following the label is more likely than paths that are not 36783 so annotated. This attribute is used in cases where 36784 '__builtin_expect' cannot be used, for instance with computed goto 36785 or 'asm goto'. 36786 36787'cold' 36788 The 'cold' attribute on labels is used to inform the compiler that 36789 the path following the label is unlikely to be executed. This 36790 attribute is used in cases where '__builtin_expect' cannot be used, 36791 for instance with computed goto or 'asm goto'. 36792 36793 36794File: gcc.info, Node: Enumerator Attributes, Next: Statement Attributes, Prev: Label Attributes, Up: C Extensions 36795 367966.37 Enumerator Attributes 36797========================== 36798 36799GCC allows attributes to be set on enumerators. *Note Attribute 36800Syntax::, for details of the exact syntax for using attributes. Other 36801attributes are available for functions (*note Function Attributes::), 36802variables (*note Variable Attributes::), labels (*note Label 36803Attributes::), statements (*note Statement Attributes::), and for types 36804(*note Type Attributes::). 36805 36806 This example uses the 'deprecated' enumerator attribute to indicate the 36807'oldval' enumerator is deprecated: 36808 36809 enum E { 36810 oldval __attribute__((deprecated)), 36811 newval 36812 }; 36813 36814 int 36815 fn (void) 36816 { 36817 return oldval; 36818 } 36819 36820'deprecated' 36821 The 'deprecated' attribute results in a warning if the enumerator 36822 is used anywhere in the source file. This is useful when 36823 identifying enumerators that are expected to be removed in a future 36824 version of a program. The warning also includes the location of 36825 the declaration of the deprecated enumerator, to enable users to 36826 easily find further information about why the enumerator is 36827 deprecated, or what they should do instead. Note that the warnings 36828 only occurs for uses. 36829 36830 36831File: gcc.info, Node: Statement Attributes, Next: Attribute Syntax, Prev: Enumerator Attributes, Up: C Extensions 36832 368336.38 Statement Attributes 36834========================= 36835 36836GCC allows attributes to be set on null statements. *Note Attribute 36837Syntax::, for details of the exact syntax for using attributes. Other 36838attributes are available for functions (*note Function Attributes::), 36839variables (*note Variable Attributes::), labels (*note Label 36840Attributes::), enumerators (*note Enumerator Attributes::), and for 36841types (*note Type Attributes::). 36842 36843 This example uses the 'fallthrough' statement attribute to indicate 36844that the '-Wimplicit-fallthrough' warning should not be emitted: 36845 36846 switch (cond) 36847 { 36848 case 1: 36849 bar (1); 36850 __attribute__((fallthrough)); 36851 case 2: 36852 ... 36853 } 36854 36855'fallthrough' 36856 The 'fallthrough' attribute with a null statement serves as a 36857 fallthrough statement. It hints to the compiler that a statement 36858 that falls through to another case label, or user-defined label in 36859 a switch statement is intentional and thus the 36860 '-Wimplicit-fallthrough' warning must not trigger. The fallthrough 36861 attribute may appear at most once in each attribute list, and may 36862 not be mixed with other attributes. It can only be used in a 36863 switch statement (the compiler will issue an error otherwise), 36864 after a preceding statement and before a logically succeeding case 36865 label, or user-defined label. 36866 36867 36868File: gcc.info, Node: Attribute Syntax, Next: Function Prototypes, Prev: Statement Attributes, Up: C Extensions 36869 368706.39 Attribute Syntax 36871===================== 36872 36873This section describes the syntax with which '__attribute__' may be 36874used, and the constructs to which attribute specifiers bind, for the C 36875language. Some details may vary for C++ and Objective-C. Because of 36876infelicities in the grammar for attributes, some forms described here 36877may not be successfully parsed in all cases. 36878 36879 There are some problems with the semantics of attributes in C++. For 36880example, there are no manglings for attributes, although they may affect 36881code generation, so problems may arise when attributed types are used in 36882conjunction with templates or overloading. Similarly, 'typeid' does not 36883distinguish between types with different attributes. Support for 36884attributes in C++ may be restricted in future to attributes on 36885declarations only, but not on nested declarators. 36886 36887 *Note Function Attributes::, for details of the semantics of attributes 36888applying to functions. *Note Variable Attributes::, for details of the 36889semantics of attributes applying to variables. *Note Type Attributes::, 36890for details of the semantics of attributes applying to structure, union 36891and enumerated types. *Note Label Attributes::, for details of the 36892semantics of attributes applying to labels. *Note Enumerator 36893Attributes::, for details of the semantics of attributes applying to 36894enumerators. *Note Statement Attributes::, for details of the semantics 36895of attributes applying to statements. 36896 36897 An "attribute specifier" is of the form '__attribute__ 36898((ATTRIBUTE-LIST))'. An "attribute list" is a possibly empty 36899comma-separated sequence of "attributes", where each attribute is one of 36900the following: 36901 36902 * Empty. Empty attributes are ignored. 36903 36904 * An attribute name (which may be an identifier such as 'unused', or 36905 a reserved word such as 'const'). 36906 36907 * An attribute name followed by a parenthesized list of parameters 36908 for the attribute. These parameters take one of the following 36909 forms: 36910 36911 * An identifier. For example, 'mode' attributes use this form. 36912 36913 * An identifier followed by a comma and a non-empty 36914 comma-separated list of expressions. For example, 'format' 36915 attributes use this form. 36916 36917 * A possibly empty comma-separated list of expressions. For 36918 example, 'format_arg' attributes use this form with the list 36919 being a single integer constant expression, and 'alias' 36920 attributes use this form with the list being a single string 36921 constant. 36922 36923 An "attribute specifier list" is a sequence of one or more attribute 36924specifiers, not separated by any other tokens. 36925 36926 You may optionally specify attribute names with '__' preceding and 36927following the name. This allows you to use them in header files without 36928being concerned about a possible macro of the same name. For example, 36929you may use the attribute name '__noreturn__' instead of 'noreturn'. 36930 36931Label Attributes 36932................ 36933 36934In GNU C, an attribute specifier list may appear after the colon 36935following a label, other than a 'case' or 'default' label. GNU C++ only 36936permits attributes on labels if the attribute specifier is immediately 36937followed by a semicolon (i.e., the label applies to an empty statement). 36938If the semicolon is missing, C++ label attributes are ambiguous, as it 36939is permissible for a declaration, which could begin with an attribute 36940list, to be labelled in C++. Declarations cannot be labelled in C90 or 36941C99, so the ambiguity does not arise there. 36942 36943Enumerator Attributes 36944..................... 36945 36946In GNU C, an attribute specifier list may appear as part of an 36947enumerator. The attribute goes after the enumeration constant, before 36948'=', if present. The optional attribute in the enumerator appertains to 36949the enumeration constant. It is not possible to place the attribute 36950after the constant expression, if present. 36951 36952Statement Attributes 36953.................... 36954 36955In GNU C, an attribute specifier list may appear as part of a null 36956statement. The attribute goes before the semicolon. 36957 36958Type Attributes 36959............... 36960 36961An attribute specifier list may appear as part of a 'struct', 'union' or 36962'enum' specifier. It may go either immediately after the 'struct', 36963'union' or 'enum' keyword, or after the closing brace. The former 36964syntax is preferred. Where attribute specifiers follow the closing 36965brace, they are considered to relate to the structure, union or 36966enumerated type defined, not to any enclosing declaration the type 36967specifier appears in, and the type defined is not complete until after 36968the attribute specifiers. 36969 36970All other attributes 36971.................... 36972 36973Otherwise, an attribute specifier appears as part of a declaration, 36974counting declarations of unnamed parameters and type names, and relates 36975to that declaration (which may be nested in another declaration, for 36976example in the case of a parameter declaration), or to a particular 36977declarator within a declaration. Where an attribute specifier is 36978applied to a parameter declared as a function or an array, it should 36979apply to the function or array rather than the pointer to which the 36980parameter is implicitly converted, but this is not yet correctly 36981implemented. 36982 36983 Any list of specifiers and qualifiers at the start of a declaration may 36984contain attribute specifiers, whether or not such a list may in that 36985context contain storage class specifiers. (Some attributes, however, 36986are essentially in the nature of storage class specifiers, and only make 36987sense where storage class specifiers may be used; for example, 36988'section'.) There is one necessary limitation to this syntax: the first 36989old-style parameter declaration in a function definition cannot begin 36990with an attribute specifier, because such an attribute applies to the 36991function instead by syntax described below (which, however, is not yet 36992implemented in this case). In some other cases, attribute specifiers 36993are permitted by this grammar but not yet supported by the compiler. 36994All attribute specifiers in this place relate to the declaration as a 36995whole. In the obsolescent usage where a type of 'int' is implied by the 36996absence of type specifiers, such a list of specifiers and qualifiers may 36997be an attribute specifier list with no other specifiers or qualifiers. 36998 36999 At present, the first parameter in a function prototype must have some 37000type specifier that is not an attribute specifier; this resolves an 37001ambiguity in the interpretation of 'void f(int (__attribute__((foo)) 37002x))', but is subject to change. At present, if the parentheses of a 37003function declarator contain only attributes then those attributes are 37004ignored, rather than yielding an error or warning or implying a single 37005parameter of type int, but this is subject to change. 37006 37007 An attribute specifier list may appear immediately before a declarator 37008(other than the first) in a comma-separated list of declarators in a 37009declaration of more than one identifier using a single list of 37010specifiers and qualifiers. Such attribute specifiers apply only to the 37011identifier before whose declarator they appear. For example, in 37012 37013 __attribute__((noreturn)) void d0 (void), 37014 __attribute__((format(printf, 1, 2))) d1 (const char *, ...), 37015 d2 (void); 37016 37017the 'noreturn' attribute applies to all the functions declared; the 37018'format' attribute only applies to 'd1'. 37019 37020 An attribute specifier list may appear immediately before the comma, 37021'=' or semicolon terminating the declaration of an identifier other than 37022a function definition. Such attribute specifiers apply to the declared 37023object or function. Where an assembler name for an object or function 37024is specified (*note Asm Labels::), the attribute must follow the 'asm' 37025specification. 37026 37027 An attribute specifier list may, in future, be permitted to appear 37028after the declarator in a function definition (before any old-style 37029parameter declarations or the function body). 37030 37031 Attribute specifiers may be mixed with type qualifiers appearing inside 37032the '[]' of a parameter array declarator, in the C99 construct by which 37033such qualifiers are applied to the pointer to which the array is 37034implicitly converted. Such attribute specifiers apply to the pointer, 37035not to the array, but at present this is not implemented and they are 37036ignored. 37037 37038 An attribute specifier list may appear at the start of a nested 37039declarator. At present, there are some limitations in this usage: the 37040attributes correctly apply to the declarator, but for most individual 37041attributes the semantics this implies are not implemented. When 37042attribute specifiers follow the '*' of a pointer declarator, they may be 37043mixed with any type qualifiers present. The following describes the 37044formal semantics of this syntax. It makes the most sense if you are 37045familiar with the formal specification of declarators in the ISO C 37046standard. 37047 37048 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration 'T D1', 37049where 'T' contains declaration specifiers that specify a type TYPE (such 37050as 'int') and 'D1' is a declarator that contains an identifier IDENT. 37051The type specified for IDENT for derived declarators whose type does not 37052include an attribute specifier is as in the ISO C standard. 37053 37054 If 'D1' has the form '( ATTRIBUTE-SPECIFIER-LIST D )', and the 37055declaration 'T D' specifies the type "DERIVED-DECLARATOR-TYPE-LIST TYPE" 37056for IDENT, then 'T D1' specifies the type "DERIVED-DECLARATOR-TYPE-LIST 37057ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT. 37058 37059 If 'D1' has the form '* TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST D', 37060and the declaration 'T D' specifies the type 37061"DERIVED-DECLARATOR-TYPE-LIST TYPE" for IDENT, then 'T D1' specifies the 37062type "DERIVED-DECLARATOR-TYPE-LIST 37063TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST pointer to TYPE" for IDENT. 37064 37065 For example, 37066 37067 void (__attribute__((noreturn)) ****f) (void); 37068 37069specifies the type "pointer to pointer to pointer to pointer to 37070non-returning function returning 'void'". As another example, 37071 37072 char *__attribute__((aligned(8))) *f; 37073 37074specifies the type "pointer to 8-byte-aligned pointer to 'char'". Note 37075again that this does not work with most attributes; for example, the 37076usage of 'aligned' and 'noreturn' attributes given above is not yet 37077supported. 37078 37079 For compatibility with existing code written for compiler versions that 37080did not implement attributes on nested declarators, some laxity is 37081allowed in the placing of attributes. If an attribute that only applies 37082to types is applied to a declaration, it is treated as applying to the 37083type of that declaration. If an attribute that only applies to 37084declarations is applied to the type of a declaration, it is treated as 37085applying to that declaration; and, for compatibility with code placing 37086the attributes immediately before the identifier declared, such an 37087attribute applied to a function return type is treated as applying to 37088the function type, and such an attribute applied to an array element 37089type is treated as applying to the array type. If an attribute that 37090only applies to function types is applied to a pointer-to-function type, 37091it is treated as applying to the pointer target type; if such an 37092attribute is applied to a function return type that is not a 37093pointer-to-function type, it is treated as applying to the function 37094type. 37095 37096 37097File: gcc.info, Node: Function Prototypes, Next: C++ Comments, Prev: Attribute Syntax, Up: C Extensions 37098 370996.40 Prototypes and Old-Style Function Definitions 37100================================================== 37101 37102GNU C extends ISO C to allow a function prototype to override a later 37103old-style non-prototype definition. Consider the following example: 37104 37105 /* Use prototypes unless the compiler is old-fashioned. */ 37106 #ifdef __STDC__ 37107 #define P(x) x 37108 #else 37109 #define P(x) () 37110 #endif 37111 37112 /* Prototype function declaration. */ 37113 int isroot P((uid_t)); 37114 37115 /* Old-style function definition. */ 37116 int 37117 isroot (x) /* ??? lossage here ??? */ 37118 uid_t x; 37119 { 37120 return x == 0; 37121 } 37122 37123 Suppose the type 'uid_t' happens to be 'short'. ISO C does not allow 37124this example, because subword arguments in old-style non-prototype 37125definitions are promoted. Therefore in this example the function 37126definition's argument is really an 'int', which does not match the 37127prototype argument type of 'short'. 37128 37129 This restriction of ISO C makes it hard to write code that is portable 37130to traditional C compilers, because the programmer does not know whether 37131the 'uid_t' type is 'short', 'int', or 'long'. Therefore, in cases like 37132these GNU C allows a prototype to override a later old-style definition. 37133More precisely, in GNU C, a function prototype argument type overrides 37134the argument type specified by a later old-style definition if the 37135former type is the same as the latter type before promotion. Thus in 37136GNU C the above example is equivalent to the following: 37137 37138 int isroot (uid_t); 37139 37140 int 37141 isroot (uid_t x) 37142 { 37143 return x == 0; 37144 } 37145 37146GNU C++ does not support old-style function definitions, so this 37147extension is irrelevant. 37148 37149 37150File: gcc.info, Node: C++ Comments, Next: Dollar Signs, Prev: Function Prototypes, Up: C Extensions 37151 371526.41 C++ Style Comments 37153======================= 37154 37155In GNU C, you may use C++ style comments, which start with '//' and 37156continue until the end of the line. Many other C implementations allow 37157such comments, and they are included in the 1999 C standard. However, 37158C++ style comments are not recognized if you specify an '-std' option 37159specifying a version of ISO C before C99, or '-ansi' (equivalent to 37160'-std=c90'). 37161 37162 37163File: gcc.info, Node: Dollar Signs, Next: Character Escapes, Prev: C++ Comments, Up: C Extensions 37164 371656.42 Dollar Signs in Identifier Names 37166===================================== 37167 37168In GNU C, you may normally use dollar signs in identifier names. This 37169is because many traditional C implementations allow such identifiers. 37170However, dollar signs in identifiers are not supported on a few target 37171machines, typically because the target assembler does not allow them. 37172 37173 37174File: gcc.info, Node: Character Escapes, Next: Alignment, Prev: Dollar Signs, Up: C Extensions 37175 371766.43 The Character <ESC> in Constants 37177===================================== 37178 37179You can use the sequence '\e' in a string or character constant to stand 37180for the ASCII character <ESC>. 37181 37182 37183File: gcc.info, Node: Alignment, Next: Inline, Prev: Character Escapes, Up: C Extensions 37184 371856.44 Determining the Alignment of Functions, Types or Variables 37186=============================================================== 37187 37188The keyword '__alignof__' determines the alignment requirement of a 37189function, object, or a type, or the minimum alignment usually required 37190by a type. Its syntax is just like 'sizeof' and C11 '_Alignof'. 37191 37192 For example, if the target machine requires a 'double' value to be 37193aligned on an 8-byte boundary, then '__alignof__ (double)' is 8. This 37194is true on many RISC machines. On more traditional machine designs, 37195'__alignof__ (double)' is 4 or even 2. 37196 37197 Some machines never actually require alignment; they allow references 37198to any data type even at an odd address. For these machines, 37199'__alignof__' reports the smallest alignment that GCC gives the data 37200type, usually as mandated by the target ABI. 37201 37202 If the operand of '__alignof__' is an lvalue rather than a type, its 37203value is the required alignment for its type, taking into account any 37204minimum alignment specified by attribute 'aligned' (*note Common 37205Variable Attributes::). For example, after this declaration: 37206 37207 struct foo { int x; char y; } foo1; 37208 37209the value of '__alignof__ (foo1.y)' is 1, even though its actual 37210alignment is probably 2 or 4, the same as '__alignof__ (int)'. It is an 37211error to ask for the alignment of an incomplete type other than 'void'. 37212 37213 If the operand of the '__alignof__' expression is a function, the 37214expression evaluates to the alignment of the function which may be 37215specified by attribute 'aligned' (*note Common Function Attributes::). 37216 37217 37218File: gcc.info, Node: Inline, Next: Volatiles, Prev: Alignment, Up: C Extensions 37219 372206.45 An Inline Function is As Fast As a Macro 37221============================================= 37222 37223By declaring a function inline, you can direct GCC to make calls to that 37224function faster. One way GCC can achieve this is to integrate that 37225function's code into the code for its callers. This makes execution 37226faster by eliminating the function-call overhead; in addition, if any of 37227the actual argument values are constant, their known values may permit 37228simplifications at compile time so that not all of the inline function's 37229code needs to be included. The effect on code size is less predictable; 37230object code may be larger or smaller with function inlining, depending 37231on the particular case. You can also direct GCC to try to integrate all 37232"simple enough" functions into their callers with the option 37233'-finline-functions'. 37234 37235 GCC implements three different semantics of declaring a function 37236inline. One is available with '-std=gnu89' or '-fgnu89-inline' or when 37237'gnu_inline' attribute is present on all inline declarations, another 37238when '-std=c99', '-std=gnu99' or an option for a later C version is used 37239(without '-fgnu89-inline'), and the third is used when compiling C++. 37240 37241 To declare a function inline, use the 'inline' keyword in its 37242declaration, like this: 37243 37244 static inline int 37245 inc (int *a) 37246 { 37247 return (*a)++; 37248 } 37249 37250 If you are writing a header file to be included in ISO C90 programs, 37251write '__inline__' instead of 'inline'. *Note Alternate Keywords::. 37252 37253 The three types of inlining behave similarly in two important cases: 37254when the 'inline' keyword is used on a 'static' function, like the 37255example above, and when a function is first declared without using the 37256'inline' keyword and then is defined with 'inline', like this: 37257 37258 extern int inc (int *a); 37259 inline int 37260 inc (int *a) 37261 { 37262 return (*a)++; 37263 } 37264 37265 In both of these common cases, the program behaves the same as if you 37266had not used the 'inline' keyword, except for its speed. 37267 37268 When a function is both inline and 'static', if all calls to the 37269function are integrated into the caller, and the function's address is 37270never used, then the function's own assembler code is never referenced. 37271In this case, GCC does not actually output assembler code for the 37272function, unless you specify the option '-fkeep-inline-functions'. If 37273there is a nonintegrated call, then the function is compiled to 37274assembler code as usual. The function must also be compiled as usual if 37275the program refers to its address, because that cannot be inlined. 37276 37277 Note that certain usages in a function definition can make it 37278unsuitable for inline substitution. Among these usages are: variadic 37279functions, use of 'alloca', use of computed goto (*note Labels as 37280Values::), use of nonlocal goto, use of nested functions, use of 37281'setjmp', use of '__builtin_longjmp' and use of '__builtin_return' or 37282'__builtin_apply_args'. Using '-Winline' warns when a function marked 37283'inline' could not be substituted, and gives the reason for the failure. 37284 37285 As required by ISO C++, GCC considers member functions defined within 37286the body of a class to be marked inline even if they are not explicitly 37287declared with the 'inline' keyword. You can override this with 37288'-fno-default-inline'; *note Options Controlling C++ Dialect: C++ 37289Dialect Options. 37290 37291 GCC does not inline any functions when not optimizing unless you 37292specify the 'always_inline' attribute for the function, like this: 37293 37294 /* Prototype. */ 37295 inline void foo (const char) __attribute__((always_inline)); 37296 37297 The remainder of this section is specific to GNU C90 inlining. 37298 37299 When an inline function is not 'static', then the compiler must assume 37300that there may be calls from other source files; since a global symbol 37301can be defined only once in any program, the function must not be 37302defined in the other source files, so the calls therein cannot be 37303integrated. Therefore, a non-'static' inline function is always 37304compiled on its own in the usual fashion. 37305 37306 If you specify both 'inline' and 'extern' in the function definition, 37307then the definition is used only for inlining. In no case is the 37308function compiled on its own, not even if you refer to its address 37309explicitly. Such an address becomes an external reference, as if you 37310had only declared the function, and had not defined it. 37311 37312 This combination of 'inline' and 'extern' has almost the effect of a 37313macro. The way to use it is to put a function definition in a header 37314file with these keywords, and put another copy of the definition 37315(lacking 'inline' and 'extern') in a library file. The definition in 37316the header file causes most calls to the function to be inlined. If any 37317uses of the function remain, they refer to the single copy in the 37318library. 37319 37320 37321File: gcc.info, Node: Volatiles, Next: Using Assembly Language with C, Prev: Inline, Up: C Extensions 37322 373236.46 When is a Volatile Object Accessed? 37324======================================== 37325 37326C has the concept of volatile objects. These are normally accessed by 37327pointers and used for accessing hardware or inter-thread communication. 37328The standard encourages compilers to refrain from optimizations 37329concerning accesses to volatile objects, but leaves it implementation 37330defined as to what constitutes a volatile access. The minimum 37331requirement is that at a sequence point all previous accesses to 37332volatile objects have stabilized and no subsequent accesses have 37333occurred. Thus an implementation is free to reorder and combine 37334volatile accesses that occur between sequence points, but cannot do so 37335for accesses across a sequence point. The use of volatile does not 37336allow you to violate the restriction on updating objects multiple times 37337between two sequence points. 37338 37339 Accesses to non-volatile objects are not ordered with respect to 37340volatile accesses. You cannot use a volatile object as a memory barrier 37341to order a sequence of writes to non-volatile memory. For instance: 37342 37343 int *ptr = SOMETHING; 37344 volatile int vobj; 37345 *ptr = SOMETHING; 37346 vobj = 1; 37347 37348Unless *PTR and VOBJ can be aliased, it is not guaranteed that the write 37349to *PTR occurs by the time the update of VOBJ happens. If you need this 37350guarantee, you must use a stronger memory barrier such as: 37351 37352 int *ptr = SOMETHING; 37353 volatile int vobj; 37354 *ptr = SOMETHING; 37355 asm volatile ("" : : : "memory"); 37356 vobj = 1; 37357 37358 A scalar volatile object is read when it is accessed in a void context: 37359 37360 volatile int *src = SOMEVALUE; 37361 *src; 37362 37363 Such expressions are rvalues, and GCC implements this as a read of the 37364volatile object being pointed to. 37365 37366 Assignments are also expressions and have an rvalue. However when 37367assigning to a scalar volatile, the volatile object is not reread, 37368regardless of whether the assignment expression's rvalue is used or not. 37369If the assignment's rvalue is used, the value is that assigned to the 37370volatile object. For instance, there is no read of VOBJ in all the 37371following cases: 37372 37373 int obj; 37374 volatile int vobj; 37375 vobj = SOMETHING; 37376 obj = vobj = SOMETHING; 37377 obj ? vobj = ONETHING : vobj = ANOTHERTHING; 37378 obj = (SOMETHING, vobj = ANOTHERTHING); 37379 37380 If you need to read the volatile object after an assignment has 37381occurred, you must use a separate expression with an intervening 37382sequence point. 37383 37384 As bit-fields are not individually addressable, volatile bit-fields may 37385be implicitly read when written to, or when adjacent bit-fields are 37386accessed. Bit-field operations may be optimized such that adjacent 37387bit-fields are only partially accessed, if they straddle a storage unit 37388boundary. For these reasons it is unwise to use volatile bit-fields to 37389access hardware. 37390 37391 37392File: gcc.info, Node: Using Assembly Language with C, Next: Alternate Keywords, Prev: Volatiles, Up: C Extensions 37393 373946.47 How to Use Inline Assembly Language in C Code 37395================================================== 37396 37397The 'asm' keyword allows you to embed assembler instructions within C 37398code. GCC provides two forms of inline 'asm' statements. A "basic 37399'asm'" statement is one with no operands (*note Basic Asm::), while an 37400"extended 'asm'" statement (*note Extended Asm::) includes one or more 37401operands. The extended form is preferred for mixing C and assembly 37402language within a function, but to include assembly language at top 37403level you must use basic 'asm'. 37404 37405 You can also use the 'asm' keyword to override the assembler name for a 37406C symbol, or to place a C variable in a specific register. 37407 37408* Menu: 37409 37410* Basic Asm:: Inline assembler without operands. 37411* Extended Asm:: Inline assembler with operands. 37412* Constraints:: Constraints for 'asm' operands 37413* Asm Labels:: Specifying the assembler name to use for a C symbol. 37414* Explicit Register Variables:: Defining variables residing in specified 37415 registers. 37416* Size of an asm:: How GCC calculates the size of an 'asm' block. 37417 37418 37419File: gcc.info, Node: Basic Asm, Next: Extended Asm, Up: Using Assembly Language with C 37420 374216.47.1 Basic Asm -- Assembler Instructions Without Operands 37422----------------------------------------------------------- 37423 37424A basic 'asm' statement has the following syntax: 37425 37426 asm ASM-QUALIFIERS ( ASSEMBLERINSTRUCTIONS ) 37427 37428 The 'asm' keyword is a GNU extension. When writing code that can be 37429compiled with '-ansi' and the various '-std' options, use '__asm__' 37430instead of 'asm' (*note Alternate Keywords::). 37431 37432Qualifiers 37433.......... 37434 37435'volatile' 37436 The optional 'volatile' qualifier has no effect. All basic 'asm' 37437 blocks are implicitly volatile. 37438 37439'inline' 37440 If you use the 'inline' qualifier, then for inlining purposes the 37441 size of the 'asm' statement is taken as the smallest size possible 37442 (*note Size of an asm::). 37443 37444Parameters 37445.......... 37446 37447ASSEMBLERINSTRUCTIONS 37448 This is a literal string that specifies the assembler code. The 37449 string can contain any instructions recognized by the assembler, 37450 including directives. GCC does not parse the assembler 37451 instructions themselves and does not know what they mean or even 37452 whether they are valid assembler input. 37453 37454 You may place multiple assembler instructions together in a single 37455 'asm' string, separated by the characters normally used in assembly 37456 code for the system. A combination that works in most places is a 37457 newline to break the line, plus a tab character (written as 37458 '\n\t'). Some assemblers allow semicolons as a line separator. 37459 However, note that some assembler dialects use semicolons to start 37460 a comment. 37461 37462Remarks 37463....... 37464 37465Using extended 'asm' (*note Extended Asm::) typically produces smaller, 37466safer, and more efficient code, and in most cases it is a better 37467solution than basic 'asm'. However, there are two situations where only 37468basic 'asm' can be used: 37469 37470 * Extended 'asm' statements have to be inside a C function, so to 37471 write inline assembly language at file scope ("top-level"), outside 37472 of C functions, you must use basic 'asm'. You can use this 37473 technique to emit assembler directives, define assembly language 37474 macros that can be invoked elsewhere in the file, or write entire 37475 functions in assembly language. Basic 'asm' statements outside of 37476 functions may not use any qualifiers. 37477 37478 * Functions declared with the 'naked' attribute also require basic 37479 'asm' (*note Function Attributes::). 37480 37481 Safely accessing C data and calling functions from basic 'asm' is more 37482complex than it may appear. To access C data, it is better to use 37483extended 'asm'. 37484 37485 Do not expect a sequence of 'asm' statements to remain perfectly 37486consecutive after compilation. If certain instructions need to remain 37487consecutive in the output, put them in a single multi-instruction 'asm' 37488statement. Note that GCC's optimizers can move 'asm' statements 37489relative to other code, including across jumps. 37490 37491 'asm' statements may not perform jumps into other 'asm' statements. 37492GCC does not know about these jumps, and therefore cannot take account 37493of them when deciding how to optimize. Jumps from 'asm' to C labels are 37494only supported in extended 'asm'. 37495 37496 Under certain circumstances, GCC may duplicate (or remove duplicates 37497of) your assembly code when optimizing. This can lead to unexpected 37498duplicate symbol errors during compilation if your assembly code defines 37499symbols or labels. 37500 37501 *Warning:* The C standards do not specify semantics for 'asm', making 37502it a potential source of incompatibilities between compilers. These 37503incompatibilities may not produce compiler warnings/errors. 37504 37505 GCC does not parse basic 'asm''s ASSEMBLERINSTRUCTIONS, which means 37506there is no way to communicate to the compiler what is happening inside 37507them. GCC has no visibility of symbols in the 'asm' and may discard 37508them as unreferenced. It also does not know about side effects of the 37509assembler code, such as modifications to memory or registers. Unlike 37510some compilers, GCC assumes that no changes to general purpose registers 37511occur. This assumption may change in a future release. 37512 37513 To avoid complications from future changes to the semantics and the 37514compatibility issues between compilers, consider replacing basic 'asm' 37515with extended 'asm'. See How to convert from basic asm to extended asm 37516(https://gcc.gnu.org/wiki/ConvertBasicAsmToExtended) for information 37517about how to perform this conversion. 37518 37519 The compiler copies the assembler instructions in a basic 'asm' 37520verbatim to the assembly language output file, without processing 37521dialects or any of the '%' operators that are available with extended 37522'asm'. This results in minor differences between basic 'asm' strings 37523and extended 'asm' templates. For example, to refer to registers you 37524might use '%eax' in basic 'asm' and '%%eax' in extended 'asm'. 37525 37526 On targets such as x86 that support multiple assembler dialects, all 37527basic 'asm' blocks use the assembler dialect specified by the '-masm' 37528command-line option (*note x86 Options::). Basic 'asm' provides no 37529mechanism to provide different assembler strings for different dialects. 37530 37531 For basic 'asm' with non-empty assembler string GCC assumes the 37532assembler block does not change any general purpose registers, but it 37533may read or write any globally accessible variable. 37534 37535 Here is an example of basic 'asm' for i386: 37536 37537 /* Note that this code will not compile with -masm=intel */ 37538 #define DebugBreak() asm("int $3") 37539 37540 37541File: gcc.info, Node: Extended Asm, Next: Constraints, Prev: Basic Asm, Up: Using Assembly Language with C 37542 375436.47.2 Extended Asm - Assembler Instructions with C Expression Operands 37544----------------------------------------------------------------------- 37545 37546With extended 'asm' you can read and write C variables from assembler 37547and perform jumps from assembler code to C labels. Extended 'asm' 37548syntax uses colons (':') to delimit the operand parameters after the 37549assembler template: 37550 37551 asm ASM-QUALIFIERS ( ASSEMBLERTEMPLATE 37552 : OUTPUTOPERANDS 37553 [ : INPUTOPERANDS 37554 [ : CLOBBERS ] ]) 37555 37556 asm ASM-QUALIFIERS ( ASSEMBLERTEMPLATE 37557 : 37558 : INPUTOPERANDS 37559 : CLOBBERS 37560 : GOTOLABELS) 37561 where in the last form, ASM-QUALIFIERS contains 'goto' (and in the 37562first form, not). 37563 37564 The 'asm' keyword is a GNU extension. When writing code that can be 37565compiled with '-ansi' and the various '-std' options, use '__asm__' 37566instead of 'asm' (*note Alternate Keywords::). 37567 37568Qualifiers 37569.......... 37570 37571'volatile' 37572 The typical use of extended 'asm' statements is to manipulate input 37573 values to produce output values. However, your 'asm' statements 37574 may also produce side effects. If so, you may need to use the 37575 'volatile' qualifier to disable certain optimizations. *Note 37576 Volatile::. 37577 37578'inline' 37579 If you use the 'inline' qualifier, then for inlining purposes the 37580 size of the 'asm' statement is taken as the smallest size possible 37581 (*note Size of an asm::). 37582 37583'goto' 37584 This qualifier informs the compiler that the 'asm' statement may 37585 perform a jump to one of the labels listed in the GOTOLABELS. 37586 *Note GotoLabels::. 37587 37588Parameters 37589.......... 37590 37591ASSEMBLERTEMPLATE 37592 This is a literal string that is the template for the assembler 37593 code. It is a combination of fixed text and tokens that refer to 37594 the input, output, and goto parameters. *Note AssemblerTemplate::. 37595 37596OUTPUTOPERANDS 37597 A comma-separated list of the C variables modified by the 37598 instructions in the ASSEMBLERTEMPLATE. An empty list is permitted. 37599 *Note OutputOperands::. 37600 37601INPUTOPERANDS 37602 A comma-separated list of C expressions read by the instructions in 37603 the ASSEMBLERTEMPLATE. An empty list is permitted. *Note 37604 InputOperands::. 37605 37606CLOBBERS 37607 A comma-separated list of registers or other values changed by the 37608 ASSEMBLERTEMPLATE, beyond those listed as outputs. An empty list 37609 is permitted. *Note Clobbers and Scratch Registers::. 37610 37611GOTOLABELS 37612 When you are using the 'goto' form of 'asm', this section contains 37613 the list of all C labels to which the code in the ASSEMBLERTEMPLATE 37614 may jump. *Note GotoLabels::. 37615 37616 'asm' statements may not perform jumps into other 'asm' statements, 37617 only to the listed GOTOLABELS. GCC's optimizers do not know about 37618 other jumps; therefore they cannot take account of them when 37619 deciding how to optimize. 37620 37621 The total number of input + output + goto operands is limited to 30. 37622 37623Remarks 37624....... 37625 37626The 'asm' statement allows you to include assembly instructions directly 37627within C code. This may help you to maximize performance in 37628time-sensitive code or to access assembly instructions that are not 37629readily available to C programs. 37630 37631 Note that extended 'asm' statements must be inside a function. Only 37632basic 'asm' may be outside functions (*note Basic Asm::). Functions 37633declared with the 'naked' attribute also require basic 'asm' (*note 37634Function Attributes::). 37635 37636 While the uses of 'asm' are many and varied, it may help to think of an 37637'asm' statement as a series of low-level instructions that convert input 37638parameters to output parameters. So a simple (if not particularly 37639useful) example for i386 using 'asm' might look like this: 37640 37641 int src = 1; 37642 int dst; 37643 37644 asm ("mov %1, %0\n\t" 37645 "add $1, %0" 37646 : "=r" (dst) 37647 : "r" (src)); 37648 37649 printf("%d\n", dst); 37650 37651 This code copies 'src' to 'dst' and add 1 to 'dst'. 37652 376536.47.2.1 Volatile 37654................. 37655 37656GCC's optimizers sometimes discard 'asm' statements if they determine 37657there is no need for the output variables. Also, the optimizers may 37658move code out of loops if they believe that the code will always return 37659the same result (i.e. none of its input values change between calls). 37660Using the 'volatile' qualifier disables these optimizations. 'asm' 37661statements that have no output operands, including 'asm goto' 37662statements, are implicitly volatile. 37663 37664 This i386 code demonstrates a case that does not use (or require) the 37665'volatile' qualifier. If it is performing assertion checking, this code 37666uses 'asm' to perform the validation. Otherwise, 'dwRes' is 37667unreferenced by any code. As a result, the optimizers can discard the 37668'asm' statement, which in turn removes the need for the entire 'DoCheck' 37669routine. By omitting the 'volatile' qualifier when it isn't needed you 37670allow the optimizers to produce the most efficient code possible. 37671 37672 void DoCheck(uint32_t dwSomeValue) 37673 { 37674 uint32_t dwRes; 37675 37676 // Assumes dwSomeValue is not zero. 37677 asm ("bsfl %1,%0" 37678 : "=r" (dwRes) 37679 : "r" (dwSomeValue) 37680 : "cc"); 37681 37682 assert(dwRes > 3); 37683 } 37684 37685 The next example shows a case where the optimizers can recognize that 37686the input ('dwSomeValue') never changes during the execution of the 37687function and can therefore move the 'asm' outside the loop to produce 37688more efficient code. Again, using the 'volatile' qualifier disables 37689this type of optimization. 37690 37691 void do_print(uint32_t dwSomeValue) 37692 { 37693 uint32_t dwRes; 37694 37695 for (uint32_t x=0; x < 5; x++) 37696 { 37697 // Assumes dwSomeValue is not zero. 37698 asm ("bsfl %1,%0" 37699 : "=r" (dwRes) 37700 : "r" (dwSomeValue) 37701 : "cc"); 37702 37703 printf("%u: %u %u\n", x, dwSomeValue, dwRes); 37704 } 37705 } 37706 37707 The following example demonstrates a case where you need to use the 37708'volatile' qualifier. It uses the x86 'rdtsc' instruction, which reads 37709the computer's time-stamp counter. Without the 'volatile' qualifier, 37710the optimizers might assume that the 'asm' block will always return the 37711same value and therefore optimize away the second call. 37712 37713 uint64_t msr; 37714 37715 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX. 37716 "shl $32, %%rdx\n\t" // Shift the upper bits left. 37717 "or %%rdx, %0" // 'Or' in the lower bits. 37718 : "=a" (msr) 37719 : 37720 : "rdx"); 37721 37722 printf("msr: %llx\n", msr); 37723 37724 // Do other work... 37725 37726 // Reprint the timestamp 37727 asm volatile ( "rdtsc\n\t" // Returns the time in EDX:EAX. 37728 "shl $32, %%rdx\n\t" // Shift the upper bits left. 37729 "or %%rdx, %0" // 'Or' in the lower bits. 37730 : "=a" (msr) 37731 : 37732 : "rdx"); 37733 37734 printf("msr: %llx\n", msr); 37735 37736 GCC's optimizers do not treat this code like the non-volatile code in 37737the earlier examples. They do not move it out of loops or omit it on 37738the assumption that the result from a previous call is still valid. 37739 37740 Note that the compiler can move even 'volatile asm' instructions 37741relative to other code, including across jump instructions. For 37742example, on many targets there is a system register that controls the 37743rounding mode of floating-point operations. Setting it with a 'volatile 37744asm' statement, as in the following PowerPC example, does not work 37745reliably. 37746 37747 asm volatile("mtfsf 255, %0" : : "f" (fpenv)); 37748 sum = x + y; 37749 37750 The compiler may move the addition back before the 'volatile asm' 37751statement. To make it work as expected, add an artificial dependency to 37752the 'asm' by referencing a variable in the subsequent code, for example: 37753 37754 asm volatile ("mtfsf 255,%1" : "=X" (sum) : "f" (fpenv)); 37755 sum = x + y; 37756 37757 Under certain circumstances, GCC may duplicate (or remove duplicates 37758of) your assembly code when optimizing. This can lead to unexpected 37759duplicate symbol errors during compilation if your 'asm' code defines 37760symbols or labels. Using '%=' (*note AssemblerTemplate::) may help 37761resolve this problem. 37762 377636.47.2.2 Assembler Template 37764........................... 37765 37766An assembler template is a literal string containing assembler 37767instructions. The compiler replaces tokens in the template that refer 37768to inputs, outputs, and goto labels, and then outputs the resulting 37769string to the assembler. The string can contain any instructions 37770recognized by the assembler, including directives. GCC does not parse 37771the assembler instructions themselves and does not know what they mean 37772or even whether they are valid assembler input. However, it does count 37773the statements (*note Size of an asm::). 37774 37775 You may place multiple assembler instructions together in a single 37776'asm' string, separated by the characters normally used in assembly code 37777for the system. A combination that works in most places is a newline to 37778break the line, plus a tab character to move to the instruction field 37779(written as '\n\t'). Some assemblers allow semicolons as a line 37780separator. However, note that some assembler dialects use semicolons to 37781start a comment. 37782 37783 Do not expect a sequence of 'asm' statements to remain perfectly 37784consecutive after compilation, even when you are using the 'volatile' 37785qualifier. If certain instructions need to remain consecutive in the 37786output, put them in a single multi-instruction 'asm' statement. 37787 37788 Accessing data from C programs without using input/output operands 37789(such as by using global symbols directly from the assembler template) 37790may not work as expected. Similarly, calling functions directly from an 37791assembler template requires a detailed understanding of the target 37792assembler and ABI. 37793 37794 Since GCC does not parse the assembler template, it has no visibility 37795of any symbols it references. This may result in GCC discarding those 37796symbols as unreferenced unless they are also listed as input, output, or 37797goto operands. 37798 37799Special format strings 37800...................... 37801 37802In addition to the tokens described by the input, output, and goto 37803operands, these tokens have special meanings in the assembler template: 37804 37805'%%' 37806 Outputs a single '%' into the assembler code. 37807 37808'%=' 37809 Outputs a number that is unique to each instance of the 'asm' 37810 statement in the entire compilation. This option is useful when 37811 creating local labels and referring to them multiple times in a 37812 single template that generates multiple assembler instructions. 37813 37814'%{' 37815'%|' 37816'%}' 37817 Outputs '{', '|', and '}' characters (respectively) into the 37818 assembler code. When unescaped, these characters have special 37819 meaning to indicate multiple assembler dialects, as described 37820 below. 37821 37822Multiple assembler dialects in 'asm' templates 37823.............................................. 37824 37825On targets such as x86, GCC supports multiple assembler dialects. The 37826'-masm' option controls which dialect GCC uses as its default for inline 37827assembler. The target-specific documentation for the '-masm' option 37828contains the list of supported dialects, as well as the default dialect 37829if the option is not specified. This information may be important to 37830understand, since assembler code that works correctly when compiled 37831using one dialect will likely fail if compiled using another. *Note x86 37832Options::. 37833 37834 If your code needs to support multiple assembler dialects (for example, 37835if you are writing public headers that need to support a variety of 37836compilation options), use constructs of this form: 37837 37838 { dialect0 | dialect1 | dialect2... } 37839 37840 This construct outputs 'dialect0' when using dialect #0 to compile the 37841code, 'dialect1' for dialect #1, etc. If there are fewer alternatives 37842within the braces than the number of dialects the compiler supports, the 37843construct outputs nothing. 37844 37845 For example, if an x86 compiler supports two dialects ('att', 'intel'), 37846an assembler template such as this: 37847 37848 "bt{l %[Offset],%[Base] | %[Base],%[Offset]}; jc %l2" 37849 37850is equivalent to one of 37851 37852 "btl %[Offset],%[Base] ; jc %l2" /* att dialect */ 37853 "bt %[Base],%[Offset]; jc %l2" /* intel dialect */ 37854 37855 Using that same compiler, this code: 37856 37857 "xchg{l}\t{%%}ebx, %1" 37858 37859corresponds to either 37860 37861 "xchgl\t%%ebx, %1" /* att dialect */ 37862 "xchg\tebx, %1" /* intel dialect */ 37863 37864 There is no support for nesting dialect alternatives. 37865 378666.47.2.3 Output Operands 37867........................ 37868 37869An 'asm' statement has zero or more output operands indicating the names 37870of C variables modified by the assembler code. 37871 37872 In this i386 example, 'old' (referred to in the template string as 37873'%0') and '*Base' (as '%1') are outputs and 'Offset' ('%2') is an input: 37874 37875 bool old; 37876 37877 __asm__ ("btsl %2,%1\n\t" // Turn on zero-based bit #Offset in Base. 37878 "sbb %0,%0" // Use the CF to calculate old. 37879 : "=r" (old), "+rm" (*Base) 37880 : "Ir" (Offset) 37881 : "cc"); 37882 37883 return old; 37884 37885 Operands are separated by commas. Each operand has this format: 37886 37887 [ [ASMSYMBOLICNAME] ] CONSTRAINT (CVARIABLENAME) 37888 37889ASMSYMBOLICNAME 37890 Specifies a symbolic name for the operand. Reference the name in 37891 the assembler template by enclosing it in square brackets (i.e. 37892 '%[Value]'). The scope of the name is the 'asm' statement that 37893 contains the definition. Any valid C variable name is acceptable, 37894 including names already defined in the surrounding code. No two 37895 operands within the same 'asm' statement can use the same symbolic 37896 name. 37897 37898 When not using an ASMSYMBOLICNAME, use the (zero-based) position of 37899 the operand in the list of operands in the assembler template. For 37900 example if there are three output operands, use '%0' in the 37901 template to refer to the first, '%1' for the second, and '%2' for 37902 the third. 37903 37904CONSTRAINT 37905 A string constant specifying constraints on the placement of the 37906 operand; *Note Constraints::, for details. 37907 37908 Output constraints must begin with either '=' (a variable 37909 overwriting an existing value) or '+' (when reading and writing). 37910 When using '=', do not assume the location contains the existing 37911 value on entry to the 'asm', except when the operand is tied to an 37912 input; *note Input Operands: InputOperands. 37913 37914 After the prefix, there must be one or more additional constraints 37915 (*note Constraints::) that describe where the value resides. 37916 Common constraints include 'r' for register and 'm' for memory. 37917 When you list more than one possible location (for example, 37918 '"=rm"'), the compiler chooses the most efficient one based on the 37919 current context. If you list as many alternates as the 'asm' 37920 statement allows, you permit the optimizers to produce the best 37921 possible code. If you must use a specific register, but your 37922 Machine Constraints do not provide sufficient control to select the 37923 specific register you want, local register variables may provide a 37924 solution (*note Local Register Variables::). 37925 37926CVARIABLENAME 37927 Specifies a C lvalue expression to hold the output, typically a 37928 variable name. The enclosing parentheses are a required part of 37929 the syntax. 37930 37931 When the compiler selects the registers to use to represent the output 37932operands, it does not use any of the clobbered registers (*note Clobbers 37933and Scratch Registers::). 37934 37935 Output operand expressions must be lvalues. The compiler cannot check 37936whether the operands have data types that are reasonable for the 37937instruction being executed. For output expressions that are not 37938directly addressable (for example a bit-field), the constraint must 37939allow a register. In that case, GCC uses the register as the output of 37940the 'asm', and then stores that register into the output. 37941 37942 Operands using the '+' constraint modifier count as two operands (that 37943is, both as input and output) towards the total maximum of 30 operands 37944per 'asm' statement. 37945 37946 Use the '&' constraint modifier (*note Modifiers::) on all output 37947operands that must not overlap an input. Otherwise, GCC may allocate 37948the output operand in the same register as an unrelated input operand, 37949on the assumption that the assembler code consumes its inputs before 37950producing outputs. This assumption may be false if the assembler code 37951actually consists of more than one instruction. 37952 37953 The same problem can occur if one output parameter (A) allows a 37954register constraint and another output parameter (B) allows a memory 37955constraint. The code generated by GCC to access the memory address in B 37956can contain registers which _might_ be shared by A, and GCC considers 37957those registers to be inputs to the asm. As above, GCC assumes that 37958such input registers are consumed before any outputs are written. This 37959assumption may result in incorrect behavior if the 'asm' statement 37960writes to A before using B. Combining the '&' modifier with the 37961register constraint on A ensures that modifying A does not affect the 37962address referenced by B. Otherwise, the location of B is undefined if A 37963is modified before using B. 37964 37965 'asm' supports operand modifiers on operands (for example '%k2' instead 37966of simply '%2'). Typically these qualifiers are hardware dependent. 37967The list of supported modifiers for x86 is found at *note x86 Operand 37968modifiers: x86Operandmodifiers. 37969 37970 If the C code that follows the 'asm' makes no use of any of the output 37971operands, use 'volatile' for the 'asm' statement to prevent the 37972optimizers from discarding the 'asm' statement as unneeded (see *note 37973Volatile::). 37974 37975 This code makes no use of the optional ASMSYMBOLICNAME. Therefore it 37976references the first output operand as '%0' (were there a second, it 37977would be '%1', etc). The number of the first input operand is one 37978greater than that of the last output operand. In this i386 example, 37979that makes 'Mask' referenced as '%1': 37980 37981 uint32_t Mask = 1234; 37982 uint32_t Index; 37983 37984 asm ("bsfl %1, %0" 37985 : "=r" (Index) 37986 : "r" (Mask) 37987 : "cc"); 37988 37989 That code overwrites the variable 'Index' ('='), placing the value in a 37990register ('r'). Using the generic 'r' constraint instead of a 37991constraint for a specific register allows the compiler to pick the 37992register to use, which can result in more efficient code. This may not 37993be possible if an assembler instruction requires a specific register. 37994 37995 The following i386 example uses the ASMSYMBOLICNAME syntax. It 37996produces the same result as the code above, but some may consider it 37997more readable or more maintainable since reordering index numbers is not 37998necessary when adding or removing operands. The names 'aIndex' and 37999'aMask' are only used in this example to emphasize which names get used 38000where. It is acceptable to reuse the names 'Index' and 'Mask'. 38001 38002 uint32_t Mask = 1234; 38003 uint32_t Index; 38004 38005 asm ("bsfl %[aMask], %[aIndex]" 38006 : [aIndex] "=r" (Index) 38007 : [aMask] "r" (Mask) 38008 : "cc"); 38009 38010 Here are some more examples of output operands. 38011 38012 uint32_t c = 1; 38013 uint32_t d; 38014 uint32_t *e = &c; 38015 38016 asm ("mov %[e], %[d]" 38017 : [d] "=rm" (d) 38018 : [e] "rm" (*e)); 38019 38020 Here, 'd' may either be in a register or in memory. Since the compiler 38021might already have the current value of the 'uint32_t' location pointed 38022to by 'e' in a register, you can enable it to choose the best location 38023for 'd' by specifying both constraints. 38024 380256.47.2.4 Flag Output Operands 38026............................. 38027 38028Some targets have a special register that holds the "flags" for the 38029result of an operation or comparison. Normally, the contents of that 38030register are either unmodifed by the asm, or the 'asm' statement is 38031considered to clobber the contents. 38032 38033 On some targets, a special form of output operand exists by which 38034conditions in the flags register may be outputs of the asm. The set of 38035conditions supported are target specific, but the general rule is that 38036the output variable must be a scalar integer, and the value is boolean. 38037When supported, the target defines the preprocessor symbol 38038'__GCC_ASM_FLAG_OUTPUTS__'. 38039 38040 Because of the special nature of the flag output operands, the 38041constraint may not include alternatives. 38042 38043 Most often, the target has only one flags register, and thus is an 38044implied operand of many instructions. In this case, the operand should 38045not be referenced within the assembler template via '%0' etc, as there's 38046no corresponding text in the assembly language. 38047 38048ARM 38049AArch64 38050 The flag output constraints for the ARM family are of the form 38051 '=@ccCOND' where COND is one of the standard conditions defined in 38052 the ARM ARM for 'ConditionHolds'. 38053 38054 'eq' 38055 Z flag set, or equal 38056 'ne' 38057 Z flag clear or not equal 38058 'cs' 38059 'hs' 38060 C flag set or unsigned greater than equal 38061 'cc' 38062 'lo' 38063 C flag clear or unsigned less than 38064 'mi' 38065 N flag set or "minus" 38066 'pl' 38067 N flag clear or "plus" 38068 'vs' 38069 V flag set or signed overflow 38070 'vc' 38071 V flag clear 38072 'hi' 38073 unsigned greater than 38074 'ls' 38075 unsigned less than equal 38076 'ge' 38077 signed greater than equal 38078 'lt' 38079 signed less than 38080 'gt' 38081 signed greater than 38082 'le' 38083 signed less than equal 38084 38085 The flag output constraints are not supported in thumb1 mode. 38086 38087x86 family 38088 The flag output constraints for the x86 family are of the form 38089 '=@ccCOND' where COND is one of the standard conditions defined in 38090 the ISA manual for 'jCC' or 'setCC'. 38091 38092 'a' 38093 "above" or unsigned greater than 38094 'ae' 38095 "above or equal" or unsigned greater than or equal 38096 'b' 38097 "below" or unsigned less than 38098 'be' 38099 "below or equal" or unsigned less than or equal 38100 'c' 38101 carry flag set 38102 'e' 38103 'z' 38104 "equal" or zero flag set 38105 'g' 38106 signed greater than 38107 'ge' 38108 signed greater than or equal 38109 'l' 38110 signed less than 38111 'le' 38112 signed less than or equal 38113 'o' 38114 overflow flag set 38115 'p' 38116 parity flag set 38117 's' 38118 sign flag set 38119 'na' 38120 'nae' 38121 'nb' 38122 'nbe' 38123 'nc' 38124 'ne' 38125 'ng' 38126 'nge' 38127 'nl' 38128 'nle' 38129 'no' 38130 'np' 38131 'ns' 38132 'nz' 38133 "not" FLAG, or inverted versions of those above 38134 381356.47.2.5 Input Operands 38136....................... 38137 38138Input operands make values from C variables and expressions available to 38139the assembly code. 38140 38141 Operands are separated by commas. Each operand has this format: 38142 38143 [ [ASMSYMBOLICNAME] ] CONSTRAINT (CEXPRESSION) 38144 38145ASMSYMBOLICNAME 38146 Specifies a symbolic name for the operand. Reference the name in 38147 the assembler template by enclosing it in square brackets (i.e. 38148 '%[Value]'). The scope of the name is the 'asm' statement that 38149 contains the definition. Any valid C variable name is acceptable, 38150 including names already defined in the surrounding code. No two 38151 operands within the same 'asm' statement can use the same symbolic 38152 name. 38153 38154 When not using an ASMSYMBOLICNAME, use the (zero-based) position of 38155 the operand in the list of operands in the assembler template. For 38156 example if there are two output operands and three inputs, use '%2' 38157 in the template to refer to the first input operand, '%3' for the 38158 second, and '%4' for the third. 38159 38160CONSTRAINT 38161 A string constant specifying constraints on the placement of the 38162 operand; *Note Constraints::, for details. 38163 38164 Input constraint strings may not begin with either '=' or '+'. 38165 When you list more than one possible location (for example, 38166 '"irm"'), the compiler chooses the most efficient one based on the 38167 current context. If you must use a specific register, but your 38168 Machine Constraints do not provide sufficient control to select the 38169 specific register you want, local register variables may provide a 38170 solution (*note Local Register Variables::). 38171 38172 Input constraints can also be digits (for example, '"0"'). This 38173 indicates that the specified input must be in the same place as the 38174 output constraint at the (zero-based) index in the output 38175 constraint list. When using ASMSYMBOLICNAME syntax for the output 38176 operands, you may use these names (enclosed in brackets '[]') 38177 instead of digits. 38178 38179CEXPRESSION 38180 This is the C variable or expression being passed to the 'asm' 38181 statement as input. The enclosing parentheses are a required part 38182 of the syntax. 38183 38184 When the compiler selects the registers to use to represent the input 38185operands, it does not use any of the clobbered registers (*note Clobbers 38186and Scratch Registers::). 38187 38188 If there are no output operands but there are input operands, place two 38189consecutive colons where the output operands would go: 38190 38191 __asm__ ("some instructions" 38192 : /* No outputs. */ 38193 : "r" (Offset / 8)); 38194 38195 *Warning:* Do _not_ modify the contents of input-only operands (except 38196for inputs tied to outputs). The compiler assumes that on exit from the 38197'asm' statement these operands contain the same values as they had 38198before executing the statement. It is _not_ possible to use clobbers to 38199inform the compiler that the values in these inputs are changing. One 38200common work-around is to tie the changing input variable to an output 38201variable that never gets used. Note, however, that if the code that 38202follows the 'asm' statement makes no use of any of the output operands, 38203the GCC optimizers may discard the 'asm' statement as unneeded (see 38204*note Volatile::). 38205 38206 'asm' supports operand modifiers on operands (for example '%k2' instead 38207of simply '%2'). Typically these qualifiers are hardware dependent. 38208The list of supported modifiers for x86 is found at *note x86 Operand 38209modifiers: x86Operandmodifiers. 38210 38211 In this example using the fictitious 'combine' instruction, the 38212constraint '"0"' for input operand 1 says that it must occupy the same 38213location as output operand 0. Only input operands may use numbers in 38214constraints, and they must each refer to an output operand. Only a 38215number (or the symbolic assembler name) in the constraint can guarantee 38216that one operand is in the same place as another. The mere fact that 38217'foo' is the value of both operands is not enough to guarantee that they 38218are in the same place in the generated assembler code. 38219 38220 asm ("combine %2, %0" 38221 : "=r" (foo) 38222 : "0" (foo), "g" (bar)); 38223 38224 Here is an example using symbolic names. 38225 38226 asm ("cmoveq %1, %2, %[result]" 38227 : [result] "=r"(result) 38228 : "r" (test), "r" (new), "[result]" (old)); 38229 382306.47.2.6 Clobbers and Scratch Registers 38231....................................... 38232 38233While the compiler is aware of changes to entries listed in the output 38234operands, the inline 'asm' code may modify more than just the outputs. 38235For example, calculations may require additional registers, or the 38236processor may overwrite a register as a side effect of a particular 38237assembler instruction. In order to inform the compiler of these 38238changes, list them in the clobber list. Clobber list items are either 38239register names or the special clobbers (listed below). Each clobber 38240list item is a string constant enclosed in double quotes and separated 38241by commas. 38242 38243 Clobber descriptions may not in any way overlap with an input or output 38244operand. For example, you may not have an operand describing a register 38245class with one member when listing that register in the clobber list. 38246Variables declared to live in specific registers (*note Explicit 38247Register Variables::) and used as 'asm' input or output operands must 38248have no part mentioned in the clobber description. In particular, there 38249is no way to specify that input operands get modified without also 38250specifying them as output operands. 38251 38252 When the compiler selects which registers to use to represent input and 38253output operands, it does not use any of the clobbered registers. As a 38254result, clobbered registers are available for any use in the assembler 38255code. 38256 38257 Another restriction is that the clobber list should not contain the 38258stack pointer register. This is because the compiler requires the value 38259of the stack pointer to be the same after an 'asm' statement as it was 38260on entry to the statement. However, previous versions of GCC did not 38261enforce this rule and allowed the stack pointer to appear in the list, 38262with unclear semantics. This behavior is deprecated and listing the 38263stack pointer may become an error in future versions of GCC. 38264 38265 Here is a realistic example for the VAX showing the use of clobbered 38266registers: 38267 38268 asm volatile ("movc3 %0, %1, %2" 38269 : /* No outputs. */ 38270 : "g" (from), "g" (to), "g" (count) 38271 : "r0", "r1", "r2", "r3", "r4", "r5", "memory"); 38272 38273 Also, there are two special clobber arguments: 38274 38275'"cc"' 38276 The '"cc"' clobber indicates that the assembler code modifies the 38277 flags register. On some machines, GCC represents the condition 38278 codes as a specific hardware register; '"cc"' serves to name this 38279 register. On other machines, condition code handling is different, 38280 and specifying '"cc"' has no effect. But it is valid no matter 38281 what the target. 38282 38283'"memory"' 38284 The '"memory"' clobber tells the compiler that the assembly code 38285 performs memory reads or writes to items other than those listed in 38286 the input and output operands (for example, accessing the memory 38287 pointed to by one of the input parameters). To ensure memory 38288 contains correct values, GCC may need to flush specific register 38289 values to memory before executing the 'asm'. Further, the compiler 38290 does not assume that any values read from memory before an 'asm' 38291 remain unchanged after that 'asm'; it reloads them as needed. 38292 Using the '"memory"' clobber effectively forms a read/write memory 38293 barrier for the compiler. 38294 38295 Note that this clobber does not prevent the _processor_ from doing 38296 speculative reads past the 'asm' statement. To prevent that, you 38297 need processor-specific fence instructions. 38298 38299 Flushing registers to memory has performance implications and may be an 38300issue for time-sensitive code. You can provide better information to 38301GCC to avoid this, as shown in the following examples. At a minimum, 38302aliasing rules allow GCC to know what memory _doesn't_ need to be 38303flushed. 38304 38305 Here is a fictitious sum of squares instruction, that takes two 38306pointers to floating point values in memory and produces a floating 38307point register output. Notice that 'x', and 'y' both appear twice in 38308the 'asm' parameters, once to specify memory accessed, and once to 38309specify a base register used by the 'asm'. You won't normally be 38310wasting a register by doing this as GCC can use the same register for 38311both purposes. However, it would be foolish to use both '%1' and '%3' 38312for 'x' in this 'asm' and expect them to be the same. In fact, '%3' may 38313well not be a register. It might be a symbolic memory reference to the 38314object pointed to by 'x'. 38315 38316 asm ("sumsq %0, %1, %2" 38317 : "+f" (result) 38318 : "r" (x), "r" (y), "m" (*x), "m" (*y)); 38319 38320 Here is a fictitious '*z++ = *x++ * *y++' instruction. Notice that the 38321'x', 'y' and 'z' pointer registers must be specified as input/output 38322because the 'asm' modifies them. 38323 38324 asm ("vecmul %0, %1, %2" 38325 : "+r" (z), "+r" (x), "+r" (y), "=m" (*z) 38326 : "m" (*x), "m" (*y)); 38327 38328 An x86 example where the string memory argument is of unknown length. 38329 38330 asm("repne scasb" 38331 : "=c" (count), "+D" (p) 38332 : "m" (*(const char (*)[]) p), "0" (-1), "a" (0)); 38333 38334 If you know the above will only be reading a ten byte array then you 38335could instead use a memory input like: '"m" (*(const char (*)[10]) p)'. 38336 38337 Here is an example of a PowerPC vector scale implemented in assembly, 38338complete with vector and condition code clobbers, and some initialized 38339offset registers that are unchanged by the 'asm'. 38340 38341 void 38342 dscal (size_t n, double *x, double alpha) 38343 { 38344 asm ("/* lots of asm here */" 38345 : "+m" (*(double (*)[n]) x), "+&r" (n), "+b" (x) 38346 : "d" (alpha), "b" (32), "b" (48), "b" (64), 38347 "b" (80), "b" (96), "b" (112) 38348 : "cr0", 38349 "vs32","vs33","vs34","vs35","vs36","vs37","vs38","vs39", 38350 "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47"); 38351 } 38352 38353 Rather than allocating fixed registers via clobbers to provide scratch 38354registers for an 'asm' statement, an alternative is to define a variable 38355and make it an early-clobber output as with 'a2' and 'a3' in the example 38356below. This gives the compiler register allocator more freedom. You 38357can also define a variable and make it an output tied to an input as 38358with 'a0' and 'a1', tied respectively to 'ap' and 'lda'. Of course, 38359with tied outputs your 'asm' can't use the input value after modifying 38360the output register since they are one and the same register. What's 38361more, if you omit the early-clobber on the output, it is possible that 38362GCC might allocate the same register to another of the inputs if GCC 38363could prove they had the same value on entry to the 'asm'. This is why 38364'a1' has an early-clobber. Its tied input, 'lda' might conceivably be 38365known to have the value 16 and without an early-clobber share the same 38366register as '%11'. On the other hand, 'ap' can't be the same as any of 38367the other inputs, so an early-clobber on 'a0' is not needed. It is also 38368not desirable in this case. An early-clobber on 'a0' would cause GCC to 38369allocate a separate register for the '"m" (*(const double (*)[]) ap)' 38370input. Note that tying an input to an output is the way to set up an 38371initialized temporary register modified by an 'asm' statement. An input 38372not tied to an output is assumed by GCC to be unchanged, for example 38373'"b" (16)' below sets up '%11' to 16, and GCC might use that register in 38374following code if the value 16 happened to be needed. You can even use 38375a normal 'asm' output for a scratch if all inputs that might share the 38376same register are consumed before the scratch is used. The VSX 38377registers clobbered by the 'asm' statement could have used this 38378technique except for GCC's limit on the number of 'asm' parameters. 38379 38380 static void 38381 dgemv_kernel_4x4 (long n, const double *ap, long lda, 38382 const double *x, double *y, double alpha) 38383 { 38384 double *a0; 38385 double *a1; 38386 double *a2; 38387 double *a3; 38388 38389 __asm__ 38390 ( 38391 /* lots of asm here */ 38392 "#n=%1 ap=%8=%12 lda=%13 x=%7=%10 y=%0=%2 alpha=%9 o16=%11\n" 38393 "#a0=%3 a1=%4 a2=%5 a3=%6" 38394 : 38395 "+m" (*(double (*)[n]) y), 38396 "+&r" (n), // 1 38397 "+b" (y), // 2 38398 "=b" (a0), // 3 38399 "=&b" (a1), // 4 38400 "=&b" (a2), // 5 38401 "=&b" (a3) // 6 38402 : 38403 "m" (*(const double (*)[n]) x), 38404 "m" (*(const double (*)[]) ap), 38405 "d" (alpha), // 9 38406 "r" (x), // 10 38407 "b" (16), // 11 38408 "3" (ap), // 12 38409 "4" (lda) // 13 38410 : 38411 "cr0", 38412 "vs32","vs33","vs34","vs35","vs36","vs37", 38413 "vs40","vs41","vs42","vs43","vs44","vs45","vs46","vs47" 38414 ); 38415 } 38416 384176.47.2.7 Goto Labels 38418.................... 38419 38420'asm goto' allows assembly code to jump to one or more C labels. The 38421GOTOLABELS section in an 'asm goto' statement contains a comma-separated 38422list of all C labels to which the assembler code may jump. GCC assumes 38423that 'asm' execution falls through to the next statement (if this is not 38424the case, consider using the '__builtin_unreachable' intrinsic after the 38425'asm' statement). Optimization of 'asm goto' may be improved by using 38426the 'hot' and 'cold' label attributes (*note Label Attributes::). 38427 38428 An 'asm goto' statement cannot have outputs. This is due to an 38429internal restriction of the compiler: control transfer instructions 38430cannot have outputs. If the assembler code does modify anything, use 38431the '"memory"' clobber to force the optimizers to flush all register 38432values to memory and reload them if necessary after the 'asm' statement. 38433 38434 Also note that an 'asm goto' statement is always implicitly considered 38435volatile. 38436 38437 To reference a label in the assembler template, prefix it with '%l' 38438(lowercase 'L') followed by its (zero-based) position in GOTOLABELS plus 38439the number of input operands. For example, if the 'asm' has three 38440inputs and references two labels, refer to the first label as '%l3' and 38441the second as '%l4'). 38442 38443 Alternately, you can reference labels using the actual C label name 38444enclosed in brackets. For example, to reference a label named 'carry', 38445you can use '%l[carry]'. The label must still be listed in the 38446GOTOLABELS section when using this approach. 38447 38448 Here is an example of 'asm goto' for i386: 38449 38450 asm goto ( 38451 "btl %1, %0\n\t" 38452 "jc %l2" 38453 : /* No outputs. */ 38454 : "r" (p1), "r" (p2) 38455 : "cc" 38456 : carry); 38457 38458 return 0; 38459 38460 carry: 38461 return 1; 38462 38463 The following example shows an 'asm goto' that uses a memory clobber. 38464 38465 int frob(int x) 38466 { 38467 int y; 38468 asm goto ("frob %%r5, %1; jc %l[error]; mov (%2), %%r5" 38469 : /* No outputs. */ 38470 : "r"(x), "r"(&y) 38471 : "r5", "memory" 38472 : error); 38473 return y; 38474 error: 38475 return -1; 38476 } 38477 384786.47.2.8 x86 Operand Modifiers 38479.............................. 38480 38481References to input, output, and goto operands in the assembler template 38482of extended 'asm' statements can use modifiers to affect the way the 38483operands are formatted in the code output to the assembler. For 38484example, the following code uses the 'h' and 'b' modifiers for x86: 38485 38486 uint16_t num; 38487 asm volatile ("xchg %h0, %b0" : "+a" (num) ); 38488 38489These modifiers generate this assembler code: 38490 38491 xchg %ah, %al 38492 38493 The rest of this discussion uses the following code for illustrative 38494purposes. 38495 38496 int main() 38497 { 38498 int iInt = 1; 38499 38500 top: 38501 38502 asm volatile goto ("some assembler instructions here" 38503 : /* No outputs. */ 38504 : "q" (iInt), "X" (sizeof(unsigned char) + 1), "i" (42) 38505 : /* No clobbers. */ 38506 : top); 38507 } 38508 38509 With no modifiers, this is what the output from the operands would be 38510for the 'att' and 'intel' dialects of assembler: 38511 38512Operand 'att' 'intel' 38513----------------------------------- 38514'%0' '%eax' 'eax' 38515'%1' '$2' '2' 38516'%3' '$.L3' 'OFFSET 38517 FLAT:.L3' 38518 38519 The table below shows the list of supported modifiers and their 38520effects. 38521 38522Modifier Description Operand 'att' 'intel' 38523------------------------------------------------------------------------------------ 38524'A' Print an absolute memory reference. '%A0' '*%rax' 'rax' 38525'b' Print the QImode name of the register. '%b0' '%al' 'al' 38526'c' Require a constant operand and print the '%c1' '2' '2' 38527 constant expression with no punctuation. 38528'E' Print the address in Double Integer '%E1' '%(rax)''[rax]' 38529 (DImode) mode (8 bytes) when the target is 38530 64-bit. Otherwise mode is unspecified 38531 (VOIDmode). 38532'h' Print the QImode name for a "high" '%h0' '%ah' 'ah' 38533 register. 38534'H' Add 8 bytes to an offsettable memory '%H0' '8(%rax)''8[rax]' 38535 reference. Useful when accessing the high 38536 8 bytes of SSE values. For a memref in 38537 (%rax), it generates 38538'k' Print the SImode name of the register. '%k0' '%eax' 'eax' 38539'l' Print the label name with no punctuation. '%l3' '.L3' '.L3' 38540'p' Print raw symbol name (without '%p2' '42' '42' 38541 syntax-specific prefixes). 38542'P' If used for a function, print the PLT 38543 suffix and generate PIC code. For 38544 example, emit 'foo@PLT' instead of 'foo' 38545 for the function foo(). If used for a 38546 constant, drop all syntax-specific 38547 prefixes and issue the bare constant. See 38548 'p' above. 38549'q' Print the DImode name of the register. '%q0' '%rax' 'rax' 38550'w' Print the HImode name of the register. '%w0' '%ax' 'ax' 38551'z' Print the opcode suffix for the size of '%z0' 'l' 38552 the current integer operand (one of 38553 'b'/'w'/'l'/'q'). 38554 38555 'V' is a special modifier which prints the name of the full integer 38556register without '%'. 38557 385586.47.2.9 x86 Floating-Point 'asm' Operands 38559.......................................... 38560 38561On x86 targets, there are several rules on the usage of stack-like 38562registers in the operands of an 'asm'. These rules apply only to the 38563operands that are stack-like registers: 38564 38565 1. Given a set of input registers that die in an 'asm', it is 38566 necessary to know which are implicitly popped by the 'asm', and 38567 which must be explicitly popped by GCC. 38568 38569 An input register that is implicitly popped by the 'asm' must be 38570 explicitly clobbered, unless it is constrained to match an output 38571 operand. 38572 38573 2. For any input register that is implicitly popped by an 'asm', it is 38574 necessary to know how to adjust the stack to compensate for the 38575 pop. If any non-popped input is closer to the top of the reg-stack 38576 than the implicitly popped register, it would not be possible to 38577 know what the stack looked like--it's not clear how the rest of the 38578 stack "slides up". 38579 38580 All implicitly popped input registers must be closer to the top of 38581 the reg-stack than any input that is not implicitly popped. 38582 38583 It is possible that if an input dies in an 'asm', the compiler 38584 might use the input register for an output reload. Consider this 38585 example: 38586 38587 asm ("foo" : "=t" (a) : "f" (b)); 38588 38589 This code says that input 'b' is not popped by the 'asm', and that 38590 the 'asm' pushes a result onto the reg-stack, i.e., the stack is 38591 one deeper after the 'asm' than it was before. But, it is possible 38592 that reload may think that it can use the same register for both 38593 the input and the output. 38594 38595 To prevent this from happening, if any input operand uses the 'f' 38596 constraint, all output register constraints must use the '&' 38597 early-clobber modifier. 38598 38599 The example above is correctly written as: 38600 38601 asm ("foo" : "=&t" (a) : "f" (b)); 38602 38603 3. Some operands need to be in particular places on the stack. All 38604 output operands fall in this category--GCC has no other way to know 38605 which registers the outputs appear in unless you indicate this in 38606 the constraints. 38607 38608 Output operands must specifically indicate which register an output 38609 appears in after an 'asm'. '=f' is not allowed: the operand 38610 constraints must select a class with a single register. 38611 38612 4. Output operands may not be "inserted" between existing stack 38613 registers. Since no 387 opcode uses a read/write operand, all 38614 output operands are dead before the 'asm', and are pushed by the 38615 'asm'. It makes no sense to push anywhere but the top of the 38616 reg-stack. 38617 38618 Output operands must start at the top of the reg-stack: output 38619 operands may not "skip" a register. 38620 38621 5. Some 'asm' statements may need extra stack space for internal 38622 calculations. This can be guaranteed by clobbering stack registers 38623 unrelated to the inputs and outputs. 38624 38625 This 'asm' takes one input, which is internally popped, and produces 38626two outputs. 38627 38628 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp)); 38629 38630This 'asm' takes two inputs, which are popped by the 'fyl2xp1' opcode, 38631and replaces them with one output. The 'st(1)' clobber is necessary for 38632the compiler to know that 'fyl2xp1' pops both inputs. 38633 38634 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)"); 38635 38636 38637File: gcc.info, Node: Constraints, Next: Asm Labels, Prev: Extended Asm, Up: Using Assembly Language with C 38638 386396.47.3 Constraints for 'asm' Operands 38640------------------------------------- 38641 38642Here are specific details on what constraint letters you can use with 38643'asm' operands. Constraints can say whether an operand may be in a 38644register, and which kinds of register; whether the operand can be a 38645memory reference, and which kinds of address; whether the operand may be 38646an immediate constant, and which possible values it may have. 38647Constraints can also require two operands to match. Side-effects aren't 38648allowed in operands of inline 'asm', unless '<' or '>' constraints are 38649used, because there is no guarantee that the side effects will happen 38650exactly once in an instruction that can update the addressing register. 38651 38652* Menu: 38653 38654* Simple Constraints:: Basic use of constraints. 38655* Multi-Alternative:: When an insn has two alternative constraint-patterns. 38656* Modifiers:: More precise control over effects of constraints. 38657* Machine Constraints:: Special constraints for some particular machines. 38658 38659 38660File: gcc.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints 38661 386626.47.3.1 Simple Constraints 38663........................... 38664 38665The simplest kind of constraint is a string full of letters, each of 38666which describes one kind of operand that is permitted. Here are the 38667letters that are allowed: 38668 38669whitespace 38670 Whitespace characters are ignored and can be inserted at any 38671 position except the first. This enables each alternative for 38672 different operands to be visually aligned in the machine 38673 description even if they have different number of constraints and 38674 modifiers. 38675 38676'm' 38677 A memory operand is allowed, with any kind of address that the 38678 machine supports in general. Note that the letter used for the 38679 general memory constraint can be re-defined by a back end using the 38680 'TARGET_MEM_CONSTRAINT' macro. 38681 38682'o' 38683 A memory operand is allowed, but only if the address is 38684 "offsettable". This means that adding a small integer (actually, 38685 the width in bytes of the operand, as determined by its machine 38686 mode) may be added to the address and the result is also a valid 38687 memory address. 38688 38689 For example, an address which is constant is offsettable; so is an 38690 address that is the sum of a register and a constant (as long as a 38691 slightly larger constant is also within the range of 38692 address-offsets supported by the machine); but an autoincrement or 38693 autodecrement address is not offsettable. More complicated 38694 indirect/indexed addresses may or may not be offsettable depending 38695 on the other addressing modes that the machine supports. 38696 38697 Note that in an output operand which can be matched by another 38698 operand, the constraint letter 'o' is valid only when accompanied 38699 by both '<' (if the target machine has predecrement addressing) and 38700 '>' (if the target machine has preincrement addressing). 38701 38702'V' 38703 A memory operand that is not offsettable. In other words, anything 38704 that would fit the 'm' constraint but not the 'o' constraint. 38705 38706'<' 38707 A memory operand with autodecrement addressing (either predecrement 38708 or postdecrement) is allowed. In inline 'asm' this constraint is 38709 only allowed if the operand is used exactly once in an instruction 38710 that can handle the side effects. Not using an operand with '<' in 38711 constraint string in the inline 'asm' pattern at all or using it in 38712 multiple instructions isn't valid, because the side effects 38713 wouldn't be performed or would be performed more than once. 38714 Furthermore, on some targets the operand with '<' in constraint 38715 string must be accompanied by special instruction suffixes like 38716 '%U0' instruction suffix on PowerPC or '%P0' on IA-64. 38717 38718'>' 38719 A memory operand with autoincrement addressing (either preincrement 38720 or postincrement) is allowed. In inline 'asm' the same 38721 restrictions as for '<' apply. 38722 38723'r' 38724 A register operand is allowed provided that it is in a general 38725 register. 38726 38727'i' 38728 An immediate integer operand (one with constant value) is allowed. 38729 This includes symbolic constants whose values will be known only at 38730 assembly time or later. 38731 38732'n' 38733 An immediate integer operand with a known numeric value is allowed. 38734 Many systems cannot support assembly-time constants for operands 38735 less than a word wide. Constraints for these operands should use 38736 'n' rather than 'i'. 38737 38738'I', 'J', 'K', ... 'P' 38739 Other letters in the range 'I' through 'P' may be defined in a 38740 machine-dependent fashion to permit immediate integer operands with 38741 explicit integer values in specified ranges. For example, on the 38742 68000, 'I' is defined to stand for the range of values 1 to 8. 38743 This is the range permitted as a shift count in the shift 38744 instructions. 38745 38746'E' 38747 An immediate floating operand (expression code 'const_double') is 38748 allowed, but only if the target floating point format is the same 38749 as that of the host machine (on which the compiler is running). 38750 38751'F' 38752 An immediate floating operand (expression code 'const_double' or 38753 'const_vector') is allowed. 38754 38755'G', 'H' 38756 'G' and 'H' may be defined in a machine-dependent fashion to permit 38757 immediate floating operands in particular ranges of values. 38758 38759's' 38760 An immediate integer operand whose value is not an explicit integer 38761 is allowed. 38762 38763 This might appear strange; if an insn allows a constant operand 38764 with a value not known at compile time, it certainly must allow any 38765 known value. So why use 's' instead of 'i'? Sometimes it allows 38766 better code to be generated. 38767 38768 For example, on the 68000 in a fullword instruction it is possible 38769 to use an immediate operand; but if the immediate value is between 38770 -128 and 127, better code results from loading the value into a 38771 register and using the register. This is because the load into the 38772 register can be done with a 'moveq' instruction. We arrange for 38773 this to happen by defining the letter 'K' to mean "any integer 38774 outside the range -128 to 127", and then specifying 'Ks' in the 38775 operand constraints. 38776 38777'g' 38778 Any register, memory or immediate integer operand is allowed, 38779 except for registers that are not general registers. 38780 38781'X' 38782 Any operand whatsoever is allowed. 38783 38784'0', '1', '2', ... '9' 38785 An operand that matches the specified operand number is allowed. 38786 If a digit is used together with letters within the same 38787 alternative, the digit should come last. 38788 38789 This number is allowed to be more than a single digit. If multiple 38790 digits are encountered consecutively, they are interpreted as a 38791 single decimal integer. There is scant chance for ambiguity, since 38792 to-date it has never been desirable that '10' be interpreted as 38793 matching either operand 1 _or_ operand 0. Should this be desired, 38794 one can use multiple alternatives instead. 38795 38796 This is called a "matching constraint" and what it really means is 38797 that the assembler has only a single operand that fills two roles 38798 which 'asm' distinguishes. For example, an add instruction uses 38799 two input operands and an output operand, but on most CISC machines 38800 an add instruction really has only two operands, one of them an 38801 input-output operand: 38802 38803 addl #35,r12 38804 38805 Matching constraints are used in these circumstances. More 38806 precisely, the two operands that match must include one input-only 38807 operand and one output-only operand. Moreover, the digit must be a 38808 smaller number than the number of the operand that uses it in the 38809 constraint. 38810 38811'p' 38812 An operand that is a valid memory address is allowed. This is for 38813 "load address" and "push address" instructions. 38814 38815 'p' in the constraint must be accompanied by 'address_operand' as 38816 the predicate in the 'match_operand'. This predicate interprets 38817 the mode specified in the 'match_operand' as the mode of the memory 38818 reference for which the address would be valid. 38819 38820OTHER-LETTERS 38821 Other letters can be defined in machine-dependent fashion to stand 38822 for particular classes of registers or other arbitrary operand 38823 types. 'd', 'a' and 'f' are defined on the 68000/68020 to stand 38824 for data, address and floating point registers. 38825 38826 38827File: gcc.info, Node: Multi-Alternative, Next: Modifiers, Prev: Simple Constraints, Up: Constraints 38828 388296.47.3.2 Multiple Alternative Constraints 38830......................................... 38831 38832Sometimes a single instruction has multiple alternative sets of possible 38833operands. For example, on the 68000, a logical-or instruction can 38834combine register or an immediate value into memory, or it can combine 38835any kind of operand into a register; but it cannot combine one memory 38836location into another. 38837 38838 These constraints are represented as multiple alternatives. An 38839alternative can be described by a series of letters for each operand. 38840The overall constraint for an operand is made from the letters for this 38841operand from the first alternative, a comma, the letters for this 38842operand from the second alternative, a comma, and so on until the last 38843alternative. All operands for a single instruction must have the same 38844number of alternatives. 38845 38846 So the first alternative for the 68000's logical-or could be written as 38847'"+m" (output) : "ir" (input)'. The second could be '"+r" (output): 38848"irm" (input)'. However, the fact that two memory locations cannot be 38849used in a single instruction prevents simply using '"+rm" (output) : 38850"irm" (input)'. Using multi-alternatives, this might be written as 38851'"+m,r" (output) : "ir,irm" (input)'. This describes all the available 38852alternatives to the compiler, allowing it to choose the most efficient 38853one for the current conditions. 38854 38855 There is no way within the template to determine which alternative was 38856chosen. However you may be able to wrap your 'asm' statements with 38857builtins such as '__builtin_constant_p' to achieve the desired results. 38858 38859 38860File: gcc.info, Node: Modifiers, Next: Machine Constraints, Prev: Multi-Alternative, Up: Constraints 38861 388626.47.3.3 Constraint Modifier Characters 38863....................................... 38864 38865Here are constraint modifier characters. 38866 38867'=' 38868 Means that this operand is written to by this instruction: the 38869 previous value is discarded and replaced by new data. 38870 38871'+' 38872 Means that this operand is both read and written by the 38873 instruction. 38874 38875 When the compiler fixes up the operands to satisfy the constraints, 38876 it needs to know which operands are read by the instruction and 38877 which are written by it. '=' identifies an operand which is only 38878 written; '+' identifies an operand that is both read and written; 38879 all other operands are assumed to only be read. 38880 38881 If you specify '=' or '+' in a constraint, you put it in the first 38882 character of the constraint string. 38883 38884'&' 38885 Means (in a particular alternative) that this operand is an 38886 "earlyclobber" operand, which is written before the instruction is 38887 finished using the input operands. Therefore, this operand may not 38888 lie in a register that is read by the instruction or as part of any 38889 memory address. 38890 38891 '&' applies only to the alternative in which it is written. In 38892 constraints with multiple alternatives, sometimes one alternative 38893 requires '&' while others do not. See, for example, the 'movdf' 38894 insn of the 68000. 38895 38896 A operand which is read by the instruction can be tied to an 38897 earlyclobber operand if its only use as an input occurs before the 38898 early result is written. Adding alternatives of this form often 38899 allows GCC to produce better code when only some of the read 38900 operands can be affected by the earlyclobber. See, for example, 38901 the 'mulsi3' insn of the ARM. 38902 38903 Furthermore, if the "earlyclobber" operand is also a read/write 38904 operand, then that operand is written only after it's used. 38905 38906 '&' does not obviate the need to write '=' or '+'. As 38907 "earlyclobber" operands are always written, a read-only 38908 "earlyclobber" operand is ill-formed and will be rejected by the 38909 compiler. 38910 38911'%' 38912 Declares the instruction to be commutative for this operand and the 38913 following operand. This means that the compiler may interchange 38914 the two operands if that is the cheapest way to make all operands 38915 fit the constraints. '%' applies to all alternatives and must 38916 appear as the first character in the constraint. Only read-only 38917 operands can use '%'. 38918 38919 GCC can only handle one commutative pair in an asm; if you use 38920 more, the compiler may fail. Note that you need not use the 38921 modifier if the two alternatives are strictly identical; this would 38922 only waste time in the reload pass. 38923 38924 38925File: gcc.info, Node: Machine Constraints, Prev: Modifiers, Up: Constraints 38926 389276.47.3.4 Constraints for Particular Machines 38928............................................ 38929 38930Whenever possible, you should use the general-purpose constraint letters 38931in 'asm' arguments, since they will convey meaning more readily to 38932people reading your code. Failing that, use the constraint letters that 38933usually have very similar meanings across architectures. The most 38934commonly used constraints are 'm' and 'r' (for memory and 38935general-purpose registers respectively; *note Simple Constraints::), and 38936'I', usually the letter indicating the most common immediate-constant 38937format. 38938 38939 Each architecture defines additional constraints. These constraints 38940are used by the compiler itself for instruction generation, as well as 38941for 'asm' statements; therefore, some of the constraints are not 38942particularly useful for 'asm'. Here is a summary of some of the 38943machine-dependent constraints available on some particular machines; it 38944includes both constraints that are useful for 'asm' and constraints that 38945aren't. The compiler source file mentioned in the table heading for 38946each architecture is the definitive reference for the meanings of that 38947architecture's constraints. 38948 38949_AArch64 family--'config/aarch64/constraints.md'_ 38950 'k' 38951 The stack pointer register ('SP') 38952 38953 'w' 38954 Floating point register, Advanced SIMD vector register or SVE 38955 vector register 38956 38957 'x' 38958 Like 'w', but restricted to registers 0 to 15 inclusive. 38959 38960 'y' 38961 Like 'w', but restricted to registers 0 to 7 inclusive. 38962 38963 'Upl' 38964 One of the low eight SVE predicate registers ('P0' to 'P7') 38965 38966 'Upa' 38967 Any of the SVE predicate registers ('P0' to 'P15') 38968 38969 'I' 38970 Integer constant that is valid as an immediate operand in an 38971 'ADD' instruction 38972 38973 'J' 38974 Integer constant that is valid as an immediate operand in a 38975 'SUB' instruction (once negated) 38976 38977 'K' 38978 Integer constant that can be used with a 32-bit logical 38979 instruction 38980 38981 'L' 38982 Integer constant that can be used with a 64-bit logical 38983 instruction 38984 38985 'M' 38986 Integer constant that is valid as an immediate operand in a 38987 32-bit 'MOV' pseudo instruction. The 'MOV' may be assembled 38988 to one of several different machine instructions depending on 38989 the value 38990 38991 'N' 38992 Integer constant that is valid as an immediate operand in a 38993 64-bit 'MOV' pseudo instruction 38994 38995 'S' 38996 An absolute symbolic address or a label reference 38997 38998 'Y' 38999 Floating point constant zero 39000 39001 'Z' 39002 Integer constant zero 39003 39004 'Ush' 39005 The high part (bits 12 and upwards) of the pc-relative address 39006 of a symbol within 4GB of the instruction 39007 39008 'Q' 39009 A memory address which uses a single base register with no 39010 offset 39011 39012 'Ump' 39013 A memory address suitable for a load/store pair instruction in 39014 SI, DI, SF and DF modes 39015 39016_AMD GCN --'config/gcn/constraints.md'_ 39017 'I' 39018 Immediate integer in the range -16 to 64 39019 39020 'J' 39021 Immediate 16-bit signed integer 39022 39023 'Kf' 39024 Immediate constant -1 39025 39026 'L' 39027 Immediate 15-bit unsigned integer 39028 39029 'A' 39030 Immediate constant that can be inlined in an instruction 39031 encoding: integer -16..64, or float 0.0, +/-0.5, +/-1.0, 39032 +/-2.0, +/-4.0, 1.0/(2.0*PI) 39033 39034 'B' 39035 Immediate 32-bit signed integer that can be attached to an 39036 instruction encoding 39037 39038 'C' 39039 Immediate 32-bit integer in range -16..4294967295 (i.e. 39040 32-bit unsigned integer or 'A' constraint) 39041 39042 'DA' 39043 Immediate 64-bit constant that can be split into two 'A' 39044 constants 39045 39046 'DB' 39047 Immediate 64-bit constant that can be split into two 'B' 39048 constants 39049 39050 'U' 39051 Any 'unspec' 39052 39053 'Y' 39054 Any 'symbol_ref' or 'label_ref' 39055 39056 'v' 39057 VGPR register 39058 39059 'Sg' 39060 SGPR register 39061 39062 'SD' 39063 SGPR registers valid for instruction destinations, including 39064 VCC, M0 and EXEC 39065 39066 'SS' 39067 SGPR registers valid for instruction sources, including VCC, 39068 M0, EXEC and SCC 39069 39070 'Sm' 39071 SGPR registers valid as a source for scalar memory 39072 instructions (excludes M0 and EXEC) 39073 39074 'Sv' 39075 SGPR registers valid as a source or destination for vector 39076 instructions (excludes EXEC) 39077 39078 'ca' 39079 All condition registers: SCC, VCCZ, EXECZ 39080 39081 'cs' 39082 Scalar condition register: SCC 39083 39084 'cV' 39085 Vector condition register: VCC, VCC_LO, VCC_HI 39086 39087 'e' 39088 EXEC register (EXEC_LO and EXEC_HI) 39089 39090 'RB' 39091 Memory operand with address space suitable for 'buffer_*' 39092 instructions 39093 39094 'RF' 39095 Memory operand with address space suitable for 'flat_*' 39096 instructions 39097 39098 'RS' 39099 Memory operand with address space suitable for 's_*' 39100 instructions 39101 39102 'RL' 39103 Memory operand with address space suitable for 'ds_*' LDS 39104 instructions 39105 39106 'RG' 39107 Memory operand with address space suitable for 'ds_*' GDS 39108 instructions 39109 39110 'RD' 39111 Memory operand with address space suitable for any 'ds_*' 39112 instructions 39113 39114 'RM' 39115 Memory operand with address space suitable for 'global_*' 39116 instructions 39117 39118_ARC --'config/arc/constraints.md'_ 39119 'q' 39120 Registers usable in ARCompact 16-bit instructions: 'r0'-'r3', 39121 'r12'-'r15'. This constraint can only match when the '-mq' 39122 option is in effect. 39123 39124 'e' 39125 Registers usable as base-regs of memory addresses in ARCompact 39126 16-bit memory instructions: 'r0'-'r3', 'r12'-'r15', 'sp'. 39127 This constraint can only match when the '-mq' option is in 39128 effect. 39129 'D' 39130 ARC FPX (dpfp) 64-bit registers. 'D0', 'D1'. 39131 39132 'I' 39133 A signed 12-bit integer constant. 39134 39135 'Cal' 39136 constant for arithmetic/logical operations. This might be any 39137 constant that can be put into a long immediate by the assmbler 39138 or linker without involving a PIC relocation. 39139 39140 'K' 39141 A 3-bit unsigned integer constant. 39142 39143 'L' 39144 A 6-bit unsigned integer constant. 39145 39146 'CnL' 39147 One's complement of a 6-bit unsigned integer constant. 39148 39149 'CmL' 39150 Two's complement of a 6-bit unsigned integer constant. 39151 39152 'M' 39153 A 5-bit unsigned integer constant. 39154 39155 'O' 39156 A 7-bit unsigned integer constant. 39157 39158 'P' 39159 A 8-bit unsigned integer constant. 39160 39161 'H' 39162 Any const_double value. 39163 39164_ARM family--'config/arm/constraints.md'_ 39165 39166 'h' 39167 In Thumb state, the core registers 'r8'-'r15'. 39168 39169 'k' 39170 The stack pointer register. 39171 39172 'l' 39173 In Thumb State the core registers 'r0'-'r7'. In ARM state 39174 this is an alias for the 'r' constraint. 39175 39176 't' 39177 VFP floating-point registers 's0'-'s31'. Used for 32 bit 39178 values. 39179 39180 'w' 39181 VFP floating-point registers 'd0'-'d31' and the appropriate 39182 subset 'd0'-'d15' based on command line options. Used for 64 39183 bit values only. Not valid for Thumb1. 39184 39185 'y' 39186 The iWMMX co-processor registers. 39187 39188 'z' 39189 The iWMMX GR registers. 39190 39191 'G' 39192 The floating-point constant 0.0 39193 39194 'I' 39195 Integer that is valid as an immediate operand in a data 39196 processing instruction. That is, an integer in the range 0 to 39197 255 rotated by a multiple of 2 39198 39199 'J' 39200 Integer in the range -4095 to 4095 39201 39202 'K' 39203 Integer that satisfies constraint 'I' when inverted (ones 39204 complement) 39205 39206 'L' 39207 Integer that satisfies constraint 'I' when negated (twos 39208 complement) 39209 39210 'M' 39211 Integer in the range 0 to 32 39212 39213 'Q' 39214 A memory reference where the exact address is in a single 39215 register (''m'' is preferable for 'asm' statements) 39216 39217 'R' 39218 An item in the constant pool 39219 39220 'S' 39221 A symbol in the text segment of the current file 39222 39223 'Uv' 39224 A memory reference suitable for VFP load/store insns 39225 (reg+constant offset) 39226 39227 'Uy' 39228 A memory reference suitable for iWMMXt load/store 39229 instructions. 39230 39231 'Uq' 39232 A memory reference suitable for the ARMv4 ldrsb instruction. 39233 39234_AVR family--'config/avr/constraints.md'_ 39235 'l' 39236 Registers from r0 to r15 39237 39238 'a' 39239 Registers from r16 to r23 39240 39241 'd' 39242 Registers from r16 to r31 39243 39244 'w' 39245 Registers from r24 to r31. These registers can be used in 39246 'adiw' command 39247 39248 'e' 39249 Pointer register (r26-r31) 39250 39251 'b' 39252 Base pointer register (r28-r31) 39253 39254 'q' 39255 Stack pointer register (SPH:SPL) 39256 39257 't' 39258 Temporary register r0 39259 39260 'x' 39261 Register pair X (r27:r26) 39262 39263 'y' 39264 Register pair Y (r29:r28) 39265 39266 'z' 39267 Register pair Z (r31:r30) 39268 39269 'I' 39270 Constant greater than -1, less than 64 39271 39272 'J' 39273 Constant greater than -64, less than 1 39274 39275 'K' 39276 Constant integer 2 39277 39278 'L' 39279 Constant integer 0 39280 39281 'M' 39282 Constant that fits in 8 bits 39283 39284 'N' 39285 Constant integer -1 39286 39287 'O' 39288 Constant integer 8, 16, or 24 39289 39290 'P' 39291 Constant integer 1 39292 39293 'G' 39294 A floating point constant 0.0 39295 39296 'Q' 39297 A memory address based on Y or Z pointer with displacement. 39298 39299_Blackfin family--'config/bfin/constraints.md'_ 39300 'a' 39301 P register 39302 39303 'd' 39304 D register 39305 39306 'z' 39307 A call clobbered P register. 39308 39309 'qN' 39310 A single register. If N is in the range 0 to 7, the 39311 corresponding D register. If it is 'A', then the register P0. 39312 39313 'D' 39314 Even-numbered D register 39315 39316 'W' 39317 Odd-numbered D register 39318 39319 'e' 39320 Accumulator register. 39321 39322 'A' 39323 Even-numbered accumulator register. 39324 39325 'B' 39326 Odd-numbered accumulator register. 39327 39328 'b' 39329 I register 39330 39331 'v' 39332 B register 39333 39334 'f' 39335 M register 39336 39337 'c' 39338 Registers used for circular buffering, i.e. I, B, or L 39339 registers. 39340 39341 'C' 39342 The CC register. 39343 39344 't' 39345 LT0 or LT1. 39346 39347 'k' 39348 LC0 or LC1. 39349 39350 'u' 39351 LB0 or LB1. 39352 39353 'x' 39354 Any D, P, B, M, I or L register. 39355 39356 'y' 39357 Additional registers typically used only in prologues and 39358 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and 39359 USP. 39360 39361 'w' 39362 Any register except accumulators or CC. 39363 39364 'Ksh' 39365 Signed 16 bit integer (in the range -32768 to 32767) 39366 39367 'Kuh' 39368 Unsigned 16 bit integer (in the range 0 to 65535) 39369 39370 'Ks7' 39371 Signed 7 bit integer (in the range -64 to 63) 39372 39373 'Ku7' 39374 Unsigned 7 bit integer (in the range 0 to 127) 39375 39376 'Ku5' 39377 Unsigned 5 bit integer (in the range 0 to 31) 39378 39379 'Ks4' 39380 Signed 4 bit integer (in the range -8 to 7) 39381 39382 'Ks3' 39383 Signed 3 bit integer (in the range -3 to 4) 39384 39385 'Ku3' 39386 Unsigned 3 bit integer (in the range 0 to 7) 39387 39388 'PN' 39389 Constant N, where N is a single-digit constant in the range 0 39390 to 4. 39391 39392 'PA' 39393 An integer equal to one of the MACFLAG_XXX constants that is 39394 suitable for use with either accumulator. 39395 39396 'PB' 39397 An integer equal to one of the MACFLAG_XXX constants that is 39398 suitable for use only with accumulator A1. 39399 39400 'M1' 39401 Constant 255. 39402 39403 'M2' 39404 Constant 65535. 39405 39406 'J' 39407 An integer constant with exactly a single bit set. 39408 39409 'L' 39410 An integer constant with all bits set except exactly one. 39411 39412 'H' 39413 39414 'Q' 39415 Any SYMBOL_REF. 39416 39417_CR16 Architecture--'config/cr16/cr16.h'_ 39418 39419 'b' 39420 Registers from r0 to r14 (registers without stack pointer) 39421 39422 't' 39423 Register from r0 to r11 (all 16-bit registers) 39424 39425 'p' 39426 Register from r12 to r15 (all 32-bit registers) 39427 39428 'I' 39429 Signed constant that fits in 4 bits 39430 39431 'J' 39432 Signed constant that fits in 5 bits 39433 39434 'K' 39435 Signed constant that fits in 6 bits 39436 39437 'L' 39438 Unsigned constant that fits in 4 bits 39439 39440 'M' 39441 Signed constant that fits in 32 bits 39442 39443 'N' 39444 Check for 64 bits wide constants for add/sub instructions 39445 39446 'G' 39447 Floating point constant that is legal for store immediate 39448 39449_C-SKY--'config/csky/constraints.md'_ 39450 39451 'a' 39452 The mini registers r0 - r7. 39453 39454 'b' 39455 The low registers r0 - r15. 39456 39457 'c' 39458 C register. 39459 39460 'y' 39461 HI and LO registers. 39462 39463 'l' 39464 LO register. 39465 39466 'h' 39467 HI register. 39468 39469 'v' 39470 Vector registers. 39471 39472 'z' 39473 Stack pointer register (SP). 39474 39475_Epiphany--'config/epiphany/constraints.md'_ 39476 'U16' 39477 An unsigned 16-bit constant. 39478 39479 'K' 39480 An unsigned 5-bit constant. 39481 39482 'L' 39483 A signed 11-bit constant. 39484 39485 'Cm1' 39486 A signed 11-bit constant added to -1. Can only match when the 39487 '-m1reg-REG' option is active. 39488 39489 'Cl1' 39490 Left-shift of -1, i.e., a bit mask with a block of leading 39491 ones, the rest being a block of trailing zeroes. Can only 39492 match when the '-m1reg-REG' option is active. 39493 39494 'Cr1' 39495 Right-shift of -1, i.e., a bit mask with a trailing block of 39496 ones, the rest being zeroes. Or to put it another way, one 39497 less than a power of two. Can only match when the 39498 '-m1reg-REG' option is active. 39499 39500 'Cal' 39501 Constant for arithmetic/logical operations. This is like 'i', 39502 except that for position independent code, no symbols / 39503 expressions needing relocations are allowed. 39504 39505 'Csy' 39506 Symbolic constant for call/jump instruction. 39507 39508 'Rcs' 39509 The register class usable in short insns. This is a register 39510 class constraint, and can thus drive register allocation. 39511 This constraint won't match unless '-mprefer-short-insn-regs' 39512 is in effect. 39513 39514 'Rsc' 39515 The the register class of registers that can be used to hold a 39516 sibcall call address. I.e., a caller-saved register. 39517 39518 'Rct' 39519 Core control register class. 39520 39521 'Rgs' 39522 The register group usable in short insns. This constraint 39523 does not use a register class, so that it only passively 39524 matches suitable registers, and doesn't drive register 39525 allocation. 39526 39527 'Rra' 39528 Matches the return address if it can be replaced with the link 39529 register. 39530 39531 'Rcc' 39532 Matches the integer condition code register. 39533 39534 'Sra' 39535 Matches the return address if it is in a stack slot. 39536 39537 'Cfm' 39538 Matches control register values to switch fp mode, which are 39539 encapsulated in 'UNSPEC_FP_MODE'. 39540 39541_FRV--'config/frv/frv.h'_ 39542 'a' 39543 Register in the class 'ACC_REGS' ('acc0' to 'acc7'). 39544 39545 'b' 39546 Register in the class 'EVEN_ACC_REGS' ('acc0' to 'acc7'). 39547 39548 'c' 39549 Register in the class 'CC_REGS' ('fcc0' to 'fcc3' and 'icc0' 39550 to 'icc3'). 39551 39552 'd' 39553 Register in the class 'GPR_REGS' ('gr0' to 'gr63'). 39554 39555 'e' 39556 Register in the class 'EVEN_REGS' ('gr0' to 'gr63'). Odd 39557 registers are excluded not in the class but through the use of 39558 a machine mode larger than 4 bytes. 39559 39560 'f' 39561 Register in the class 'FPR_REGS' ('fr0' to 'fr63'). 39562 39563 'h' 39564 Register in the class 'FEVEN_REGS' ('fr0' to 'fr63'). Odd 39565 registers are excluded not in the class but through the use of 39566 a machine mode larger than 4 bytes. 39567 39568 'l' 39569 Register in the class 'LR_REG' (the 'lr' register). 39570 39571 'q' 39572 Register in the class 'QUAD_REGS' ('gr2' to 'gr63'). Register 39573 numbers not divisible by 4 are excluded not in the class but 39574 through the use of a machine mode larger than 8 bytes. 39575 39576 't' 39577 Register in the class 'ICC_REGS' ('icc0' to 'icc3'). 39578 39579 'u' 39580 Register in the class 'FCC_REGS' ('fcc0' to 'fcc3'). 39581 39582 'v' 39583 Register in the class 'ICR_REGS' ('cc4' to 'cc7'). 39584 39585 'w' 39586 Register in the class 'FCR_REGS' ('cc0' to 'cc3'). 39587 39588 'x' 39589 Register in the class 'QUAD_FPR_REGS' ('fr0' to 'fr63'). 39590 Register numbers not divisible by 4 are excluded not in the 39591 class but through the use of a machine mode larger than 8 39592 bytes. 39593 39594 'z' 39595 Register in the class 'SPR_REGS' ('lcr' and 'lr'). 39596 39597 'A' 39598 Register in the class 'QUAD_ACC_REGS' ('acc0' to 'acc7'). 39599 39600 'B' 39601 Register in the class 'ACCG_REGS' ('accg0' to 'accg7'). 39602 39603 'C' 39604 Register in the class 'CR_REGS' ('cc0' to 'cc7'). 39605 39606 'G' 39607 Floating point constant zero 39608 39609 'I' 39610 6-bit signed integer constant 39611 39612 'J' 39613 10-bit signed integer constant 39614 39615 'L' 39616 16-bit signed integer constant 39617 39618 'M' 39619 16-bit unsigned integer constant 39620 39621 'N' 39622 12-bit signed integer constant that is negative--i.e. in the 39623 range of -2048 to -1 39624 39625 'O' 39626 Constant zero 39627 39628 'P' 39629 12-bit signed integer constant that is greater than zero--i.e. 39630 in the range of 1 to 2047. 39631 39632_FT32--'config/ft32/constraints.md'_ 39633 'A' 39634 An absolute address 39635 39636 'B' 39637 An offset address 39638 39639 'W' 39640 A register indirect memory operand 39641 39642 'e' 39643 An offset address. 39644 39645 'f' 39646 An offset address. 39647 39648 'O' 39649 The constant zero or one 39650 39651 'I' 39652 A 16-bit signed constant (-32768 ... 32767) 39653 39654 'w' 39655 A bitfield mask suitable for bext or bins 39656 39657 'x' 39658 An inverted bitfield mask suitable for bext or bins 39659 39660 'L' 39661 A 16-bit unsigned constant, multiple of 4 (0 ... 65532) 39662 39663 'S' 39664 A 20-bit signed constant (-524288 ... 524287) 39665 39666 'b' 39667 A constant for a bitfield width (1 ... 16) 39668 39669 'KA' 39670 A 10-bit signed constant (-512 ... 511) 39671 39672_Hewlett-Packard PA-RISC--'config/pa/pa.h'_ 39673 'a' 39674 General register 1 39675 39676 'f' 39677 Floating point register 39678 39679 'q' 39680 Shift amount register 39681 39682 'x' 39683 Floating point register (deprecated) 39684 39685 'y' 39686 Upper floating point register (32-bit), floating point 39687 register (64-bit) 39688 39689 'Z' 39690 Any register 39691 39692 'I' 39693 Signed 11-bit integer constant 39694 39695 'J' 39696 Signed 14-bit integer constant 39697 39698 'K' 39699 Integer constant that can be deposited with a 'zdepi' 39700 instruction 39701 39702 'L' 39703 Signed 5-bit integer constant 39704 39705 'M' 39706 Integer constant 0 39707 39708 'N' 39709 Integer constant that can be loaded with a 'ldil' instruction 39710 39711 'O' 39712 Integer constant whose value plus one is a power of 2 39713 39714 'P' 39715 Integer constant that can be used for 'and' operations in 39716 'depi' and 'extru' instructions 39717 39718 'S' 39719 Integer constant 31 39720 39721 'U' 39722 Integer constant 63 39723 39724 'G' 39725 Floating-point constant 0.0 39726 39727 'A' 39728 A 'lo_sum' data-linkage-table memory operand 39729 39730 'Q' 39731 A memory operand that can be used as the destination operand 39732 of an integer store instruction 39733 39734 'R' 39735 A scaled or unscaled indexed memory operand 39736 39737 'T' 39738 A memory operand for floating-point loads and stores 39739 39740 'W' 39741 A register indirect memory operand 39742 39743_Intel IA-64--'config/ia64/ia64.h'_ 39744 'a' 39745 General register 'r0' to 'r3' for 'addl' instruction 39746 39747 'b' 39748 Branch register 39749 39750 'c' 39751 Predicate register ('c' as in "conditional") 39752 39753 'd' 39754 Application register residing in M-unit 39755 39756 'e' 39757 Application register residing in I-unit 39758 39759 'f' 39760 Floating-point register 39761 39762 'm' 39763 Memory operand. If used together with '<' or '>', the operand 39764 can have postincrement and postdecrement which require 39765 printing with '%Pn' on IA-64. 39766 39767 'G' 39768 Floating-point constant 0.0 or 1.0 39769 39770 'I' 39771 14-bit signed integer constant 39772 39773 'J' 39774 22-bit signed integer constant 39775 39776 'K' 39777 8-bit signed integer constant for logical instructions 39778 39779 'L' 39780 8-bit adjusted signed integer constant for compare pseudo-ops 39781 39782 'M' 39783 6-bit unsigned integer constant for shift counts 39784 39785 'N' 39786 9-bit signed integer constant for load and store 39787 postincrements 39788 39789 'O' 39790 The constant zero 39791 39792 'P' 39793 0 or -1 for 'dep' instruction 39794 39795 'Q' 39796 Non-volatile memory for floating-point loads and stores 39797 39798 'R' 39799 Integer constant in the range 1 to 4 for 'shladd' instruction 39800 39801 'S' 39802 Memory operand except postincrement and postdecrement. This 39803 is now roughly the same as 'm' when not used together with '<' 39804 or '>'. 39805 39806_M32C--'config/m32c/m32c.c'_ 39807 'Rsp' 39808 'Rfb' 39809 'Rsb' 39810 '$sp', '$fb', '$sb'. 39811 39812 'Rcr' 39813 Any control register, when they're 16 bits wide (nothing if 39814 control registers are 24 bits wide) 39815 39816 'Rcl' 39817 Any control register, when they're 24 bits wide. 39818 39819 'R0w' 39820 'R1w' 39821 'R2w' 39822 'R3w' 39823 $r0, $r1, $r2, $r3. 39824 39825 'R02' 39826 $r0 or $r2, or $r2r0 for 32 bit values. 39827 39828 'R13' 39829 $r1 or $r3, or $r3r1 for 32 bit values. 39830 39831 'Rdi' 39832 A register that can hold a 64 bit value. 39833 39834 'Rhl' 39835 $r0 or $r1 (registers with addressable high/low bytes) 39836 39837 'R23' 39838 $r2 or $r3 39839 39840 'Raa' 39841 Address registers 39842 39843 'Raw' 39844 Address registers when they're 16 bits wide. 39845 39846 'Ral' 39847 Address registers when they're 24 bits wide. 39848 39849 'Rqi' 39850 Registers that can hold QI values. 39851 39852 'Rad' 39853 Registers that can be used with displacements ($a0, $a1, $sb). 39854 39855 'Rsi' 39856 Registers that can hold 32 bit values. 39857 39858 'Rhi' 39859 Registers that can hold 16 bit values. 39860 39861 'Rhc' 39862 Registers chat can hold 16 bit values, including all control 39863 registers. 39864 39865 'Rra' 39866 $r0 through R1, plus $a0 and $a1. 39867 39868 'Rfl' 39869 The flags register. 39870 39871 'Rmm' 39872 The memory-based pseudo-registers $mem0 through $mem15. 39873 39874 'Rpi' 39875 Registers that can hold pointers (16 bit registers for r8c, 39876 m16c; 24 bit registers for m32cm, m32c). 39877 39878 'Rpa' 39879 Matches multiple registers in a PARALLEL to form a larger 39880 register. Used to match function return values. 39881 39882 'Is3' 39883 -8 ... 7 39884 39885 'IS1' 39886 -128 ... 127 39887 39888 'IS2' 39889 -32768 ... 32767 39890 39891 'IU2' 39892 0 ... 65535 39893 39894 'In4' 39895 -8 ... -1 or 1 ... 8 39896 39897 'In5' 39898 -16 ... -1 or 1 ... 16 39899 39900 'In6' 39901 -32 ... -1 or 1 ... 32 39902 39903 'IM2' 39904 -65536 ... -1 39905 39906 'Ilb' 39907 An 8 bit value with exactly one bit set. 39908 39909 'Ilw' 39910 A 16 bit value with exactly one bit set. 39911 39912 'Sd' 39913 The common src/dest memory addressing modes. 39914 39915 'Sa' 39916 Memory addressed using $a0 or $a1. 39917 39918 'Si' 39919 Memory addressed with immediate addresses. 39920 39921 'Ss' 39922 Memory addressed using the stack pointer ($sp). 39923 39924 'Sf' 39925 Memory addressed using the frame base register ($fb). 39926 39927 'Ss' 39928 Memory addressed using the small base register ($sb). 39929 39930 'S1' 39931 $r1h 39932 39933_MicroBlaze--'config/microblaze/constraints.md'_ 39934 'd' 39935 A general register ('r0' to 'r31'). 39936 39937 'z' 39938 A status register ('rmsr', '$fcc1' to '$fcc7'). 39939 39940_MIPS--'config/mips/constraints.md'_ 39941 'd' 39942 A general-purpose register. This is equivalent to 'r' unless 39943 generating MIPS16 code, in which case the MIPS16 register set 39944 is used. 39945 39946 'f' 39947 A floating-point register (if available). 39948 39949 'h' 39950 Formerly the 'hi' register. This constraint is no longer 39951 supported. 39952 39953 'l' 39954 The 'lo' register. Use this register to store values that are 39955 no bigger than a word. 39956 39957 'x' 39958 The concatenated 'hi' and 'lo' registers. Use this register 39959 to store doubleword values. 39960 39961 'c' 39962 A register suitable for use in an indirect jump. This will 39963 always be '$25' for '-mabicalls'. 39964 39965 'v' 39966 Register '$3'. Do not use this constraint in new code; it is 39967 retained only for compatibility with glibc. 39968 39969 'y' 39970 Equivalent to 'r'; retained for backwards compatibility. 39971 39972 'z' 39973 A floating-point condition code register. 39974 39975 'I' 39976 A signed 16-bit constant (for arithmetic instructions). 39977 39978 'J' 39979 Integer zero. 39980 39981 'K' 39982 An unsigned 16-bit constant (for logic instructions). 39983 39984 'L' 39985 A signed 32-bit constant in which the lower 16 bits are zero. 39986 Such constants can be loaded using 'lui'. 39987 39988 'M' 39989 A constant that cannot be loaded using 'lui', 'addiu' or 39990 'ori'. 39991 39992 'N' 39993 A constant in the range -65535 to -1 (inclusive). 39994 39995 'O' 39996 A signed 15-bit constant. 39997 39998 'P' 39999 A constant in the range 1 to 65535 (inclusive). 40000 40001 'G' 40002 Floating-point zero. 40003 40004 'R' 40005 An address that can be used in a non-macro load or store. 40006 40007 'ZC' 40008 A memory operand whose address is formed by a base register 40009 and offset that is suitable for use in instructions with the 40010 same addressing mode as 'll' and 'sc'. 40011 40012 'ZD' 40013 An address suitable for a 'prefetch' instruction, or for any 40014 other instruction with the same addressing mode as 'prefetch'. 40015 40016_Motorola 680x0--'config/m68k/constraints.md'_ 40017 'a' 40018 Address register 40019 40020 'd' 40021 Data register 40022 40023 'f' 40024 68881 floating-point register, if available 40025 40026 'I' 40027 Integer in the range 1 to 8 40028 40029 'J' 40030 16-bit signed number 40031 40032 'K' 40033 Signed number whose magnitude is greater than 0x80 40034 40035 'L' 40036 Integer in the range -8 to -1 40037 40038 'M' 40039 Signed number whose magnitude is greater than 0x100 40040 40041 'N' 40042 Range 24 to 31, rotatert:SI 8 to 1 expressed as rotate 40043 40044 'O' 40045 16 (for rotate using swap) 40046 40047 'P' 40048 Range 8 to 15, rotatert:HI 8 to 1 expressed as rotate 40049 40050 'R' 40051 Numbers that mov3q can handle 40052 40053 'G' 40054 Floating point constant that is not a 68881 constant 40055 40056 'S' 40057 Operands that satisfy 'm' when -mpcrel is in effect 40058 40059 'T' 40060 Operands that satisfy 's' when -mpcrel is not in effect 40061 40062 'Q' 40063 Address register indirect addressing mode 40064 40065 'U' 40066 Register offset addressing 40067 40068 'W' 40069 const_call_operand 40070 40071 'Cs' 40072 symbol_ref or const 40073 40074 'Ci' 40075 const_int 40076 40077 'C0' 40078 const_int 0 40079 40080 'Cj' 40081 Range of signed numbers that don't fit in 16 bits 40082 40083 'Cmvq' 40084 Integers valid for mvq 40085 40086 'Capsw' 40087 Integers valid for a moveq followed by a swap 40088 40089 'Cmvz' 40090 Integers valid for mvz 40091 40092 'Cmvs' 40093 Integers valid for mvs 40094 40095 'Ap' 40096 push_operand 40097 40098 'Ac' 40099 Non-register operands allowed in clr 40100 40101_Moxie--'config/moxie/constraints.md'_ 40102 'A' 40103 An absolute address 40104 40105 'B' 40106 An offset address 40107 40108 'W' 40109 A register indirect memory operand 40110 40111 'I' 40112 A constant in the range of 0 to 255. 40113 40114 'N' 40115 A constant in the range of 0 to -255. 40116 40117_MSP430-'config/msp430/constraints.md'_ 40118 40119 'R12' 40120 Register R12. 40121 40122 'R13' 40123 Register R13. 40124 40125 'K' 40126 Integer constant 1. 40127 40128 'L' 40129 Integer constant -1^20..1^19. 40130 40131 'M' 40132 Integer constant 1-4. 40133 40134 'Ya' 40135 Memory references which do not require an extended MOVX 40136 instruction. 40137 40138 'Yl' 40139 Memory reference, labels only. 40140 40141 'Ys' 40142 Memory reference, stack only. 40143 40144_NDS32--'config/nds32/constraints.md'_ 40145 'w' 40146 LOW register class $r0 to $r7 constraint for V3/V3M ISA. 40147 'l' 40148 LOW register class $r0 to $r7. 40149 'd' 40150 MIDDLE register class $r0 to $r11, $r16 to $r19. 40151 'h' 40152 HIGH register class $r12 to $r14, $r20 to $r31. 40153 't' 40154 Temporary assist register $ta (i.e. $r15). 40155 'k' 40156 Stack register $sp. 40157 'Iu03' 40158 Unsigned immediate 3-bit value. 40159 'In03' 40160 Negative immediate 3-bit value in the range of -7-0. 40161 'Iu04' 40162 Unsigned immediate 4-bit value. 40163 'Is05' 40164 Signed immediate 5-bit value. 40165 'Iu05' 40166 Unsigned immediate 5-bit value. 40167 'In05' 40168 Negative immediate 5-bit value in the range of -31-0. 40169 'Ip05' 40170 Unsigned immediate 5-bit value for movpi45 instruction with 40171 range 16-47. 40172 'Iu06' 40173 Unsigned immediate 6-bit value constraint for addri36.sp 40174 instruction. 40175 'Iu08' 40176 Unsigned immediate 8-bit value. 40177 'Iu09' 40178 Unsigned immediate 9-bit value. 40179 'Is10' 40180 Signed immediate 10-bit value. 40181 'Is11' 40182 Signed immediate 11-bit value. 40183 'Is15' 40184 Signed immediate 15-bit value. 40185 'Iu15' 40186 Unsigned immediate 15-bit value. 40187 'Ic15' 40188 A constant which is not in the range of imm15u but ok for bclr 40189 instruction. 40190 'Ie15' 40191 A constant which is not in the range of imm15u but ok for bset 40192 instruction. 40193 'It15' 40194 A constant which is not in the range of imm15u but ok for btgl 40195 instruction. 40196 'Ii15' 40197 A constant whose compliment value is in the range of imm15u 40198 and ok for bitci instruction. 40199 'Is16' 40200 Signed immediate 16-bit value. 40201 'Is17' 40202 Signed immediate 17-bit value. 40203 'Is19' 40204 Signed immediate 19-bit value. 40205 'Is20' 40206 Signed immediate 20-bit value. 40207 'Ihig' 40208 The immediate value that can be simply set high 20-bit. 40209 'Izeb' 40210 The immediate value 0xff. 40211 'Izeh' 40212 The immediate value 0xffff. 40213 'Ixls' 40214 The immediate value 0x01. 40215 'Ix11' 40216 The immediate value 0x7ff. 40217 'Ibms' 40218 The immediate value with power of 2. 40219 'Ifex' 40220 The immediate value with power of 2 minus 1. 40221 'U33' 40222 Memory constraint for 333 format. 40223 'U45' 40224 Memory constraint for 45 format. 40225 'U37' 40226 Memory constraint for 37 format. 40227 40228_Nios II family--'config/nios2/constraints.md'_ 40229 40230 'I' 40231 Integer that is valid as an immediate operand in an 40232 instruction taking a signed 16-bit number. Range -32768 to 40233 32767. 40234 40235 'J' 40236 Integer that is valid as an immediate operand in an 40237 instruction taking an unsigned 16-bit number. Range 0 to 40238 65535. 40239 40240 'K' 40241 Integer that is valid as an immediate operand in an 40242 instruction taking only the upper 16-bits of a 32-bit number. 40243 Range 32-bit numbers with the lower 16-bits being 0. 40244 40245 'L' 40246 Integer that is valid as an immediate operand for a shift 40247 instruction. Range 0 to 31. 40248 40249 'M' 40250 Integer that is valid as an immediate operand for only the 40251 value 0. Can be used in conjunction with the format modifier 40252 'z' to use 'r0' instead of '0' in the assembly output. 40253 40254 'N' 40255 Integer that is valid as an immediate operand for a custom 40256 instruction opcode. Range 0 to 255. 40257 40258 'P' 40259 An immediate operand for R2 andchi/andci instructions. 40260 40261 'S' 40262 Matches immediates which are addresses in the small data 40263 section and therefore can be added to 'gp' as a 16-bit 40264 immediate to re-create their 32-bit value. 40265 40266 'U' 40267 Matches constants suitable as an operand for the rdprs and 40268 cache instructions. 40269 40270 'v' 40271 A memory operand suitable for Nios II R2 load/store exclusive 40272 instructions. 40273 40274 'w' 40275 A memory operand suitable for load/store IO and cache 40276 instructions. 40277 40278_OpenRISC--'config/or1k/constraints.md'_ 40279 'I' 40280 Integer that is valid as an immediate operand in an 40281 instruction taking a signed 16-bit number. Range -32768 to 40282 32767. 40283 40284 'K' 40285 Integer that is valid as an immediate operand in an 40286 instruction taking an unsigned 16-bit number. Range 0 to 40287 65535. 40288 40289 'M' 40290 Signed 16-bit constant shifted left 16 bits. (Used with 40291 'l.movhi') 40292 40293 'O' 40294 Zero 40295 40296_PDP-11--'config/pdp11/constraints.md'_ 40297 'a' 40298 Floating point registers AC0 through AC3. These can be loaded 40299 from/to memory with a single instruction. 40300 40301 'd' 40302 Odd numbered general registers (R1, R3, R5). These are used 40303 for 16-bit multiply operations. 40304 40305 'D' 40306 A memory reference that is encoded within the opcode, but not 40307 auto-increment or auto-decrement. 40308 40309 'f' 40310 Any of the floating point registers (AC0 through AC5). 40311 40312 'G' 40313 Floating point constant 0. 40314 40315 'h' 40316 Floating point registers AC4 and AC5. These cannot be loaded 40317 from/to memory with a single instruction. 40318 40319 'I' 40320 An integer constant that fits in 16 bits. 40321 40322 'J' 40323 An integer constant whose low order 16 bits are zero. 40324 40325 'K' 40326 An integer constant that does not meet the constraints for 40327 codes 'I' or 'J'. 40328 40329 'L' 40330 The integer constant 1. 40331 40332 'M' 40333 The integer constant -1. 40334 40335 'N' 40336 The integer constant 0. 40337 40338 'O' 40339 Integer constants 0 through 3; shifts by these amounts are 40340 handled as multiple single-bit shifts rather than a single 40341 variable-length shift. 40342 40343 'Q' 40344 A memory reference which requires an additional word (address 40345 or offset) after the opcode. 40346 40347 'R' 40348 A memory reference that is encoded within the opcode. 40349 40350_PowerPC and IBM RS6000--'config/rs6000/constraints.md'_ 40351 'r' 40352 A general purpose register (GPR), 'r0'...'r31'. 40353 40354 'b' 40355 A base register. Like 'r', but 'r0' is not allowed, so 40356 'r1'...'r31'. 40357 40358 'f' 40359 A floating point register (FPR), 'f0'...'f31'. 40360 40361 'd' 40362 A floating point register. This is the same as 'f' nowadays; 40363 historically 'f' was for single-precision and 'd' was for 40364 double-precision floating point. 40365 40366 'v' 40367 An Altivec vector register (VR), 'v0'...'v31'. 40368 40369 'wa' 40370 A VSX register (VSR), 'vs0'...'vs63'. This is either an FPR 40371 ('vs0'...'vs31' are 'f0'...'f31') or a VR ('vs32'...'vs63' are 40372 'v0'...'v31'). 40373 40374 When using 'wa', you should use the '%x' output modifier, so 40375 that the correct register number is printed. For example: 40376 40377 asm ("xvadddp %x0,%x1,%x2" 40378 : "=wa" (v1) 40379 : "wa" (v2), "wa" (v3)); 40380 40381 You should not use '%x' for 'v' operands: 40382 40383 asm ("xsaddqp %0,%1,%2" 40384 : "=v" (v1) 40385 : "v" (v2), "v" (v3)); 40386 40387 'c' 40388 The count register, 'ctr'. 40389 40390 'l' 40391 The link register, 'lr'. 40392 40393 'x' 40394 Condition register field 0, 'cr0'. 40395 40396 'y' 40397 Any condition register field, 'cr0'...'cr7'. 40398 40399 'I' 40400 A signed 16-bit constant. 40401 40402 'J' 40403 An unsigned 16-bit constant shifted left 16 bits (use 'L' 40404 instead for 'SImode' constants). 40405 40406 'K' 40407 An unsigned 16-bit constant. 40408 40409 'L' 40410 A signed 16-bit constant shifted left 16 bits. 40411 40412 'eI' 40413 A signed 34-bit integer constant if prefixed instructions are 40414 supported. 40415 40416 'm' 40417 A memory operand. Normally, 'm' does not allow addresses that 40418 update the base register. If the '<' or '>' constraint is 40419 also used, they are allowed and therefore on PowerPC targets 40420 in that case it is only safe to use 'm<>' in an 'asm' 40421 statement if that 'asm' statement accesses the operand exactly 40422 once. The 'asm' statement must also use '%U<OPNO>' as a 40423 placeholder for the "update" flag in the corresponding load or 40424 store instruction. For example: 40425 40426 asm ("st%U0 %1,%0" : "=m<>" (mem) : "r" (val)); 40427 40428 is correct but: 40429 40430 asm ("st %1,%0" : "=m<>" (mem) : "r" (val)); 40431 40432 is not. 40433 40434 'Q' 40435 A memory operand addressed by just a base register. 40436 40437 'Z' 40438 A memory operand accessed with indexed or indirect addressing. 40439 40440 'a' 40441 An indexed or indirect address. 40442 40443_PRU--'config/pru/constraints.md'_ 40444 'I' 40445 An unsigned 8-bit integer constant. 40446 40447 'J' 40448 An unsigned 16-bit integer constant. 40449 40450 'L' 40451 An unsigned 5-bit integer constant (for shift counts). 40452 40453 'T' 40454 A text segment (program memory) constant label. 40455 40456 'Z' 40457 Integer constant zero. 40458 40459_RL78--'config/rl78/constraints.md'_ 40460 40461 'Int3' 40462 An integer constant in the range 1 ... 7. 40463 'Int8' 40464 An integer constant in the range 0 ... 255. 40465 'J' 40466 An integer constant in the range -255 ... 0 40467 'K' 40468 The integer constant 1. 40469 'L' 40470 The integer constant -1. 40471 'M' 40472 The integer constant 0. 40473 'N' 40474 The integer constant 2. 40475 'O' 40476 The integer constant -2. 40477 'P' 40478 An integer constant in the range 1 ... 15. 40479 'Qbi' 40480 The built-in compare types-eq, ne, gtu, ltu, geu, and leu. 40481 'Qsc' 40482 The synthetic compare types-gt, lt, ge, and le. 40483 'Wab' 40484 A memory reference with an absolute address. 40485 'Wbc' 40486 A memory reference using 'BC' as a base register, with an 40487 optional offset. 40488 'Wca' 40489 A memory reference using 'AX', 'BC', 'DE', or 'HL' for the 40490 address, for calls. 40491 'Wcv' 40492 A memory reference using any 16-bit register pair for the 40493 address, for calls. 40494 'Wd2' 40495 A memory reference using 'DE' as a base register, with an 40496 optional offset. 40497 'Wde' 40498 A memory reference using 'DE' as a base register, without any 40499 offset. 40500 'Wfr' 40501 Any memory reference to an address in the far address space. 40502 'Wh1' 40503 A memory reference using 'HL' as a base register, with an 40504 optional one-byte offset. 40505 'Whb' 40506 A memory reference using 'HL' as a base register, with 'B' or 40507 'C' as the index register. 40508 'Whl' 40509 A memory reference using 'HL' as a base register, without any 40510 offset. 40511 'Ws1' 40512 A memory reference using 'SP' as a base register, with an 40513 optional one-byte offset. 40514 'Y' 40515 Any memory reference to an address in the near address space. 40516 'A' 40517 The 'AX' register. 40518 'B' 40519 The 'BC' register. 40520 'D' 40521 The 'DE' register. 40522 'R' 40523 'A' through 'L' registers. 40524 'S' 40525 The 'SP' register. 40526 'T' 40527 The 'HL' register. 40528 'Z08W' 40529 The 16-bit 'R8' register. 40530 'Z10W' 40531 The 16-bit 'R10' register. 40532 'Zint' 40533 The registers reserved for interrupts ('R24' to 'R31'). 40534 'a' 40535 The 'A' register. 40536 'b' 40537 The 'B' register. 40538 'c' 40539 The 'C' register. 40540 'd' 40541 The 'D' register. 40542 'e' 40543 The 'E' register. 40544 'h' 40545 The 'H' register. 40546 'l' 40547 The 'L' register. 40548 'v' 40549 The virtual registers. 40550 'w' 40551 The 'PSW' register. 40552 'x' 40553 The 'X' register. 40554 40555_RISC-V--'config/riscv/constraints.md'_ 40556 40557 'f' 40558 A floating-point register (if available). 40559 40560 'I' 40561 An I-type 12-bit signed immediate. 40562 40563 'J' 40564 Integer zero. 40565 40566 'K' 40567 A 5-bit unsigned immediate for CSR access instructions. 40568 40569 'A' 40570 An address that is held in a general-purpose register. 40571 40572_RX--'config/rx/constraints.md'_ 40573 'Q' 40574 An address which does not involve register indirect addressing 40575 or pre/post increment/decrement addressing. 40576 40577 'Symbol' 40578 A symbol reference. 40579 40580 'Int08' 40581 A constant in the range -256 to 255, inclusive. 40582 40583 'Sint08' 40584 A constant in the range -128 to 127, inclusive. 40585 40586 'Sint16' 40587 A constant in the range -32768 to 32767, inclusive. 40588 40589 'Sint24' 40590 A constant in the range -8388608 to 8388607, inclusive. 40591 40592 'Uint04' 40593 A constant in the range 0 to 15, inclusive. 40594 40595_S/390 and zSeries--'config/s390/s390.h'_ 40596 'a' 40597 Address register (general purpose register except r0) 40598 40599 'c' 40600 Condition code register 40601 40602 'd' 40603 Data register (arbitrary general purpose register) 40604 40605 'f' 40606 Floating-point register 40607 40608 'I' 40609 Unsigned 8-bit constant (0-255) 40610 40611 'J' 40612 Unsigned 12-bit constant (0-4095) 40613 40614 'K' 40615 Signed 16-bit constant (-32768-32767) 40616 40617 'L' 40618 Value appropriate as displacement. 40619 '(0..4095)' 40620 for short displacement 40621 '(-524288..524287)' 40622 for long displacement 40623 40624 'M' 40625 Constant integer with a value of 0x7fffffff. 40626 40627 'N' 40628 Multiple letter constraint followed by 4 parameter letters. 40629 '0..9:' 40630 number of the part counting from most to least 40631 significant 40632 'H,Q:' 40633 mode of the part 40634 'D,S,H:' 40635 mode of the containing operand 40636 '0,F:' 40637 value of the other parts (F--all bits set) 40638 The constraint matches if the specified part of a constant has 40639 a value different from its other parts. 40640 40641 'Q' 40642 Memory reference without index register and with short 40643 displacement. 40644 40645 'R' 40646 Memory reference with index register and short displacement. 40647 40648 'S' 40649 Memory reference without index register but with long 40650 displacement. 40651 40652 'T' 40653 Memory reference with index register and long displacement. 40654 40655 'U' 40656 Pointer with short displacement. 40657 40658 'W' 40659 Pointer with long displacement. 40660 40661 'Y' 40662 Shift count operand. 40663 40664_SPARC--'config/sparc/sparc.h'_ 40665 'f' 40666 Floating-point register on the SPARC-V8 architecture and lower 40667 floating-point register on the SPARC-V9 architecture. 40668 40669 'e' 40670 Floating-point register. It is equivalent to 'f' on the 40671 SPARC-V8 architecture and contains both lower and upper 40672 floating-point registers on the SPARC-V9 architecture. 40673 40674 'c' 40675 Floating-point condition code register. 40676 40677 'd' 40678 Lower floating-point register. It is only valid on the 40679 SPARC-V9 architecture when the Visual Instruction Set is 40680 available. 40681 40682 'b' 40683 Floating-point register. It is only valid on the SPARC-V9 40684 architecture when the Visual Instruction Set is available. 40685 40686 'h' 40687 64-bit global or out register for the SPARC-V8+ architecture. 40688 40689 'C' 40690 The constant all-ones, for floating-point. 40691 40692 'A' 40693 Signed 5-bit constant 40694 40695 'D' 40696 A vector constant 40697 40698 'I' 40699 Signed 13-bit constant 40700 40701 'J' 40702 Zero 40703 40704 'K' 40705 32-bit constant with the low 12 bits clear (a constant that 40706 can be loaded with the 'sethi' instruction) 40707 40708 'L' 40709 A constant in the range supported by 'movcc' instructions 40710 (11-bit signed immediate) 40711 40712 'M' 40713 A constant in the range supported by 'movrcc' instructions 40714 (10-bit signed immediate) 40715 40716 'N' 40717 Same as 'K', except that it verifies that bits that are not in 40718 the lower 32-bit range are all zero. Must be used instead of 40719 'K' for modes wider than 'SImode' 40720 40721 'O' 40722 The constant 4096 40723 40724 'G' 40725 Floating-point zero 40726 40727 'H' 40728 Signed 13-bit constant, sign-extended to 32 or 64 bits 40729 40730 'P' 40731 The constant -1 40732 40733 'Q' 40734 Floating-point constant whose integral representation can be 40735 moved into an integer register using a single sethi 40736 instruction 40737 40738 'R' 40739 Floating-point constant whose integral representation can be 40740 moved into an integer register using a single mov instruction 40741 40742 'S' 40743 Floating-point constant whose integral representation can be 40744 moved into an integer register using a high/lo_sum instruction 40745 sequence 40746 40747 'T' 40748 Memory address aligned to an 8-byte boundary 40749 40750 'U' 40751 Even register 40752 40753 'W' 40754 Memory address for 'e' constraint registers 40755 40756 'w' 40757 Memory address with only a base register 40758 40759 'Y' 40760 Vector zero 40761 40762_TI C6X family--'config/c6x/constraints.md'_ 40763 'a' 40764 Register file A (A0-A31). 40765 40766 'b' 40767 Register file B (B0-B31). 40768 40769 'A' 40770 Predicate registers in register file A (A0-A2 on C64X and 40771 higher, A1 and A2 otherwise). 40772 40773 'B' 40774 Predicate registers in register file B (B0-B2). 40775 40776 'C' 40777 A call-used register in register file B (B0-B9, B16-B31). 40778 40779 'Da' 40780 Register file A, excluding predicate registers (A3-A31, plus 40781 A0 if not C64X or higher). 40782 40783 'Db' 40784 Register file B, excluding predicate registers (B3-B31). 40785 40786 'Iu4' 40787 Integer constant in the range 0 ... 15. 40788 40789 'Iu5' 40790 Integer constant in the range 0 ... 31. 40791 40792 'In5' 40793 Integer constant in the range -31 ... 0. 40794 40795 'Is5' 40796 Integer constant in the range -16 ... 15. 40797 40798 'I5x' 40799 Integer constant that can be the operand of an ADDA or a SUBA 40800 insn. 40801 40802 'IuB' 40803 Integer constant in the range 0 ... 65535. 40804 40805 'IsB' 40806 Integer constant in the range -32768 ... 32767. 40807 40808 'IsC' 40809 Integer constant in the range -2^{20} ... 2^{20} - 1. 40810 40811 'Jc' 40812 Integer constant that is a valid mask for the clr instruction. 40813 40814 'Js' 40815 Integer constant that is a valid mask for the set instruction. 40816 40817 'Q' 40818 Memory location with A base register. 40819 40820 'R' 40821 Memory location with B base register. 40822 40823 'Z' 40824 Register B14 (aka DP). 40825 40826_TILE-Gx--'config/tilegx/constraints.md'_ 40827 'R00' 40828 'R01' 40829 'R02' 40830 'R03' 40831 'R04' 40832 'R05' 40833 'R06' 40834 'R07' 40835 'R08' 40836 'R09' 40837 'R10' 40838 Each of these represents a register constraint for an 40839 individual register, from r0 to r10. 40840 40841 'I' 40842 Signed 8-bit integer constant. 40843 40844 'J' 40845 Signed 16-bit integer constant. 40846 40847 'K' 40848 Unsigned 16-bit integer constant. 40849 40850 'L' 40851 Integer constant that fits in one signed byte when incremented 40852 by one (-129 ... 126). 40853 40854 'm' 40855 Memory operand. If used together with '<' or '>', the operand 40856 can have postincrement which requires printing with '%In' and 40857 '%in' on TILE-Gx. For example: 40858 40859 asm ("st_add %I0,%1,%i0" : "=m<>" (*mem) : "r" (val)); 40860 40861 'M' 40862 A bit mask suitable for the BFINS instruction. 40863 40864 'N' 40865 Integer constant that is a byte tiled out eight times. 40866 40867 'O' 40868 The integer zero constant. 40869 40870 'P' 40871 Integer constant that is a sign-extended byte tiled out as 40872 four shorts. 40873 40874 'Q' 40875 Integer constant that fits in one signed byte when incremented 40876 (-129 ... 126), but excluding -1. 40877 40878 'S' 40879 Integer constant that has all 1 bits consecutive and starting 40880 at bit 0. 40881 40882 'T' 40883 A 16-bit fragment of a got, tls, or pc-relative reference. 40884 40885 'U' 40886 Memory operand except postincrement. This is roughly the same 40887 as 'm' when not used together with '<' or '>'. 40888 40889 'W' 40890 An 8-element vector constant with identical elements. 40891 40892 'Y' 40893 A 4-element vector constant with identical elements. 40894 40895 'Z0' 40896 The integer constant 0xffffffff. 40897 40898 'Z1' 40899 The integer constant 0xffffffff00000000. 40900 40901_TILEPro--'config/tilepro/constraints.md'_ 40902 'R00' 40903 'R01' 40904 'R02' 40905 'R03' 40906 'R04' 40907 'R05' 40908 'R06' 40909 'R07' 40910 'R08' 40911 'R09' 40912 'R10' 40913 Each of these represents a register constraint for an 40914 individual register, from r0 to r10. 40915 40916 'I' 40917 Signed 8-bit integer constant. 40918 40919 'J' 40920 Signed 16-bit integer constant. 40921 40922 'K' 40923 Nonzero integer constant with low 16 bits zero. 40924 40925 'L' 40926 Integer constant that fits in one signed byte when incremented 40927 by one (-129 ... 126). 40928 40929 'm' 40930 Memory operand. If used together with '<' or '>', the operand 40931 can have postincrement which requires printing with '%In' and 40932 '%in' on TILEPro. For example: 40933 40934 asm ("swadd %I0,%1,%i0" : "=m<>" (mem) : "r" (val)); 40935 40936 'M' 40937 A bit mask suitable for the MM instruction. 40938 40939 'N' 40940 Integer constant that is a byte tiled out four times. 40941 40942 'O' 40943 The integer zero constant. 40944 40945 'P' 40946 Integer constant that is a sign-extended byte tiled out as two 40947 shorts. 40948 40949 'Q' 40950 Integer constant that fits in one signed byte when incremented 40951 (-129 ... 126), but excluding -1. 40952 40953 'T' 40954 A symbolic operand, or a 16-bit fragment of a got, tls, or 40955 pc-relative reference. 40956 40957 'U' 40958 Memory operand except postincrement. This is roughly the same 40959 as 'm' when not used together with '<' or '>'. 40960 40961 'W' 40962 A 4-element vector constant with identical elements. 40963 40964 'Y' 40965 A 2-element vector constant with identical elements. 40966 40967_Visium--'config/visium/constraints.md'_ 40968 'b' 40969 EAM register 'mdb' 40970 40971 'c' 40972 EAM register 'mdc' 40973 40974 'f' 40975 Floating point register 40976 40977 'l' 40978 General register, but not 'r29', 'r30' and 'r31' 40979 40980 't' 40981 Register 'r1' 40982 40983 'u' 40984 Register 'r2' 40985 40986 'v' 40987 Register 'r3' 40988 40989 'G' 40990 Floating-point constant 0.0 40991 40992 'J' 40993 Integer constant in the range 0 .. 65535 (16-bit immediate) 40994 40995 'K' 40996 Integer constant in the range 1 .. 31 (5-bit immediate) 40997 40998 'L' 40999 Integer constant in the range -65535 .. -1 (16-bit negative 41000 immediate) 41001 41002 'M' 41003 Integer constant -1 41004 41005 'O' 41006 Integer constant 0 41007 41008 'P' 41009 Integer constant 32 41010 41011_x86 family--'config/i386/constraints.md'_ 41012 'R' 41013 Legacy register--the eight integer registers available on all 41014 i386 processors ('a', 'b', 'c', 'd', 'si', 'di', 'bp', 'sp'). 41015 41016 'q' 41017 Any register accessible as 'Rl'. In 32-bit mode, 'a', 'b', 41018 'c', and 'd'; in 64-bit mode, any integer register. 41019 41020 'Q' 41021 Any register accessible as 'Rh': 'a', 'b', 'c', and 'd'. 41022 41023 'a' 41024 The 'a' register. 41025 41026 'b' 41027 The 'b' register. 41028 41029 'c' 41030 The 'c' register. 41031 41032 'd' 41033 The 'd' register. 41034 41035 'S' 41036 The 'si' register. 41037 41038 'D' 41039 The 'di' register. 41040 41041 'A' 41042 The 'a' and 'd' registers. This class is used for 41043 instructions that return double word results in the 'ax:dx' 41044 register pair. Single word values will be allocated either in 41045 'ax' or 'dx'. For example on i386 the following implements 41046 'rdtsc': 41047 41048 unsigned long long rdtsc (void) 41049 { 41050 unsigned long long tick; 41051 __asm__ __volatile__("rdtsc":"=A"(tick)); 41052 return tick; 41053 } 41054 41055 This is not correct on x86-64 as it would allocate tick in 41056 either 'ax' or 'dx'. You have to use the following variant 41057 instead: 41058 41059 unsigned long long rdtsc (void) 41060 { 41061 unsigned int tickl, tickh; 41062 __asm__ __volatile__("rdtsc":"=a"(tickl),"=d"(tickh)); 41063 return ((unsigned long long)tickh << 32)|tickl; 41064 } 41065 41066 'U' 41067 The call-clobbered integer registers. 41068 41069 'f' 41070 Any 80387 floating-point (stack) register. 41071 41072 't' 41073 Top of 80387 floating-point stack ('%st(0)'). 41074 41075 'u' 41076 Second from top of 80387 floating-point stack ('%st(1)'). 41077 41078 'y' 41079 Any MMX register. 41080 41081 'x' 41082 Any SSE register. 41083 41084 'v' 41085 Any EVEX encodable SSE register ('%xmm0-%xmm31'). 41086 41087 'Yz' 41088 First SSE register ('%xmm0'). 41089 41090 'I' 41091 Integer constant in the range 0 ... 31, for 32-bit shifts. 41092 41093 'J' 41094 Integer constant in the range 0 ... 63, for 64-bit shifts. 41095 41096 'K' 41097 Signed 8-bit integer constant. 41098 41099 'L' 41100 '0xFF' or '0xFFFF', for andsi as a zero-extending move. 41101 41102 'M' 41103 0, 1, 2, or 3 (shifts for the 'lea' instruction). 41104 41105 'N' 41106 Unsigned 8-bit integer constant (for 'in' and 'out' 41107 instructions). 41108 41109 'G' 41110 Standard 80387 floating point constant. 41111 41112 'C' 41113 SSE constant zero operand. 41114 41115 'e' 41116 32-bit signed integer constant, or a symbolic reference known 41117 to fit that range (for immediate operands in sign-extending 41118 x86-64 instructions). 41119 41120 'We' 41121 32-bit signed integer constant, or a symbolic reference known 41122 to fit that range (for sign-extending conversion operations 41123 that require non-'VOIDmode' immediate operands). 41124 41125 'Wz' 41126 32-bit unsigned integer constant, or a symbolic reference 41127 known to fit that range (for zero-extending conversion 41128 operations that require non-'VOIDmode' immediate operands). 41129 41130 'Wd' 41131 128-bit integer constant where both the high and low 64-bit 41132 word satisfy the 'e' constraint. 41133 41134 'Z' 41135 32-bit unsigned integer constant, or a symbolic reference 41136 known to fit that range (for immediate operands in 41137 zero-extending x86-64 instructions). 41138 41139 'Tv' 41140 VSIB address operand. 41141 41142 'Ts' 41143 Address operand without segment register. 41144 41145_Xstormy16--'config/stormy16/stormy16.h'_ 41146 'a' 41147 Register r0. 41148 41149 'b' 41150 Register r1. 41151 41152 'c' 41153 Register r2. 41154 41155 'd' 41156 Register r8. 41157 41158 'e' 41159 Registers r0 through r7. 41160 41161 't' 41162 Registers r0 and r1. 41163 41164 'y' 41165 The carry register. 41166 41167 'z' 41168 Registers r8 and r9. 41169 41170 'I' 41171 A constant between 0 and 3 inclusive. 41172 41173 'J' 41174 A constant that has exactly one bit set. 41175 41176 'K' 41177 A constant that has exactly one bit clear. 41178 41179 'L' 41180 A constant between 0 and 255 inclusive. 41181 41182 'M' 41183 A constant between -255 and 0 inclusive. 41184 41185 'N' 41186 A constant between -3 and 0 inclusive. 41187 41188 'O' 41189 A constant between 1 and 4 inclusive. 41190 41191 'P' 41192 A constant between -4 and -1 inclusive. 41193 41194 'Q' 41195 A memory reference that is a stack push. 41196 41197 'R' 41198 A memory reference that is a stack pop. 41199 41200 'S' 41201 A memory reference that refers to a constant address of known 41202 value. 41203 41204 'T' 41205 The register indicated by Rx (not implemented yet). 41206 41207 'U' 41208 A constant that is not between 2 and 15 inclusive. 41209 41210 'Z' 41211 The constant 0. 41212 41213_Xtensa--'config/xtensa/constraints.md'_ 41214 'a' 41215 General-purpose 32-bit register 41216 41217 'b' 41218 One-bit boolean register 41219 41220 'A' 41221 MAC16 40-bit accumulator register 41222 41223 'I' 41224 Signed 12-bit integer constant, for use in MOVI instructions 41225 41226 'J' 41227 Signed 8-bit integer constant, for use in ADDI instructions 41228 41229 'K' 41230 Integer constant valid for BccI instructions 41231 41232 'L' 41233 Unsigned constant valid for BccUI instructions 41234 41235 41236File: gcc.info, Node: Asm Labels, Next: Explicit Register Variables, Prev: Constraints, Up: Using Assembly Language with C 41237 412386.47.4 Controlling Names Used in Assembler Code 41239----------------------------------------------- 41240 41241You can specify the name to be used in the assembler code for a C 41242function or variable by writing the 'asm' (or '__asm__') keyword after 41243the declarator. It is up to you to make sure that the assembler names 41244you choose do not conflict with any other assembler symbols, or 41245reference registers. 41246 41247Assembler names for data: 41248......................... 41249 41250This sample shows how to specify the assembler name for data: 41251 41252 int foo asm ("myfoo") = 2; 41253 41254This specifies that the name to be used for the variable 'foo' in the 41255assembler code should be 'myfoo' rather than the usual '_foo'. 41256 41257 On systems where an underscore is normally prepended to the name of a C 41258variable, this feature allows you to define names for the linker that do 41259not start with an underscore. 41260 41261 GCC does not support using this feature with a non-static local 41262variable since such variables do not have assembler names. If you are 41263trying to put the variable in a particular register, see *note Explicit 41264Register Variables::. 41265 41266Assembler names for functions: 41267.............................. 41268 41269To specify the assembler name for functions, write a declaration for the 41270function before its definition and put 'asm' there, like this: 41271 41272 int func (int x, int y) asm ("MYFUNC"); 41273 41274 int func (int x, int y) 41275 { 41276 /* ... */ 41277 41278This specifies that the name to be used for the function 'func' in the 41279assembler code should be 'MYFUNC'. 41280 41281 41282File: gcc.info, Node: Explicit Register Variables, Next: Size of an asm, Prev: Asm Labels, Up: Using Assembly Language with C 41283 412846.47.5 Variables in Specified Registers 41285--------------------------------------- 41286 41287GNU C allows you to associate specific hardware registers with C 41288variables. In almost all cases, allowing the compiler to assign 41289registers produces the best code. However under certain unusual 41290circumstances, more precise control over the variable storage is 41291required. 41292 41293 Both global and local variables can be associated with a register. The 41294consequences of performing this association are very different between 41295the two, as explained in the sections below. 41296 41297* Menu: 41298 41299* Global Register Variables:: Variables declared at global scope. 41300* Local Register Variables:: Variables declared within a function. 41301 41302 41303File: gcc.info, Node: Global Register Variables, Next: Local Register Variables, Up: Explicit Register Variables 41304 413056.47.5.1 Defining Global Register Variables 41306........................................... 41307 41308You can define a global register variable and associate it with a 41309specified register like this: 41310 41311 register int *foo asm ("r12"); 41312 41313Here 'r12' is the name of the register that should be used. Note that 41314this is the same syntax used for defining local register variables, but 41315for a global variable the declaration appears outside a function. The 41316'register' keyword is required, and cannot be combined with 'static'. 41317The register name must be a valid register name for the target platform. 41318 41319 Do not use type qualifiers such as 'const' and 'volatile', as the 41320outcome may be contrary to expectations. In particular, using the 41321'volatile' qualifier does not fully prevent the compiler from optimizing 41322accesses to the register. 41323 41324 Registers are a scarce resource on most systems and allowing the 41325compiler to manage their usage usually results in the best code. 41326However, under special circumstances it can make sense to reserve some 41327globally. For example this may be useful in programs such as 41328programming language interpreters that have a couple of global variables 41329that are accessed very often. 41330 41331 After defining a global register variable, for the current compilation 41332unit: 41333 41334 * If the register is a call-saved register, call ABI is affected: the 41335 register will not be restored in function epilogue sequences after 41336 the variable has been assigned. Therefore, functions cannot safely 41337 return to callers that assume standard ABI. 41338 * Conversely, if the register is a call-clobbered register, making 41339 calls to functions that use standard ABI may lose contents of the 41340 variable. Such calls may be created by the compiler even if none 41341 are evident in the original program, for example when libgcc 41342 functions are used to make up for unavailable instructions. 41343 * Accesses to the variable may be optimized as usual and the register 41344 remains available for allocation and use in any computations, 41345 provided that observable values of the variable are not affected. 41346 * If the variable is referenced in inline assembly, the type of 41347 access must be provided to the compiler via constraints (*note 41348 Constraints::). Accesses from basic asms are not supported. 41349 41350 Note that these points _only_ apply to code that is compiled with the 41351definition. The behavior of code that is merely linked in (for example 41352code from libraries) is not affected. 41353 41354 If you want to recompile source files that do not actually use your 41355global register variable so they do not use the specified register for 41356any other purpose, you need not actually add the global register 41357declaration to their source code. It suffices to specify the compiler 41358option '-ffixed-REG' (*note Code Gen Options::) to reserve the register. 41359 41360Declaring the variable 41361...................... 41362 41363Global register variables cannot have initial values, because an 41364executable file has no means to supply initial contents for a register. 41365 41366 When selecting a register, choose one that is normally saved and 41367restored by function calls on your machine. This ensures that code 41368which is unaware of this reservation (such as library routines) will 41369restore it before returning. 41370 41371 On machines with register windows, be sure to choose a global register 41372that is not affected magically by the function call mechanism. 41373 41374Using the variable 41375.................. 41376 41377When calling routines that are not aware of the reservation, be cautious 41378if those routines call back into code which uses them. As an example, 41379if you call the system library version of 'qsort', it may clobber your 41380registers during execution, but (if you have selected appropriate 41381registers) it will restore them before returning. However it will _not_ 41382restore them before calling 'qsort''s comparison function. As a result, 41383global values will not reliably be available to the comparison function 41384unless the 'qsort' function itself is rebuilt. 41385 41386 Similarly, it is not safe to access the global register variables from 41387signal handlers or from more than one thread of control. Unless you 41388recompile them specially for the task at hand, the system library 41389routines may temporarily use the register for other things. 41390Furthermore, since the register is not reserved exclusively for the 41391variable, accessing it from handlers of asynchronous signals may observe 41392unrelated temporary values residing in the register. 41393 41394 On most machines, 'longjmp' restores to each global register variable 41395the value it had at the time of the 'setjmp'. On some machines, 41396however, 'longjmp' does not change the value of global register 41397variables. To be portable, the function that called 'setjmp' should 41398make other arrangements to save the values of the global register 41399variables, and to restore them in a 'longjmp'. This way, the same thing 41400happens regardless of what 'longjmp' does. 41401 41402 41403File: gcc.info, Node: Local Register Variables, Prev: Global Register Variables, Up: Explicit Register Variables 41404 414056.47.5.2 Specifying Registers for Local Variables 41406................................................. 41407 41408You can define a local register variable and associate it with a 41409specified register like this: 41410 41411 register int *foo asm ("r12"); 41412 41413Here 'r12' is the name of the register that should be used. Note that 41414this is the same syntax used for defining global register variables, but 41415for a local variable the declaration appears within a function. The 41416'register' keyword is required, and cannot be combined with 'static'. 41417The register name must be a valid register name for the target platform. 41418 41419 Do not use type qualifiers such as 'const' and 'volatile', as the 41420outcome may be contrary to expectations. In particular, when the 41421'const' qualifier is used, the compiler may substitute the variable with 41422its initializer in 'asm' statements, which may cause the corresponding 41423operand to appear in a different register. 41424 41425 As with global register variables, it is recommended that you choose a 41426register that is normally saved and restored by function calls on your 41427machine, so that calls to library routines will not clobber it. 41428 41429 The only supported use for this feature is to specify registers for 41430input and output operands when calling Extended 'asm' (*note Extended 41431Asm::). This may be necessary if the constraints for a particular 41432machine don't provide sufficient control to select the desired register. 41433To force an operand into a register, create a local variable and specify 41434the register name after the variable's declaration. Then use the local 41435variable for the 'asm' operand and specify any constraint letter that 41436matches the register: 41437 41438 register int *p1 asm ("r0") = ...; 41439 register int *p2 asm ("r1") = ...; 41440 register int *result asm ("r0"); 41441 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2)); 41442 41443 _Warning:_ In the above example, be aware that a register (for example 41444'r0') can be call-clobbered by subsequent code, including function calls 41445and library calls for arithmetic operators on other variables (for 41446example the initialization of 'p2'). In this case, use temporary 41447variables for expressions between the register assignments: 41448 41449 int t1 = ...; 41450 register int *p1 asm ("r0") = ...; 41451 register int *p2 asm ("r1") = t1; 41452 register int *result asm ("r0"); 41453 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2)); 41454 41455 Defining a register variable does not reserve the register. Other than 41456when invoking the Extended 'asm', the contents of the specified register 41457are not guaranteed. For this reason, the following uses are explicitly 41458_not_ supported. If they appear to work, it is only happenstance, and 41459may stop working as intended due to (seemingly) unrelated changes in 41460surrounding code, or even minor changes in the optimization of a future 41461version of gcc: 41462 41463 * Passing parameters to or from Basic 'asm' 41464 * Passing parameters to or from Extended 'asm' without using input or 41465 output operands. 41466 * Passing parameters to or from routines written in assembler (or 41467 other languages) using non-standard calling conventions. 41468 41469 Some developers use Local Register Variables in an attempt to improve 41470gcc's allocation of registers, especially in large functions. In this 41471case the register name is essentially a hint to the register allocator. 41472While in some instances this can generate better code, improvements are 41473subject to the whims of the allocator/optimizers. Since there are no 41474guarantees that your improvements won't be lost, this usage of Local 41475Register Variables is discouraged. 41476 41477 On the MIPS platform, there is related use for local register variables 41478with slightly different characteristics (*note Defining coprocessor 41479specifics for MIPS targets: (gccint)MIPS Coprocessors.). 41480 41481 41482File: gcc.info, Node: Size of an asm, Prev: Explicit Register Variables, Up: Using Assembly Language with C 41483 414846.47.6 Size of an 'asm' 41485----------------------- 41486 41487Some targets require that GCC track the size of each instruction used in 41488order to generate correct code. Because the final length of the code 41489produced by an 'asm' statement is only known by the assembler, GCC must 41490make an estimate as to how big it will be. It does this by counting the 41491number of instructions in the pattern of the 'asm' and multiplying that 41492by the length of the longest instruction supported by that processor. 41493(When working out the number of instructions, it assumes that any 41494occurrence of a newline or of whatever statement separator character is 41495supported by the assembler -- typically ';' -- indicates the end of an 41496instruction.) 41497 41498 Normally, GCC's estimate is adequate to ensure that correct code is 41499generated, but it is possible to confuse the compiler if you use pseudo 41500instructions or assembler macros that expand into multiple real 41501instructions, or if you use assembler directives that expand to more 41502space in the object file than is needed for a single instruction. If 41503this happens then the assembler may produce a diagnostic saying that a 41504label is unreachable. 41505 41506 This size is also used for inlining decisions. If you use 'asm inline' 41507instead of just 'asm', then for inlining purposes the size of the asm is 41508taken as the minimum size, ignoring how many instructions GCC thinks it 41509is. 41510 41511 41512File: gcc.info, Node: Alternate Keywords, Next: Incomplete Enums, Prev: Using Assembly Language with C, Up: C Extensions 41513 415146.48 Alternate Keywords 41515======================= 41516 41517'-ansi' and the various '-std' options disable certain keywords. This 41518causes trouble when you want to use GNU C extensions, or a 41519general-purpose header file that should be usable by all programs, 41520including ISO C programs. The keywords 'asm', 'typeof' and 'inline' are 41521not available in programs compiled with '-ansi' or '-std' (although 41522'inline' can be used in a program compiled with '-std=c99' or a later 41523standard). The ISO C99 keyword 'restrict' is only available when 41524'-std=gnu99' (which will eventually be the default) or '-std=c99' (or 41525the equivalent '-std=iso9899:1999'), or an option for a later standard 41526version, is used. 41527 41528 The way to solve these problems is to put '__' at the beginning and end 41529of each problematical keyword. For example, use '__asm__' instead of 41530'asm', and '__inline__' instead of 'inline'. 41531 41532 Other C compilers won't accept these alternative keywords; if you want 41533to compile with another compiler, you can define the alternate keywords 41534as macros to replace them with the customary keywords. It looks like 41535this: 41536 41537 #ifndef __GNUC__ 41538 #define __asm__ asm 41539 #endif 41540 41541 '-pedantic' and other options cause warnings for many GNU C extensions. 41542You can prevent such warnings within one expression by writing 41543'__extension__' before the expression. '__extension__' has no effect 41544aside from this. 41545 41546 41547File: gcc.info, Node: Incomplete Enums, Next: Function Names, Prev: Alternate Keywords, Up: C Extensions 41548 415496.49 Incomplete 'enum' Types 41550============================ 41551 41552You can define an 'enum' tag without specifying its possible values. 41553This results in an incomplete type, much like what you get if you write 41554'struct foo' without describing the elements. A later declaration that 41555does specify the possible values completes the type. 41556 41557 You cannot allocate variables or storage using the type while it is 41558incomplete. However, you can work with pointers to that type. 41559 41560 This extension may not be very useful, but it makes the handling of 41561'enum' more consistent with the way 'struct' and 'union' are handled. 41562 41563 This extension is not supported by GNU C++. 41564 41565 41566File: gcc.info, Node: Function Names, Next: Return Address, Prev: Incomplete Enums, Up: C Extensions 41567 415686.50 Function Names as Strings 41569============================== 41570 41571GCC provides three magic constants that hold the name of the current 41572function as a string. In C++11 and later modes, all three are treated 41573as constant expressions and can be used in 'constexpr' constexts. The 41574first of these constants is '__func__', which is part of the C99 41575standard: 41576 41577 The identifier '__func__' is implicitly declared by the translator as 41578if, immediately following the opening brace of each function definition, 41579the declaration 41580 41581 static const char __func__[] = "function-name"; 41582 41583appeared, where function-name is the name of the lexically-enclosing 41584function. This name is the unadorned name of the function. As an 41585extension, at file (or, in C++, namespace scope), '__func__' evaluates 41586to the empty string. 41587 41588 '__FUNCTION__' is another name for '__func__', provided for backward 41589compatibility with old versions of GCC. 41590 41591 In C, '__PRETTY_FUNCTION__' is yet another name for '__func__', except 41592that at file scope (or, in C++, namespace scope), it evaluates to the 41593string '"top level"'. In addition, in C++, '__PRETTY_FUNCTION__' 41594contains the signature of the function as well as its bare name. For 41595example, this program: 41596 41597 extern "C" int printf (const char *, ...); 41598 41599 class a { 41600 public: 41601 void sub (int i) 41602 { 41603 printf ("__FUNCTION__ = %s\n", __FUNCTION__); 41604 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__); 41605 } 41606 }; 41607 41608 int 41609 main (void) 41610 { 41611 a ax; 41612 ax.sub (0); 41613 return 0; 41614 } 41615 41616gives this output: 41617 41618 __FUNCTION__ = sub 41619 __PRETTY_FUNCTION__ = void a::sub(int) 41620 41621 These identifiers are variables, not preprocessor macros, and may not 41622be used to initialize 'char' arrays or be concatenated with string 41623literals. 41624 41625 41626File: gcc.info, Node: Return Address, Next: Vector Extensions, Prev: Function Names, Up: C Extensions 41627 416286.51 Getting the Return or Frame Address of a Function 41629====================================================== 41630 41631These functions may be used to get information about the callers of a 41632function. 41633 41634 -- Built-in Function: void * __builtin_return_address (unsigned int 41635 LEVEL) 41636 This function returns the return address of the current function, 41637 or of one of its callers. The LEVEL argument is number of frames 41638 to scan up the call stack. A value of '0' yields the return 41639 address of the current function, a value of '1' yields the return 41640 address of the caller of the current function, and so forth. When 41641 inlining the expected behavior is that the function returns the 41642 address of the function that is returned to. To work around this 41643 behavior use the 'noinline' function attribute. 41644 41645 The LEVEL argument must be a constant integer. 41646 41647 On some machines it may be impossible to determine the return 41648 address of any function other than the current one; in such cases, 41649 or when the top of the stack has been reached, this function 41650 returns an unspecified value. In addition, 41651 '__builtin_frame_address' may be used to determine if the top of 41652 the stack has been reached. 41653 41654 Additional post-processing of the returned value may be needed, see 41655 '__builtin_extract_return_addr'. 41656 41657 The stored representation of the return address in memory may be 41658 different from the address returned by '__builtin_return_address'. 41659 For example, on AArch64 the stored address may be mangled with 41660 return address signing whereas the address returned by 41661 '__builtin_return_address' is not. 41662 41663 Calling this function with a nonzero argument can have 41664 unpredictable effects, including crashing the calling program. As 41665 a result, calls that are considered unsafe are diagnosed when the 41666 '-Wframe-address' option is in effect. Such calls should only be 41667 made in debugging situations. 41668 41669 On targets where code addresses are representable as 'void *', 41670 void *addr = __builtin_extract_return_addr (__builtin_return_address (0)); 41671 gives the code address where the current function would return. 41672 For example, such an address may be used with 'dladdr' or other 41673 interfaces that work with code addresses. 41674 41675 -- Built-in Function: void * __builtin_extract_return_addr (void *ADDR) 41676 The address as returned by '__builtin_return_address' may have to 41677 be fed through this function to get the actual encoded address. 41678 For example, on the 31-bit S/390 platform the highest bit has to be 41679 masked out, or on SPARC platforms an offset has to be added for the 41680 true next instruction to be executed. 41681 41682 If no fixup is needed, this function simply passes through ADDR. 41683 41684 -- Built-in Function: void * __builtin_frob_return_addr (void *ADDR) 41685 This function does the reverse of '__builtin_extract_return_addr'. 41686 41687 -- Built-in Function: void * __builtin_frame_address (unsigned int 41688 LEVEL) 41689 This function is similar to '__builtin_return_address', but it 41690 returns the address of the function frame rather than the return 41691 address of the function. Calling '__builtin_frame_address' with a 41692 value of '0' yields the frame address of the current function, a 41693 value of '1' yields the frame address of the caller of the current 41694 function, and so forth. 41695 41696 The frame is the area on the stack that holds local variables and 41697 saved registers. The frame address is normally the address of the 41698 first word pushed on to the stack by the function. However, the 41699 exact definition depends upon the processor and the calling 41700 convention. If the processor has a dedicated frame pointer 41701 register, and the function has a frame, then 41702 '__builtin_frame_address' returns the value of the frame pointer 41703 register. 41704 41705 On some machines it may be impossible to determine the frame 41706 address of any function other than the current one; in such cases, 41707 or when the top of the stack has been reached, this function 41708 returns '0' if the first frame pointer is properly initialized by 41709 the startup code. 41710 41711 Calling this function with a nonzero argument can have 41712 unpredictable effects, including crashing the calling program. As 41713 a result, calls that are considered unsafe are diagnosed when the 41714 '-Wframe-address' option is in effect. Such calls should only be 41715 made in debugging situations. 41716 41717 41718File: gcc.info, Node: Vector Extensions, Next: Offsetof, Prev: Return Address, Up: C Extensions 41719 417206.52 Using Vector Instructions through Built-in Functions 41721========================================================= 41722 41723On some targets, the instruction set contains SIMD vector instructions 41724which operate on multiple values contained in one large register at the 41725same time. For example, on the x86 the MMX, 3DNow! and SSE extensions 41726can be used this way. 41727 41728 The first step in using these extensions is to provide the necessary 41729data types. This should be done using an appropriate 'typedef': 41730 41731 typedef int v4si __attribute__ ((vector_size (16))); 41732 41733The 'int' type specifies the "base type", while the attribute specifies 41734the vector size for the variable, measured in bytes. For example, the 41735declaration above causes the compiler to set the mode for the 'v4si' 41736type to be 16 bytes wide and divided into 'int' sized units. For a 4173732-bit 'int' this means a vector of 4 units of 4 bytes, and the 41738corresponding mode of 'foo' is V4SI. 41739 41740 The 'vector_size' attribute is only applicable to integral and floating 41741scalars, although arrays, pointers, and function return values are 41742allowed in conjunction with this construct. Only sizes that are 41743positive power-of-two multiples of the base type size are currently 41744allowed. 41745 41746 All the basic integer types can be used as base types, both as signed 41747and as unsigned: 'char', 'short', 'int', 'long', 'long long'. In 41748addition, 'float' and 'double' can be used to build floating-point 41749vector types. 41750 41751 Specifying a combination that is not valid for the current architecture 41752causes GCC to synthesize the instructions using a narrower mode. For 41753example, if you specify a variable of type 'V4SI' and your architecture 41754does not allow for this specific SIMD type, GCC produces code that uses 417554 'SIs'. 41756 41757 The types defined in this manner can be used with a subset of normal C 41758operations. Currently, GCC allows using the following operators on 41759these types: '+, -, *, /, unary minus, ^, |, &, ~, %'. 41760 41761 The operations behave like C++ 'valarrays'. Addition is defined as the 41762addition of the corresponding elements of the operands. For example, in 41763the code below, each of the 4 elements in A is added to the 41764corresponding 4 elements in B and the resulting vector is stored in C. 41765 41766 typedef int v4si __attribute__ ((vector_size (16))); 41767 41768 v4si a, b, c; 41769 41770 c = a + b; 41771 41772 Subtraction, multiplication, division, and the logical operations 41773operate in a similar manner. Likewise, the result of using the unary 41774minus or complement operators on a vector type is a vector whose 41775elements are the negative or complemented values of the corresponding 41776elements in the operand. 41777 41778 It is possible to use shifting operators '<<', '>>' on integer-type 41779vectors. The operation is defined as following: '{a0, a1, ..., an} >> 41780{b0, b1, ..., bn} == {a0 >> b0, a1 >> b1, ..., an >> bn}'. Vector 41781operands must have the same number of elements. 41782 41783 For convenience, it is allowed to use a binary vector operation where 41784one operand is a scalar. In that case the compiler transforms the 41785scalar operand into a vector where each element is the scalar from the 41786operation. The transformation happens only if the scalar could be 41787safely converted to the vector-element type. Consider the following 41788code. 41789 41790 typedef int v4si __attribute__ ((vector_size (16))); 41791 41792 v4si a, b, c; 41793 long l; 41794 41795 a = b + 1; /* a = b + {1,1,1,1}; */ 41796 a = 2 * b; /* a = {2,2,2,2} * b; */ 41797 41798 a = l + a; /* Error, cannot convert long to int. */ 41799 41800 Vectors can be subscripted as if the vector were an array with the same 41801number of elements and base type. Out of bound accesses invoke 41802undefined behavior at run time. Warnings for out of bound accesses for 41803vector subscription can be enabled with '-Warray-bounds'. 41804 41805 Vector comparison is supported with standard comparison operators: '==, 41806!=, <, <=, >, >='. Comparison operands can be vector expressions of 41807integer-type or real-type. Comparison between integer-type vectors and 41808real-type vectors are not supported. The result of the comparison is a 41809vector of the same width and number of elements as the comparison 41810operands with a signed integral element type. 41811 41812 Vectors are compared element-wise producing 0 when comparison is false 41813and -1 (constant of the appropriate type where all bits are set) 41814otherwise. Consider the following example. 41815 41816 typedef int v4si __attribute__ ((vector_size (16))); 41817 41818 v4si a = {1,2,3,4}; 41819 v4si b = {3,2,1,4}; 41820 v4si c; 41821 41822 c = a > b; /* The result would be {0, 0,-1, 0} */ 41823 c = a == b; /* The result would be {0,-1, 0,-1} */ 41824 41825 In C++, the ternary operator '?:' is available. 'a?b:c', where 'b' and 41826'c' are vectors of the same type and 'a' is an integer vector with the 41827same number of elements of the same size as 'b' and 'c', computes all 41828three arguments and creates a vector '{a[0]?b[0]:c[0], a[1]?b[1]:c[1], 41829...}'. Note that unlike in OpenCL, 'a' is thus interpreted as 'a != 0' 41830and not 'a < 0'. As in the case of binary operations, this syntax is 41831also accepted when one of 'b' or 'c' is a scalar that is then 41832transformed into a vector. If both 'b' and 'c' are scalars and the type 41833of 'true?b:c' has the same size as the element type of 'a', then 'b' and 41834'c' are converted to a vector type whose elements have this type and 41835with the same number of elements as 'a'. 41836 41837 In C++, the logic operators '!, &&, ||' are available for vectors. 41838'!v' is equivalent to 'v == 0', 'a && b' is equivalent to 'a!=0 & b!=0' 41839and 'a || b' is equivalent to 'a!=0 | b!=0'. For mixed operations 41840between a scalar 's' and a vector 'v', 's && v' is equivalent to 41841's?v!=0:0' (the evaluation is short-circuit) and 'v && s' is equivalent 41842to 'v!=0 & (s?-1:0)'. 41843 41844 Vector shuffling is available using functions '__builtin_shuffle (vec, 41845mask)' and '__builtin_shuffle (vec0, vec1, mask)'. Both functions 41846construct a permutation of elements from one or two vectors and return a 41847vector of the same type as the input vector(s). The MASK is an integral 41848vector with the same width (W) and element count (N) as the output 41849vector. 41850 41851 The elements of the input vectors are numbered in memory ordering of 41852VEC0 beginning at 0 and VEC1 beginning at N. The elements of MASK are 41853considered modulo N in the single-operand case and modulo 2*N in the 41854two-operand case. 41855 41856 Consider the following example, 41857 41858 typedef int v4si __attribute__ ((vector_size (16))); 41859 41860 v4si a = {1,2,3,4}; 41861 v4si b = {5,6,7,8}; 41862 v4si mask1 = {0,1,1,3}; 41863 v4si mask2 = {0,4,2,5}; 41864 v4si res; 41865 41866 res = __builtin_shuffle (a, mask1); /* res is {1,2,2,4} */ 41867 res = __builtin_shuffle (a, b, mask2); /* res is {1,5,3,6} */ 41868 41869 Note that '__builtin_shuffle' is intentionally semantically compatible 41870with the OpenCL 'shuffle' and 'shuffle2' functions. 41871 41872 You can declare variables and use them in function calls and returns, 41873as well as in assignments and some casts. You can specify a vector type 41874as a return type for a function. Vector types can also be used as 41875function arguments. It is possible to cast from one vector type to 41876another, provided they are of the same size (in fact, you can also cast 41877vectors to and from other datatypes of the same size). 41878 41879 You cannot operate between vectors of different lengths or different 41880signedness without a cast. 41881 41882 Vector conversion is available using the '__builtin_convertvector (vec, 41883vectype)' function. VEC must be an expression with integral or floating 41884vector type and VECTYPE an integral or floating vector type with the 41885same number of elements. The result has VECTYPE type and value of a C 41886cast of every element of VEC to the element type of VECTYPE. 41887 41888 Consider the following example, 41889 typedef int v4si __attribute__ ((vector_size (16))); 41890 typedef float v4sf __attribute__ ((vector_size (16))); 41891 typedef double v4df __attribute__ ((vector_size (32))); 41892 typedef unsigned long long v4di __attribute__ ((vector_size (32))); 41893 41894 v4si a = {1,-2,3,-4}; 41895 v4sf b = {1.5f,-2.5f,3.f,7.f}; 41896 v4di c = {1ULL,5ULL,0ULL,10ULL}; 41897 v4sf d = __builtin_convertvector (a, v4sf); /* d is {1.f,-2.f,3.f,-4.f} */ 41898 /* Equivalent of: 41899 v4sf d = { (float)a[0], (float)a[1], (float)a[2], (float)a[3] }; */ 41900 v4df e = __builtin_convertvector (a, v4df); /* e is {1.,-2.,3.,-4.} */ 41901 v4df f = __builtin_convertvector (b, v4df); /* f is {1.5,-2.5,3.,7.} */ 41902 v4si g = __builtin_convertvector (f, v4si); /* g is {1,-2,3,7} */ 41903 v4si h = __builtin_convertvector (c, v4si); /* h is {1,5,0,10} */ 41904 41905 Sometimes it is desirable to write code using a mix of generic vector 41906operations (for clarity) and machine-specific vector intrinsics (to 41907access vector instructions that are not exposed via generic built-ins). 41908On x86, intrinsic functions for integer vectors typically use the same 41909vector type '__m128i' irrespective of how they interpret the vector, 41910making it necessary to cast their arguments and return values from/to 41911other vector types. In C, you can make use of a 'union' type: 41912 #include <immintrin.h> 41913 41914 typedef unsigned char u8x16 __attribute__ ((vector_size (16))); 41915 typedef unsigned int u32x4 __attribute__ ((vector_size (16))); 41916 41917 typedef union { 41918 __m128i mm; 41919 u8x16 u8; 41920 u32x4 u32; 41921 } v128; 41922 41923for variables that can be used with both built-in operators and x86 41924intrinsics: 41925 41926 v128 x, y = { 0 }; 41927 memcpy (&x, ptr, sizeof x); 41928 y.u8 += 0x80; 41929 x.mm = _mm_adds_epu8 (x.mm, y.mm); 41930 x.u32 &= 0xffffff; 41931 41932 /* Instead of a variable, a compound literal may be used to pass the 41933 return value of an intrinsic call to a function expecting the union: */ 41934 v128 foo (v128); 41935 x = foo ((v128) {_mm_adds_epu8 (x.mm, y.mm)}); 41936 41937 41938File: gcc.info, Node: Offsetof, Next: __sync Builtins, Prev: Vector Extensions, Up: C Extensions 41939 419406.53 Support for 'offsetof' 41941=========================== 41942 41943GCC implements for both C and C++ a syntactic extension to implement the 41944'offsetof' macro. 41945 41946 primary: 41947 "__builtin_offsetof" "(" typename "," offsetof_member_designator ")" 41948 41949 offsetof_member_designator: 41950 identifier 41951 | offsetof_member_designator "." identifier 41952 | offsetof_member_designator "[" expr "]" 41953 41954 This extension is sufficient such that 41955 41956 #define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER) 41957 41958is a suitable definition of the 'offsetof' macro. In C++, TYPE may be 41959dependent. In either case, MEMBER may consist of a single identifier, 41960or a sequence of member accesses and array references. 41961 41962 41963File: gcc.info, Node: __sync Builtins, Next: __atomic Builtins, Prev: Offsetof, Up: C Extensions 41964 419656.54 Legacy '__sync' Built-in Functions for Atomic Memory Access 41966================================================================ 41967 41968The following built-in functions are intended to be compatible with 41969those described in the 'Intel Itanium Processor-specific Application 41970Binary Interface', section 7.4. As such, they depart from normal GCC 41971practice by not using the '__builtin_' prefix and also by being 41972overloaded so that they work on multiple types. 41973 41974 The definition given in the Intel documentation allows only for the use 41975of the types 'int', 'long', 'long long' or their unsigned counterparts. 41976GCC allows any scalar type that is 1, 2, 4 or 8 bytes in size other than 41977the C type '_Bool' or the C++ type 'bool'. Operations on pointer 41978arguments are performed as if the operands were of the 'uintptr_t' type. 41979That is, they are not scaled by the size of the type to which the 41980pointer points. 41981 41982 These functions are implemented in terms of the '__atomic' builtins 41983(*note __atomic Builtins::). They should not be used for new code which 41984should use the '__atomic' builtins instead. 41985 41986 Not all operations are supported by all target processors. If a 41987particular operation cannot be implemented on the target processor, a 41988warning is generated and a call to an external function is generated. 41989The external function carries the same name as the built-in version, 41990with an additional suffix '_N' where N is the size of the data type. 41991 41992 In most cases, these built-in functions are considered a "full 41993barrier". That is, no memory operand is moved across the operation, 41994either forward or backward. Further, instructions are issued as 41995necessary to prevent the processor from speculating loads across the 41996operation and from queuing stores after the operation. 41997 41998 All of the routines are described in the Intel documentation to take 41999"an optional list of variables protected by the memory barrier". It's 42000not clear what is meant by that; it could mean that _only_ the listed 42001variables are protected, or it could mean a list of additional variables 42002to be protected. The list is ignored by GCC which treats it as empty. 42003GCC interprets an empty list as meaning that all globally accessible 42004variables should be protected. 42005 42006'TYPE __sync_fetch_and_add (TYPE *ptr, TYPE value, ...)' 42007'TYPE __sync_fetch_and_sub (TYPE *ptr, TYPE value, ...)' 42008'TYPE __sync_fetch_and_or (TYPE *ptr, TYPE value, ...)' 42009'TYPE __sync_fetch_and_and (TYPE *ptr, TYPE value, ...)' 42010'TYPE __sync_fetch_and_xor (TYPE *ptr, TYPE value, ...)' 42011'TYPE __sync_fetch_and_nand (TYPE *ptr, TYPE value, ...)' 42012 These built-in functions perform the operation suggested by the 42013 name, and returns the value that had previously been in memory. 42014 That is, operations on integer operands have the following 42015 semantics. Operations on pointer arguments are performed as if the 42016 operands were of the 'uintptr_t' type. That is, they are not 42017 scaled by the size of the type to which the pointer points. 42018 42019 { tmp = *ptr; *ptr OP= value; return tmp; } 42020 { tmp = *ptr; *ptr = ~(tmp & value); return tmp; } // nand 42021 42022 The object pointed to by the first argument must be of integer or 42023 pointer type. It must not be a boolean type. 42024 42025 _Note:_ GCC 4.4 and later implement '__sync_fetch_and_nand' as 42026 '*ptr = ~(tmp & value)' instead of '*ptr = ~tmp & value'. 42027 42028'TYPE __sync_add_and_fetch (TYPE *ptr, TYPE value, ...)' 42029'TYPE __sync_sub_and_fetch (TYPE *ptr, TYPE value, ...)' 42030'TYPE __sync_or_and_fetch (TYPE *ptr, TYPE value, ...)' 42031'TYPE __sync_and_and_fetch (TYPE *ptr, TYPE value, ...)' 42032'TYPE __sync_xor_and_fetch (TYPE *ptr, TYPE value, ...)' 42033'TYPE __sync_nand_and_fetch (TYPE *ptr, TYPE value, ...)' 42034 These built-in functions perform the operation suggested by the 42035 name, and return the new value. That is, operations on integer 42036 operands have the following semantics. Operations on pointer 42037 operands are performed as if the operand's type were 'uintptr_t'. 42038 42039 { *ptr OP= value; return *ptr; } 42040 { *ptr = ~(*ptr & value); return *ptr; } // nand 42041 42042 The same constraints on arguments apply as for the corresponding 42043 '__sync_op_and_fetch' built-in functions. 42044 42045 _Note:_ GCC 4.4 and later implement '__sync_nand_and_fetch' as 42046 '*ptr = ~(*ptr & value)' instead of '*ptr = ~*ptr & value'. 42047 42048'bool __sync_bool_compare_and_swap (TYPE *ptr, TYPE oldval, TYPE newval, ...)' 42049'TYPE __sync_val_compare_and_swap (TYPE *ptr, TYPE oldval, TYPE newval, ...)' 42050 These built-in functions perform an atomic compare and swap. That 42051 is, if the current value of '*PTR' is OLDVAL, then write NEWVAL 42052 into '*PTR'. 42053 42054 The "bool" version returns 'true' if the comparison is successful 42055 and NEWVAL is written. The "val" version returns the contents of 42056 '*PTR' before the operation. 42057 42058'__sync_synchronize (...)' 42059 This built-in function issues a full memory barrier. 42060 42061'TYPE __sync_lock_test_and_set (TYPE *ptr, TYPE value, ...)' 42062 This built-in function, as described by Intel, is not a traditional 42063 test-and-set operation, but rather an atomic exchange operation. 42064 It writes VALUE into '*PTR', and returns the previous contents of 42065 '*PTR'. 42066 42067 Many targets have only minimal support for such locks, and do not 42068 support a full exchange operation. In this case, a target may 42069 support reduced functionality here by which the _only_ valid value 42070 to store is the immediate constant 1. The exact value actually 42071 stored in '*PTR' is implementation defined. 42072 42073 This built-in function is not a full barrier, but rather an 42074 "acquire barrier". This means that references after the operation 42075 cannot move to (or be speculated to) before the operation, but 42076 previous memory stores may not be globally visible yet, and 42077 previous memory loads may not yet be satisfied. 42078 42079'void __sync_lock_release (TYPE *ptr, ...)' 42080 This built-in function releases the lock acquired by 42081 '__sync_lock_test_and_set'. Normally this means writing the 42082 constant 0 to '*PTR'. 42083 42084 This built-in function is not a full barrier, but rather a "release 42085 barrier". This means that all previous memory stores are globally 42086 visible, and all previous memory loads have been satisfied, but 42087 following memory reads are not prevented from being speculated to 42088 before the barrier. 42089 42090 42091File: gcc.info, Node: __atomic Builtins, Next: Integer Overflow Builtins, Prev: __sync Builtins, Up: C Extensions 42092 420936.55 Built-in Functions for Memory Model Aware Atomic Operations 42094================================================================ 42095 42096The following built-in functions approximately match the requirements 42097for the C++11 memory model. They are all identified by being prefixed 42098with '__atomic' and most are overloaded so that they work with multiple 42099types. 42100 42101 These functions are intended to replace the legacy '__sync' builtins. 42102The main difference is that the memory order that is requested is a 42103parameter to the functions. New code should always use the '__atomic' 42104builtins rather than the '__sync' builtins. 42105 42106 Note that the '__atomic' builtins assume that programs will conform to 42107the C++11 memory model. In particular, they assume that programs are 42108free of data races. See the C++11 standard for detailed requirements. 42109 42110 The '__atomic' builtins can be used with any integral scalar or pointer 42111type that is 1, 2, 4, or 8 bytes in length. 16-byte integral types are 42112also allowed if '__int128' (*note __int128::) is supported by the 42113architecture. 42114 42115 The four non-arithmetic functions (load, store, exchange, and 42116compare_exchange) all have a generic version as well. This generic 42117version works on any data type. It uses the lock-free built-in function 42118if the specific data type size makes that possible; otherwise, an 42119external call is left to be resolved at run time. This external call is 42120the same format with the addition of a 'size_t' parameter inserted as 42121the first parameter indicating the size of the object being pointed to. 42122All objects must be the same size. 42123 42124 There are 6 different memory orders that can be specified. These map 42125to the C++11 memory orders with the same names, see the C++11 standard 42126or the GCC wiki on atomic synchronization 42127(http://gcc.gnu.org/wiki/Atomic/GCCMM/AtomicSync) for detailed 42128definitions. Individual targets may also support additional memory 42129orders for use on specific architectures. Refer to the target 42130documentation for details of these. 42131 42132 An atomic operation can both constrain code motion and be mapped to 42133hardware instructions for synchronization between threads (e.g., a 42134fence). To which extent this happens is controlled by the memory 42135orders, which are listed here in approximately ascending order of 42136strength. The description of each memory order is only meant to roughly 42137illustrate the effects and is not a specification; see the C++11 memory 42138model for precise semantics. 42139 42140'__ATOMIC_RELAXED' 42141 Implies no inter-thread ordering constraints. 42142'__ATOMIC_CONSUME' 42143 This is currently implemented using the stronger '__ATOMIC_ACQUIRE' 42144 memory order because of a deficiency in C++11's semantics for 42145 'memory_order_consume'. 42146'__ATOMIC_ACQUIRE' 42147 Creates an inter-thread happens-before constraint from the release 42148 (or stronger) semantic store to this acquire load. Can prevent 42149 hoisting of code to before the operation. 42150'__ATOMIC_RELEASE' 42151 Creates an inter-thread happens-before constraint to acquire (or 42152 stronger) semantic loads that read from this release store. Can 42153 prevent sinking of code to after the operation. 42154'__ATOMIC_ACQ_REL' 42155 Combines the effects of both '__ATOMIC_ACQUIRE' and 42156 '__ATOMIC_RELEASE'. 42157'__ATOMIC_SEQ_CST' 42158 Enforces total ordering with all other '__ATOMIC_SEQ_CST' 42159 operations. 42160 42161 Note that in the C++11 memory model, _fences_ (e.g., 42162'__atomic_thread_fence') take effect in combination with other atomic 42163operations on specific memory locations (e.g., atomic loads); operations 42164on specific memory locations do not necessarily affect other operations 42165in the same way. 42166 42167 Target architectures are encouraged to provide their own patterns for 42168each of the atomic built-in functions. If no target is provided, the 42169original non-memory model set of '__sync' atomic built-in functions are 42170used, along with any required synchronization fences surrounding it in 42171order to achieve the proper behavior. Execution in this case is subject 42172to the same restrictions as those built-in functions. 42173 42174 If there is no pattern or mechanism to provide a lock-free instruction 42175sequence, a call is made to an external routine with the same parameters 42176to be resolved at run time. 42177 42178 When implementing patterns for these built-in functions, the memory 42179order parameter can be ignored as long as the pattern implements the 42180most restrictive '__ATOMIC_SEQ_CST' memory order. Any of the other 42181memory orders execute correctly with this memory order but they may not 42182execute as efficiently as they could with a more appropriate 42183implementation of the relaxed requirements. 42184 42185 Note that the C++11 standard allows for the memory order parameter to 42186be determined at run time rather than at compile time. These built-in 42187functions map any run-time value to '__ATOMIC_SEQ_CST' rather than 42188invoke a runtime library call or inline a switch statement. This is 42189standard compliant, safe, and the simplest approach for now. 42190 42191 The memory order parameter is a signed int, but only the lower 16 bits 42192are reserved for the memory order. The remainder of the signed int is 42193reserved for target use and should be 0. Use of the predefined atomic 42194values ensures proper usage. 42195 42196 -- Built-in Function: TYPE __atomic_load_n (TYPE *ptr, int memorder) 42197 This built-in function implements an atomic load operation. It 42198 returns the contents of '*PTR'. 42199 42200 The valid memory order variants are '__ATOMIC_RELAXED', 42201 '__ATOMIC_SEQ_CST', '__ATOMIC_ACQUIRE', and '__ATOMIC_CONSUME'. 42202 42203 -- Built-in Function: void __atomic_load (TYPE *ptr, TYPE *ret, int 42204 memorder) 42205 This is the generic version of an atomic load. It returns the 42206 contents of '*PTR' in '*RET'. 42207 42208 -- Built-in Function: void __atomic_store_n (TYPE *ptr, TYPE val, int 42209 memorder) 42210 This built-in function implements an atomic store operation. It 42211 writes 'VAL' into '*PTR'. 42212 42213 The valid memory order variants are '__ATOMIC_RELAXED', 42214 '__ATOMIC_SEQ_CST', and '__ATOMIC_RELEASE'. 42215 42216 -- Built-in Function: void __atomic_store (TYPE *ptr, TYPE *val, int 42217 memorder) 42218 This is the generic version of an atomic store. It stores the 42219 value of '*VAL' into '*PTR'. 42220 42221 -- Built-in Function: TYPE __atomic_exchange_n (TYPE *ptr, TYPE val, 42222 int memorder) 42223 This built-in function implements an atomic exchange operation. It 42224 writes VAL into '*PTR', and returns the previous contents of 42225 '*PTR'. 42226 42227 The valid memory order variants are '__ATOMIC_RELAXED', 42228 '__ATOMIC_SEQ_CST', '__ATOMIC_ACQUIRE', '__ATOMIC_RELEASE', and 42229 '__ATOMIC_ACQ_REL'. 42230 42231 -- Built-in Function: void __atomic_exchange (TYPE *ptr, TYPE *val, 42232 TYPE *ret, int memorder) 42233 This is the generic version of an atomic exchange. It stores the 42234 contents of '*VAL' into '*PTR'. The original value of '*PTR' is 42235 copied into '*RET'. 42236 42237 -- Built-in Function: bool __atomic_compare_exchange_n (TYPE *ptr, TYPE 42238 *expected, TYPE desired, bool weak, int success_memorder, int 42239 failure_memorder) 42240 This built-in function implements an atomic compare and exchange 42241 operation. This compares the contents of '*PTR' with the contents 42242 of '*EXPECTED'. If equal, the operation is a _read-modify-write_ 42243 operation that writes DESIRED into '*PTR'. If they are not equal, 42244 the operation is a _read_ and the current contents of '*PTR' are 42245 written into '*EXPECTED'. WEAK is 'true' for weak 42246 compare_exchange, which may fail spuriously, and 'false' for the 42247 strong variation, which never fails spuriously. Many targets only 42248 offer the strong variation and ignore the parameter. When in 42249 doubt, use the strong variation. 42250 42251 If DESIRED is written into '*PTR' then 'true' is returned and 42252 memory is affected according to the memory order specified by 42253 SUCCESS_MEMORDER. There are no restrictions on what memory order 42254 can be used here. 42255 42256 Otherwise, 'false' is returned and memory is affected according to 42257 FAILURE_MEMORDER. This memory order cannot be '__ATOMIC_RELEASE' 42258 nor '__ATOMIC_ACQ_REL'. It also cannot be a stronger order than 42259 that specified by SUCCESS_MEMORDER. 42260 42261 -- Built-in Function: bool __atomic_compare_exchange (TYPE *ptr, TYPE 42262 *expected, TYPE *desired, bool weak, int success_memorder, int 42263 failure_memorder) 42264 This built-in function implements the generic version of 42265 '__atomic_compare_exchange'. The function is virtually identical 42266 to '__atomic_compare_exchange_n', except the desired value is also 42267 a pointer. 42268 42269 -- Built-in Function: TYPE __atomic_add_fetch (TYPE *ptr, TYPE val, int 42270 memorder) 42271 -- Built-in Function: TYPE __atomic_sub_fetch (TYPE *ptr, TYPE val, int 42272 memorder) 42273 -- Built-in Function: TYPE __atomic_and_fetch (TYPE *ptr, TYPE val, int 42274 memorder) 42275 -- Built-in Function: TYPE __atomic_xor_fetch (TYPE *ptr, TYPE val, int 42276 memorder) 42277 -- Built-in Function: TYPE __atomic_or_fetch (TYPE *ptr, TYPE val, int 42278 memorder) 42279 -- Built-in Function: TYPE __atomic_nand_fetch (TYPE *ptr, TYPE val, 42280 int memorder) 42281 These built-in functions perform the operation suggested by the 42282 name, and return the result of the operation. Operations on 42283 pointer arguments are performed as if the operands were of the 42284 'uintptr_t' type. That is, they are not scaled by the size of the 42285 type to which the pointer points. 42286 42287 { *ptr OP= val; return *ptr; } 42288 { *ptr = ~(*ptr & val); return *ptr; } // nand 42289 42290 The object pointed to by the first argument must be of integer or 42291 pointer type. It must not be a boolean type. All memory orders 42292 are valid. 42293 42294 -- Built-in Function: TYPE __atomic_fetch_add (TYPE *ptr, TYPE val, int 42295 memorder) 42296 -- Built-in Function: TYPE __atomic_fetch_sub (TYPE *ptr, TYPE val, int 42297 memorder) 42298 -- Built-in Function: TYPE __atomic_fetch_and (TYPE *ptr, TYPE val, int 42299 memorder) 42300 -- Built-in Function: TYPE __atomic_fetch_xor (TYPE *ptr, TYPE val, int 42301 memorder) 42302 -- Built-in Function: TYPE __atomic_fetch_or (TYPE *ptr, TYPE val, int 42303 memorder) 42304 -- Built-in Function: TYPE __atomic_fetch_nand (TYPE *ptr, TYPE val, 42305 int memorder) 42306 These built-in functions perform the operation suggested by the 42307 name, and return the value that had previously been in '*PTR'. 42308 Operations on pointer arguments are performed as if the operands 42309 were of the 'uintptr_t' type. That is, they are not scaled by the 42310 size of the type to which the pointer points. 42311 42312 { tmp = *ptr; *ptr OP= val; return tmp; } 42313 { tmp = *ptr; *ptr = ~(*ptr & val); return tmp; } // nand 42314 42315 The same constraints on arguments apply as for the corresponding 42316 '__atomic_op_fetch' built-in functions. All memory orders are 42317 valid. 42318 42319 -- Built-in Function: bool __atomic_test_and_set (void *ptr, int 42320 memorder) 42321 42322 This built-in function performs an atomic test-and-set operation on 42323 the byte at '*PTR'. The byte is set to some implementation defined 42324 nonzero "set" value and the return value is 'true' if and only if 42325 the previous contents were "set". It should be only used for 42326 operands of type 'bool' or 'char'. For other types only part of 42327 the value may be set. 42328 42329 All memory orders are valid. 42330 42331 -- Built-in Function: void __atomic_clear (bool *ptr, int memorder) 42332 42333 This built-in function performs an atomic clear operation on 42334 '*PTR'. After the operation, '*PTR' contains 0. It should be only 42335 used for operands of type 'bool' or 'char' and in conjunction with 42336 '__atomic_test_and_set'. For other types it may only clear 42337 partially. If the type is not 'bool' prefer using 42338 '__atomic_store'. 42339 42340 The valid memory order variants are '__ATOMIC_RELAXED', 42341 '__ATOMIC_SEQ_CST', and '__ATOMIC_RELEASE'. 42342 42343 -- Built-in Function: void __atomic_thread_fence (int memorder) 42344 42345 This built-in function acts as a synchronization fence between 42346 threads based on the specified memory order. 42347 42348 All memory orders are valid. 42349 42350 -- Built-in Function: void __atomic_signal_fence (int memorder) 42351 42352 This built-in function acts as a synchronization fence between a 42353 thread and signal handlers based in the same thread. 42354 42355 All memory orders are valid. 42356 42357 -- Built-in Function: bool __atomic_always_lock_free (size_t size, void 42358 *ptr) 42359 42360 This built-in function returns 'true' if objects of SIZE bytes 42361 always generate lock-free atomic instructions for the target 42362 architecture. SIZE must resolve to a compile-time constant and the 42363 result also resolves to a compile-time constant. 42364 42365 PTR is an optional pointer to the object that may be used to 42366 determine alignment. A value of 0 indicates typical alignment 42367 should be used. The compiler may also ignore this parameter. 42368 42369 if (__atomic_always_lock_free (sizeof (long long), 0)) 42370 42371 -- Built-in Function: bool __atomic_is_lock_free (size_t size, void 42372 *ptr) 42373 42374 This built-in function returns 'true' if objects of SIZE bytes 42375 always generate lock-free atomic instructions for the target 42376 architecture. If the built-in function is not known to be 42377 lock-free, a call is made to a runtime routine named 42378 '__atomic_is_lock_free'. 42379 42380 PTR is an optional pointer to the object that may be used to 42381 determine alignment. A value of 0 indicates typical alignment 42382 should be used. The compiler may also ignore this parameter. 42383 42384 42385File: gcc.info, Node: Integer Overflow Builtins, Next: x86 specific memory model extensions for transactional memory, Prev: __atomic Builtins, Up: C Extensions 42386 423876.56 Built-in Functions to Perform Arithmetic with Overflow Checking 42388==================================================================== 42389 42390The following built-in functions allow performing simple arithmetic 42391operations together with checking whether the operations overflowed. 42392 42393 -- Built-in Function: bool __builtin_add_overflow (TYPE1 a, TYPE2 b, 42394 TYPE3 *res) 42395 -- Built-in Function: bool __builtin_sadd_overflow (int a, int b, int 42396 *res) 42397 -- Built-in Function: bool __builtin_saddl_overflow (long int a, long 42398 int b, long int *res) 42399 -- Built-in Function: bool __builtin_saddll_overflow (long long int a, 42400 long long int b, long long int *res) 42401 -- Built-in Function: bool __builtin_uadd_overflow (unsigned int a, 42402 unsigned int b, unsigned int *res) 42403 -- Built-in Function: bool __builtin_uaddl_overflow (unsigned long int 42404 a, unsigned long int b, unsigned long int *res) 42405 -- Built-in Function: bool __builtin_uaddll_overflow (unsigned long 42406 long int a, unsigned long long int b, unsigned long long int 42407 *res) 42408 42409 These built-in functions promote the first two operands into 42410 infinite precision signed type and perform addition on those 42411 promoted operands. The result is then cast to the type the third 42412 pointer argument points to and stored there. If the stored result 42413 is equal to the infinite precision result, the built-in functions 42414 return 'false', otherwise they return 'true'. As the addition is 42415 performed in infinite signed precision, these built-in functions 42416 have fully defined behavior for all argument values. 42417 42418 The first built-in function allows arbitrary integral types for 42419 operands and the result type must be pointer to some integral type 42420 other than enumerated or boolean type, the rest of the built-in 42421 functions have explicit integer types. 42422 42423 The compiler will attempt to use hardware instructions to implement 42424 these built-in functions where possible, like conditional jump on 42425 overflow after addition, conditional jump on carry etc. 42426 42427 -- Built-in Function: bool __builtin_sub_overflow (TYPE1 a, TYPE2 b, 42428 TYPE3 *res) 42429 -- Built-in Function: bool __builtin_ssub_overflow (int a, int b, int 42430 *res) 42431 -- Built-in Function: bool __builtin_ssubl_overflow (long int a, long 42432 int b, long int *res) 42433 -- Built-in Function: bool __builtin_ssubll_overflow (long long int a, 42434 long long int b, long long int *res) 42435 -- Built-in Function: bool __builtin_usub_overflow (unsigned int a, 42436 unsigned int b, unsigned int *res) 42437 -- Built-in Function: bool __builtin_usubl_overflow (unsigned long int 42438 a, unsigned long int b, unsigned long int *res) 42439 -- Built-in Function: bool __builtin_usubll_overflow (unsigned long 42440 long int a, unsigned long long int b, unsigned long long int 42441 *res) 42442 42443 These built-in functions are similar to the add overflow checking 42444 built-in functions above, except they perform subtraction, subtract 42445 the second argument from the first one, instead of addition. 42446 42447 -- Built-in Function: bool __builtin_mul_overflow (TYPE1 a, TYPE2 b, 42448 TYPE3 *res) 42449 -- Built-in Function: bool __builtin_smul_overflow (int a, int b, int 42450 *res) 42451 -- Built-in Function: bool __builtin_smull_overflow (long int a, long 42452 int b, long int *res) 42453 -- Built-in Function: bool __builtin_smulll_overflow (long long int a, 42454 long long int b, long long int *res) 42455 -- Built-in Function: bool __builtin_umul_overflow (unsigned int a, 42456 unsigned int b, unsigned int *res) 42457 -- Built-in Function: bool __builtin_umull_overflow (unsigned long int 42458 a, unsigned long int b, unsigned long int *res) 42459 -- Built-in Function: bool __builtin_umulll_overflow (unsigned long 42460 long int a, unsigned long long int b, unsigned long long int 42461 *res) 42462 42463 These built-in functions are similar to the add overflow checking 42464 built-in functions above, except they perform multiplication, 42465 instead of addition. 42466 42467 The following built-in functions allow checking if simple arithmetic 42468operation would overflow. 42469 42470 -- Built-in Function: bool __builtin_add_overflow_p (TYPE1 a, TYPE2 b, 42471 TYPE3 c) 42472 -- Built-in Function: bool __builtin_sub_overflow_p (TYPE1 a, TYPE2 b, 42473 TYPE3 c) 42474 -- Built-in Function: bool __builtin_mul_overflow_p (TYPE1 a, TYPE2 b, 42475 TYPE3 c) 42476 42477 These built-in functions are similar to '__builtin_add_overflow', 42478 '__builtin_sub_overflow', or '__builtin_mul_overflow', except that 42479 they don't store the result of the arithmetic operation anywhere 42480 and the last argument is not a pointer, but some expression with 42481 integral type other than enumerated or boolean type. 42482 42483 The built-in functions promote the first two operands into infinite 42484 precision signed type and perform addition on those promoted 42485 operands. The result is then cast to the type of the third 42486 argument. If the cast result is equal to the infinite precision 42487 result, the built-in functions return 'false', otherwise they 42488 return 'true'. The value of the third argument is ignored, just 42489 the side effects in the third argument are evaluated, and no 42490 integral argument promotions are performed on the last argument. 42491 If the third argument is a bit-field, the type used for the result 42492 cast has the precision and signedness of the given bit-field, 42493 rather than precision and signedness of the underlying type. 42494 42495 For example, the following macro can be used to portably check, at 42496 compile-time, whether or not adding two constant integers will 42497 overflow, and perform the addition only when it is known to be safe 42498 and not to trigger a '-Woverflow' warning. 42499 42500 #define INT_ADD_OVERFLOW_P(a, b) \ 42501 __builtin_add_overflow_p (a, b, (__typeof__ ((a) + (b))) 0) 42502 42503 enum { 42504 A = INT_MAX, B = 3, 42505 C = INT_ADD_OVERFLOW_P (A, B) ? 0 : A + B, 42506 D = __builtin_add_overflow_p (1, SCHAR_MAX, (signed char) 0) 42507 }; 42508 42509 The compiler will attempt to use hardware instructions to implement 42510 these built-in functions where possible, like conditional jump on 42511 overflow after addition, conditional jump on carry etc. 42512 42513 42514File: gcc.info, Node: x86 specific memory model extensions for transactional memory, Next: Object Size Checking, Prev: Integer Overflow Builtins, Up: C Extensions 42515 425166.57 x86-Specific Memory Model Extensions for Transactional Memory 42517================================================================== 42518 42519The x86 architecture supports additional memory ordering flags to mark 42520critical sections for hardware lock elision. These must be specified in 42521addition to an existing memory order to atomic intrinsics. 42522 42523'__ATOMIC_HLE_ACQUIRE' 42524 Start lock elision on a lock variable. Memory order must be 42525 '__ATOMIC_ACQUIRE' or stronger. 42526'__ATOMIC_HLE_RELEASE' 42527 End lock elision on a lock variable. Memory order must be 42528 '__ATOMIC_RELEASE' or stronger. 42529 42530 When a lock acquire fails, it is required for good performance to abort 42531the transaction quickly. This can be done with a '_mm_pause'. 42532 42533 #include <immintrin.h> // For _mm_pause 42534 42535 int lockvar; 42536 42537 /* Acquire lock with lock elision */ 42538 while (__atomic_exchange_n(&lockvar, 1, __ATOMIC_ACQUIRE|__ATOMIC_HLE_ACQUIRE)) 42539 _mm_pause(); /* Abort failed transaction */ 42540 ... 42541 /* Free lock with lock elision */ 42542 __atomic_store_n(&lockvar, 0, __ATOMIC_RELEASE|__ATOMIC_HLE_RELEASE); 42543 42544 42545File: gcc.info, Node: Object Size Checking, Next: Other Builtins, Prev: x86 specific memory model extensions for transactional memory, Up: C Extensions 42546 425476.58 Object Size Checking Built-in Functions 42548============================================ 42549 42550GCC implements a limited buffer overflow protection mechanism that can 42551prevent some buffer overflow attacks by determining the sizes of objects 42552into which data is about to be written and preventing the writes when 42553the size isn't sufficient. The built-in functions described below yield 42554the best results when used together and when optimization is enabled. 42555For example, to detect object sizes across function boundaries or to 42556follow pointer assignments through non-trivial control flow they rely on 42557various optimization passes enabled with '-O2'. However, to a limited 42558extent, they can be used without optimization as well. 42559 42560 -- Built-in Function: size_t __builtin_object_size (const void * PTR, 42561 int TYPE) 42562 is a built-in construct that returns a constant number of bytes 42563 from PTR to the end of the object PTR pointer points to (if known 42564 at compile time). To determine the sizes of dynamically allocated 42565 objects the function relies on the allocation functions called to 42566 obtain the storage to be declared with the 'alloc_size' attribute 42567 (*note Common Function Attributes::). '__builtin_object_size' 42568 never evaluates its arguments for side effects. If there are any 42569 side effects in them, it returns '(size_t) -1' for TYPE 0 or 1 and 42570 '(size_t) 0' for TYPE 2 or 3. If there are multiple objects PTR 42571 can point to and all of them are known at compile time, the 42572 returned number is the maximum of remaining byte counts in those 42573 objects if TYPE & 2 is 0 and minimum if nonzero. If it is not 42574 possible to determine which objects PTR points to at compile time, 42575 '__builtin_object_size' should return '(size_t) -1' for TYPE 0 or 1 42576 and '(size_t) 0' for TYPE 2 or 3. 42577 42578 TYPE is an integer constant from 0 to 3. If the least significant 42579 bit is clear, objects are whole variables, if it is set, a closest 42580 surrounding subobject is considered the object a pointer points to. 42581 The second bit determines if maximum or minimum of remaining bytes 42582 is computed. 42583 42584 struct V { char buf1[10]; int b; char buf2[10]; } var; 42585 char *p = &var.buf1[1], *q = &var.b; 42586 42587 /* Here the object p points to is var. */ 42588 assert (__builtin_object_size (p, 0) == sizeof (var) - 1); 42589 /* The subobject p points to is var.buf1. */ 42590 assert (__builtin_object_size (p, 1) == sizeof (var.buf1) - 1); 42591 /* The object q points to is var. */ 42592 assert (__builtin_object_size (q, 0) 42593 == (char *) (&var + 1) - (char *) &var.b); 42594 /* The subobject q points to is var.b. */ 42595 assert (__builtin_object_size (q, 1) == sizeof (var.b)); 42596 42597 There are built-in functions added for many common string operation 42598functions, e.g., for 'memcpy' '__builtin___memcpy_chk' built-in is 42599provided. This built-in has an additional last argument, which is the 42600number of bytes remaining in the object the DEST argument points to or 42601'(size_t) -1' if the size is not known. 42602 42603 The built-in functions are optimized into the normal string functions 42604like 'memcpy' if the last argument is '(size_t) -1' or if it is known at 42605compile time that the destination object will not be overflowed. If the 42606compiler can determine at compile time that the object will always be 42607overflowed, it issues a warning. 42608 42609 The intended use can be e.g. 42610 42611 #undef memcpy 42612 #define bos0(dest) __builtin_object_size (dest, 0) 42613 #define memcpy(dest, src, n) \ 42614 __builtin___memcpy_chk (dest, src, n, bos0 (dest)) 42615 42616 char *volatile p; 42617 char buf[10]; 42618 /* It is unknown what object p points to, so this is optimized 42619 into plain memcpy - no checking is possible. */ 42620 memcpy (p, "abcde", n); 42621 /* Destination is known and length too. It is known at compile 42622 time there will be no overflow. */ 42623 memcpy (&buf[5], "abcde", 5); 42624 /* Destination is known, but the length is not known at compile time. 42625 This will result in __memcpy_chk call that can check for overflow 42626 at run time. */ 42627 memcpy (&buf[5], "abcde", n); 42628 /* Destination is known and it is known at compile time there will 42629 be overflow. There will be a warning and __memcpy_chk call that 42630 will abort the program at run time. */ 42631 memcpy (&buf[6], "abcde", 5); 42632 42633 Such built-in functions are provided for 'memcpy', 'mempcpy', 42634'memmove', 'memset', 'strcpy', 'stpcpy', 'strncpy', 'strcat' and 42635'strncat'. 42636 42637 There are also checking built-in functions for formatted output 42638functions. 42639 int __builtin___sprintf_chk (char *s, int flag, size_t os, const char *fmt, ...); 42640 int __builtin___snprintf_chk (char *s, size_t maxlen, int flag, size_t os, 42641 const char *fmt, ...); 42642 int __builtin___vsprintf_chk (char *s, int flag, size_t os, const char *fmt, 42643 va_list ap); 42644 int __builtin___vsnprintf_chk (char *s, size_t maxlen, int flag, size_t os, 42645 const char *fmt, va_list ap); 42646 42647 The added FLAG argument is passed unchanged to '__sprintf_chk' etc. 42648functions and can contain implementation specific flags on what 42649additional security measures the checking function might take, such as 42650handling '%n' differently. 42651 42652 The OS argument is the object size S points to, like in the other 42653built-in functions. There is a small difference in the behavior though, 42654if OS is '(size_t) -1', the built-in functions are optimized into the 42655non-checking functions only if FLAG is 0, otherwise the checking 42656function is called with OS argument set to '(size_t) -1'. 42657 42658 In addition to this, there are checking built-in functions 42659'__builtin___printf_chk', '__builtin___vprintf_chk', 42660'__builtin___fprintf_chk' and '__builtin___vfprintf_chk'. These have 42661just one additional argument, FLAG, right before format string FMT. If 42662the compiler is able to optimize them to 'fputc' etc. functions, it 42663does, otherwise the checking function is called and the FLAG argument 42664passed to it. 42665 42666 42667File: gcc.info, Node: Other Builtins, Next: Target Builtins, Prev: Object Size Checking, Up: C Extensions 42668 426696.59 Other Built-in Functions Provided by GCC 42670============================================= 42671 42672GCC provides a large number of built-in functions other than the ones 42673mentioned above. Some of these are for internal use in the processing 42674of exceptions or variable-length argument lists and are not documented 42675here because they may change from time to time; we do not recommend 42676general use of these functions. 42677 42678 The remaining functions are provided for optimization purposes. 42679 42680 With the exception of built-ins that have library equivalents such as 42681the standard C library functions discussed below, or that expand to 42682library calls, GCC built-in functions are always expanded inline and 42683thus do not have corresponding entry points and their address cannot be 42684obtained. Attempting to use them in an expression other than a function 42685call results in a compile-time error. 42686 42687 GCC includes built-in versions of many of the functions in the standard 42688C library. These functions come in two forms: one whose names start 42689with the '__builtin_' prefix, and the other without. Both forms have 42690the same type (including prototype), the same address (when their 42691address is taken), and the same meaning as the C library functions even 42692if you specify the '-fno-builtin' option *note C Dialect Options::). 42693Many of these functions are only optimized in certain cases; if they are 42694not optimized in a particular case, a call to the library function is 42695emitted. 42696 42697 Outside strict ISO C mode ('-ansi', '-std=c90', '-std=c99' or 42698'-std=c11'), the functions '_exit', 'alloca', 'bcmp', 'bzero', 42699'dcgettext', 'dgettext', 'dremf', 'dreml', 'drem', 'exp10f', 'exp10l', 42700'exp10', 'ffsll', 'ffsl', 'ffs', 'fprintf_unlocked', 'fputs_unlocked', 42701'gammaf', 'gammal', 'gamma', 'gammaf_r', 'gammal_r', 'gamma_r', 42702'gettext', 'index', 'isascii', 'j0f', 'j0l', 'j0', 'j1f', 'j1l', 'j1', 42703'jnf', 'jnl', 'jn', 'lgammaf_r', 'lgammal_r', 'lgamma_r', 'mempcpy', 42704'pow10f', 'pow10l', 'pow10', 'printf_unlocked', 'rindex', 'roundeven', 42705'roundevenf', 'roudnevenl', 'scalbf', 'scalbl', 'scalb', 'signbit', 42706'signbitf', 'signbitl', 'signbitd32', 'signbitd64', 'signbitd128', 42707'significandf', 'significandl', 'significand', 'sincosf', 'sincosl', 42708'sincos', 'stpcpy', 'stpncpy', 'strcasecmp', 'strdup', 'strfmon', 42709'strncasecmp', 'strndup', 'strnlen', 'toascii', 'y0f', 'y0l', 'y0', 42710'y1f', 'y1l', 'y1', 'ynf', 'ynl' and 'yn' may be handled as built-in 42711functions. All these functions have corresponding versions prefixed 42712with '__builtin_', which may be used even in strict C90 mode. 42713 42714 The ISO C99 functions '_Exit', 'acoshf', 'acoshl', 'acosh', 'asinhf', 42715'asinhl', 'asinh', 'atanhf', 'atanhl', 'atanh', 'cabsf', 'cabsl', 42716'cabs', 'cacosf', 'cacoshf', 'cacoshl', 'cacosh', 'cacosl', 'cacos', 42717'cargf', 'cargl', 'carg', 'casinf', 'casinhf', 'casinhl', 'casinh', 42718'casinl', 'casin', 'catanf', 'catanhf', 'catanhl', 'catanh', 'catanl', 42719'catan', 'cbrtf', 'cbrtl', 'cbrt', 'ccosf', 'ccoshf', 'ccoshl', 'ccosh', 42720'ccosl', 'ccos', 'cexpf', 'cexpl', 'cexp', 'cimagf', 'cimagl', 'cimag', 42721'clogf', 'clogl', 'clog', 'conjf', 'conjl', 'conj', 'copysignf', 42722'copysignl', 'copysign', 'cpowf', 'cpowl', 'cpow', 'cprojf', 'cprojl', 42723'cproj', 'crealf', 'creall', 'creal', 'csinf', 'csinhf', 'csinhl', 42724'csinh', 'csinl', 'csin', 'csqrtf', 'csqrtl', 'csqrt', 'ctanf', 42725'ctanhf', 'ctanhl', 'ctanh', 'ctanl', 'ctan', 'erfcf', 'erfcl', 'erfc', 42726'erff', 'erfl', 'erf', 'exp2f', 'exp2l', 'exp2', 'expm1f', 'expm1l', 42727'expm1', 'fdimf', 'fdiml', 'fdim', 'fmaf', 'fmal', 'fmaxf', 'fmaxl', 42728'fmax', 'fma', 'fminf', 'fminl', 'fmin', 'hypotf', 'hypotl', 'hypot', 42729'ilogbf', 'ilogbl', 'ilogb', 'imaxabs', 'isblank', 'iswblank', 42730'lgammaf', 'lgammal', 'lgamma', 'llabs', 'llrintf', 'llrintl', 'llrint', 42731'llroundf', 'llroundl', 'llround', 'log1pf', 'log1pl', 'log1p', 'log2f', 42732'log2l', 'log2', 'logbf', 'logbl', 'logb', 'lrintf', 'lrintl', 'lrint', 42733'lroundf', 'lroundl', 'lround', 'nearbyintf', 'nearbyintl', 'nearbyint', 42734'nextafterf', 'nextafterl', 'nextafter', 'nexttowardf', 'nexttowardl', 42735'nexttoward', 'remainderf', 'remainderl', 'remainder', 'remquof', 42736'remquol', 'remquo', 'rintf', 'rintl', 'rint', 'roundf', 'roundl', 42737'round', 'scalblnf', 'scalblnl', 'scalbln', 'scalbnf', 'scalbnl', 42738'scalbn', 'snprintf', 'tgammaf', 'tgammal', 'tgamma', 'truncf', 42739'truncl', 'trunc', 'vfscanf', 'vscanf', 'vsnprintf' and 'vsscanf' are 42740handled as built-in functions except in strict ISO C90 mode ('-ansi' or 42741'-std=c90'). 42742 42743 There are also built-in versions of the ISO C99 functions 'acosf', 42744'acosl', 'asinf', 'asinl', 'atan2f', 'atan2l', 'atanf', 'atanl', 42745'ceilf', 'ceill', 'cosf', 'coshf', 'coshl', 'cosl', 'expf', 'expl', 42746'fabsf', 'fabsl', 'floorf', 'floorl', 'fmodf', 'fmodl', 'frexpf', 42747'frexpl', 'ldexpf', 'ldexpl', 'log10f', 'log10l', 'logf', 'logl', 42748'modfl', 'modf', 'powf', 'powl', 'sinf', 'sinhf', 'sinhl', 'sinl', 42749'sqrtf', 'sqrtl', 'tanf', 'tanhf', 'tanhl' and 'tanl' that are 42750recognized in any mode since ISO C90 reserves these names for the 42751purpose to which ISO C99 puts them. All these functions have 42752corresponding versions prefixed with '__builtin_'. 42753 42754 There are also built-in functions '__builtin_fabsfN', 42755'__builtin_fabsfNx', '__builtin_copysignfN' and '__builtin_copysignfNx', 42756corresponding to the TS 18661-3 functions 'fabsfN', 'fabsfNx', 42757'copysignfN' and 'copysignfNx', for supported types '_FloatN' and 42758'_FloatNx'. 42759 42760 There are also GNU extension functions 'clog10', 'clog10f' and 42761'clog10l' which names are reserved by ISO C99 for future use. All these 42762functions have versions prefixed with '__builtin_'. 42763 42764 The ISO C94 functions 'iswalnum', 'iswalpha', 'iswcntrl', 'iswdigit', 42765'iswgraph', 'iswlower', 'iswprint', 'iswpunct', 'iswspace', 'iswupper', 42766'iswxdigit', 'towlower' and 'towupper' are handled as built-in functions 42767except in strict ISO C90 mode ('-ansi' or '-std=c90'). 42768 42769 The ISO C90 functions 'abort', 'abs', 'acos', 'asin', 'atan2', 'atan', 42770'calloc', 'ceil', 'cosh', 'cos', 'exit', 'exp', 'fabs', 'floor', 'fmod', 42771'fprintf', 'fputs', 'free', 'frexp', 'fscanf', 'isalnum', 'isalpha', 42772'iscntrl', 'isdigit', 'isgraph', 'islower', 'isprint', 'ispunct', 42773'isspace', 'isupper', 'isxdigit', 'tolower', 'toupper', 'labs', 'ldexp', 42774'log10', 'log', 'malloc', 'memchr', 'memcmp', 'memcpy', 'memset', 42775'modf', 'pow', 'printf', 'putchar', 'puts', 'realloc', 'scanf', 'sinh', 42776'sin', 'snprintf', 'sprintf', 'sqrt', 'sscanf', 'strcat', 'strchr', 42777'strcmp', 'strcpy', 'strcspn', 'strlen', 'strncat', 'strncmp', 42778'strncpy', 'strpbrk', 'strrchr', 'strspn', 'strstr', 'tanh', 'tan', 42779'vfprintf', 'vprintf' and 'vsprintf' are all recognized as built-in 42780functions unless '-fno-builtin' is specified (or '-fno-builtin-FUNCTION' 42781is specified for an individual function). All of these functions have 42782corresponding versions prefixed with '__builtin_'. 42783 42784 GCC provides built-in versions of the ISO C99 floating-point comparison 42785macros that avoid raising exceptions for unordered operands. They have 42786the same names as the standard macros ( 'isgreater', 'isgreaterequal', 42787'isless', 'islessequal', 'islessgreater', and 'isunordered') , with 42788'__builtin_' prefixed. We intend for a library implementor to be able 42789to simply '#define' each standard macro to its built-in equivalent. In 42790the same fashion, GCC provides 'fpclassify', 'isfinite', 'isinf_sign', 42791'isnormal' and 'signbit' built-ins used with '__builtin_' prefixed. The 42792'isinf' and 'isnan' built-in functions appear both with and without the 42793'__builtin_' prefix. 42794 42795 -- Built-in Function: void *__builtin_alloca (size_t size) 42796 The '__builtin_alloca' function must be called at block scope. The 42797 function allocates an object SIZE bytes large on the stack of the 42798 calling function. The object is aligned on the default stack 42799 alignment boundary for the target determined by the 42800 '__BIGGEST_ALIGNMENT__' macro. The '__builtin_alloca' function 42801 returns a pointer to the first byte of the allocated object. The 42802 lifetime of the allocated object ends just before the calling 42803 function returns to its caller. This is so even when 42804 '__builtin_alloca' is called within a nested block. 42805 42806 For example, the following function allocates eight objects of 'n' 42807 bytes each on the stack, storing a pointer to each in consecutive 42808 elements of the array 'a'. It then passes the array to function 42809 'g' which can safely use the storage pointed to by each of the 42810 array elements. 42811 42812 void f (unsigned n) 42813 { 42814 void *a [8]; 42815 for (int i = 0; i != 8; ++i) 42816 a [i] = __builtin_alloca (n); 42817 42818 g (a, n); // safe 42819 } 42820 42821 Since the '__builtin_alloca' function doesn't validate its argument 42822 it is the responsibility of its caller to make sure the argument 42823 doesn't cause it to exceed the stack size limit. The 42824 '__builtin_alloca' function is provided to make it possible to 42825 allocate on the stack arrays of bytes with an upper bound that may 42826 be computed at run time. Since C99 Variable Length Arrays offer 42827 similar functionality under a portable, more convenient, and safer 42828 interface they are recommended instead, in both C99 and C++ 42829 programs where GCC provides them as an extension. *Note Variable 42830 Length::, for details. 42831 42832 -- Built-in Function: void *__builtin_alloca_with_align (size_t size, 42833 size_t alignment) 42834 The '__builtin_alloca_with_align' function must be called at block 42835 scope. The function allocates an object SIZE bytes large on the 42836 stack of the calling function. The allocated object is aligned on 42837 the boundary specified by the argument ALIGNMENT whose unit is 42838 given in bits (not bytes). The SIZE argument must be positive and 42839 not exceed the stack size limit. The ALIGNMENT argument must be a 42840 constant integer expression that evaluates to a power of 2 greater 42841 than or equal to 'CHAR_BIT' and less than some unspecified maximum. 42842 Invocations with other values are rejected with an error indicating 42843 the valid bounds. The function returns a pointer to the first byte 42844 of the allocated object. The lifetime of the allocated object ends 42845 at the end of the block in which the function was called. The 42846 allocated storage is released no later than just before the calling 42847 function returns to its caller, but may be released at the end of 42848 the block in which the function was called. 42849 42850 For example, in the following function the call to 'g' is unsafe 42851 because when 'overalign' is non-zero, the space allocated by 42852 '__builtin_alloca_with_align' may have been released at the end of 42853 the 'if' statement in which it was called. 42854 42855 void f (unsigned n, bool overalign) 42856 { 42857 void *p; 42858 if (overalign) 42859 p = __builtin_alloca_with_align (n, 64 /* bits */); 42860 else 42861 p = __builtin_alloc (n); 42862 42863 g (p, n); // unsafe 42864 } 42865 42866 Since the '__builtin_alloca_with_align' function doesn't validate 42867 its SIZE argument it is the responsibility of its caller to make 42868 sure the argument doesn't cause it to exceed the stack size limit. 42869 The '__builtin_alloca_with_align' function is provided to make it 42870 possible to allocate on the stack overaligned arrays of bytes with 42871 an upper bound that may be computed at run time. Since C99 42872 Variable Length Arrays offer the same functionality under a 42873 portable, more convenient, and safer interface they are recommended 42874 instead, in both C99 and C++ programs where GCC provides them as an 42875 extension. *Note Variable Length::, for details. 42876 42877 -- Built-in Function: void *__builtin_alloca_with_align_and_max (size_t 42878 size, size_t alignment, size_t max_size) 42879 Similar to '__builtin_alloca_with_align' but takes an extra 42880 argument specifying an upper bound for SIZE in case its value 42881 cannot be computed at compile time, for use by '-fstack-usage', 42882 '-Wstack-usage' and '-Walloca-larger-than'. MAX_SIZE must be a 42883 constant integer expression, it has no effect on code generation 42884 and no attempt is made to check its compatibility with SIZE. 42885 42886 -- Built-in Function: bool __builtin_has_attribute (TYPE-OR-EXPRESSION, 42887 ATTRIBUTE) 42888 The '__builtin_has_attribute' function evaluates to an integer 42889 constant expression equal to 'true' if the symbol or type 42890 referenced by the TYPE-OR-EXPRESSION argument has been declared 42891 with the ATTRIBUTE referenced by the second argument. For an 42892 TYPE-OR-EXPRESSION argument that does not reference a symbol, since 42893 attributes do not apply to expressions the built-in consider the 42894 type of the argument. Neither argument is evaluated. The 42895 TYPE-OR-EXPRESSION argument is subject to the same restrictions as 42896 the argument to 'typeof' (*note Typeof::). The ATTRIBUTE argument 42897 is an attribute name optionally followed by a comma-separated list 42898 of arguments enclosed in parentheses. Both forms of attribute 42899 names--with and without double leading and trailing 42900 underscores--are recognized. *Note Attribute Syntax::, for 42901 details. When no attribute arguments are specified for an 42902 attribute that expects one or more arguments the function returns 42903 'true' if TYPE-OR-EXPRESSION has been declared with the attribute 42904 regardless of the attribute argument values. Arguments provided 42905 for an attribute that expects some are validated and matched up to 42906 the provided number. The function returns 'true' if all provided 42907 arguments match. For example, the first call to the function below 42908 evaluates to 'true' because 'x' is declared with the 'aligned' 42909 attribute but the second call evaluates to 'false' because 'x' is 42910 declared 'aligned (8)' and not 'aligned (4)'. 42911 42912 __attribute__ ((aligned (8))) int x; 42913 _Static_assert (__builtin_has_attribute (x, aligned), "aligned"); 42914 _Static_assert (!__builtin_has_attribute (x, aligned (4)), "aligned (4)"); 42915 42916 Due to a limitation the '__builtin_has_attribute' function returns 42917 'false' for the 'mode' attribute even if the type or variable 42918 referenced by the TYPE-OR-EXPRESSION argument was declared with 42919 one. The function is also not supported with labels, and in C with 42920 enumerators. 42921 42922 Note that unlike the '__has_attribute' preprocessor operator which 42923 is suitable for use in '#if' preprocessing directives 42924 '__builtin_has_attribute' is an intrinsic function that is not 42925 recognized in such contexts. 42926 42927 -- Built-in Function: TYPE __builtin_speculation_safe_value (TYPE val, 42928 TYPE failval) 42929 42930 This built-in function can be used to help mitigate against unsafe 42931 speculative execution. TYPE may be any integral type or any 42932 pointer type. 42933 42934 1. If the CPU is not speculatively executing the code, then VAL 42935 is returned. 42936 2. If the CPU is executing speculatively then either: 42937 * The function may cause execution to pause until it is 42938 known that the code is no-longer being executed 42939 speculatively (in which case VAL can be returned, as 42940 above); or 42941 * The function may use target-dependent speculation 42942 tracking state to cause FAILVAL to be returned when it is 42943 known that speculative execution has incorrectly 42944 predicted a conditional branch operation. 42945 42946 The second argument, FAILVAL, is optional and defaults to zero if 42947 omitted. 42948 42949 GCC defines the preprocessor macro 42950 '__HAVE_BUILTIN_SPECULATION_SAFE_VALUE' for targets that have been 42951 updated to support this builtin. 42952 42953 The built-in function can be used where a variable appears to be 42954 used in a safe way, but the CPU, due to speculative execution may 42955 temporarily ignore the bounds checks. Consider, for example, the 42956 following function: 42957 42958 int array[500]; 42959 int f (unsigned untrusted_index) 42960 { 42961 if (untrusted_index < 500) 42962 return array[untrusted_index]; 42963 return 0; 42964 } 42965 42966 If the function is called repeatedly with 'untrusted_index' less 42967 than the limit of 500, then a branch predictor will learn that the 42968 block of code that returns a value stored in 'array' will be 42969 executed. If the function is subsequently called with an 42970 out-of-range value it will still try to execute that block of code 42971 first until the CPU determines that the prediction was incorrect 42972 (the CPU will unwind any incorrect operations at that point). 42973 However, depending on how the result of the function is used, it 42974 might be possible to leave traces in the cache that can reveal what 42975 was stored at the out-of-bounds location. The built-in function 42976 can be used to provide some protection against leaking data in this 42977 way by changing the code to: 42978 42979 int array[500]; 42980 int f (unsigned untrusted_index) 42981 { 42982 if (untrusted_index < 500) 42983 return array[__builtin_speculation_safe_value (untrusted_index)]; 42984 return 0; 42985 } 42986 42987 The built-in function will either cause execution to stall until 42988 the conditional branch has been fully resolved, or it may permit 42989 speculative execution to continue, but using 0 instead of 42990 'untrusted_value' if that exceeds the limit. 42991 42992 If accessing any memory location is potentially unsafe when 42993 speculative execution is incorrect, then the code can be rewritten 42994 as 42995 42996 int array[500]; 42997 int f (unsigned untrusted_index) 42998 { 42999 if (untrusted_index < 500) 43000 return *__builtin_speculation_safe_value (&array[untrusted_index], NULL); 43001 return 0; 43002 } 43003 43004 which will cause a 'NULL' pointer to be used for the unsafe case. 43005 43006 -- Built-in Function: int __builtin_types_compatible_p (TYPE1, TYPE2) 43007 43008 You can use the built-in function '__builtin_types_compatible_p' to 43009 determine whether two types are the same. 43010 43011 This built-in function returns 1 if the unqualified versions of the 43012 types TYPE1 and TYPE2 (which are types, not expressions) are 43013 compatible, 0 otherwise. The result of this built-in function can 43014 be used in integer constant expressions. 43015 43016 This built-in function ignores top level qualifiers (e.g., 'const', 43017 'volatile'). For example, 'int' is equivalent to 'const int'. 43018 43019 The type 'int[]' and 'int[5]' are compatible. On the other hand, 43020 'int' and 'char *' are not compatible, even if the size of their 43021 types, on the particular architecture are the same. Also, the 43022 amount of pointer indirection is taken into account when 43023 determining similarity. Consequently, 'short *' is not similar to 43024 'short **'. Furthermore, two types that are typedefed are 43025 considered compatible if their underlying types are compatible. 43026 43027 An 'enum' type is not considered to be compatible with another 43028 'enum' type even if both are compatible with the same integer type; 43029 this is what the C standard specifies. For example, 'enum {foo, 43030 bar}' is not similar to 'enum {hot, dog}'. 43031 43032 You typically use this function in code whose execution varies 43033 depending on the arguments' types. For example: 43034 43035 #define foo(x) \ 43036 ({ \ 43037 typeof (x) tmp = (x); \ 43038 if (__builtin_types_compatible_p (typeof (x), long double)) \ 43039 tmp = foo_long_double (tmp); \ 43040 else if (__builtin_types_compatible_p (typeof (x), double)) \ 43041 tmp = foo_double (tmp); \ 43042 else if (__builtin_types_compatible_p (typeof (x), float)) \ 43043 tmp = foo_float (tmp); \ 43044 else \ 43045 abort (); \ 43046 tmp; \ 43047 }) 43048 43049 _Note:_ This construct is only available for C. 43050 43051 -- Built-in Function: TYPE __builtin_call_with_static_chain (CALL_EXP, 43052 POINTER_EXP) 43053 43054 The CALL_EXP expression must be a function call, and the 43055 POINTER_EXP expression must be a pointer. The POINTER_EXP is 43056 passed to the function call in the target's static chain location. 43057 The result of builtin is the result of the function call. 43058 43059 _Note:_ This builtin is only available for C. This builtin can be 43060 used to call Go closures from C. 43061 43062 -- Built-in Function: TYPE __builtin_choose_expr (CONST_EXP, EXP1, 43063 EXP2) 43064 43065 You can use the built-in function '__builtin_choose_expr' to 43066 evaluate code depending on the value of a constant expression. 43067 This built-in function returns EXP1 if CONST_EXP, which is an 43068 integer constant expression, is nonzero. Otherwise it returns 43069 EXP2. 43070 43071 This built-in function is analogous to the '? :' operator in C, 43072 except that the expression returned has its type unaltered by 43073 promotion rules. Also, the built-in function does not evaluate the 43074 expression that is not chosen. For example, if CONST_EXP evaluates 43075 to 'true', EXP2 is not evaluated even if it has side effects. 43076 43077 This built-in function can return an lvalue if the chosen argument 43078 is an lvalue. 43079 43080 If EXP1 is returned, the return type is the same as EXP1's type. 43081 Similarly, if EXP2 is returned, its return type is the same as 43082 EXP2. 43083 43084 Example: 43085 43086 #define foo(x) \ 43087 __builtin_choose_expr ( \ 43088 __builtin_types_compatible_p (typeof (x), double), \ 43089 foo_double (x), \ 43090 __builtin_choose_expr ( \ 43091 __builtin_types_compatible_p (typeof (x), float), \ 43092 foo_float (x), \ 43093 /* The void expression results in a compile-time error \ 43094 when assigning the result to something. */ \ 43095 (void)0)) 43096 43097 _Note:_ This construct is only available for C. Furthermore, the 43098 unused expression (EXP1 or EXP2 depending on the value of 43099 CONST_EXP) may still generate syntax errors. This may change in 43100 future revisions. 43101 43102 -- Built-in Function: TYPE __builtin_tgmath (FUNCTIONS, ARGUMENTS) 43103 43104 The built-in function '__builtin_tgmath', available only for C and 43105 Objective-C, calls a function determined according to the rules of 43106 '<tgmath.h>' macros. It is intended to be used in implementations 43107 of that header, so that expansions of macros from that header only 43108 expand each of their arguments once, to avoid problems when calls 43109 to such macros are nested inside the arguments of other calls to 43110 such macros; in addition, it results in better diagnostics for 43111 invalid calls to '<tgmath.h>' macros than implementations using 43112 other GNU C language features. For example, the 'pow' type-generic 43113 macro might be defined as: 43114 43115 #define pow(a, b) __builtin_tgmath (powf, pow, powl, \ 43116 cpowf, cpow, cpowl, a, b) 43117 43118 The arguments to '__builtin_tgmath' are at least two pointers to 43119 functions, followed by the arguments to the type-generic macro 43120 (which will be passed as arguments to the selected function). All 43121 the pointers to functions must be pointers to prototyped functions, 43122 none of which may have variable arguments, and all of which must 43123 have the same number of parameters; the number of parameters of the 43124 first function determines how many arguments to '__builtin_tgmath' 43125 are interpreted as function pointers, and how many as the arguments 43126 to the called function. 43127 43128 The types of the specified functions must all be different, but 43129 related to each other in the same way as a set of functions that 43130 may be selected between by a macro in '<tgmath.h>'. This means 43131 that the functions are parameterized by a floating-point type T, 43132 different for each such function. The function return types may 43133 all be the same type, or they may be T for each function, or they 43134 may be the real type corresponding to T for each function (if some 43135 of the types T are complex). Likewise, for each parameter 43136 position, the type of the parameter in that position may always be 43137 the same type, or may be T for each function (this case must apply 43138 for at least one parameter position), or may be the real type 43139 corresponding to T for each function. 43140 43141 The standard rules for '<tgmath.h>' macros are used to find a 43142 common type U from the types of the arguments for parameters whose 43143 types vary between the functions; complex integer types (a GNU 43144 extension) are treated like '_Complex double' for this purpose (or 43145 '_Complex _Float64' if all the function return types are the same 43146 '_FloatN' or '_FloatNx' type). If the function return types vary, 43147 or are all the same integer type, the function called is the one 43148 for which T is U, and it is an error if there is no such function. 43149 If the function return types are all the same floating-point type, 43150 the type-generic macro is taken to be one of those from TS 18661 43151 that rounds the result to a narrower type; if there is a function 43152 for which T is U, it is called, and otherwise the first function, 43153 if any, for which T has at least the range and precision of U is 43154 called, and it is an error if there is no such function. 43155 43156 -- Built-in Function: TYPE __builtin_complex (REAL, IMAG) 43157 43158 The built-in function '__builtin_complex' is provided for use in 43159 implementing the ISO C11 macros 'CMPLXF', 'CMPLX' and 'CMPLXL'. 43160 REAL and IMAG must have the same type, a real binary floating-point 43161 type, and the result has the corresponding complex type with real 43162 and imaginary parts REAL and IMAG. Unlike 'REAL + I * IMAG', this 43163 works even when infinities, NaNs and negative zeros are involved. 43164 43165 -- Built-in Function: int __builtin_constant_p (EXP) 43166 You can use the built-in function '__builtin_constant_p' to 43167 determine if a value is known to be constant at compile time and 43168 hence that GCC can perform constant-folding on expressions 43169 involving that value. The argument of the function is the value to 43170 test. The function returns the integer 1 if the argument is known 43171 to be a compile-time constant and 0 if it is not known to be a 43172 compile-time constant. A return of 0 does not indicate that the 43173 value is _not_ a constant, but merely that GCC cannot prove it is a 43174 constant with the specified value of the '-O' option. 43175 43176 You typically use this function in an embedded application where 43177 memory is a critical resource. If you have some complex 43178 calculation, you may want it to be folded if it involves constants, 43179 but need to call a function if it does not. For example: 43180 43181 #define Scale_Value(X) \ 43182 (__builtin_constant_p (X) \ 43183 ? ((X) * SCALE + OFFSET) : Scale (X)) 43184 43185 You may use this built-in function in either a macro or an inline 43186 function. However, if you use it in an inlined function and pass 43187 an argument of the function as the argument to the built-in, GCC 43188 never returns 1 when you call the inline function with a string 43189 constant or compound literal (*note Compound Literals::) and does 43190 not return 1 when you pass a constant numeric value to the inline 43191 function unless you specify the '-O' option. 43192 43193 You may also use '__builtin_constant_p' in initializers for static 43194 data. For instance, you can write 43195 43196 static const int table[] = { 43197 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1, 43198 /* ... */ 43199 }; 43200 43201 This is an acceptable initializer even if EXPRESSION is not a 43202 constant expression, including the case where 43203 '__builtin_constant_p' returns 1 because EXPRESSION can be folded 43204 to a constant but EXPRESSION contains operands that are not 43205 otherwise permitted in a static initializer (for example, '0 && foo 43206 ()'). GCC must be more conservative about evaluating the built-in 43207 in this case, because it has no opportunity to perform 43208 optimization. 43209 43210 -- Built-in Function: bool __builtin_is_constant_evaluated (void) 43211 The '__builtin_is_constant_evaluated' function is available only in 43212 C++. The built-in is intended to be used by implementations of the 43213 'std::is_constant_evaluated' C++ function. Programs should make 43214 use of the latter function rather than invoking the built-in 43215 directly. 43216 43217 The main use case of the built-in is to determine whether a 43218 'constexpr' function is being called in a 'constexpr' context. A 43219 call to the function evaluates to a core constant expression with 43220 the value 'true' if and only if it occurs within the evaluation of 43221 an expression or conversion that is manifestly constant-evaluated 43222 as defined in the C++ standard. Manifestly constant-evaluated 43223 contexts include constant-expressions, the conditions of 'constexpr 43224 if' statements, constraint-expressions, and initializers of 43225 variables usable in constant expressions. For more details refer 43226 to the latest revision of the C++ standard. 43227 43228 -- Built-in Function: long __builtin_expect (long EXP, long C) 43229 You may use '__builtin_expect' to provide the compiler with branch 43230 prediction information. In general, you should prefer to use 43231 actual profile feedback for this ('-fprofile-arcs'), as programmers 43232 are notoriously bad at predicting how their programs actually 43233 perform. However, there are applications in which this data is 43234 hard to collect. 43235 43236 The return value is the value of EXP, which should be an integral 43237 expression. The semantics of the built-in are that it is expected 43238 that EXP == C. For example: 43239 43240 if (__builtin_expect (x, 0)) 43241 foo (); 43242 43243 indicates that we do not expect to call 'foo', since we expect 'x' 43244 to be zero. Since you are limited to integral expressions for EXP, 43245 you should use constructions such as 43246 43247 if (__builtin_expect (ptr != NULL, 1)) 43248 foo (*ptr); 43249 43250 when testing pointer or floating-point values. 43251 43252 For the purposes of branch prediction optimizations, the 43253 probability that a '__builtin_expect' expression is 'true' is 43254 controlled by GCC's 'builtin-expect-probability' parameter, which 43255 defaults to 90%. 43256 43257 You can also use '__builtin_expect_with_probability' to explicitly 43258 assign a probability value to individual expressions. If the 43259 built-in is used in a loop construct, the provided probability will 43260 influence the expected number of iterations made by loop 43261 optimizations. 43262 43263 -- Built-in Function: long __builtin_expect_with_probability 43264 (long EXP, long C, double PROBABILITY) 43265 43266 This function has the same semantics as '__builtin_expect', but the 43267 caller provides the expected probability that EXP == C. The last 43268 argument, PROBABILITY, is a floating-point value in the range 0.0 43269 to 1.0, inclusive. The PROBABILITY argument must be constant 43270 floating-point expression. 43271 43272 -- Built-in Function: void __builtin_trap (void) 43273 This function causes the program to exit abnormally. GCC 43274 implements this function by using a target-dependent mechanism 43275 (such as intentionally executing an illegal instruction) or by 43276 calling 'abort'. The mechanism used may vary from release to 43277 release so you should not rely on any particular implementation. 43278 43279 -- Built-in Function: void __builtin_unreachable (void) 43280 If control flow reaches the point of the '__builtin_unreachable', 43281 the program is undefined. It is useful in situations where the 43282 compiler cannot deduce the unreachability of the code. 43283 43284 One such case is immediately following an 'asm' statement that 43285 either never terminates, or one that transfers control elsewhere 43286 and never returns. In this example, without the 43287 '__builtin_unreachable', GCC issues a warning that control reaches 43288 the end of a non-void function. It also generates code to return 43289 after the 'asm'. 43290 43291 int f (int c, int v) 43292 { 43293 if (c) 43294 { 43295 return v; 43296 } 43297 else 43298 { 43299 asm("jmp error_handler"); 43300 __builtin_unreachable (); 43301 } 43302 } 43303 43304 Because the 'asm' statement unconditionally transfers control out 43305 of the function, control never reaches the end of the function 43306 body. The '__builtin_unreachable' is in fact unreachable and 43307 communicates this fact to the compiler. 43308 43309 Another use for '__builtin_unreachable' is following a call a 43310 function that never returns but that is not declared 43311 '__attribute__((noreturn))', as in this example: 43312 43313 void function_that_never_returns (void); 43314 43315 int g (int c) 43316 { 43317 if (c) 43318 { 43319 return 1; 43320 } 43321 else 43322 { 43323 function_that_never_returns (); 43324 __builtin_unreachable (); 43325 } 43326 } 43327 43328 -- Built-in Function: void * __builtin_assume_aligned (const void *EXP, 43329 size_t ALIGN, ...) 43330 This function returns its first argument, and allows the compiler 43331 to assume that the returned pointer is at least ALIGN bytes 43332 aligned. This built-in can have either two or three arguments, if 43333 it has three, the third argument should have integer type, and if 43334 it is nonzero means misalignment offset. For example: 43335 43336 void *x = __builtin_assume_aligned (arg, 16); 43337 43338 means that the compiler can assume 'x', set to 'arg', is at least 43339 16-byte aligned, while: 43340 43341 void *x = __builtin_assume_aligned (arg, 32, 8); 43342 43343 means that the compiler can assume for 'x', set to 'arg', that 43344 '(char *) x - 8' is 32-byte aligned. 43345 43346 -- Built-in Function: int __builtin_LINE () 43347 This function is the equivalent of the preprocessor '__LINE__' 43348 macro and returns a constant integer expression that evaluates to 43349 the line number of the invocation of the built-in. When used as a 43350 C++ default argument for a function F, it returns the line number 43351 of the call to F. 43352 43353 -- Built-in Function: const char * __builtin_FUNCTION () 43354 This function is the equivalent of the '__FUNCTION__' symbol and 43355 returns an address constant pointing to the name of the function 43356 from which the built-in was invoked, or the empty string if the 43357 invocation is not at function scope. When used as a C++ default 43358 argument for a function F, it returns the name of F's caller or the 43359 empty string if the call was not made at function scope. 43360 43361 -- Built-in Function: const char * __builtin_FILE () 43362 This function is the equivalent of the preprocessor '__FILE__' 43363 macro and returns an address constant pointing to the file name 43364 containing the invocation of the built-in, or the empty string if 43365 the invocation is not at function scope. When used as a C++ 43366 default argument for a function F, it returns the file name of the 43367 call to F or the empty string if the call was not made at function 43368 scope. 43369 43370 For example, in the following, each call to function 'foo' will 43371 print a line similar to '"file.c:123: foo: message"' with the name 43372 of the file and the line number of the 'printf' call, the name of 43373 the function 'foo', followed by the word 'message'. 43374 43375 const char* 43376 function (const char *func = __builtin_FUNCTION ()) 43377 { 43378 return func; 43379 } 43380 43381 void foo (void) 43382 { 43383 printf ("%s:%i: %s: message\n", file (), line (), function ()); 43384 } 43385 43386 -- Built-in Function: void __builtin___clear_cache (void *BEGIN, void 43387 *END) 43388 This function is used to flush the processor's instruction cache 43389 for the region of memory between BEGIN inclusive and END exclusive. 43390 Some targets require that the instruction cache be flushed, after 43391 modifying memory containing code, in order to obtain deterministic 43392 behavior. 43393 43394 If the target does not require instruction cache flushes, 43395 '__builtin___clear_cache' has no effect. Otherwise either 43396 instructions are emitted in-line to clear the instruction cache or 43397 a call to the '__clear_cache' function in libgcc is made. 43398 43399 -- Built-in Function: void __builtin_prefetch (const void *ADDR, ...) 43400 This function is used to minimize cache-miss latency by moving data 43401 into a cache before it is accessed. You can insert calls to 43402 '__builtin_prefetch' into code for which you know addresses of data 43403 in memory that is likely to be accessed soon. If the target 43404 supports them, data prefetch instructions are generated. If the 43405 prefetch is done early enough before the access then the data will 43406 be in the cache by the time it is accessed. 43407 43408 The value of ADDR is the address of the memory to prefetch. There 43409 are two optional arguments, RW and LOCALITY. The value of RW is a 43410 compile-time constant one or zero; one means that the prefetch is 43411 preparing for a write to the memory address and zero, the default, 43412 means that the prefetch is preparing for a read. The value 43413 LOCALITY must be a compile-time constant integer between zero and 43414 three. A value of zero means that the data has no temporal 43415 locality, so it need not be left in the cache after the access. A 43416 value of three means that the data has a high degree of temporal 43417 locality and should be left in all levels of cache possible. 43418 Values of one and two mean, respectively, a low or moderate degree 43419 of temporal locality. The default is three. 43420 43421 for (i = 0; i < n; i++) 43422 { 43423 a[i] = a[i] + b[i]; 43424 __builtin_prefetch (&a[i+j], 1, 1); 43425 __builtin_prefetch (&b[i+j], 0, 1); 43426 /* ... */ 43427 } 43428 43429 Data prefetch does not generate faults if ADDR is invalid, but the 43430 address expression itself must be valid. For example, a prefetch 43431 of 'p->next' does not fault if 'p->next' is not a valid address, 43432 but evaluation faults if 'p' is not a valid address. 43433 43434 If the target does not support data prefetch, the address 43435 expression is evaluated if it includes side effects but no other 43436 code is generated and GCC does not issue a warning. 43437 43438 -- Built-in Function: size_t __builtin_object_size (const void * PTR, 43439 int TYPE) 43440 Returns the size of an object pointed to by PTR. *Note Object Size 43441 Checking::, for a detailed description of the function. 43442 43443 -- Built-in Function: double __builtin_huge_val (void) 43444 Returns a positive infinity, if supported by the floating-point 43445 format, else 'DBL_MAX'. This function is suitable for implementing 43446 the ISO C macro 'HUGE_VAL'. 43447 43448 -- Built-in Function: float __builtin_huge_valf (void) 43449 Similar to '__builtin_huge_val', except the return type is 'float'. 43450 43451 -- Built-in Function: long double __builtin_huge_vall (void) 43452 Similar to '__builtin_huge_val', except the return type is 'long 43453 double'. 43454 43455 -- Built-in Function: _FloatN __builtin_huge_valfN (void) 43456 Similar to '__builtin_huge_val', except the return type is 43457 '_FloatN'. 43458 43459 -- Built-in Function: _FloatNx __builtin_huge_valfNx (void) 43460 Similar to '__builtin_huge_val', except the return type is 43461 '_FloatNx'. 43462 43463 -- Built-in Function: int __builtin_fpclassify (int, int, int, int, 43464 int, ...) 43465 This built-in implements the C99 fpclassify functionality. The 43466 first five int arguments should be the target library's notion of 43467 the possible FP classes and are used for return values. They must 43468 be constant values and they must appear in this order: 'FP_NAN', 43469 'FP_INFINITE', 'FP_NORMAL', 'FP_SUBNORMAL' and 'FP_ZERO'. The 43470 ellipsis is for exactly one floating-point value to classify. GCC 43471 treats the last argument as type-generic, which means it does not 43472 do default promotion from float to double. 43473 43474 -- Built-in Function: double __builtin_inf (void) 43475 Similar to '__builtin_huge_val', except a warning is generated if 43476 the target floating-point format does not support infinities. 43477 43478 -- Built-in Function: _Decimal32 __builtin_infd32 (void) 43479 Similar to '__builtin_inf', except the return type is '_Decimal32'. 43480 43481 -- Built-in Function: _Decimal64 __builtin_infd64 (void) 43482 Similar to '__builtin_inf', except the return type is '_Decimal64'. 43483 43484 -- Built-in Function: _Decimal128 __builtin_infd128 (void) 43485 Similar to '__builtin_inf', except the return type is 43486 '_Decimal128'. 43487 43488 -- Built-in Function: float __builtin_inff (void) 43489 Similar to '__builtin_inf', except the return type is 'float'. 43490 This function is suitable for implementing the ISO C99 macro 43491 'INFINITY'. 43492 43493 -- Built-in Function: long double __builtin_infl (void) 43494 Similar to '__builtin_inf', except the return type is 'long 43495 double'. 43496 43497 -- Built-in Function: _FloatN __builtin_inffN (void) 43498 Similar to '__builtin_inf', except the return type is '_FloatN'. 43499 43500 -- Built-in Function: _FloatN __builtin_inffNx (void) 43501 Similar to '__builtin_inf', except the return type is '_FloatNx'. 43502 43503 -- Built-in Function: int __builtin_isinf_sign (...) 43504 Similar to 'isinf', except the return value is -1 for an argument 43505 of '-Inf' and 1 for an argument of '+Inf'. Note while the 43506 parameter list is an ellipsis, this function only accepts exactly 43507 one floating-point argument. GCC treats this parameter as 43508 type-generic, which means it does not do default promotion from 43509 float to double. 43510 43511 -- Built-in Function: double __builtin_nan (const char *str) 43512 This is an implementation of the ISO C99 function 'nan'. 43513 43514 Since ISO C99 defines this function in terms of 'strtod', which we 43515 do not implement, a description of the parsing is in order. The 43516 string is parsed as by 'strtol'; that is, the base is recognized by 43517 leading '0' or '0x' prefixes. The number parsed is placed in the 43518 significand such that the least significant bit of the number is at 43519 the least significant bit of the significand. The number is 43520 truncated to fit the significand field provided. The significand 43521 is forced to be a quiet NaN. 43522 43523 This function, if given a string literal all of which would have 43524 been consumed by 'strtol', is evaluated early enough that it is 43525 considered a compile-time constant. 43526 43527 -- Built-in Function: _Decimal32 __builtin_nand32 (const char *str) 43528 Similar to '__builtin_nan', except the return type is '_Decimal32'. 43529 43530 -- Built-in Function: _Decimal64 __builtin_nand64 (const char *str) 43531 Similar to '__builtin_nan', except the return type is '_Decimal64'. 43532 43533 -- Built-in Function: _Decimal128 __builtin_nand128 (const char *str) 43534 Similar to '__builtin_nan', except the return type is 43535 '_Decimal128'. 43536 43537 -- Built-in Function: float __builtin_nanf (const char *str) 43538 Similar to '__builtin_nan', except the return type is 'float'. 43539 43540 -- Built-in Function: long double __builtin_nanl (const char *str) 43541 Similar to '__builtin_nan', except the return type is 'long 43542 double'. 43543 43544 -- Built-in Function: _FloatN __builtin_nanfN (const char *str) 43545 Similar to '__builtin_nan', except the return type is '_FloatN'. 43546 43547 -- Built-in Function: _FloatNx __builtin_nanfNx (const char *str) 43548 Similar to '__builtin_nan', except the return type is '_FloatNx'. 43549 43550 -- Built-in Function: double __builtin_nans (const char *str) 43551 Similar to '__builtin_nan', except the significand is forced to be 43552 a signaling NaN. The 'nans' function is proposed by WG14 N965. 43553 43554 -- Built-in Function: float __builtin_nansf (const char *str) 43555 Similar to '__builtin_nans', except the return type is 'float'. 43556 43557 -- Built-in Function: long double __builtin_nansl (const char *str) 43558 Similar to '__builtin_nans', except the return type is 'long 43559 double'. 43560 43561 -- Built-in Function: _FloatN __builtin_nansfN (const char *str) 43562 Similar to '__builtin_nans', except the return type is '_FloatN'. 43563 43564 -- Built-in Function: _FloatNx __builtin_nansfNx (const char *str) 43565 Similar to '__builtin_nans', except the return type is '_FloatNx'. 43566 43567 -- Built-in Function: int __builtin_ffs (int x) 43568 Returns one plus the index of the least significant 1-bit of X, or 43569 if X is zero, returns zero. 43570 43571 -- Built-in Function: int __builtin_clz (unsigned int x) 43572 Returns the number of leading 0-bits in X, starting at the most 43573 significant bit position. If X is 0, the result is undefined. 43574 43575 -- Built-in Function: int __builtin_ctz (unsigned int x) 43576 Returns the number of trailing 0-bits in X, starting at the least 43577 significant bit position. If X is 0, the result is undefined. 43578 43579 -- Built-in Function: int __builtin_clrsb (int x) 43580 Returns the number of leading redundant sign bits in X, i.e. the 43581 number of bits following the most significant bit that are 43582 identical to it. There are no special cases for 0 or other values. 43583 43584 -- Built-in Function: int __builtin_popcount (unsigned int x) 43585 Returns the number of 1-bits in X. 43586 43587 -- Built-in Function: int __builtin_parity (unsigned int x) 43588 Returns the parity of X, i.e. the number of 1-bits in X modulo 2. 43589 43590 -- Built-in Function: int __builtin_ffsl (long) 43591 Similar to '__builtin_ffs', except the argument type is 'long'. 43592 43593 -- Built-in Function: int __builtin_clzl (unsigned long) 43594 Similar to '__builtin_clz', except the argument type is 'unsigned 43595 long'. 43596 43597 -- Built-in Function: int __builtin_ctzl (unsigned long) 43598 Similar to '__builtin_ctz', except the argument type is 'unsigned 43599 long'. 43600 43601 -- Built-in Function: int __builtin_clrsbl (long) 43602 Similar to '__builtin_clrsb', except the argument type is 'long'. 43603 43604 -- Built-in Function: int __builtin_popcountl (unsigned long) 43605 Similar to '__builtin_popcount', except the argument type is 43606 'unsigned long'. 43607 43608 -- Built-in Function: int __builtin_parityl (unsigned long) 43609 Similar to '__builtin_parity', except the argument type is 43610 'unsigned long'. 43611 43612 -- Built-in Function: int __builtin_ffsll (long long) 43613 Similar to '__builtin_ffs', except the argument type is 'long 43614 long'. 43615 43616 -- Built-in Function: int __builtin_clzll (unsigned long long) 43617 Similar to '__builtin_clz', except the argument type is 'unsigned 43618 long long'. 43619 43620 -- Built-in Function: int __builtin_ctzll (unsigned long long) 43621 Similar to '__builtin_ctz', except the argument type is 'unsigned 43622 long long'. 43623 43624 -- Built-in Function: int __builtin_clrsbll (long long) 43625 Similar to '__builtin_clrsb', except the argument type is 'long 43626 long'. 43627 43628 -- Built-in Function: int __builtin_popcountll (unsigned long long) 43629 Similar to '__builtin_popcount', except the argument type is 43630 'unsigned long long'. 43631 43632 -- Built-in Function: int __builtin_parityll (unsigned long long) 43633 Similar to '__builtin_parity', except the argument type is 43634 'unsigned long long'. 43635 43636 -- Built-in Function: double __builtin_powi (double, int) 43637 Returns the first argument raised to the power of the second. 43638 Unlike the 'pow' function no guarantees about precision and 43639 rounding are made. 43640 43641 -- Built-in Function: float __builtin_powif (float, int) 43642 Similar to '__builtin_powi', except the argument and return types 43643 are 'float'. 43644 43645 -- Built-in Function: long double __builtin_powil (long double, int) 43646 Similar to '__builtin_powi', except the argument and return types 43647 are 'long double'. 43648 43649 -- Built-in Function: uint16_t __builtin_bswap16 (uint16_t x) 43650 Returns X with the order of the bytes reversed; for example, 43651 '0xaabb' becomes '0xbbaa'. Byte here always means exactly 8 bits. 43652 43653 -- Built-in Function: uint32_t __builtin_bswap32 (uint32_t x) 43654 Similar to '__builtin_bswap16', except the argument and return 43655 types are 32 bit. 43656 43657 -- Built-in Function: uint64_t __builtin_bswap64 (uint64_t x) 43658 Similar to '__builtin_bswap32', except the argument and return 43659 types are 64 bit. 43660 43661 -- Built-in Function: Pmode __builtin_extend_pointer (void * x) 43662 On targets where the user visible pointer size is smaller than the 43663 size of an actual hardware address this function returns the 43664 extended user pointer. Targets where this is true included ILP32 43665 mode on x86_64 or Aarch64. This function is mainly useful when 43666 writing inline assembly code. 43667 43668 -- Built-in Function: int __builtin_goacc_parlevel_id (int x) 43669 Returns the openacc gang, worker or vector id depending on whether 43670 X is 0, 1 or 2. 43671 43672 -- Built-in Function: int __builtin_goacc_parlevel_size (int x) 43673 Returns the openacc gang, worker or vector size depending on 43674 whether X is 0, 1 or 2. 43675 43676 43677File: gcc.info, Node: Target Builtins, Next: Target Format Checks, Prev: Other Builtins, Up: C Extensions 43678 436796.60 Built-in Functions Specific to Particular Target Machines 43680============================================================== 43681 43682On some target machines, GCC supports many built-in functions specific 43683to those machines. Generally these generate calls to specific machine 43684instructions, but allow the compiler to schedule those calls. 43685 43686* Menu: 43687 43688* AArch64 Built-in Functions:: 43689* Alpha Built-in Functions:: 43690* Altera Nios II Built-in Functions:: 43691* ARC Built-in Functions:: 43692* ARC SIMD Built-in Functions:: 43693* ARM iWMMXt Built-in Functions:: 43694* ARM C Language Extensions (ACLE):: 43695* ARM Floating Point Status and Control Intrinsics:: 43696* ARM ARMv8-M Security Extensions:: 43697* AVR Built-in Functions:: 43698* Blackfin Built-in Functions:: 43699* BPF Built-in Functions:: 43700* FR-V Built-in Functions:: 43701* MIPS DSP Built-in Functions:: 43702* MIPS Paired-Single Support:: 43703* MIPS Loongson Built-in Functions:: 43704* MIPS SIMD Architecture (MSA) Support:: 43705* Other MIPS Built-in Functions:: 43706* MSP430 Built-in Functions:: 43707* NDS32 Built-in Functions:: 43708* picoChip Built-in Functions:: 43709* Basic PowerPC Built-in Functions:: 43710* PowerPC AltiVec/VSX Built-in Functions:: 43711* PowerPC Hardware Transactional Memory Built-in Functions:: 43712* PowerPC Atomic Memory Operation Functions:: 43713* PowerPC Matrix-Multiply Assist Built-in Functions:: 43714* RISC-V Built-in Functions:: 43715* RX Built-in Functions:: 43716* S/390 System z Built-in Functions:: 43717* SH Built-in Functions:: 43718* SPARC VIS Built-in Functions:: 43719* TI C6X Built-in Functions:: 43720* TILE-Gx Built-in Functions:: 43721* TILEPro Built-in Functions:: 43722* x86 Built-in Functions:: 43723* x86 transactional memory intrinsics:: 43724* x86 control-flow protection intrinsics:: 43725 43726 43727File: gcc.info, Node: AArch64 Built-in Functions, Next: Alpha Built-in Functions, Up: Target Builtins 43728 437296.60.1 AArch64 Built-in Functions 43730--------------------------------- 43731 43732These built-in functions are available for the AArch64 family of 43733processors. 43734 unsigned int __builtin_aarch64_get_fpcr () 43735 void __builtin_aarch64_set_fpcr (unsigned int) 43736 unsigned int __builtin_aarch64_get_fpsr () 43737 void __builtin_aarch64_set_fpsr (unsigned int) 43738 43739 43740File: gcc.info, Node: Alpha Built-in Functions, Next: Altera Nios II Built-in Functions, Prev: AArch64 Built-in Functions, Up: Target Builtins 43741 437426.60.2 Alpha Built-in Functions 43743------------------------------- 43744 43745These built-in functions are available for the Alpha family of 43746processors, depending on the command-line switches used. 43747 43748 The following built-in functions are always available. They all 43749generate the machine instruction that is part of the name. 43750 43751 long __builtin_alpha_implver (void) 43752 long __builtin_alpha_rpcc (void) 43753 long __builtin_alpha_amask (long) 43754 long __builtin_alpha_cmpbge (long, long) 43755 long __builtin_alpha_extbl (long, long) 43756 long __builtin_alpha_extwl (long, long) 43757 long __builtin_alpha_extll (long, long) 43758 long __builtin_alpha_extql (long, long) 43759 long __builtin_alpha_extwh (long, long) 43760 long __builtin_alpha_extlh (long, long) 43761 long __builtin_alpha_extqh (long, long) 43762 long __builtin_alpha_insbl (long, long) 43763 long __builtin_alpha_inswl (long, long) 43764 long __builtin_alpha_insll (long, long) 43765 long __builtin_alpha_insql (long, long) 43766 long __builtin_alpha_inswh (long, long) 43767 long __builtin_alpha_inslh (long, long) 43768 long __builtin_alpha_insqh (long, long) 43769 long __builtin_alpha_mskbl (long, long) 43770 long __builtin_alpha_mskwl (long, long) 43771 long __builtin_alpha_mskll (long, long) 43772 long __builtin_alpha_mskql (long, long) 43773 long __builtin_alpha_mskwh (long, long) 43774 long __builtin_alpha_msklh (long, long) 43775 long __builtin_alpha_mskqh (long, long) 43776 long __builtin_alpha_umulh (long, long) 43777 long __builtin_alpha_zap (long, long) 43778 long __builtin_alpha_zapnot (long, long) 43779 43780 The following built-in functions are always with '-mmax' or '-mcpu=CPU' 43781where CPU is 'pca56' or later. They all generate the machine 43782instruction that is part of the name. 43783 43784 long __builtin_alpha_pklb (long) 43785 long __builtin_alpha_pkwb (long) 43786 long __builtin_alpha_unpkbl (long) 43787 long __builtin_alpha_unpkbw (long) 43788 long __builtin_alpha_minub8 (long, long) 43789 long __builtin_alpha_minsb8 (long, long) 43790 long __builtin_alpha_minuw4 (long, long) 43791 long __builtin_alpha_minsw4 (long, long) 43792 long __builtin_alpha_maxub8 (long, long) 43793 long __builtin_alpha_maxsb8 (long, long) 43794 long __builtin_alpha_maxuw4 (long, long) 43795 long __builtin_alpha_maxsw4 (long, long) 43796 long __builtin_alpha_perr (long, long) 43797 43798 The following built-in functions are always with '-mcix' or '-mcpu=CPU' 43799where CPU is 'ev67' or later. They all generate the machine instruction 43800that is part of the name. 43801 43802 long __builtin_alpha_cttz (long) 43803 long __builtin_alpha_ctlz (long) 43804 long __builtin_alpha_ctpop (long) 43805 43806 The following built-in functions are available on systems that use the 43807OSF/1 PALcode. Normally they invoke the 'rduniq' and 'wruniq' PAL 43808calls, but when invoked with '-mtls-kernel', they invoke 'rdval' and 43809'wrval'. 43810 43811 void *__builtin_thread_pointer (void) 43812 void __builtin_set_thread_pointer (void *) 43813 43814 43815File: gcc.info, Node: Altera Nios II Built-in Functions, Next: ARC Built-in Functions, Prev: Alpha Built-in Functions, Up: Target Builtins 43816 438176.60.3 Altera Nios II Built-in Functions 43818---------------------------------------- 43819 43820These built-in functions are available for the Altera Nios II family of 43821processors. 43822 43823 The following built-in functions are always available. They all 43824generate the machine instruction that is part of the name. 43825 43826 int __builtin_ldbio (volatile const void *) 43827 int __builtin_ldbuio (volatile const void *) 43828 int __builtin_ldhio (volatile const void *) 43829 int __builtin_ldhuio (volatile const void *) 43830 int __builtin_ldwio (volatile const void *) 43831 void __builtin_stbio (volatile void *, int) 43832 void __builtin_sthio (volatile void *, int) 43833 void __builtin_stwio (volatile void *, int) 43834 void __builtin_sync (void) 43835 int __builtin_rdctl (int) 43836 int __builtin_rdprs (int, int) 43837 void __builtin_wrctl (int, int) 43838 void __builtin_flushd (volatile void *) 43839 void __builtin_flushda (volatile void *) 43840 int __builtin_wrpie (int); 43841 void __builtin_eni (int); 43842 int __builtin_ldex (volatile const void *) 43843 int __builtin_stex (volatile void *, int) 43844 int __builtin_ldsex (volatile const void *) 43845 int __builtin_stsex (volatile void *, int) 43846 43847 The following built-in functions are always available. They all 43848generate a Nios II Custom Instruction. The name of the function 43849represents the types that the function takes and returns. The letter 43850before the 'n' is the return type or void if absent. The 'n' represents 43851the first parameter to all the custom instructions, the custom 43852instruction number. The two letters after the 'n' represent the up to 43853two parameters to the function. 43854 43855 The letters represent the following data types: 43856'<no letter>' 43857 'void' for return type and no parameter for parameter types. 43858 43859'i' 43860 'int' for return type and parameter type 43861 43862'f' 43863 'float' for return type and parameter type 43864 43865'p' 43866 'void *' for return type and parameter type 43867 43868 And the function names are: 43869 void __builtin_custom_n (void) 43870 void __builtin_custom_ni (int) 43871 void __builtin_custom_nf (float) 43872 void __builtin_custom_np (void *) 43873 void __builtin_custom_nii (int, int) 43874 void __builtin_custom_nif (int, float) 43875 void __builtin_custom_nip (int, void *) 43876 void __builtin_custom_nfi (float, int) 43877 void __builtin_custom_nff (float, float) 43878 void __builtin_custom_nfp (float, void *) 43879 void __builtin_custom_npi (void *, int) 43880 void __builtin_custom_npf (void *, float) 43881 void __builtin_custom_npp (void *, void *) 43882 int __builtin_custom_in (void) 43883 int __builtin_custom_ini (int) 43884 int __builtin_custom_inf (float) 43885 int __builtin_custom_inp (void *) 43886 int __builtin_custom_inii (int, int) 43887 int __builtin_custom_inif (int, float) 43888 int __builtin_custom_inip (int, void *) 43889 int __builtin_custom_infi (float, int) 43890 int __builtin_custom_inff (float, float) 43891 int __builtin_custom_infp (float, void *) 43892 int __builtin_custom_inpi (void *, int) 43893 int __builtin_custom_inpf (void *, float) 43894 int __builtin_custom_inpp (void *, void *) 43895 float __builtin_custom_fn (void) 43896 float __builtin_custom_fni (int) 43897 float __builtin_custom_fnf (float) 43898 float __builtin_custom_fnp (void *) 43899 float __builtin_custom_fnii (int, int) 43900 float __builtin_custom_fnif (int, float) 43901 float __builtin_custom_fnip (int, void *) 43902 float __builtin_custom_fnfi (float, int) 43903 float __builtin_custom_fnff (float, float) 43904 float __builtin_custom_fnfp (float, void *) 43905 float __builtin_custom_fnpi (void *, int) 43906 float __builtin_custom_fnpf (void *, float) 43907 float __builtin_custom_fnpp (void *, void *) 43908 void * __builtin_custom_pn (void) 43909 void * __builtin_custom_pni (int) 43910 void * __builtin_custom_pnf (float) 43911 void * __builtin_custom_pnp (void *) 43912 void * __builtin_custom_pnii (int, int) 43913 void * __builtin_custom_pnif (int, float) 43914 void * __builtin_custom_pnip (int, void *) 43915 void * __builtin_custom_pnfi (float, int) 43916 void * __builtin_custom_pnff (float, float) 43917 void * __builtin_custom_pnfp (float, void *) 43918 void * __builtin_custom_pnpi (void *, int) 43919 void * __builtin_custom_pnpf (void *, float) 43920 void * __builtin_custom_pnpp (void *, void *) 43921 43922 43923File: gcc.info, Node: ARC Built-in Functions, Next: ARC SIMD Built-in Functions, Prev: Altera Nios II Built-in Functions, Up: Target Builtins 43924 439256.60.4 ARC Built-in Functions 43926----------------------------- 43927 43928The following built-in functions are provided for ARC targets. The 43929built-ins generate the corresponding assembly instructions. In the 43930examples given below, the generated code often requires an operand or 43931result to be in a register. Where necessary further code will be 43932generated to ensure this is true, but for brevity this is not described 43933in each case. 43934 43935 _Note:_ Using a built-in to generate an instruction not supported by a 43936target may cause problems. At present the compiler is not guaranteed to 43937detect such misuse, and as a result an internal compiler error may be 43938generated. 43939 43940 -- Built-in Function: int __builtin_arc_aligned (void *VAL, int 43941 ALIGNVAL) 43942 Return 1 if VAL is known to have the byte alignment given by 43943 ALIGNVAL, otherwise return 0. Note that this is different from 43944 __alignof__(*(char *)VAL) >= alignval 43945 because __alignof__ sees only the type of the dereference, whereas 43946 __builtin_arc_align uses alignment information from the pointer as 43947 well as from the pointed-to type. The information available will 43948 depend on optimization level. 43949 43950 -- Built-in Function: void __builtin_arc_brk (void) 43951 Generates 43952 brk 43953 43954 -- Built-in Function: unsigned int __builtin_arc_core_read (unsigned 43955 int REGNO) 43956 The operand is the number of a register to be read. Generates: 43957 mov DEST, rREGNO 43958 where the value in DEST will be the result returned from the 43959 built-in. 43960 43961 -- Built-in Function: void __builtin_arc_core_write (unsigned int 43962 REGNO, unsigned int VAL) 43963 The first operand is the number of a register to be written, the 43964 second operand is a compile time constant to write into that 43965 register. Generates: 43966 mov rREGNO, VAL 43967 43968 -- Built-in Function: int __builtin_arc_divaw (int A, int B) 43969 Only available if either '-mcpu=ARC700' or '-meA' is set. 43970 Generates: 43971 divaw DEST, A, B 43972 where the value in DEST will be the result returned from the 43973 built-in. 43974 43975 -- Built-in Function: void __builtin_arc_flag (unsigned int A) 43976 Generates 43977 flag A 43978 43979 -- Built-in Function: unsigned int __builtin_arc_lr (unsigned int AUXR) 43980 The operand, AUXV, is the address of an auxiliary register and must 43981 be a compile time constant. Generates: 43982 lr DEST, [AUXR] 43983 Where the value in DEST will be the result returned from the 43984 built-in. 43985 43986 -- Built-in Function: void __builtin_arc_mul64 (int A, int B) 43987 Only available with '-mmul64'. Generates: 43988 mul64 A, B 43989 43990 -- Built-in Function: void __builtin_arc_mulu64 (unsigned int A, 43991 unsigned int B) 43992 Only available with '-mmul64'. Generates: 43993 mulu64 A, B 43994 43995 -- Built-in Function: void __builtin_arc_nop (void) 43996 Generates: 43997 nop 43998 43999 -- Built-in Function: int __builtin_arc_norm (int SRC) 44000 Only valid if the 'norm' instruction is available through the 44001 '-mnorm' option or by default with '-mcpu=ARC700'. Generates: 44002 norm DEST, SRC 44003 Where the value in DEST will be the result returned from the 44004 built-in. 44005 44006 -- Built-in Function: short int __builtin_arc_normw (short int SRC) 44007 Only valid if the 'normw' instruction is available through the 44008 '-mnorm' option or by default with '-mcpu=ARC700'. Generates: 44009 normw DEST, SRC 44010 Where the value in DEST will be the result returned from the 44011 built-in. 44012 44013 -- Built-in Function: void __builtin_arc_rtie (void) 44014 Generates: 44015 rtie 44016 44017 -- Built-in Function: void __builtin_arc_sleep (int A 44018 Generates: 44019 sleep A 44020 44021 -- Built-in Function: void __builtin_arc_sr (unsigned int AUXR, 44022 unsigned int VAL) 44023 The first argument, AUXV, is the address of an auxiliary register, 44024 the second argument, VAL, is a compile time constant to be written 44025 to the register. Generates: 44026 sr AUXR, [VAL] 44027 44028 -- Built-in Function: int __builtin_arc_swap (int SRC) 44029 Only valid with '-mswap'. Generates: 44030 swap DEST, SRC 44031 Where the value in DEST will be the result returned from the 44032 built-in. 44033 44034 -- Built-in Function: void __builtin_arc_swi (void) 44035 Generates: 44036 swi 44037 44038 -- Built-in Function: void __builtin_arc_sync (void) 44039 Only available with '-mcpu=ARC700'. Generates: 44040 sync 44041 44042 -- Built-in Function: void __builtin_arc_trap_s (unsigned int C) 44043 Only available with '-mcpu=ARC700'. Generates: 44044 trap_s C 44045 44046 -- Built-in Function: void __builtin_arc_unimp_s (void) 44047 Only available with '-mcpu=ARC700'. Generates: 44048 unimp_s 44049 44050 The instructions generated by the following builtins are not considered 44051as candidates for scheduling. They are not moved around by the compiler 44052during scheduling, and thus can be expected to appear where they are put 44053in the C code: 44054 __builtin_arc_brk() 44055 __builtin_arc_core_read() 44056 __builtin_arc_core_write() 44057 __builtin_arc_flag() 44058 __builtin_arc_lr() 44059 __builtin_arc_sleep() 44060 __builtin_arc_sr() 44061 __builtin_arc_swi() 44062 44063 44064File: gcc.info, Node: ARC SIMD Built-in Functions, Next: ARM iWMMXt Built-in Functions, Prev: ARC Built-in Functions, Up: Target Builtins 44065 440666.60.5 ARC SIMD Built-in Functions 44067---------------------------------- 44068 44069SIMD builtins provided by the compiler can be used to generate the 44070vector instructions. This section describes the available builtins and 44071their usage in programs. With the '-msimd' option, the compiler 44072provides 128-bit vector types, which can be specified using the 44073'vector_size' attribute. The header file 'arc-simd.h' can be included 44074to use the following predefined types: 44075 typedef int __v4si __attribute__((vector_size(16))); 44076 typedef short __v8hi __attribute__((vector_size(16))); 44077 44078 These types can be used to define 128-bit variables. The built-in 44079functions listed in the following section can be used on these variables 44080to generate the vector operations. 44081 44082 For all builtins, '__builtin_arc_SOMEINSN', the header file 44083'arc-simd.h' also provides equivalent macros called '_SOMEINSN' that can 44084be used for programming ease and improved readability. The following 44085macros for DMA control are also provided: 44086 #define _setup_dma_in_channel_reg _vdiwr 44087 #define _setup_dma_out_channel_reg _vdowr 44088 44089 The following is a complete list of all the SIMD built-ins provided for 44090ARC, grouped by calling signature. 44091 44092 The following take two '__v8hi' arguments and return a '__v8hi' result: 44093 __v8hi __builtin_arc_vaddaw (__v8hi, __v8hi) 44094 __v8hi __builtin_arc_vaddw (__v8hi, __v8hi) 44095 __v8hi __builtin_arc_vand (__v8hi, __v8hi) 44096 __v8hi __builtin_arc_vandaw (__v8hi, __v8hi) 44097 __v8hi __builtin_arc_vavb (__v8hi, __v8hi) 44098 __v8hi __builtin_arc_vavrb (__v8hi, __v8hi) 44099 __v8hi __builtin_arc_vbic (__v8hi, __v8hi) 44100 __v8hi __builtin_arc_vbicaw (__v8hi, __v8hi) 44101 __v8hi __builtin_arc_vdifaw (__v8hi, __v8hi) 44102 __v8hi __builtin_arc_vdifw (__v8hi, __v8hi) 44103 __v8hi __builtin_arc_veqw (__v8hi, __v8hi) 44104 __v8hi __builtin_arc_vh264f (__v8hi, __v8hi) 44105 __v8hi __builtin_arc_vh264ft (__v8hi, __v8hi) 44106 __v8hi __builtin_arc_vh264fw (__v8hi, __v8hi) 44107 __v8hi __builtin_arc_vlew (__v8hi, __v8hi) 44108 __v8hi __builtin_arc_vltw (__v8hi, __v8hi) 44109 __v8hi __builtin_arc_vmaxaw (__v8hi, __v8hi) 44110 __v8hi __builtin_arc_vmaxw (__v8hi, __v8hi) 44111 __v8hi __builtin_arc_vminaw (__v8hi, __v8hi) 44112 __v8hi __builtin_arc_vminw (__v8hi, __v8hi) 44113 __v8hi __builtin_arc_vmr1aw (__v8hi, __v8hi) 44114 __v8hi __builtin_arc_vmr1w (__v8hi, __v8hi) 44115 __v8hi __builtin_arc_vmr2aw (__v8hi, __v8hi) 44116 __v8hi __builtin_arc_vmr2w (__v8hi, __v8hi) 44117 __v8hi __builtin_arc_vmr3aw (__v8hi, __v8hi) 44118 __v8hi __builtin_arc_vmr3w (__v8hi, __v8hi) 44119 __v8hi __builtin_arc_vmr4aw (__v8hi, __v8hi) 44120 __v8hi __builtin_arc_vmr4w (__v8hi, __v8hi) 44121 __v8hi __builtin_arc_vmr5aw (__v8hi, __v8hi) 44122 __v8hi __builtin_arc_vmr5w (__v8hi, __v8hi) 44123 __v8hi __builtin_arc_vmr6aw (__v8hi, __v8hi) 44124 __v8hi __builtin_arc_vmr6w (__v8hi, __v8hi) 44125 __v8hi __builtin_arc_vmr7aw (__v8hi, __v8hi) 44126 __v8hi __builtin_arc_vmr7w (__v8hi, __v8hi) 44127 __v8hi __builtin_arc_vmrb (__v8hi, __v8hi) 44128 __v8hi __builtin_arc_vmulaw (__v8hi, __v8hi) 44129 __v8hi __builtin_arc_vmulfaw (__v8hi, __v8hi) 44130 __v8hi __builtin_arc_vmulfw (__v8hi, __v8hi) 44131 __v8hi __builtin_arc_vmulw (__v8hi, __v8hi) 44132 __v8hi __builtin_arc_vnew (__v8hi, __v8hi) 44133 __v8hi __builtin_arc_vor (__v8hi, __v8hi) 44134 __v8hi __builtin_arc_vsubaw (__v8hi, __v8hi) 44135 __v8hi __builtin_arc_vsubw (__v8hi, __v8hi) 44136 __v8hi __builtin_arc_vsummw (__v8hi, __v8hi) 44137 __v8hi __builtin_arc_vvc1f (__v8hi, __v8hi) 44138 __v8hi __builtin_arc_vvc1ft (__v8hi, __v8hi) 44139 __v8hi __builtin_arc_vxor (__v8hi, __v8hi) 44140 __v8hi __builtin_arc_vxoraw (__v8hi, __v8hi) 44141 44142 The following take one '__v8hi' and one 'int' argument and return a 44143'__v8hi' result: 44144 44145 __v8hi __builtin_arc_vbaddw (__v8hi, int) 44146 __v8hi __builtin_arc_vbmaxw (__v8hi, int) 44147 __v8hi __builtin_arc_vbminw (__v8hi, int) 44148 __v8hi __builtin_arc_vbmulaw (__v8hi, int) 44149 __v8hi __builtin_arc_vbmulfw (__v8hi, int) 44150 __v8hi __builtin_arc_vbmulw (__v8hi, int) 44151 __v8hi __builtin_arc_vbrsubw (__v8hi, int) 44152 __v8hi __builtin_arc_vbsubw (__v8hi, int) 44153 44154 The following take one '__v8hi' argument and one 'int' argument which 44155must be a 3-bit compile time constant indicating a register number 44156I0-I7. They return a '__v8hi' result. 44157 __v8hi __builtin_arc_vasrw (__v8hi, const int) 44158 __v8hi __builtin_arc_vsr8 (__v8hi, const int) 44159 __v8hi __builtin_arc_vsr8aw (__v8hi, const int) 44160 44161 The following take one '__v8hi' argument and one 'int' argument which 44162must be a 6-bit compile time constant. They return a '__v8hi' result. 44163 __v8hi __builtin_arc_vasrpwbi (__v8hi, const int) 44164 __v8hi __builtin_arc_vasrrpwbi (__v8hi, const int) 44165 __v8hi __builtin_arc_vasrrwi (__v8hi, const int) 44166 __v8hi __builtin_arc_vasrsrwi (__v8hi, const int) 44167 __v8hi __builtin_arc_vasrwi (__v8hi, const int) 44168 __v8hi __builtin_arc_vsr8awi (__v8hi, const int) 44169 __v8hi __builtin_arc_vsr8i (__v8hi, const int) 44170 44171 The following take one '__v8hi' argument and one 'int' argument which 44172must be a 8-bit compile time constant. They return a '__v8hi' result. 44173 __v8hi __builtin_arc_vd6tapf (__v8hi, const int) 44174 __v8hi __builtin_arc_vmvaw (__v8hi, const int) 44175 __v8hi __builtin_arc_vmvw (__v8hi, const int) 44176 __v8hi __builtin_arc_vmvzw (__v8hi, const int) 44177 44178 The following take two 'int' arguments, the second of which which must 44179be a 8-bit compile time constant. They return a '__v8hi' result: 44180 __v8hi __builtin_arc_vmovaw (int, const int) 44181 __v8hi __builtin_arc_vmovw (int, const int) 44182 __v8hi __builtin_arc_vmovzw (int, const int) 44183 44184 The following take a single '__v8hi' argument and return a '__v8hi' 44185result: 44186 __v8hi __builtin_arc_vabsaw (__v8hi) 44187 __v8hi __builtin_arc_vabsw (__v8hi) 44188 __v8hi __builtin_arc_vaddsuw (__v8hi) 44189 __v8hi __builtin_arc_vexch1 (__v8hi) 44190 __v8hi __builtin_arc_vexch2 (__v8hi) 44191 __v8hi __builtin_arc_vexch4 (__v8hi) 44192 __v8hi __builtin_arc_vsignw (__v8hi) 44193 __v8hi __builtin_arc_vupbaw (__v8hi) 44194 __v8hi __builtin_arc_vupbw (__v8hi) 44195 __v8hi __builtin_arc_vupsbaw (__v8hi) 44196 __v8hi __builtin_arc_vupsbw (__v8hi) 44197 44198 The following take two 'int' arguments and return no result: 44199 void __builtin_arc_vdirun (int, int) 44200 void __builtin_arc_vdorun (int, int) 44201 44202 The following take two 'int' arguments and return no result. The first 44203argument must a 3-bit compile time constant indicating one of the 44204DR0-DR7 DMA setup channels: 44205 void __builtin_arc_vdiwr (const int, int) 44206 void __builtin_arc_vdowr (const int, int) 44207 44208 The following take an 'int' argument and return no result: 44209 void __builtin_arc_vendrec (int) 44210 void __builtin_arc_vrec (int) 44211 void __builtin_arc_vrecrun (int) 44212 void __builtin_arc_vrun (int) 44213 44214 The following take a '__v8hi' argument and two 'int' arguments and 44215return a '__v8hi' result. The second argument must be a 3-bit compile 44216time constants, indicating one the registers I0-I7, and the third 44217argument must be an 8-bit compile time constant. 44218 44219 _Note:_ Although the equivalent hardware instructions do not take an 44220SIMD register as an operand, these builtins overwrite the relevant bits 44221of the '__v8hi' register provided as the first argument with the value 44222loaded from the '[Ib, u8]' location in the SDM. 44223 44224 __v8hi __builtin_arc_vld32 (__v8hi, const int, const int) 44225 __v8hi __builtin_arc_vld32wh (__v8hi, const int, const int) 44226 __v8hi __builtin_arc_vld32wl (__v8hi, const int, const int) 44227 __v8hi __builtin_arc_vld64 (__v8hi, const int, const int) 44228 44229 The following take two 'int' arguments and return a '__v8hi' result. 44230The first argument must be a 3-bit compile time constants, indicating 44231one the registers I0-I7, and the second argument must be an 8-bit 44232compile time constant. 44233 44234 __v8hi __builtin_arc_vld128 (const int, const int) 44235 __v8hi __builtin_arc_vld64w (const int, const int) 44236 44237 The following take a '__v8hi' argument and two 'int' arguments and 44238return no result. The second argument must be a 3-bit compile time 44239constants, indicating one the registers I0-I7, and the third argument 44240must be an 8-bit compile time constant. 44241 44242 void __builtin_arc_vst128 (__v8hi, const int, const int) 44243 void __builtin_arc_vst64 (__v8hi, const int, const int) 44244 44245 The following take a '__v8hi' argument and three 'int' arguments and 44246return no result. The second argument must be a 3-bit compile-time 44247constant, identifying the 16-bit sub-register to be stored, the third 44248argument must be a 3-bit compile time constants, indicating one the 44249registers I0-I7, and the fourth argument must be an 8-bit compile time 44250constant. 44251 44252 void __builtin_arc_vst16_n (__v8hi, const int, const int, const int) 44253 void __builtin_arc_vst32_n (__v8hi, const int, const int, const int) 44254 44255 44256File: gcc.info, Node: ARM iWMMXt Built-in Functions, Next: ARM C Language Extensions (ACLE), Prev: ARC SIMD Built-in Functions, Up: Target Builtins 44257 442586.60.6 ARM iWMMXt Built-in Functions 44259------------------------------------ 44260 44261These built-in functions are available for the ARM family of processors 44262when the '-mcpu=iwmmxt' switch is used: 44263 44264 typedef int v2si __attribute__ ((vector_size (8))); 44265 typedef short v4hi __attribute__ ((vector_size (8))); 44266 typedef char v8qi __attribute__ ((vector_size (8))); 44267 44268 int __builtin_arm_getwcgr0 (void) 44269 void __builtin_arm_setwcgr0 (int) 44270 int __builtin_arm_getwcgr1 (void) 44271 void __builtin_arm_setwcgr1 (int) 44272 int __builtin_arm_getwcgr2 (void) 44273 void __builtin_arm_setwcgr2 (int) 44274 int __builtin_arm_getwcgr3 (void) 44275 void __builtin_arm_setwcgr3 (int) 44276 int __builtin_arm_textrmsb (v8qi, int) 44277 int __builtin_arm_textrmsh (v4hi, int) 44278 int __builtin_arm_textrmsw (v2si, int) 44279 int __builtin_arm_textrmub (v8qi, int) 44280 int __builtin_arm_textrmuh (v4hi, int) 44281 int __builtin_arm_textrmuw (v2si, int) 44282 v8qi __builtin_arm_tinsrb (v8qi, int, int) 44283 v4hi __builtin_arm_tinsrh (v4hi, int, int) 44284 v2si __builtin_arm_tinsrw (v2si, int, int) 44285 long long __builtin_arm_tmia (long long, int, int) 44286 long long __builtin_arm_tmiabb (long long, int, int) 44287 long long __builtin_arm_tmiabt (long long, int, int) 44288 long long __builtin_arm_tmiaph (long long, int, int) 44289 long long __builtin_arm_tmiatb (long long, int, int) 44290 long long __builtin_arm_tmiatt (long long, int, int) 44291 int __builtin_arm_tmovmskb (v8qi) 44292 int __builtin_arm_tmovmskh (v4hi) 44293 int __builtin_arm_tmovmskw (v2si) 44294 long long __builtin_arm_waccb (v8qi) 44295 long long __builtin_arm_wacch (v4hi) 44296 long long __builtin_arm_waccw (v2si) 44297 v8qi __builtin_arm_waddb (v8qi, v8qi) 44298 v8qi __builtin_arm_waddbss (v8qi, v8qi) 44299 v8qi __builtin_arm_waddbus (v8qi, v8qi) 44300 v4hi __builtin_arm_waddh (v4hi, v4hi) 44301 v4hi __builtin_arm_waddhss (v4hi, v4hi) 44302 v4hi __builtin_arm_waddhus (v4hi, v4hi) 44303 v2si __builtin_arm_waddw (v2si, v2si) 44304 v2si __builtin_arm_waddwss (v2si, v2si) 44305 v2si __builtin_arm_waddwus (v2si, v2si) 44306 v8qi __builtin_arm_walign (v8qi, v8qi, int) 44307 long long __builtin_arm_wand(long long, long long) 44308 long long __builtin_arm_wandn (long long, long long) 44309 v8qi __builtin_arm_wavg2b (v8qi, v8qi) 44310 v8qi __builtin_arm_wavg2br (v8qi, v8qi) 44311 v4hi __builtin_arm_wavg2h (v4hi, v4hi) 44312 v4hi __builtin_arm_wavg2hr (v4hi, v4hi) 44313 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi) 44314 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi) 44315 v2si __builtin_arm_wcmpeqw (v2si, v2si) 44316 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi) 44317 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi) 44318 v2si __builtin_arm_wcmpgtsw (v2si, v2si) 44319 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi) 44320 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi) 44321 v2si __builtin_arm_wcmpgtuw (v2si, v2si) 44322 long long __builtin_arm_wmacs (long long, v4hi, v4hi) 44323 long long __builtin_arm_wmacsz (v4hi, v4hi) 44324 long long __builtin_arm_wmacu (long long, v4hi, v4hi) 44325 long long __builtin_arm_wmacuz (v4hi, v4hi) 44326 v4hi __builtin_arm_wmadds (v4hi, v4hi) 44327 v4hi __builtin_arm_wmaddu (v4hi, v4hi) 44328 v8qi __builtin_arm_wmaxsb (v8qi, v8qi) 44329 v4hi __builtin_arm_wmaxsh (v4hi, v4hi) 44330 v2si __builtin_arm_wmaxsw (v2si, v2si) 44331 v8qi __builtin_arm_wmaxub (v8qi, v8qi) 44332 v4hi __builtin_arm_wmaxuh (v4hi, v4hi) 44333 v2si __builtin_arm_wmaxuw (v2si, v2si) 44334 v8qi __builtin_arm_wminsb (v8qi, v8qi) 44335 v4hi __builtin_arm_wminsh (v4hi, v4hi) 44336 v2si __builtin_arm_wminsw (v2si, v2si) 44337 v8qi __builtin_arm_wminub (v8qi, v8qi) 44338 v4hi __builtin_arm_wminuh (v4hi, v4hi) 44339 v2si __builtin_arm_wminuw (v2si, v2si) 44340 v4hi __builtin_arm_wmulsm (v4hi, v4hi) 44341 v4hi __builtin_arm_wmulul (v4hi, v4hi) 44342 v4hi __builtin_arm_wmulum (v4hi, v4hi) 44343 long long __builtin_arm_wor (long long, long long) 44344 v2si __builtin_arm_wpackdss (long long, long long) 44345 v2si __builtin_arm_wpackdus (long long, long long) 44346 v8qi __builtin_arm_wpackhss (v4hi, v4hi) 44347 v8qi __builtin_arm_wpackhus (v4hi, v4hi) 44348 v4hi __builtin_arm_wpackwss (v2si, v2si) 44349 v4hi __builtin_arm_wpackwus (v2si, v2si) 44350 long long __builtin_arm_wrord (long long, long long) 44351 long long __builtin_arm_wrordi (long long, int) 44352 v4hi __builtin_arm_wrorh (v4hi, long long) 44353 v4hi __builtin_arm_wrorhi (v4hi, int) 44354 v2si __builtin_arm_wrorw (v2si, long long) 44355 v2si __builtin_arm_wrorwi (v2si, int) 44356 v2si __builtin_arm_wsadb (v2si, v8qi, v8qi) 44357 v2si __builtin_arm_wsadbz (v8qi, v8qi) 44358 v2si __builtin_arm_wsadh (v2si, v4hi, v4hi) 44359 v2si __builtin_arm_wsadhz (v4hi, v4hi) 44360 v4hi __builtin_arm_wshufh (v4hi, int) 44361 long long __builtin_arm_wslld (long long, long long) 44362 long long __builtin_arm_wslldi (long long, int) 44363 v4hi __builtin_arm_wsllh (v4hi, long long) 44364 v4hi __builtin_arm_wsllhi (v4hi, int) 44365 v2si __builtin_arm_wsllw (v2si, long long) 44366 v2si __builtin_arm_wsllwi (v2si, int) 44367 long long __builtin_arm_wsrad (long long, long long) 44368 long long __builtin_arm_wsradi (long long, int) 44369 v4hi __builtin_arm_wsrah (v4hi, long long) 44370 v4hi __builtin_arm_wsrahi (v4hi, int) 44371 v2si __builtin_arm_wsraw (v2si, long long) 44372 v2si __builtin_arm_wsrawi (v2si, int) 44373 long long __builtin_arm_wsrld (long long, long long) 44374 long long __builtin_arm_wsrldi (long long, int) 44375 v4hi __builtin_arm_wsrlh (v4hi, long long) 44376 v4hi __builtin_arm_wsrlhi (v4hi, int) 44377 v2si __builtin_arm_wsrlw (v2si, long long) 44378 v2si __builtin_arm_wsrlwi (v2si, int) 44379 v8qi __builtin_arm_wsubb (v8qi, v8qi) 44380 v8qi __builtin_arm_wsubbss (v8qi, v8qi) 44381 v8qi __builtin_arm_wsubbus (v8qi, v8qi) 44382 v4hi __builtin_arm_wsubh (v4hi, v4hi) 44383 v4hi __builtin_arm_wsubhss (v4hi, v4hi) 44384 v4hi __builtin_arm_wsubhus (v4hi, v4hi) 44385 v2si __builtin_arm_wsubw (v2si, v2si) 44386 v2si __builtin_arm_wsubwss (v2si, v2si) 44387 v2si __builtin_arm_wsubwus (v2si, v2si) 44388 v4hi __builtin_arm_wunpckehsb (v8qi) 44389 v2si __builtin_arm_wunpckehsh (v4hi) 44390 long long __builtin_arm_wunpckehsw (v2si) 44391 v4hi __builtin_arm_wunpckehub (v8qi) 44392 v2si __builtin_arm_wunpckehuh (v4hi) 44393 long long __builtin_arm_wunpckehuw (v2si) 44394 v4hi __builtin_arm_wunpckelsb (v8qi) 44395 v2si __builtin_arm_wunpckelsh (v4hi) 44396 long long __builtin_arm_wunpckelsw (v2si) 44397 v4hi __builtin_arm_wunpckelub (v8qi) 44398 v2si __builtin_arm_wunpckeluh (v4hi) 44399 long long __builtin_arm_wunpckeluw (v2si) 44400 v8qi __builtin_arm_wunpckihb (v8qi, v8qi) 44401 v4hi __builtin_arm_wunpckihh (v4hi, v4hi) 44402 v2si __builtin_arm_wunpckihw (v2si, v2si) 44403 v8qi __builtin_arm_wunpckilb (v8qi, v8qi) 44404 v4hi __builtin_arm_wunpckilh (v4hi, v4hi) 44405 v2si __builtin_arm_wunpckilw (v2si, v2si) 44406 long long __builtin_arm_wxor (long long, long long) 44407 long long __builtin_arm_wzero () 44408 44409 44410File: gcc.info, Node: ARM C Language Extensions (ACLE), Next: ARM Floating Point Status and Control Intrinsics, Prev: ARM iWMMXt Built-in Functions, Up: Target Builtins 44411 444126.60.7 ARM C Language Extensions (ACLE) 44413--------------------------------------- 44414 44415GCC implements extensions for C as described in the ARM C Language 44416Extensions (ACLE) specification, which can be found at 44417<https://developer.arm.com/documentation/ihi0053/latest/>. 44418 44419 As a part of ACLE, GCC implements extensions for Advanced SIMD as 44420described in the ARM C Language Extensions Specification. The complete 44421list of Advanced SIMD intrinsics can be found at 44422<https://developer.arm.com/documentation/ihi0073/latest/>. The built-in 44423intrinsics for the Advanced SIMD extension are available when NEON is 44424enabled. 44425 44426 Currently, ARM and AArch64 back ends do not support ACLE 2.0 fully. 44427Both back ends support CRC32 intrinsics and the ARM back end supports 44428the Coprocessor intrinsics, all from 'arm_acle.h'. The ARM back end's 4442916-bit floating-point Advanced SIMD intrinsics currently comply to ACLE 44430v1.1. AArch64's back end does not have support for 16-bit floating 44431point Advanced SIMD intrinsics yet. 44432 44433 See *note ARM Options:: and *note AArch64 Options:: for more 44434information on the availability of extensions. 44435 44436 44437File: gcc.info, Node: ARM Floating Point Status and Control Intrinsics, Next: ARM ARMv8-M Security Extensions, Prev: ARM C Language Extensions (ACLE), Up: Target Builtins 44438 444396.60.8 ARM Floating Point Status and Control Intrinsics 44440------------------------------------------------------- 44441 44442These built-in functions are available for the ARM family of processors 44443with floating-point unit. 44444 44445 unsigned int __builtin_arm_get_fpscr () 44446 void __builtin_arm_set_fpscr (unsigned int) 44447 44448 44449File: gcc.info, Node: ARM ARMv8-M Security Extensions, Next: AVR Built-in Functions, Prev: ARM Floating Point Status and Control Intrinsics, Up: Target Builtins 44450 444516.60.9 ARM ARMv8-M Security Extensions 44452-------------------------------------- 44453 44454GCC implements the ARMv8-M Security Extensions as described in the 44455ARMv8-M Security Extensions: Requirements on Development Tools 44456Engineering Specification, which can be found at 44457<https://developer.arm.com/documentation/ecm0359818/latest/>. 44458 44459 As part of the Security Extensions GCC implements two new function 44460attributes: 'cmse_nonsecure_entry' and 'cmse_nonsecure_call'. 44461 44462 As part of the Security Extensions GCC implements the intrinsics below. 44463FPTR is used here to mean any function pointer type. 44464 44465 cmse_address_info_t cmse_TT (void *) 44466 cmse_address_info_t cmse_TT_fptr (FPTR) 44467 cmse_address_info_t cmse_TTT (void *) 44468 cmse_address_info_t cmse_TTT_fptr (FPTR) 44469 cmse_address_info_t cmse_TTA (void *) 44470 cmse_address_info_t cmse_TTA_fptr (FPTR) 44471 cmse_address_info_t cmse_TTAT (void *) 44472 cmse_address_info_t cmse_TTAT_fptr (FPTR) 44473 void * cmse_check_address_range (void *, size_t, int) 44474 typeof(p) cmse_nsfptr_create (FPTR p) 44475 intptr_t cmse_is_nsfptr (FPTR) 44476 int cmse_nonsecure_caller (void) 44477 44478 44479File: gcc.info, Node: AVR Built-in Functions, Next: Blackfin Built-in Functions, Prev: ARM ARMv8-M Security Extensions, Up: Target Builtins 44480 444816.60.10 AVR Built-in Functions 44482------------------------------ 44483 44484For each built-in function for AVR, there is an equally named, uppercase 44485built-in macro defined. That way users can easily query if or if not a 44486specific built-in is implemented or not. For example, if 44487'__builtin_avr_nop' is available the macro '__BUILTIN_AVR_NOP' is 44488defined to '1' and undefined otherwise. 44489 44490'void __builtin_avr_nop (void)' 44491'void __builtin_avr_sei (void)' 44492'void __builtin_avr_cli (void)' 44493'void __builtin_avr_sleep (void)' 44494'void __builtin_avr_wdr (void)' 44495'unsigned char __builtin_avr_swap (unsigned char)' 44496'unsigned int __builtin_avr_fmul (unsigned char, unsigned char)' 44497'int __builtin_avr_fmuls (char, char)' 44498'int __builtin_avr_fmulsu (char, unsigned char)' 44499 These built-in functions map to the respective machine instruction, 44500 i.e. 'nop', 'sei', 'cli', 'sleep', 'wdr', 'swap', 'fmul', 'fmuls' 44501 resp. 'fmulsu'. The three 'fmul*' built-ins are implemented as 44502 library call if no hardware multiplier is available. 44503 44504'void __builtin_avr_delay_cycles (unsigned long ticks)' 44505 Delay execution for TICKS cycles. Note that this built-in does not 44506 take into account the effect of interrupts that might increase 44507 delay time. TICKS must be a compile-time integer constant; delays 44508 with a variable number of cycles are not supported. 44509 44510'char __builtin_avr_flash_segment (const __memx void*)' 44511 This built-in takes a byte address to the 24-bit *note address 44512 space: AVR Named Address Spaces. '__memx' and returns the number of 44513 the flash segment (the 64 KiB chunk) where the address points to. 44514 Counting starts at '0'. If the address does not point to flash 44515 memory, return '-1'. 44516 44517'uint8_t __builtin_avr_insert_bits (uint32_t map, uint8_t bits, uint8_t val)' 44518 Insert bits from BITS into VAL and return the resulting value. The 44519 nibbles of MAP determine how the insertion is performed: Let X be 44520 the N-th nibble of MAP 44521 1. If X is '0xf', then the N-th bit of VAL is returned unaltered. 44522 44523 2. If X is in the range 0...7, then the N-th result bit is set to 44524 the X-th bit of BITS 44525 44526 3. If X is in the range 8...'0xe', then the N-th result bit is 44527 undefined. 44528 44529 One typical use case for this built-in is adjusting input and 44530 output values to non-contiguous port layouts. Some examples: 44531 44532 // same as val, bits is unused 44533 __builtin_avr_insert_bits (0xffffffff, bits, val) 44534 44535 // same as bits, val is unused 44536 __builtin_avr_insert_bits (0x76543210, bits, val) 44537 44538 // same as rotating bits by 4 44539 __builtin_avr_insert_bits (0x32107654, bits, 0) 44540 44541 // high nibble of result is the high nibble of val 44542 // low nibble of result is the low nibble of bits 44543 __builtin_avr_insert_bits (0xffff3210, bits, val) 44544 44545 // reverse the bit order of bits 44546 __builtin_avr_insert_bits (0x01234567, bits, 0) 44547 44548'void __builtin_avr_nops (unsigned count)' 44549 Insert COUNT 'NOP' instructions. The number of instructions must 44550 be a compile-time integer constant. 44551 44552There are many more AVR-specific built-in functions that are used to 44553implement the ISO/IEC TR 18037 "Embedded C" fixed-point functions of 44554section 7.18a.6. You don't need to use these built-ins directly. 44555Instead, use the declarations as supplied by the 'stdfix.h' header with 44556GNU-C99: 44557 44558 #include <stdfix.h> 44559 44560 // Re-interpret the bit representation of unsigned 16-bit 44561 // integer UVAL as Q-format 0.16 value. 44562 unsigned fract get_bits (uint_ur_t uval) 44563 { 44564 return urbits (uval); 44565 } 44566 44567 44568File: gcc.info, Node: Blackfin Built-in Functions, Next: BPF Built-in Functions, Prev: AVR Built-in Functions, Up: Target Builtins 44569 445706.60.11 Blackfin Built-in Functions 44571----------------------------------- 44572 44573Currently, there are two Blackfin-specific built-in functions. These 44574are used for generating 'CSYNC' and 'SSYNC' machine insns without using 44575inline assembly; by using these built-in functions the compiler can 44576automatically add workarounds for hardware errata involving these 44577instructions. These functions are named as follows: 44578 44579 void __builtin_bfin_csync (void) 44580 void __builtin_bfin_ssync (void) 44581 44582 44583File: gcc.info, Node: BPF Built-in Functions, Next: FR-V Built-in Functions, Prev: Blackfin Built-in Functions, Up: Target Builtins 44584 445856.60.12 BPF Built-in Functions 44586------------------------------ 44587 44588The following built-in functions are available for eBPF targets. 44589 44590 -- Built-in Function: unsigned long long __builtin_bpf_load_byte 44591 (unsigned long long OFFSET) 44592 Load a byte from the 'struct sk_buff' packet data pointed by the 44593 register '%r6' and return it. 44594 44595 -- Built-in Function: unsigned long long __builtin_bpf_load_half 44596 (unsigned long long OFFSET) 44597 Load 16-bits from the 'struct sk_buff' packet data pointed by the 44598 register '%r6' and return it. 44599 44600 -- Built-in Function: unsigned long long __builtin_bpf_load_word 44601 (unsigned long long OFFSET) 44602 Load 32-bits from the 'struct sk_buff' packet data pointed by the 44603 register '%r6' and return it. 44604 44605 44606File: gcc.info, Node: FR-V Built-in Functions, Next: MIPS DSP Built-in Functions, Prev: BPF Built-in Functions, Up: Target Builtins 44607 446086.60.13 FR-V Built-in Functions 44609------------------------------- 44610 44611GCC provides many FR-V-specific built-in functions. In general, these 44612functions are intended to be compatible with those described by 'FR-V 44613Family, Softune C/C++ Compiler Manual (V6), Fujitsu Semiconductor'. The 44614two exceptions are '__MDUNPACKH' and '__MBTOHE', the GCC forms of which 44615pass 128-bit values by pointer rather than by value. 44616 44617 Most of the functions are named after specific FR-V instructions. Such 44618functions are said to be "directly mapped" and are summarized here in 44619tabular form. 44620 44621* Menu: 44622 44623* Argument Types:: 44624* Directly-mapped Integer Functions:: 44625* Directly-mapped Media Functions:: 44626* Raw read/write Functions:: 44627* Other Built-in Functions:: 44628 44629 44630File: gcc.info, Node: Argument Types, Next: Directly-mapped Integer Functions, Up: FR-V Built-in Functions 44631 446326.60.13.1 Argument Types 44633........................ 44634 44635The arguments to the built-in functions can be divided into three 44636groups: register numbers, compile-time constants and run-time values. 44637In order to make this classification clear at a glance, the arguments 44638and return values are given the following pseudo types: 44639 44640Pseudo type Real C type Constant? Description 44641'uh' 'unsigned short' No an unsigned halfword 44642'uw1' 'unsigned int' No an unsigned word 44643'sw1' 'int' No a signed word 44644'uw2' 'unsigned long long' No an unsigned doubleword 44645'sw2' 'long long' No a signed doubleword 44646'const' 'int' Yes an integer constant 44647'acc' 'int' Yes an ACC register number 44648'iacc' 'int' Yes an IACC register number 44649 44650 These pseudo types are not defined by GCC, they are simply a notational 44651convenience used in this manual. 44652 44653 Arguments of type 'uh', 'uw1', 'sw1', 'uw2' and 'sw2' are evaluated at 44654run time. They correspond to register operands in the underlying FR-V 44655instructions. 44656 44657 'const' arguments represent immediate operands in the underlying FR-V 44658instructions. They must be compile-time constants. 44659 44660 'acc' arguments are evaluated at compile time and specify the number of 44661an accumulator register. For example, an 'acc' argument of 2 selects 44662the ACC2 register. 44663 44664 'iacc' arguments are similar to 'acc' arguments but specify the number 44665of an IACC register. See *note Other Built-in Functions:: for more 44666details. 44667 44668 44669File: gcc.info, Node: Directly-mapped Integer Functions, Next: Directly-mapped Media Functions, Prev: Argument Types, Up: FR-V Built-in Functions 44670 446716.60.13.2 Directly-Mapped Integer Functions 44672........................................... 44673 44674The functions listed below map directly to FR-V I-type instructions. 44675 44676Function prototype Example usage Assembly output 44677'sw1 __ADDSS (sw1, sw1)' 'C = __ADDSS (A, B)' 'ADDSS A,B,C' 44678'sw1 __SCAN (sw1, sw1)' 'C = __SCAN (A, B)' 'SCAN A,B,C' 44679'sw1 __SCUTSS (sw1)' 'B = __SCUTSS (A)' 'SCUTSS A,B' 44680'sw1 __SLASS (sw1, sw1)' 'C = __SLASS (A, B)' 'SLASS A,B,C' 44681'void __SMASS (sw1, sw1)' '__SMASS (A, B)' 'SMASS A,B' 44682'void __SMSSS (sw1, sw1)' '__SMSSS (A, B)' 'SMSSS A,B' 44683'void __SMU (sw1, sw1)' '__SMU (A, B)' 'SMU A,B' 44684'sw2 __SMUL (sw1, sw1)' 'C = __SMUL (A, B)' 'SMUL A,B,C' 44685'sw1 __SUBSS (sw1, sw1)' 'C = __SUBSS (A, B)' 'SUBSS A,B,C' 44686'uw2 __UMUL (uw1, uw1)' 'C = __UMUL (A, B)' 'UMUL A,B,C' 44687 44688 44689File: gcc.info, Node: Directly-mapped Media Functions, Next: Raw read/write Functions, Prev: Directly-mapped Integer Functions, Up: FR-V Built-in Functions 44690 446916.60.13.3 Directly-Mapped Media Functions 44692......................................... 44693 44694The functions listed below map directly to FR-V M-type instructions. 44695 44696Function prototype Example usage Assembly output 44697'uw1 __MABSHS (sw1)' 'B = __MABSHS (A)' 'MABSHS A,B' 44698'void __MADDACCS (acc, acc)' '__MADDACCS (B, A)' 'MADDACCS A,B' 44699'sw1 __MADDHSS (sw1, sw1)' 'C = __MADDHSS (A, 'MADDHSS A,B,C' 44700 B)' 44701'uw1 __MADDHUS (uw1, uw1)' 'C = __MADDHUS (A, 'MADDHUS A,B,C' 44702 B)' 44703'uw1 __MAND (uw1, uw1)' 'C = __MAND (A, B)' 'MAND A,B,C' 44704'void __MASACCS (acc, acc)' '__MASACCS (B, A)' 'MASACCS A,B' 44705'uw1 __MAVEH (uw1, uw1)' 'C = __MAVEH (A, B)' 'MAVEH A,B,C' 44706'uw2 __MBTOH (uw1)' 'B = __MBTOH (A)' 'MBTOH A,B' 44707'void __MBTOHE (uw1 *, uw1)' '__MBTOHE (&B, A)' 'MBTOHE A,B' 44708'void __MCLRACC (acc)' '__MCLRACC (A)' 'MCLRACC A' 44709'void __MCLRACCA (void)' '__MCLRACCA ()' 'MCLRACCA' 44710'uw1 __Mcop1 (uw1, uw1)' 'C = __Mcop1 (A, B)' 'Mcop1 A,B,C' 44711'uw1 __Mcop2 (uw1, uw1)' 'C = __Mcop2 (A, B)' 'Mcop2 A,B,C' 44712'uw1 __MCPLHI (uw2, const)' 'C = __MCPLHI (A, B)' 'MCPLHI A,#B,C' 44713'uw1 __MCPLI (uw2, const)' 'C = __MCPLI (A, B)' 'MCPLI A,#B,C' 44714'void __MCPXIS (acc, sw1, '__MCPXIS (C, A, B)' 'MCPXIS A,B,C' 44715sw1)' 44716'void __MCPXIU (acc, uw1, '__MCPXIU (C, A, B)' 'MCPXIU A,B,C' 44717uw1)' 44718'void __MCPXRS (acc, sw1, '__MCPXRS (C, A, B)' 'MCPXRS A,B,C' 44719sw1)' 44720'void __MCPXRU (acc, uw1, '__MCPXRU (C, A, B)' 'MCPXRU A,B,C' 44721uw1)' 44722'uw1 __MCUT (acc, uw1)' 'C = __MCUT (A, B)' 'MCUT A,B,C' 44723'uw1 __MCUTSS (acc, sw1)' 'C = __MCUTSS (A, B)' 'MCUTSS A,B,C' 44724'void __MDADDACCS (acc, acc)' '__MDADDACCS (B, A)' 'MDADDACCS A,B' 44725'void __MDASACCS (acc, acc)' '__MDASACCS (B, A)' 'MDASACCS A,B' 44726'uw2 __MDCUTSSI (acc, const)' 'C = __MDCUTSSI (A, 'MDCUTSSI 44727 B)' A,#B,C' 44728'uw2 __MDPACKH (uw2, uw2)' 'C = __MDPACKH (A, 'MDPACKH A,B,C' 44729 B)' 44730'uw2 __MDROTLI (uw2, const)' 'C = __MDROTLI (A, 'MDROTLI 44731 B)' A,#B,C' 44732'void __MDSUBACCS (acc, acc)' '__MDSUBACCS (B, A)' 'MDSUBACCS A,B' 44733'void __MDUNPACKH (uw1 *, '__MDUNPACKH (&B, A)' 'MDUNPACKH A,B' 44734uw2)' 44735'uw2 __MEXPDHD (uw1, const)' 'C = __MEXPDHD (A, 'MEXPDHD 44736 B)' A,#B,C' 44737'uw1 __MEXPDHW (uw1, const)' 'C = __MEXPDHW (A, 'MEXPDHW 44738 B)' A,#B,C' 44739'uw1 __MHDSETH (uw1, const)' 'C = __MHDSETH (A, 'MHDSETH 44740 B)' A,#B,C' 44741'sw1 __MHDSETS (const)' 'B = __MHDSETS (A)' 'MHDSETS #A,B' 44742'uw1 __MHSETHIH (uw1, const)' 'B = __MHSETHIH (B, 'MHSETHIH #A,B' 44743 A)' 44744'sw1 __MHSETHIS (sw1, const)' 'B = __MHSETHIS (B, 'MHSETHIS #A,B' 44745 A)' 44746'uw1 __MHSETLOH (uw1, const)' 'B = __MHSETLOH (B, 'MHSETLOH #A,B' 44747 A)' 44748'sw1 __MHSETLOS (sw1, const)' 'B = __MHSETLOS (B, 'MHSETLOS #A,B' 44749 A)' 44750'uw1 __MHTOB (uw2)' 'B = __MHTOB (A)' 'MHTOB A,B' 44751'void __MMACHS (acc, sw1, '__MMACHS (C, A, B)' 'MMACHS A,B,C' 44752sw1)' 44753'void __MMACHU (acc, uw1, '__MMACHU (C, A, B)' 'MMACHU A,B,C' 44754uw1)' 44755'void __MMRDHS (acc, sw1, '__MMRDHS (C, A, B)' 'MMRDHS A,B,C' 44756sw1)' 44757'void __MMRDHU (acc, uw1, '__MMRDHU (C, A, B)' 'MMRDHU A,B,C' 44758uw1)' 44759'void __MMULHS (acc, sw1, '__MMULHS (C, A, B)' 'MMULHS A,B,C' 44760sw1)' 44761'void __MMULHU (acc, uw1, '__MMULHU (C, A, B)' 'MMULHU A,B,C' 44762uw1)' 44763'void __MMULXHS (acc, sw1, '__MMULXHS (C, A, B)' 'MMULXHS A,B,C' 44764sw1)' 44765'void __MMULXHU (acc, uw1, '__MMULXHU (C, A, B)' 'MMULXHU A,B,C' 44766uw1)' 44767'uw1 __MNOT (uw1)' 'B = __MNOT (A)' 'MNOT A,B' 44768'uw1 __MOR (uw1, uw1)' 'C = __MOR (A, B)' 'MOR A,B,C' 44769'uw1 __MPACKH (uh, uh)' 'C = __MPACKH (A, B)' 'MPACKH A,B,C' 44770'sw2 __MQADDHSS (sw2, sw2)' 'C = __MQADDHSS (A, 'MQADDHSS 44771 B)' A,B,C' 44772'uw2 __MQADDHUS (uw2, uw2)' 'C = __MQADDHUS (A, 'MQADDHUS 44773 B)' A,B,C' 44774'void __MQCPXIS (acc, sw2, '__MQCPXIS (C, A, B)' 'MQCPXIS A,B,C' 44775sw2)' 44776'void __MQCPXIU (acc, uw2, '__MQCPXIU (C, A, B)' 'MQCPXIU A,B,C' 44777uw2)' 44778'void __MQCPXRS (acc, sw2, '__MQCPXRS (C, A, B)' 'MQCPXRS A,B,C' 44779sw2)' 44780'void __MQCPXRU (acc, uw2, '__MQCPXRU (C, A, B)' 'MQCPXRU A,B,C' 44781uw2)' 44782'sw2 __MQLCLRHS (sw2, sw2)' 'C = __MQLCLRHS (A, 'MQLCLRHS 44783 B)' A,B,C' 44784'sw2 __MQLMTHS (sw2, sw2)' 'C = __MQLMTHS (A, 'MQLMTHS A,B,C' 44785 B)' 44786'void __MQMACHS (acc, sw2, '__MQMACHS (C, A, B)' 'MQMACHS A,B,C' 44787sw2)' 44788'void __MQMACHU (acc, uw2, '__MQMACHU (C, A, B)' 'MQMACHU A,B,C' 44789uw2)' 44790'void __MQMACXHS (acc, sw2, '__MQMACXHS (C, A, 'MQMACXHS 44791sw2)' B)' A,B,C' 44792'void __MQMULHS (acc, sw2, '__MQMULHS (C, A, B)' 'MQMULHS A,B,C' 44793sw2)' 44794'void __MQMULHU (acc, uw2, '__MQMULHU (C, A, B)' 'MQMULHU A,B,C' 44795uw2)' 44796'void __MQMULXHS (acc, sw2, '__MQMULXHS (C, A, 'MQMULXHS 44797sw2)' B)' A,B,C' 44798'void __MQMULXHU (acc, uw2, '__MQMULXHU (C, A, 'MQMULXHU 44799uw2)' B)' A,B,C' 44800'sw2 __MQSATHS (sw2, sw2)' 'C = __MQSATHS (A, 'MQSATHS A,B,C' 44801 B)' 44802'uw2 __MQSLLHI (uw2, int)' 'C = __MQSLLHI (A, 'MQSLLHI A,B,C' 44803 B)' 44804'sw2 __MQSRAHI (sw2, int)' 'C = __MQSRAHI (A, 'MQSRAHI A,B,C' 44805 B)' 44806'sw2 __MQSUBHSS (sw2, sw2)' 'C = __MQSUBHSS (A, 'MQSUBHSS 44807 B)' A,B,C' 44808'uw2 __MQSUBHUS (uw2, uw2)' 'C = __MQSUBHUS (A, 'MQSUBHUS 44809 B)' A,B,C' 44810'void __MQXMACHS (acc, sw2, '__MQXMACHS (C, A, 'MQXMACHS 44811sw2)' B)' A,B,C' 44812'void __MQXMACXHS (acc, sw2, '__MQXMACXHS (C, A, 'MQXMACXHS 44813sw2)' B)' A,B,C' 44814'uw1 __MRDACC (acc)' 'B = __MRDACC (A)' 'MRDACC A,B' 44815'uw1 __MRDACCG (acc)' 'B = __MRDACCG (A)' 'MRDACCG A,B' 44816'uw1 __MROTLI (uw1, const)' 'C = __MROTLI (A, B)' 'MROTLI A,#B,C' 44817'uw1 __MROTRI (uw1, const)' 'C = __MROTRI (A, B)' 'MROTRI A,#B,C' 44818'sw1 __MSATHS (sw1, sw1)' 'C = __MSATHS (A, B)' 'MSATHS A,B,C' 44819'uw1 __MSATHU (uw1, uw1)' 'C = __MSATHU (A, B)' 'MSATHU A,B,C' 44820'uw1 __MSLLHI (uw1, const)' 'C = __MSLLHI (A, B)' 'MSLLHI A,#B,C' 44821'sw1 __MSRAHI (sw1, const)' 'C = __MSRAHI (A, B)' 'MSRAHI A,#B,C' 44822'uw1 __MSRLHI (uw1, const)' 'C = __MSRLHI (A, B)' 'MSRLHI A,#B,C' 44823'void __MSUBACCS (acc, acc)' '__MSUBACCS (B, A)' 'MSUBACCS A,B' 44824'sw1 __MSUBHSS (sw1, sw1)' 'C = __MSUBHSS (A, 'MSUBHSS A,B,C' 44825 B)' 44826'uw1 __MSUBHUS (uw1, uw1)' 'C = __MSUBHUS (A, 'MSUBHUS A,B,C' 44827 B)' 44828'void __MTRAP (void)' '__MTRAP ()' 'MTRAP' 44829'uw2 __MUNPACKH (uw1)' 'B = __MUNPACKH (A)' 'MUNPACKH A,B' 44830'uw1 __MWCUT (uw2, uw1)' 'C = __MWCUT (A, B)' 'MWCUT A,B,C' 44831'void __MWTACC (acc, uw1)' '__MWTACC (B, A)' 'MWTACC A,B' 44832'void __MWTACCG (acc, uw1)' '__MWTACCG (B, A)' 'MWTACCG A,B' 44833'uw1 __MXOR (uw1, uw1)' 'C = __MXOR (A, B)' 'MXOR A,B,C' 44834 44835 44836File: gcc.info, Node: Raw read/write Functions, Next: Other Built-in Functions, Prev: Directly-mapped Media Functions, Up: FR-V Built-in Functions 44837 448386.60.13.4 Raw Read/Write Functions 44839.................................. 44840 44841This sections describes built-in functions related to read and write 44842instructions to access memory. These functions generate 'membar' 44843instructions to flush the I/O load and stores where appropriate, as 44844described in Fujitsu's manual described above. 44845 44846'unsigned char __builtin_read8 (void *DATA)' 44847'unsigned short __builtin_read16 (void *DATA)' 44848'unsigned long __builtin_read32 (void *DATA)' 44849'unsigned long long __builtin_read64 (void *DATA)' 44850 44851'void __builtin_write8 (void *DATA, unsigned char DATUM)' 44852'void __builtin_write16 (void *DATA, unsigned short DATUM)' 44853'void __builtin_write32 (void *DATA, unsigned long DATUM)' 44854'void __builtin_write64 (void *DATA, unsigned long long DATUM)' 44855 44856 44857File: gcc.info, Node: Other Built-in Functions, Prev: Raw read/write Functions, Up: FR-V Built-in Functions 44858 448596.60.13.5 Other Built-in Functions 44860.................................. 44861 44862This section describes built-in functions that are not named after a 44863specific FR-V instruction. 44864 44865'sw2 __IACCreadll (iacc REG)' 44866 Return the full 64-bit value of IACC0. The REG argument is 44867 reserved for future expansion and must be 0. 44868 44869'sw1 __IACCreadl (iacc REG)' 44870 Return the value of IACC0H if REG is 0 and IACC0L if REG is 1. 44871 Other values of REG are rejected as invalid. 44872 44873'void __IACCsetll (iacc REG, sw2 X)' 44874 Set the full 64-bit value of IACC0 to X. The REG argument is 44875 reserved for future expansion and must be 0. 44876 44877'void __IACCsetl (iacc REG, sw1 X)' 44878 Set IACC0H to X if REG is 0 and IACC0L to X if REG is 1. Other 44879 values of REG are rejected as invalid. 44880 44881'void __data_prefetch0 (const void *X)' 44882 Use the 'dcpl' instruction to load the contents of address X into 44883 the data cache. 44884 44885'void __data_prefetch (const void *X)' 44886 Use the 'nldub' instruction to load the contents of address X into 44887 the data cache. The instruction is issued in slot I1. 44888 44889 44890File: gcc.info, Node: MIPS DSP Built-in Functions, Next: MIPS Paired-Single Support, Prev: FR-V Built-in Functions, Up: Target Builtins 44891 448926.60.14 MIPS DSP Built-in Functions 44893----------------------------------- 44894 44895The MIPS DSP Application-Specific Extension (ASE) includes new 44896instructions that are designed to improve the performance of DSP and 44897media applications. It provides instructions that operate on packed 448988-bit/16-bit integer data, Q7, Q15 and Q31 fractional data. 44899 44900 GCC supports MIPS DSP operations using both the generic vector 44901extensions (*note Vector Extensions::) and a collection of MIPS-specific 44902built-in functions. Both kinds of support are enabled by the '-mdsp' 44903command-line option. 44904 44905 Revision 2 of the ASE was introduced in the second half of 2006. This 44906revision adds extra instructions to the original ASE, but is otherwise 44907backwards-compatible with it. You can select revision 2 using the 44908command-line option '-mdspr2'; this option implies '-mdsp'. 44909 44910 The SCOUNT and POS bits of the DSP control register are global. The 44911WRDSP, EXTPDP, EXTPDPV and MTHLIP instructions modify the SCOUNT and POS 44912bits. During optimization, the compiler does not delete these 44913instructions and it does not delete calls to functions containing these 44914instructions. 44915 44916 At present, GCC only provides support for operations on 32-bit vectors. 44917The vector type associated with 8-bit integer data is usually called 44918'v4i8', the vector type associated with Q7 is usually called 'v4q7', the 44919vector type associated with 16-bit integer data is usually called 44920'v2i16', and the vector type associated with Q15 is usually called 44921'v2q15'. They can be defined in C as follows: 44922 44923 typedef signed char v4i8 __attribute__ ((vector_size(4))); 44924 typedef signed char v4q7 __attribute__ ((vector_size(4))); 44925 typedef short v2i16 __attribute__ ((vector_size(4))); 44926 typedef short v2q15 __attribute__ ((vector_size(4))); 44927 44928 'v4i8', 'v4q7', 'v2i16' and 'v2q15' values are initialized in the same 44929way as aggregates. For example: 44930 44931 v4i8 a = {1, 2, 3, 4}; 44932 v4i8 b; 44933 b = (v4i8) {5, 6, 7, 8}; 44934 44935 v2q15 c = {0x0fcb, 0x3a75}; 44936 v2q15 d; 44937 d = (v2q15) {0.1234 * 0x1.0p15, 0.4567 * 0x1.0p15}; 44938 44939 _Note:_ The CPU's endianness determines the order in which values are 44940packed. On little-endian targets, the first value is the least 44941significant and the last value is the most significant. The opposite 44942order applies to big-endian targets. For example, the code above sets 44943the lowest byte of 'a' to '1' on little-endian targets and '4' on 44944big-endian targets. 44945 44946 _Note:_ Q7, Q15 and Q31 values must be initialized with their integer 44947representation. As shown in this example, the integer representation of 44948a Q7 value can be obtained by multiplying the fractional value by 44949'0x1.0p7'. The equivalent for Q15 values is to multiply by '0x1.0p15'. 44950The equivalent for Q31 values is to multiply by '0x1.0p31'. 44951 44952 The table below lists the 'v4i8' and 'v2q15' operations for which 44953hardware support exists. 'a' and 'b' are 'v4i8' values, and 'c' and 'd' 44954are 'v2q15' values. 44955 44956C code MIPS instruction 44957'a + b' 'addu.qb' 44958'c + d' 'addq.ph' 44959'a - b' 'subu.qb' 44960'c - d' 'subq.ph' 44961 44962 The table below lists the 'v2i16' operation for which hardware support 44963exists for the DSP ASE REV 2. 'e' and 'f' are 'v2i16' values. 44964 44965C code MIPS instruction 44966'e * f' 'mul.ph' 44967 44968 It is easier to describe the DSP built-in functions if we first define 44969the following types: 44970 44971 typedef int q31; 44972 typedef int i32; 44973 typedef unsigned int ui32; 44974 typedef long long a64; 44975 44976 'q31' and 'i32' are actually the same as 'int', but we use 'q31' to 44977indicate a Q31 fractional value and 'i32' to indicate a 32-bit integer 44978value. Similarly, 'a64' is the same as 'long long', but we use 'a64' to 44979indicate values that are placed in one of the four DSP accumulators 44980('$ac0', '$ac1', '$ac2' or '$ac3'). 44981 44982 Also, some built-in functions prefer or require immediate numbers as 44983parameters, because the corresponding DSP instructions accept both 44984immediate numbers and register operands, or accept immediate numbers 44985only. The immediate parameters are listed as follows. 44986 44987 imm0_3: 0 to 3. 44988 imm0_7: 0 to 7. 44989 imm0_15: 0 to 15. 44990 imm0_31: 0 to 31. 44991 imm0_63: 0 to 63. 44992 imm0_255: 0 to 255. 44993 imm_n32_31: -32 to 31. 44994 imm_n512_511: -512 to 511. 44995 44996 The following built-in functions map directly to a particular MIPS DSP 44997instruction. Please refer to the architecture specification for details 44998on what each instruction does. 44999 45000 v2q15 __builtin_mips_addq_ph (v2q15, v2q15) 45001 v2q15 __builtin_mips_addq_s_ph (v2q15, v2q15) 45002 q31 __builtin_mips_addq_s_w (q31, q31) 45003 v4i8 __builtin_mips_addu_qb (v4i8, v4i8) 45004 v4i8 __builtin_mips_addu_s_qb (v4i8, v4i8) 45005 v2q15 __builtin_mips_subq_ph (v2q15, v2q15) 45006 v2q15 __builtin_mips_subq_s_ph (v2q15, v2q15) 45007 q31 __builtin_mips_subq_s_w (q31, q31) 45008 v4i8 __builtin_mips_subu_qb (v4i8, v4i8) 45009 v4i8 __builtin_mips_subu_s_qb (v4i8, v4i8) 45010 i32 __builtin_mips_addsc (i32, i32) 45011 i32 __builtin_mips_addwc (i32, i32) 45012 i32 __builtin_mips_modsub (i32, i32) 45013 i32 __builtin_mips_raddu_w_qb (v4i8) 45014 v2q15 __builtin_mips_absq_s_ph (v2q15) 45015 q31 __builtin_mips_absq_s_w (q31) 45016 v4i8 __builtin_mips_precrq_qb_ph (v2q15, v2q15) 45017 v2q15 __builtin_mips_precrq_ph_w (q31, q31) 45018 v2q15 __builtin_mips_precrq_rs_ph_w (q31, q31) 45019 v4i8 __builtin_mips_precrqu_s_qb_ph (v2q15, v2q15) 45020 q31 __builtin_mips_preceq_w_phl (v2q15) 45021 q31 __builtin_mips_preceq_w_phr (v2q15) 45022 v2q15 __builtin_mips_precequ_ph_qbl (v4i8) 45023 v2q15 __builtin_mips_precequ_ph_qbr (v4i8) 45024 v2q15 __builtin_mips_precequ_ph_qbla (v4i8) 45025 v2q15 __builtin_mips_precequ_ph_qbra (v4i8) 45026 v2q15 __builtin_mips_preceu_ph_qbl (v4i8) 45027 v2q15 __builtin_mips_preceu_ph_qbr (v4i8) 45028 v2q15 __builtin_mips_preceu_ph_qbla (v4i8) 45029 v2q15 __builtin_mips_preceu_ph_qbra (v4i8) 45030 v4i8 __builtin_mips_shll_qb (v4i8, imm0_7) 45031 v4i8 __builtin_mips_shll_qb (v4i8, i32) 45032 v2q15 __builtin_mips_shll_ph (v2q15, imm0_15) 45033 v2q15 __builtin_mips_shll_ph (v2q15, i32) 45034 v2q15 __builtin_mips_shll_s_ph (v2q15, imm0_15) 45035 v2q15 __builtin_mips_shll_s_ph (v2q15, i32) 45036 q31 __builtin_mips_shll_s_w (q31, imm0_31) 45037 q31 __builtin_mips_shll_s_w (q31, i32) 45038 v4i8 __builtin_mips_shrl_qb (v4i8, imm0_7) 45039 v4i8 __builtin_mips_shrl_qb (v4i8, i32) 45040 v2q15 __builtin_mips_shra_ph (v2q15, imm0_15) 45041 v2q15 __builtin_mips_shra_ph (v2q15, i32) 45042 v2q15 __builtin_mips_shra_r_ph (v2q15, imm0_15) 45043 v2q15 __builtin_mips_shra_r_ph (v2q15, i32) 45044 q31 __builtin_mips_shra_r_w (q31, imm0_31) 45045 q31 __builtin_mips_shra_r_w (q31, i32) 45046 v2q15 __builtin_mips_muleu_s_ph_qbl (v4i8, v2q15) 45047 v2q15 __builtin_mips_muleu_s_ph_qbr (v4i8, v2q15) 45048 v2q15 __builtin_mips_mulq_rs_ph (v2q15, v2q15) 45049 q31 __builtin_mips_muleq_s_w_phl (v2q15, v2q15) 45050 q31 __builtin_mips_muleq_s_w_phr (v2q15, v2q15) 45051 a64 __builtin_mips_dpau_h_qbl (a64, v4i8, v4i8) 45052 a64 __builtin_mips_dpau_h_qbr (a64, v4i8, v4i8) 45053 a64 __builtin_mips_dpsu_h_qbl (a64, v4i8, v4i8) 45054 a64 __builtin_mips_dpsu_h_qbr (a64, v4i8, v4i8) 45055 a64 __builtin_mips_dpaq_s_w_ph (a64, v2q15, v2q15) 45056 a64 __builtin_mips_dpaq_sa_l_w (a64, q31, q31) 45057 a64 __builtin_mips_dpsq_s_w_ph (a64, v2q15, v2q15) 45058 a64 __builtin_mips_dpsq_sa_l_w (a64, q31, q31) 45059 a64 __builtin_mips_mulsaq_s_w_ph (a64, v2q15, v2q15) 45060 a64 __builtin_mips_maq_s_w_phl (a64, v2q15, v2q15) 45061 a64 __builtin_mips_maq_s_w_phr (a64, v2q15, v2q15) 45062 a64 __builtin_mips_maq_sa_w_phl (a64, v2q15, v2q15) 45063 a64 __builtin_mips_maq_sa_w_phr (a64, v2q15, v2q15) 45064 i32 __builtin_mips_bitrev (i32) 45065 i32 __builtin_mips_insv (i32, i32) 45066 v4i8 __builtin_mips_repl_qb (imm0_255) 45067 v4i8 __builtin_mips_repl_qb (i32) 45068 v2q15 __builtin_mips_repl_ph (imm_n512_511) 45069 v2q15 __builtin_mips_repl_ph (i32) 45070 void __builtin_mips_cmpu_eq_qb (v4i8, v4i8) 45071 void __builtin_mips_cmpu_lt_qb (v4i8, v4i8) 45072 void __builtin_mips_cmpu_le_qb (v4i8, v4i8) 45073 i32 __builtin_mips_cmpgu_eq_qb (v4i8, v4i8) 45074 i32 __builtin_mips_cmpgu_lt_qb (v4i8, v4i8) 45075 i32 __builtin_mips_cmpgu_le_qb (v4i8, v4i8) 45076 void __builtin_mips_cmp_eq_ph (v2q15, v2q15) 45077 void __builtin_mips_cmp_lt_ph (v2q15, v2q15) 45078 void __builtin_mips_cmp_le_ph (v2q15, v2q15) 45079 v4i8 __builtin_mips_pick_qb (v4i8, v4i8) 45080 v2q15 __builtin_mips_pick_ph (v2q15, v2q15) 45081 v2q15 __builtin_mips_packrl_ph (v2q15, v2q15) 45082 i32 __builtin_mips_extr_w (a64, imm0_31) 45083 i32 __builtin_mips_extr_w (a64, i32) 45084 i32 __builtin_mips_extr_r_w (a64, imm0_31) 45085 i32 __builtin_mips_extr_s_h (a64, i32) 45086 i32 __builtin_mips_extr_rs_w (a64, imm0_31) 45087 i32 __builtin_mips_extr_rs_w (a64, i32) 45088 i32 __builtin_mips_extr_s_h (a64, imm0_31) 45089 i32 __builtin_mips_extr_r_w (a64, i32) 45090 i32 __builtin_mips_extp (a64, imm0_31) 45091 i32 __builtin_mips_extp (a64, i32) 45092 i32 __builtin_mips_extpdp (a64, imm0_31) 45093 i32 __builtin_mips_extpdp (a64, i32) 45094 a64 __builtin_mips_shilo (a64, imm_n32_31) 45095 a64 __builtin_mips_shilo (a64, i32) 45096 a64 __builtin_mips_mthlip (a64, i32) 45097 void __builtin_mips_wrdsp (i32, imm0_63) 45098 i32 __builtin_mips_rddsp (imm0_63) 45099 i32 __builtin_mips_lbux (void *, i32) 45100 i32 __builtin_mips_lhx (void *, i32) 45101 i32 __builtin_mips_lwx (void *, i32) 45102 a64 __builtin_mips_ldx (void *, i32) [MIPS64 only] 45103 i32 __builtin_mips_bposge32 (void) 45104 a64 __builtin_mips_madd (a64, i32, i32); 45105 a64 __builtin_mips_maddu (a64, ui32, ui32); 45106 a64 __builtin_mips_msub (a64, i32, i32); 45107 a64 __builtin_mips_msubu (a64, ui32, ui32); 45108 a64 __builtin_mips_mult (i32, i32); 45109 a64 __builtin_mips_multu (ui32, ui32); 45110 45111 The following built-in functions map directly to a particular MIPS DSP 45112REV 2 instruction. Please refer to the architecture specification for 45113details on what each instruction does. 45114 45115 v4q7 __builtin_mips_absq_s_qb (v4q7); 45116 v2i16 __builtin_mips_addu_ph (v2i16, v2i16); 45117 v2i16 __builtin_mips_addu_s_ph (v2i16, v2i16); 45118 v4i8 __builtin_mips_adduh_qb (v4i8, v4i8); 45119 v4i8 __builtin_mips_adduh_r_qb (v4i8, v4i8); 45120 i32 __builtin_mips_append (i32, i32, imm0_31); 45121 i32 __builtin_mips_balign (i32, i32, imm0_3); 45122 i32 __builtin_mips_cmpgdu_eq_qb (v4i8, v4i8); 45123 i32 __builtin_mips_cmpgdu_lt_qb (v4i8, v4i8); 45124 i32 __builtin_mips_cmpgdu_le_qb (v4i8, v4i8); 45125 a64 __builtin_mips_dpa_w_ph (a64, v2i16, v2i16); 45126 a64 __builtin_mips_dps_w_ph (a64, v2i16, v2i16); 45127 v2i16 __builtin_mips_mul_ph (v2i16, v2i16); 45128 v2i16 __builtin_mips_mul_s_ph (v2i16, v2i16); 45129 q31 __builtin_mips_mulq_rs_w (q31, q31); 45130 v2q15 __builtin_mips_mulq_s_ph (v2q15, v2q15); 45131 q31 __builtin_mips_mulq_s_w (q31, q31); 45132 a64 __builtin_mips_mulsa_w_ph (a64, v2i16, v2i16); 45133 v4i8 __builtin_mips_precr_qb_ph (v2i16, v2i16); 45134 v2i16 __builtin_mips_precr_sra_ph_w (i32, i32, imm0_31); 45135 v2i16 __builtin_mips_precr_sra_r_ph_w (i32, i32, imm0_31); 45136 i32 __builtin_mips_prepend (i32, i32, imm0_31); 45137 v4i8 __builtin_mips_shra_qb (v4i8, imm0_7); 45138 v4i8 __builtin_mips_shra_r_qb (v4i8, imm0_7); 45139 v4i8 __builtin_mips_shra_qb (v4i8, i32); 45140 v4i8 __builtin_mips_shra_r_qb (v4i8, i32); 45141 v2i16 __builtin_mips_shrl_ph (v2i16, imm0_15); 45142 v2i16 __builtin_mips_shrl_ph (v2i16, i32); 45143 v2i16 __builtin_mips_subu_ph (v2i16, v2i16); 45144 v2i16 __builtin_mips_subu_s_ph (v2i16, v2i16); 45145 v4i8 __builtin_mips_subuh_qb (v4i8, v4i8); 45146 v4i8 __builtin_mips_subuh_r_qb (v4i8, v4i8); 45147 v2q15 __builtin_mips_addqh_ph (v2q15, v2q15); 45148 v2q15 __builtin_mips_addqh_r_ph (v2q15, v2q15); 45149 q31 __builtin_mips_addqh_w (q31, q31); 45150 q31 __builtin_mips_addqh_r_w (q31, q31); 45151 v2q15 __builtin_mips_subqh_ph (v2q15, v2q15); 45152 v2q15 __builtin_mips_subqh_r_ph (v2q15, v2q15); 45153 q31 __builtin_mips_subqh_w (q31, q31); 45154 q31 __builtin_mips_subqh_r_w (q31, q31); 45155 a64 __builtin_mips_dpax_w_ph (a64, v2i16, v2i16); 45156 a64 __builtin_mips_dpsx_w_ph (a64, v2i16, v2i16); 45157 a64 __builtin_mips_dpaqx_s_w_ph (a64, v2q15, v2q15); 45158 a64 __builtin_mips_dpaqx_sa_w_ph (a64, v2q15, v2q15); 45159 a64 __builtin_mips_dpsqx_s_w_ph (a64, v2q15, v2q15); 45160 a64 __builtin_mips_dpsqx_sa_w_ph (a64, v2q15, v2q15); 45161 45162 45163File: gcc.info, Node: MIPS Paired-Single Support, Next: MIPS Loongson Built-in Functions, Prev: MIPS DSP Built-in Functions, Up: Target Builtins 45164 451656.60.15 MIPS Paired-Single Support 45166---------------------------------- 45167 45168The MIPS64 architecture includes a number of instructions that operate 45169on pairs of single-precision floating-point values. Each pair is packed 45170into a 64-bit floating-point register, with one element being designated 45171the "upper half" and the other being designated the "lower half". 45172 45173 GCC supports paired-single operations using both the generic vector 45174extensions (*note Vector Extensions::) and a collection of MIPS-specific 45175built-in functions. Both kinds of support are enabled by the 45176'-mpaired-single' command-line option. 45177 45178 The vector type associated with paired-single values is usually called 45179'v2sf'. It can be defined in C as follows: 45180 45181 typedef float v2sf __attribute__ ((vector_size (8))); 45182 45183 'v2sf' values are initialized in the same way as aggregates. For 45184example: 45185 45186 v2sf a = {1.5, 9.1}; 45187 v2sf b; 45188 float e, f; 45189 b = (v2sf) {e, f}; 45190 45191 _Note:_ The CPU's endianness determines which value is stored in the 45192upper half of a register and which value is stored in the lower half. 45193On little-endian targets, the first value is the lower one and the 45194second value is the upper one. The opposite order applies to big-endian 45195targets. For example, the code above sets the lower half of 'a' to 45196'1.5' on little-endian targets and '9.1' on big-endian targets. 45197 45198 45199File: gcc.info, Node: MIPS Loongson Built-in Functions, Next: MIPS SIMD Architecture (MSA) Support, Prev: MIPS Paired-Single Support, Up: Target Builtins 45200 452016.60.16 MIPS Loongson Built-in Functions 45202---------------------------------------- 45203 45204GCC provides intrinsics to access the SIMD instructions provided by the 45205ST Microelectronics Loongson-2E and -2F processors. These intrinsics, 45206available after inclusion of the 'loongson.h' header file, operate on 45207the following 64-bit vector types: 45208 45209 * 'uint8x8_t', a vector of eight unsigned 8-bit integers; 45210 * 'uint16x4_t', a vector of four unsigned 16-bit integers; 45211 * 'uint32x2_t', a vector of two unsigned 32-bit integers; 45212 * 'int8x8_t', a vector of eight signed 8-bit integers; 45213 * 'int16x4_t', a vector of four signed 16-bit integers; 45214 * 'int32x2_t', a vector of two signed 32-bit integers. 45215 45216 The intrinsics provided are listed below; each is named after the 45217machine instruction to which it corresponds, with suffixes added as 45218appropriate to distinguish intrinsics that expand to the same machine 45219instruction yet have different argument types. Refer to the 45220architecture documentation for a description of the functionality of 45221each instruction. 45222 45223 int16x4_t packsswh (int32x2_t s, int32x2_t t); 45224 int8x8_t packsshb (int16x4_t s, int16x4_t t); 45225 uint8x8_t packushb (uint16x4_t s, uint16x4_t t); 45226 uint32x2_t paddw_u (uint32x2_t s, uint32x2_t t); 45227 uint16x4_t paddh_u (uint16x4_t s, uint16x4_t t); 45228 uint8x8_t paddb_u (uint8x8_t s, uint8x8_t t); 45229 int32x2_t paddw_s (int32x2_t s, int32x2_t t); 45230 int16x4_t paddh_s (int16x4_t s, int16x4_t t); 45231 int8x8_t paddb_s (int8x8_t s, int8x8_t t); 45232 uint64_t paddd_u (uint64_t s, uint64_t t); 45233 int64_t paddd_s (int64_t s, int64_t t); 45234 int16x4_t paddsh (int16x4_t s, int16x4_t t); 45235 int8x8_t paddsb (int8x8_t s, int8x8_t t); 45236 uint16x4_t paddush (uint16x4_t s, uint16x4_t t); 45237 uint8x8_t paddusb (uint8x8_t s, uint8x8_t t); 45238 uint64_t pandn_ud (uint64_t s, uint64_t t); 45239 uint32x2_t pandn_uw (uint32x2_t s, uint32x2_t t); 45240 uint16x4_t pandn_uh (uint16x4_t s, uint16x4_t t); 45241 uint8x8_t pandn_ub (uint8x8_t s, uint8x8_t t); 45242 int64_t pandn_sd (int64_t s, int64_t t); 45243 int32x2_t pandn_sw (int32x2_t s, int32x2_t t); 45244 int16x4_t pandn_sh (int16x4_t s, int16x4_t t); 45245 int8x8_t pandn_sb (int8x8_t s, int8x8_t t); 45246 uint16x4_t pavgh (uint16x4_t s, uint16x4_t t); 45247 uint8x8_t pavgb (uint8x8_t s, uint8x8_t t); 45248 uint32x2_t pcmpeqw_u (uint32x2_t s, uint32x2_t t); 45249 uint16x4_t pcmpeqh_u (uint16x4_t s, uint16x4_t t); 45250 uint8x8_t pcmpeqb_u (uint8x8_t s, uint8x8_t t); 45251 int32x2_t pcmpeqw_s (int32x2_t s, int32x2_t t); 45252 int16x4_t pcmpeqh_s (int16x4_t s, int16x4_t t); 45253 int8x8_t pcmpeqb_s (int8x8_t s, int8x8_t t); 45254 uint32x2_t pcmpgtw_u (uint32x2_t s, uint32x2_t t); 45255 uint16x4_t pcmpgth_u (uint16x4_t s, uint16x4_t t); 45256 uint8x8_t pcmpgtb_u (uint8x8_t s, uint8x8_t t); 45257 int32x2_t pcmpgtw_s (int32x2_t s, int32x2_t t); 45258 int16x4_t pcmpgth_s (int16x4_t s, int16x4_t t); 45259 int8x8_t pcmpgtb_s (int8x8_t s, int8x8_t t); 45260 uint16x4_t pextrh_u (uint16x4_t s, int field); 45261 int16x4_t pextrh_s (int16x4_t s, int field); 45262 uint16x4_t pinsrh_0_u (uint16x4_t s, uint16x4_t t); 45263 uint16x4_t pinsrh_1_u (uint16x4_t s, uint16x4_t t); 45264 uint16x4_t pinsrh_2_u (uint16x4_t s, uint16x4_t t); 45265 uint16x4_t pinsrh_3_u (uint16x4_t s, uint16x4_t t); 45266 int16x4_t pinsrh_0_s (int16x4_t s, int16x4_t t); 45267 int16x4_t pinsrh_1_s (int16x4_t s, int16x4_t t); 45268 int16x4_t pinsrh_2_s (int16x4_t s, int16x4_t t); 45269 int16x4_t pinsrh_3_s (int16x4_t s, int16x4_t t); 45270 int32x2_t pmaddhw (int16x4_t s, int16x4_t t); 45271 int16x4_t pmaxsh (int16x4_t s, int16x4_t t); 45272 uint8x8_t pmaxub (uint8x8_t s, uint8x8_t t); 45273 int16x4_t pminsh (int16x4_t s, int16x4_t t); 45274 uint8x8_t pminub (uint8x8_t s, uint8x8_t t); 45275 uint8x8_t pmovmskb_u (uint8x8_t s); 45276 int8x8_t pmovmskb_s (int8x8_t s); 45277 uint16x4_t pmulhuh (uint16x4_t s, uint16x4_t t); 45278 int16x4_t pmulhh (int16x4_t s, int16x4_t t); 45279 int16x4_t pmullh (int16x4_t s, int16x4_t t); 45280 int64_t pmuluw (uint32x2_t s, uint32x2_t t); 45281 uint8x8_t pasubub (uint8x8_t s, uint8x8_t t); 45282 uint16x4_t biadd (uint8x8_t s); 45283 uint16x4_t psadbh (uint8x8_t s, uint8x8_t t); 45284 uint16x4_t pshufh_u (uint16x4_t dest, uint16x4_t s, uint8_t order); 45285 int16x4_t pshufh_s (int16x4_t dest, int16x4_t s, uint8_t order); 45286 uint16x4_t psllh_u (uint16x4_t s, uint8_t amount); 45287 int16x4_t psllh_s (int16x4_t s, uint8_t amount); 45288 uint32x2_t psllw_u (uint32x2_t s, uint8_t amount); 45289 int32x2_t psllw_s (int32x2_t s, uint8_t amount); 45290 uint16x4_t psrlh_u (uint16x4_t s, uint8_t amount); 45291 int16x4_t psrlh_s (int16x4_t s, uint8_t amount); 45292 uint32x2_t psrlw_u (uint32x2_t s, uint8_t amount); 45293 int32x2_t psrlw_s (int32x2_t s, uint8_t amount); 45294 uint16x4_t psrah_u (uint16x4_t s, uint8_t amount); 45295 int16x4_t psrah_s (int16x4_t s, uint8_t amount); 45296 uint32x2_t psraw_u (uint32x2_t s, uint8_t amount); 45297 int32x2_t psraw_s (int32x2_t s, uint8_t amount); 45298 uint32x2_t psubw_u (uint32x2_t s, uint32x2_t t); 45299 uint16x4_t psubh_u (uint16x4_t s, uint16x4_t t); 45300 uint8x8_t psubb_u (uint8x8_t s, uint8x8_t t); 45301 int32x2_t psubw_s (int32x2_t s, int32x2_t t); 45302 int16x4_t psubh_s (int16x4_t s, int16x4_t t); 45303 int8x8_t psubb_s (int8x8_t s, int8x8_t t); 45304 uint64_t psubd_u (uint64_t s, uint64_t t); 45305 int64_t psubd_s (int64_t s, int64_t t); 45306 int16x4_t psubsh (int16x4_t s, int16x4_t t); 45307 int8x8_t psubsb (int8x8_t s, int8x8_t t); 45308 uint16x4_t psubush (uint16x4_t s, uint16x4_t t); 45309 uint8x8_t psubusb (uint8x8_t s, uint8x8_t t); 45310 uint32x2_t punpckhwd_u (uint32x2_t s, uint32x2_t t); 45311 uint16x4_t punpckhhw_u (uint16x4_t s, uint16x4_t t); 45312 uint8x8_t punpckhbh_u (uint8x8_t s, uint8x8_t t); 45313 int32x2_t punpckhwd_s (int32x2_t s, int32x2_t t); 45314 int16x4_t punpckhhw_s (int16x4_t s, int16x4_t t); 45315 int8x8_t punpckhbh_s (int8x8_t s, int8x8_t t); 45316 uint32x2_t punpcklwd_u (uint32x2_t s, uint32x2_t t); 45317 uint16x4_t punpcklhw_u (uint16x4_t s, uint16x4_t t); 45318 uint8x8_t punpcklbh_u (uint8x8_t s, uint8x8_t t); 45319 int32x2_t punpcklwd_s (int32x2_t s, int32x2_t t); 45320 int16x4_t punpcklhw_s (int16x4_t s, int16x4_t t); 45321 int8x8_t punpcklbh_s (int8x8_t s, int8x8_t t); 45322 45323* Menu: 45324 45325* Paired-Single Arithmetic:: 45326* Paired-Single Built-in Functions:: 45327* MIPS-3D Built-in Functions:: 45328 45329 45330File: gcc.info, Node: Paired-Single Arithmetic, Next: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions 45331 453326.60.16.1 Paired-Single Arithmetic 45333.................................. 45334 45335The table below lists the 'v2sf' operations for which hardware support 45336exists. 'a', 'b' and 'c' are 'v2sf' values and 'x' is an integral 45337value. 45338 45339C code MIPS instruction 45340'a + b' 'add.ps' 45341'a - b' 'sub.ps' 45342'-a' 'neg.ps' 45343'a * b' 'mul.ps' 45344'a * b + c' 'madd.ps' 45345'a * b - c' 'msub.ps' 45346'-(a * b + c)' 'nmadd.ps' 45347'-(a * b - c)' 'nmsub.ps' 45348'x ? a : b' 'movn.ps'/'movz.ps' 45349 45350 Note that the multiply-accumulate instructions can be disabled using 45351the command-line option '-mno-fused-madd'. 45352 45353 45354File: gcc.info, Node: Paired-Single Built-in Functions, Next: MIPS-3D Built-in Functions, Prev: Paired-Single Arithmetic, Up: MIPS Loongson Built-in Functions 45355 453566.60.16.2 Paired-Single Built-in Functions 45357.......................................... 45358 45359The following paired-single functions map directly to a particular MIPS 45360instruction. Please refer to the architecture specification for details 45361on what each instruction does. 45362 45363'v2sf __builtin_mips_pll_ps (v2sf, v2sf)' 45364 Pair lower lower ('pll.ps'). 45365 45366'v2sf __builtin_mips_pul_ps (v2sf, v2sf)' 45367 Pair upper lower ('pul.ps'). 45368 45369'v2sf __builtin_mips_plu_ps (v2sf, v2sf)' 45370 Pair lower upper ('plu.ps'). 45371 45372'v2sf __builtin_mips_puu_ps (v2sf, v2sf)' 45373 Pair upper upper ('puu.ps'). 45374 45375'v2sf __builtin_mips_cvt_ps_s (float, float)' 45376 Convert pair to paired single ('cvt.ps.s'). 45377 45378'float __builtin_mips_cvt_s_pl (v2sf)' 45379 Convert pair lower to single ('cvt.s.pl'). 45380 45381'float __builtin_mips_cvt_s_pu (v2sf)' 45382 Convert pair upper to single ('cvt.s.pu'). 45383 45384'v2sf __builtin_mips_abs_ps (v2sf)' 45385 Absolute value ('abs.ps'). 45386 45387'v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)' 45388 Align variable ('alnv.ps'). 45389 45390 _Note:_ The value of the third parameter must be 0 or 4 modulo 8, 45391 otherwise the result is unpredictable. Please read the instruction 45392 description for details. 45393 45394 The following multi-instruction functions are also available. In each 45395case, COND can be any of the 16 floating-point conditions: 'f', 'un', 45396'eq', 'ueq', 'olt', 'ult', 'ole', 'ule', 'sf', 'ngle', 'seq', 'ngl', 45397'lt', 'nge', 'le' or 'ngt'. 45398 45399'v2sf __builtin_mips_movt_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)' 45400'v2sf __builtin_mips_movf_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)' 45401 Conditional move based on floating-point comparison ('c.COND.ps', 45402 'movt.ps'/'movf.ps'). 45403 45404 The 'movt' functions return the value X computed by: 45405 45406 c.COND.ps CC,A,B 45407 mov.ps X,C 45408 movt.ps X,D,CC 45409 45410 The 'movf' functions are similar but use 'movf.ps' instead of 45411 'movt.ps'. 45412 45413'int __builtin_mips_upper_c_COND_ps (v2sf A, v2sf B)' 45414'int __builtin_mips_lower_c_COND_ps (v2sf A, v2sf B)' 45415 Comparison of two paired-single values ('c.COND.ps', 45416 'bc1t'/'bc1f'). 45417 45418 These functions compare A and B using 'c.COND.ps' and return either 45419 the upper or lower half of the result. For example: 45420 45421 v2sf a, b; 45422 if (__builtin_mips_upper_c_eq_ps (a, b)) 45423 upper_halves_are_equal (); 45424 else 45425 upper_halves_are_unequal (); 45426 45427 if (__builtin_mips_lower_c_eq_ps (a, b)) 45428 lower_halves_are_equal (); 45429 else 45430 lower_halves_are_unequal (); 45431 45432 45433File: gcc.info, Node: MIPS-3D Built-in Functions, Prev: Paired-Single Built-in Functions, Up: MIPS Loongson Built-in Functions 45434 454356.60.16.3 MIPS-3D Built-in Functions 45436.................................... 45437 45438The MIPS-3D Application-Specific Extension (ASE) includes additional 45439paired-single instructions that are designed to improve the performance 45440of 3D graphics operations. Support for these instructions is controlled 45441by the '-mips3d' command-line option. 45442 45443 The functions listed below map directly to a particular MIPS-3D 45444instruction. Please refer to the architecture specification for more 45445details on what each instruction does. 45446 45447'v2sf __builtin_mips_addr_ps (v2sf, v2sf)' 45448 Reduction add ('addr.ps'). 45449 45450'v2sf __builtin_mips_mulr_ps (v2sf, v2sf)' 45451 Reduction multiply ('mulr.ps'). 45452 45453'v2sf __builtin_mips_cvt_pw_ps (v2sf)' 45454 Convert paired single to paired word ('cvt.pw.ps'). 45455 45456'v2sf __builtin_mips_cvt_ps_pw (v2sf)' 45457 Convert paired word to paired single ('cvt.ps.pw'). 45458 45459'float __builtin_mips_recip1_s (float)' 45460'double __builtin_mips_recip1_d (double)' 45461'v2sf __builtin_mips_recip1_ps (v2sf)' 45462 Reduced-precision reciprocal (sequence step 1) ('recip1.FMT'). 45463 45464'float __builtin_mips_recip2_s (float, float)' 45465'double __builtin_mips_recip2_d (double, double)' 45466'v2sf __builtin_mips_recip2_ps (v2sf, v2sf)' 45467 Reduced-precision reciprocal (sequence step 2) ('recip2.FMT'). 45468 45469'float __builtin_mips_rsqrt1_s (float)' 45470'double __builtin_mips_rsqrt1_d (double)' 45471'v2sf __builtin_mips_rsqrt1_ps (v2sf)' 45472 Reduced-precision reciprocal square root (sequence step 1) 45473 ('rsqrt1.FMT'). 45474 45475'float __builtin_mips_rsqrt2_s (float, float)' 45476'double __builtin_mips_rsqrt2_d (double, double)' 45477'v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)' 45478 Reduced-precision reciprocal square root (sequence step 2) 45479 ('rsqrt2.FMT'). 45480 45481 The following multi-instruction functions are also available. In each 45482case, COND can be any of the 16 floating-point conditions: 'f', 'un', 45483'eq', 'ueq', 'olt', 'ult', 'ole', 'ule', 'sf', 'ngle', 'seq', 'ngl', 45484'lt', 'nge', 'le' or 'ngt'. 45485 45486'int __builtin_mips_cabs_COND_s (float A, float B)' 45487'int __builtin_mips_cabs_COND_d (double A, double B)' 45488 Absolute comparison of two scalar values ('cabs.COND.FMT', 45489 'bc1t'/'bc1f'). 45490 45491 These functions compare A and B using 'cabs.COND.s' or 45492 'cabs.COND.d' and return the result as a boolean value. For 45493 example: 45494 45495 float a, b; 45496 if (__builtin_mips_cabs_eq_s (a, b)) 45497 true (); 45498 else 45499 false (); 45500 45501'int __builtin_mips_upper_cabs_COND_ps (v2sf A, v2sf B)' 45502'int __builtin_mips_lower_cabs_COND_ps (v2sf A, v2sf B)' 45503 Absolute comparison of two paired-single values ('cabs.COND.ps', 45504 'bc1t'/'bc1f'). 45505 45506 These functions compare A and B using 'cabs.COND.ps' and return 45507 either the upper or lower half of the result. For example: 45508 45509 v2sf a, b; 45510 if (__builtin_mips_upper_cabs_eq_ps (a, b)) 45511 upper_halves_are_equal (); 45512 else 45513 upper_halves_are_unequal (); 45514 45515 if (__builtin_mips_lower_cabs_eq_ps (a, b)) 45516 lower_halves_are_equal (); 45517 else 45518 lower_halves_are_unequal (); 45519 45520'v2sf __builtin_mips_movt_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)' 45521'v2sf __builtin_mips_movf_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)' 45522 Conditional move based on absolute comparison ('cabs.COND.ps', 45523 'movt.ps'/'movf.ps'). 45524 45525 The 'movt' functions return the value X computed by: 45526 45527 cabs.COND.ps CC,A,B 45528 mov.ps X,C 45529 movt.ps X,D,CC 45530 45531 The 'movf' functions are similar but use 'movf.ps' instead of 45532 'movt.ps'. 45533 45534'int __builtin_mips_any_c_COND_ps (v2sf A, v2sf B)' 45535'int __builtin_mips_all_c_COND_ps (v2sf A, v2sf B)' 45536'int __builtin_mips_any_cabs_COND_ps (v2sf A, v2sf B)' 45537'int __builtin_mips_all_cabs_COND_ps (v2sf A, v2sf B)' 45538 Comparison of two paired-single values ('c.COND.ps'/'cabs.COND.ps', 45539 'bc1any2t'/'bc1any2f'). 45540 45541 These functions compare A and B using 'c.COND.ps' or 45542 'cabs.COND.ps'. The 'any' forms return 'true' if either result is 45543 'true' and the 'all' forms return 'true' if both results are 45544 'true'. For example: 45545 45546 v2sf a, b; 45547 if (__builtin_mips_any_c_eq_ps (a, b)) 45548 one_is_true (); 45549 else 45550 both_are_false (); 45551 45552 if (__builtin_mips_all_c_eq_ps (a, b)) 45553 both_are_true (); 45554 else 45555 one_is_false (); 45556 45557'int __builtin_mips_any_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)' 45558'int __builtin_mips_all_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)' 45559'int __builtin_mips_any_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)' 45560'int __builtin_mips_all_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)' 45561 Comparison of four paired-single values 45562 ('c.COND.ps'/'cabs.COND.ps', 'bc1any4t'/'bc1any4f'). 45563 45564 These functions use 'c.COND.ps' or 'cabs.COND.ps' to compare A with 45565 B and to compare C with D. The 'any' forms return 'true' if any of 45566 the four results are 'true' and the 'all' forms return 'true' if 45567 all four results are 'true'. For example: 45568 45569 v2sf a, b, c, d; 45570 if (__builtin_mips_any_c_eq_4s (a, b, c, d)) 45571 some_are_true (); 45572 else 45573 all_are_false (); 45574 45575 if (__builtin_mips_all_c_eq_4s (a, b, c, d)) 45576 all_are_true (); 45577 else 45578 some_are_false (); 45579 45580 45581File: gcc.info, Node: MIPS SIMD Architecture (MSA) Support, Next: Other MIPS Built-in Functions, Prev: MIPS Loongson Built-in Functions, Up: Target Builtins 45582 455836.60.17 MIPS SIMD Architecture (MSA) Support 45584-------------------------------------------- 45585 45586* Menu: 45587 45588* MIPS SIMD Architecture Built-in Functions:: 45589 45590GCC provides intrinsics to access the SIMD instructions provided by the 45591MSA MIPS SIMD Architecture. The interface is made available by 45592including '<msa.h>' and using '-mmsa -mhard-float -mfp64 -mnan=2008'. 45593For each '__builtin_msa_*', there is a shortened name of the intrinsic, 45594'__msa_*'. 45595 45596 MSA implements 128-bit wide vector registers, operating on 8-, 16-, 32- 45597and 64-bit integer, 16- and 32-bit fixed-point, or 32- and 64-bit 45598floating point data elements. The following vectors typedefs are 45599included in 'msa.h': 45600 * 'v16i8', a vector of sixteen signed 8-bit integers; 45601 * 'v16u8', a vector of sixteen unsigned 8-bit integers; 45602 * 'v8i16', a vector of eight signed 16-bit integers; 45603 * 'v8u16', a vector of eight unsigned 16-bit integers; 45604 * 'v4i32', a vector of four signed 32-bit integers; 45605 * 'v4u32', a vector of four unsigned 32-bit integers; 45606 * 'v2i64', a vector of two signed 64-bit integers; 45607 * 'v2u64', a vector of two unsigned 64-bit integers; 45608 * 'v4f32', a vector of four 32-bit floats; 45609 * 'v2f64', a vector of two 64-bit doubles. 45610 45611 Instructions and corresponding built-ins may have additional 45612restrictions and/or input/output values manipulated: 45613 * 'imm0_1', an integer literal in range 0 to 1; 45614 * 'imm0_3', an integer literal in range 0 to 3; 45615 * 'imm0_7', an integer literal in range 0 to 7; 45616 * 'imm0_15', an integer literal in range 0 to 15; 45617 * 'imm0_31', an integer literal in range 0 to 31; 45618 * 'imm0_63', an integer literal in range 0 to 63; 45619 * 'imm0_255', an integer literal in range 0 to 255; 45620 * 'imm_n16_15', an integer literal in range -16 to 15; 45621 * 'imm_n512_511', an integer literal in range -512 to 511; 45622 * 'imm_n1024_1022', an integer literal in range -512 to 511 left 45623 shifted by 1 bit, i.e., -1024, -1022, ..., 1020, 1022; 45624 * 'imm_n2048_2044', an integer literal in range -512 to 511 left 45625 shifted by 2 bits, i.e., -2048, -2044, ..., 2040, 2044; 45626 * 'imm_n4096_4088', an integer literal in range -512 to 511 left 45627 shifted by 3 bits, i.e., -4096, -4088, ..., 4080, 4088; 45628 * 'imm1_4', an integer literal in range 1 to 4; 45629 * 'i32, i64, u32, u64, f32, f64', defined as follows: 45630 45631 { 45632 typedef int i32; 45633 #if __LONG_MAX__ == __LONG_LONG_MAX__ 45634 typedef long i64; 45635 #else 45636 typedef long long i64; 45637 #endif 45638 45639 typedef unsigned int u32; 45640 #if __LONG_MAX__ == __LONG_LONG_MAX__ 45641 typedef unsigned long u64; 45642 #else 45643 typedef unsigned long long u64; 45644 #endif 45645 45646 typedef double f64; 45647 typedef float f32; 45648 } 45649 45650 45651File: gcc.info, Node: MIPS SIMD Architecture Built-in Functions, Up: MIPS SIMD Architecture (MSA) Support 45652 456536.60.17.1 MIPS SIMD Architecture Built-in Functions 45654................................................... 45655 45656The intrinsics provided are listed below; each is named after the 45657machine instruction. 45658 45659 v16i8 __builtin_msa_add_a_b (v16i8, v16i8); 45660 v8i16 __builtin_msa_add_a_h (v8i16, v8i16); 45661 v4i32 __builtin_msa_add_a_w (v4i32, v4i32); 45662 v2i64 __builtin_msa_add_a_d (v2i64, v2i64); 45663 45664 v16i8 __builtin_msa_adds_a_b (v16i8, v16i8); 45665 v8i16 __builtin_msa_adds_a_h (v8i16, v8i16); 45666 v4i32 __builtin_msa_adds_a_w (v4i32, v4i32); 45667 v2i64 __builtin_msa_adds_a_d (v2i64, v2i64); 45668 45669 v16i8 __builtin_msa_adds_s_b (v16i8, v16i8); 45670 v8i16 __builtin_msa_adds_s_h (v8i16, v8i16); 45671 v4i32 __builtin_msa_adds_s_w (v4i32, v4i32); 45672 v2i64 __builtin_msa_adds_s_d (v2i64, v2i64); 45673 45674 v16u8 __builtin_msa_adds_u_b (v16u8, v16u8); 45675 v8u16 __builtin_msa_adds_u_h (v8u16, v8u16); 45676 v4u32 __builtin_msa_adds_u_w (v4u32, v4u32); 45677 v2u64 __builtin_msa_adds_u_d (v2u64, v2u64); 45678 45679 v16i8 __builtin_msa_addv_b (v16i8, v16i8); 45680 v8i16 __builtin_msa_addv_h (v8i16, v8i16); 45681 v4i32 __builtin_msa_addv_w (v4i32, v4i32); 45682 v2i64 __builtin_msa_addv_d (v2i64, v2i64); 45683 45684 v16i8 __builtin_msa_addvi_b (v16i8, imm0_31); 45685 v8i16 __builtin_msa_addvi_h (v8i16, imm0_31); 45686 v4i32 __builtin_msa_addvi_w (v4i32, imm0_31); 45687 v2i64 __builtin_msa_addvi_d (v2i64, imm0_31); 45688 45689 v16u8 __builtin_msa_and_v (v16u8, v16u8); 45690 45691 v16u8 __builtin_msa_andi_b (v16u8, imm0_255); 45692 45693 v16i8 __builtin_msa_asub_s_b (v16i8, v16i8); 45694 v8i16 __builtin_msa_asub_s_h (v8i16, v8i16); 45695 v4i32 __builtin_msa_asub_s_w (v4i32, v4i32); 45696 v2i64 __builtin_msa_asub_s_d (v2i64, v2i64); 45697 45698 v16u8 __builtin_msa_asub_u_b (v16u8, v16u8); 45699 v8u16 __builtin_msa_asub_u_h (v8u16, v8u16); 45700 v4u32 __builtin_msa_asub_u_w (v4u32, v4u32); 45701 v2u64 __builtin_msa_asub_u_d (v2u64, v2u64); 45702 45703 v16i8 __builtin_msa_ave_s_b (v16i8, v16i8); 45704 v8i16 __builtin_msa_ave_s_h (v8i16, v8i16); 45705 v4i32 __builtin_msa_ave_s_w (v4i32, v4i32); 45706 v2i64 __builtin_msa_ave_s_d (v2i64, v2i64); 45707 45708 v16u8 __builtin_msa_ave_u_b (v16u8, v16u8); 45709 v8u16 __builtin_msa_ave_u_h (v8u16, v8u16); 45710 v4u32 __builtin_msa_ave_u_w (v4u32, v4u32); 45711 v2u64 __builtin_msa_ave_u_d (v2u64, v2u64); 45712 45713 v16i8 __builtin_msa_aver_s_b (v16i8, v16i8); 45714 v8i16 __builtin_msa_aver_s_h (v8i16, v8i16); 45715 v4i32 __builtin_msa_aver_s_w (v4i32, v4i32); 45716 v2i64 __builtin_msa_aver_s_d (v2i64, v2i64); 45717 45718 v16u8 __builtin_msa_aver_u_b (v16u8, v16u8); 45719 v8u16 __builtin_msa_aver_u_h (v8u16, v8u16); 45720 v4u32 __builtin_msa_aver_u_w (v4u32, v4u32); 45721 v2u64 __builtin_msa_aver_u_d (v2u64, v2u64); 45722 45723 v16u8 __builtin_msa_bclr_b (v16u8, v16u8); 45724 v8u16 __builtin_msa_bclr_h (v8u16, v8u16); 45725 v4u32 __builtin_msa_bclr_w (v4u32, v4u32); 45726 v2u64 __builtin_msa_bclr_d (v2u64, v2u64); 45727 45728 v16u8 __builtin_msa_bclri_b (v16u8, imm0_7); 45729 v8u16 __builtin_msa_bclri_h (v8u16, imm0_15); 45730 v4u32 __builtin_msa_bclri_w (v4u32, imm0_31); 45731 v2u64 __builtin_msa_bclri_d (v2u64, imm0_63); 45732 45733 v16u8 __builtin_msa_binsl_b (v16u8, v16u8, v16u8); 45734 v8u16 __builtin_msa_binsl_h (v8u16, v8u16, v8u16); 45735 v4u32 __builtin_msa_binsl_w (v4u32, v4u32, v4u32); 45736 v2u64 __builtin_msa_binsl_d (v2u64, v2u64, v2u64); 45737 45738 v16u8 __builtin_msa_binsli_b (v16u8, v16u8, imm0_7); 45739 v8u16 __builtin_msa_binsli_h (v8u16, v8u16, imm0_15); 45740 v4u32 __builtin_msa_binsli_w (v4u32, v4u32, imm0_31); 45741 v2u64 __builtin_msa_binsli_d (v2u64, v2u64, imm0_63); 45742 45743 v16u8 __builtin_msa_binsr_b (v16u8, v16u8, v16u8); 45744 v8u16 __builtin_msa_binsr_h (v8u16, v8u16, v8u16); 45745 v4u32 __builtin_msa_binsr_w (v4u32, v4u32, v4u32); 45746 v2u64 __builtin_msa_binsr_d (v2u64, v2u64, v2u64); 45747 45748 v16u8 __builtin_msa_binsri_b (v16u8, v16u8, imm0_7); 45749 v8u16 __builtin_msa_binsri_h (v8u16, v8u16, imm0_15); 45750 v4u32 __builtin_msa_binsri_w (v4u32, v4u32, imm0_31); 45751 v2u64 __builtin_msa_binsri_d (v2u64, v2u64, imm0_63); 45752 45753 v16u8 __builtin_msa_bmnz_v (v16u8, v16u8, v16u8); 45754 45755 v16u8 __builtin_msa_bmnzi_b (v16u8, v16u8, imm0_255); 45756 45757 v16u8 __builtin_msa_bmz_v (v16u8, v16u8, v16u8); 45758 45759 v16u8 __builtin_msa_bmzi_b (v16u8, v16u8, imm0_255); 45760 45761 v16u8 __builtin_msa_bneg_b (v16u8, v16u8); 45762 v8u16 __builtin_msa_bneg_h (v8u16, v8u16); 45763 v4u32 __builtin_msa_bneg_w (v4u32, v4u32); 45764 v2u64 __builtin_msa_bneg_d (v2u64, v2u64); 45765 45766 v16u8 __builtin_msa_bnegi_b (v16u8, imm0_7); 45767 v8u16 __builtin_msa_bnegi_h (v8u16, imm0_15); 45768 v4u32 __builtin_msa_bnegi_w (v4u32, imm0_31); 45769 v2u64 __builtin_msa_bnegi_d (v2u64, imm0_63); 45770 45771 i32 __builtin_msa_bnz_b (v16u8); 45772 i32 __builtin_msa_bnz_h (v8u16); 45773 i32 __builtin_msa_bnz_w (v4u32); 45774 i32 __builtin_msa_bnz_d (v2u64); 45775 45776 i32 __builtin_msa_bnz_v (v16u8); 45777 45778 v16u8 __builtin_msa_bsel_v (v16u8, v16u8, v16u8); 45779 45780 v16u8 __builtin_msa_bseli_b (v16u8, v16u8, imm0_255); 45781 45782 v16u8 __builtin_msa_bset_b (v16u8, v16u8); 45783 v8u16 __builtin_msa_bset_h (v8u16, v8u16); 45784 v4u32 __builtin_msa_bset_w (v4u32, v4u32); 45785 v2u64 __builtin_msa_bset_d (v2u64, v2u64); 45786 45787 v16u8 __builtin_msa_bseti_b (v16u8, imm0_7); 45788 v8u16 __builtin_msa_bseti_h (v8u16, imm0_15); 45789 v4u32 __builtin_msa_bseti_w (v4u32, imm0_31); 45790 v2u64 __builtin_msa_bseti_d (v2u64, imm0_63); 45791 45792 i32 __builtin_msa_bz_b (v16u8); 45793 i32 __builtin_msa_bz_h (v8u16); 45794 i32 __builtin_msa_bz_w (v4u32); 45795 i32 __builtin_msa_bz_d (v2u64); 45796 45797 i32 __builtin_msa_bz_v (v16u8); 45798 45799 v16i8 __builtin_msa_ceq_b (v16i8, v16i8); 45800 v8i16 __builtin_msa_ceq_h (v8i16, v8i16); 45801 v4i32 __builtin_msa_ceq_w (v4i32, v4i32); 45802 v2i64 __builtin_msa_ceq_d (v2i64, v2i64); 45803 45804 v16i8 __builtin_msa_ceqi_b (v16i8, imm_n16_15); 45805 v8i16 __builtin_msa_ceqi_h (v8i16, imm_n16_15); 45806 v4i32 __builtin_msa_ceqi_w (v4i32, imm_n16_15); 45807 v2i64 __builtin_msa_ceqi_d (v2i64, imm_n16_15); 45808 45809 i32 __builtin_msa_cfcmsa (imm0_31); 45810 45811 v16i8 __builtin_msa_cle_s_b (v16i8, v16i8); 45812 v8i16 __builtin_msa_cle_s_h (v8i16, v8i16); 45813 v4i32 __builtin_msa_cle_s_w (v4i32, v4i32); 45814 v2i64 __builtin_msa_cle_s_d (v2i64, v2i64); 45815 45816 v16i8 __builtin_msa_cle_u_b (v16u8, v16u8); 45817 v8i16 __builtin_msa_cle_u_h (v8u16, v8u16); 45818 v4i32 __builtin_msa_cle_u_w (v4u32, v4u32); 45819 v2i64 __builtin_msa_cle_u_d (v2u64, v2u64); 45820 45821 v16i8 __builtin_msa_clei_s_b (v16i8, imm_n16_15); 45822 v8i16 __builtin_msa_clei_s_h (v8i16, imm_n16_15); 45823 v4i32 __builtin_msa_clei_s_w (v4i32, imm_n16_15); 45824 v2i64 __builtin_msa_clei_s_d (v2i64, imm_n16_15); 45825 45826 v16i8 __builtin_msa_clei_u_b (v16u8, imm0_31); 45827 v8i16 __builtin_msa_clei_u_h (v8u16, imm0_31); 45828 v4i32 __builtin_msa_clei_u_w (v4u32, imm0_31); 45829 v2i64 __builtin_msa_clei_u_d (v2u64, imm0_31); 45830 45831 v16i8 __builtin_msa_clt_s_b (v16i8, v16i8); 45832 v8i16 __builtin_msa_clt_s_h (v8i16, v8i16); 45833 v4i32 __builtin_msa_clt_s_w (v4i32, v4i32); 45834 v2i64 __builtin_msa_clt_s_d (v2i64, v2i64); 45835 45836 v16i8 __builtin_msa_clt_u_b (v16u8, v16u8); 45837 v8i16 __builtin_msa_clt_u_h (v8u16, v8u16); 45838 v4i32 __builtin_msa_clt_u_w (v4u32, v4u32); 45839 v2i64 __builtin_msa_clt_u_d (v2u64, v2u64); 45840 45841 v16i8 __builtin_msa_clti_s_b (v16i8, imm_n16_15); 45842 v8i16 __builtin_msa_clti_s_h (v8i16, imm_n16_15); 45843 v4i32 __builtin_msa_clti_s_w (v4i32, imm_n16_15); 45844 v2i64 __builtin_msa_clti_s_d (v2i64, imm_n16_15); 45845 45846 v16i8 __builtin_msa_clti_u_b (v16u8, imm0_31); 45847 v8i16 __builtin_msa_clti_u_h (v8u16, imm0_31); 45848 v4i32 __builtin_msa_clti_u_w (v4u32, imm0_31); 45849 v2i64 __builtin_msa_clti_u_d (v2u64, imm0_31); 45850 45851 i32 __builtin_msa_copy_s_b (v16i8, imm0_15); 45852 i32 __builtin_msa_copy_s_h (v8i16, imm0_7); 45853 i32 __builtin_msa_copy_s_w (v4i32, imm0_3); 45854 i64 __builtin_msa_copy_s_d (v2i64, imm0_1); 45855 45856 u32 __builtin_msa_copy_u_b (v16i8, imm0_15); 45857 u32 __builtin_msa_copy_u_h (v8i16, imm0_7); 45858 u32 __builtin_msa_copy_u_w (v4i32, imm0_3); 45859 u64 __builtin_msa_copy_u_d (v2i64, imm0_1); 45860 45861 void __builtin_msa_ctcmsa (imm0_31, i32); 45862 45863 v16i8 __builtin_msa_div_s_b (v16i8, v16i8); 45864 v8i16 __builtin_msa_div_s_h (v8i16, v8i16); 45865 v4i32 __builtin_msa_div_s_w (v4i32, v4i32); 45866 v2i64 __builtin_msa_div_s_d (v2i64, v2i64); 45867 45868 v16u8 __builtin_msa_div_u_b (v16u8, v16u8); 45869 v8u16 __builtin_msa_div_u_h (v8u16, v8u16); 45870 v4u32 __builtin_msa_div_u_w (v4u32, v4u32); 45871 v2u64 __builtin_msa_div_u_d (v2u64, v2u64); 45872 45873 v8i16 __builtin_msa_dotp_s_h (v16i8, v16i8); 45874 v4i32 __builtin_msa_dotp_s_w (v8i16, v8i16); 45875 v2i64 __builtin_msa_dotp_s_d (v4i32, v4i32); 45876 45877 v8u16 __builtin_msa_dotp_u_h (v16u8, v16u8); 45878 v4u32 __builtin_msa_dotp_u_w (v8u16, v8u16); 45879 v2u64 __builtin_msa_dotp_u_d (v4u32, v4u32); 45880 45881 v8i16 __builtin_msa_dpadd_s_h (v8i16, v16i8, v16i8); 45882 v4i32 __builtin_msa_dpadd_s_w (v4i32, v8i16, v8i16); 45883 v2i64 __builtin_msa_dpadd_s_d (v2i64, v4i32, v4i32); 45884 45885 v8u16 __builtin_msa_dpadd_u_h (v8u16, v16u8, v16u8); 45886 v4u32 __builtin_msa_dpadd_u_w (v4u32, v8u16, v8u16); 45887 v2u64 __builtin_msa_dpadd_u_d (v2u64, v4u32, v4u32); 45888 45889 v8i16 __builtin_msa_dpsub_s_h (v8i16, v16i8, v16i8); 45890 v4i32 __builtin_msa_dpsub_s_w (v4i32, v8i16, v8i16); 45891 v2i64 __builtin_msa_dpsub_s_d (v2i64, v4i32, v4i32); 45892 45893 v8i16 __builtin_msa_dpsub_u_h (v8i16, v16u8, v16u8); 45894 v4i32 __builtin_msa_dpsub_u_w (v4i32, v8u16, v8u16); 45895 v2i64 __builtin_msa_dpsub_u_d (v2i64, v4u32, v4u32); 45896 45897 v4f32 __builtin_msa_fadd_w (v4f32, v4f32); 45898 v2f64 __builtin_msa_fadd_d (v2f64, v2f64); 45899 45900 v4i32 __builtin_msa_fcaf_w (v4f32, v4f32); 45901 v2i64 __builtin_msa_fcaf_d (v2f64, v2f64); 45902 45903 v4i32 __builtin_msa_fceq_w (v4f32, v4f32); 45904 v2i64 __builtin_msa_fceq_d (v2f64, v2f64); 45905 45906 v4i32 __builtin_msa_fclass_w (v4f32); 45907 v2i64 __builtin_msa_fclass_d (v2f64); 45908 45909 v4i32 __builtin_msa_fcle_w (v4f32, v4f32); 45910 v2i64 __builtin_msa_fcle_d (v2f64, v2f64); 45911 45912 v4i32 __builtin_msa_fclt_w (v4f32, v4f32); 45913 v2i64 __builtin_msa_fclt_d (v2f64, v2f64); 45914 45915 v4i32 __builtin_msa_fcne_w (v4f32, v4f32); 45916 v2i64 __builtin_msa_fcne_d (v2f64, v2f64); 45917 45918 v4i32 __builtin_msa_fcor_w (v4f32, v4f32); 45919 v2i64 __builtin_msa_fcor_d (v2f64, v2f64); 45920 45921 v4i32 __builtin_msa_fcueq_w (v4f32, v4f32); 45922 v2i64 __builtin_msa_fcueq_d (v2f64, v2f64); 45923 45924 v4i32 __builtin_msa_fcule_w (v4f32, v4f32); 45925 v2i64 __builtin_msa_fcule_d (v2f64, v2f64); 45926 45927 v4i32 __builtin_msa_fcult_w (v4f32, v4f32); 45928 v2i64 __builtin_msa_fcult_d (v2f64, v2f64); 45929 45930 v4i32 __builtin_msa_fcun_w (v4f32, v4f32); 45931 v2i64 __builtin_msa_fcun_d (v2f64, v2f64); 45932 45933 v4i32 __builtin_msa_fcune_w (v4f32, v4f32); 45934 v2i64 __builtin_msa_fcune_d (v2f64, v2f64); 45935 45936 v4f32 __builtin_msa_fdiv_w (v4f32, v4f32); 45937 v2f64 __builtin_msa_fdiv_d (v2f64, v2f64); 45938 45939 v8i16 __builtin_msa_fexdo_h (v4f32, v4f32); 45940 v4f32 __builtin_msa_fexdo_w (v2f64, v2f64); 45941 45942 v4f32 __builtin_msa_fexp2_w (v4f32, v4i32); 45943 v2f64 __builtin_msa_fexp2_d (v2f64, v2i64); 45944 45945 v4f32 __builtin_msa_fexupl_w (v8i16); 45946 v2f64 __builtin_msa_fexupl_d (v4f32); 45947 45948 v4f32 __builtin_msa_fexupr_w (v8i16); 45949 v2f64 __builtin_msa_fexupr_d (v4f32); 45950 45951 v4f32 __builtin_msa_ffint_s_w (v4i32); 45952 v2f64 __builtin_msa_ffint_s_d (v2i64); 45953 45954 v4f32 __builtin_msa_ffint_u_w (v4u32); 45955 v2f64 __builtin_msa_ffint_u_d (v2u64); 45956 45957 v4f32 __builtin_msa_ffql_w (v8i16); 45958 v2f64 __builtin_msa_ffql_d (v4i32); 45959 45960 v4f32 __builtin_msa_ffqr_w (v8i16); 45961 v2f64 __builtin_msa_ffqr_d (v4i32); 45962 45963 v16i8 __builtin_msa_fill_b (i32); 45964 v8i16 __builtin_msa_fill_h (i32); 45965 v4i32 __builtin_msa_fill_w (i32); 45966 v2i64 __builtin_msa_fill_d (i64); 45967 45968 v4f32 __builtin_msa_flog2_w (v4f32); 45969 v2f64 __builtin_msa_flog2_d (v2f64); 45970 45971 v4f32 __builtin_msa_fmadd_w (v4f32, v4f32, v4f32); 45972 v2f64 __builtin_msa_fmadd_d (v2f64, v2f64, v2f64); 45973 45974 v4f32 __builtin_msa_fmax_w (v4f32, v4f32); 45975 v2f64 __builtin_msa_fmax_d (v2f64, v2f64); 45976 45977 v4f32 __builtin_msa_fmax_a_w (v4f32, v4f32); 45978 v2f64 __builtin_msa_fmax_a_d (v2f64, v2f64); 45979 45980 v4f32 __builtin_msa_fmin_w (v4f32, v4f32); 45981 v2f64 __builtin_msa_fmin_d (v2f64, v2f64); 45982 45983 v4f32 __builtin_msa_fmin_a_w (v4f32, v4f32); 45984 v2f64 __builtin_msa_fmin_a_d (v2f64, v2f64); 45985 45986 v4f32 __builtin_msa_fmsub_w (v4f32, v4f32, v4f32); 45987 v2f64 __builtin_msa_fmsub_d (v2f64, v2f64, v2f64); 45988 45989 v4f32 __builtin_msa_fmul_w (v4f32, v4f32); 45990 v2f64 __builtin_msa_fmul_d (v2f64, v2f64); 45991 45992 v4f32 __builtin_msa_frint_w (v4f32); 45993 v2f64 __builtin_msa_frint_d (v2f64); 45994 45995 v4f32 __builtin_msa_frcp_w (v4f32); 45996 v2f64 __builtin_msa_frcp_d (v2f64); 45997 45998 v4f32 __builtin_msa_frsqrt_w (v4f32); 45999 v2f64 __builtin_msa_frsqrt_d (v2f64); 46000 46001 v4i32 __builtin_msa_fsaf_w (v4f32, v4f32); 46002 v2i64 __builtin_msa_fsaf_d (v2f64, v2f64); 46003 46004 v4i32 __builtin_msa_fseq_w (v4f32, v4f32); 46005 v2i64 __builtin_msa_fseq_d (v2f64, v2f64); 46006 46007 v4i32 __builtin_msa_fsle_w (v4f32, v4f32); 46008 v2i64 __builtin_msa_fsle_d (v2f64, v2f64); 46009 46010 v4i32 __builtin_msa_fslt_w (v4f32, v4f32); 46011 v2i64 __builtin_msa_fslt_d (v2f64, v2f64); 46012 46013 v4i32 __builtin_msa_fsne_w (v4f32, v4f32); 46014 v2i64 __builtin_msa_fsne_d (v2f64, v2f64); 46015 46016 v4i32 __builtin_msa_fsor_w (v4f32, v4f32); 46017 v2i64 __builtin_msa_fsor_d (v2f64, v2f64); 46018 46019 v4f32 __builtin_msa_fsqrt_w (v4f32); 46020 v2f64 __builtin_msa_fsqrt_d (v2f64); 46021 46022 v4f32 __builtin_msa_fsub_w (v4f32, v4f32); 46023 v2f64 __builtin_msa_fsub_d (v2f64, v2f64); 46024 46025 v4i32 __builtin_msa_fsueq_w (v4f32, v4f32); 46026 v2i64 __builtin_msa_fsueq_d (v2f64, v2f64); 46027 46028 v4i32 __builtin_msa_fsule_w (v4f32, v4f32); 46029 v2i64 __builtin_msa_fsule_d (v2f64, v2f64); 46030 46031 v4i32 __builtin_msa_fsult_w (v4f32, v4f32); 46032 v2i64 __builtin_msa_fsult_d (v2f64, v2f64); 46033 46034 v4i32 __builtin_msa_fsun_w (v4f32, v4f32); 46035 v2i64 __builtin_msa_fsun_d (v2f64, v2f64); 46036 46037 v4i32 __builtin_msa_fsune_w (v4f32, v4f32); 46038 v2i64 __builtin_msa_fsune_d (v2f64, v2f64); 46039 46040 v4i32 __builtin_msa_ftint_s_w (v4f32); 46041 v2i64 __builtin_msa_ftint_s_d (v2f64); 46042 46043 v4u32 __builtin_msa_ftint_u_w (v4f32); 46044 v2u64 __builtin_msa_ftint_u_d (v2f64); 46045 46046 v8i16 __builtin_msa_ftq_h (v4f32, v4f32); 46047 v4i32 __builtin_msa_ftq_w (v2f64, v2f64); 46048 46049 v4i32 __builtin_msa_ftrunc_s_w (v4f32); 46050 v2i64 __builtin_msa_ftrunc_s_d (v2f64); 46051 46052 v4u32 __builtin_msa_ftrunc_u_w (v4f32); 46053 v2u64 __builtin_msa_ftrunc_u_d (v2f64); 46054 46055 v8i16 __builtin_msa_hadd_s_h (v16i8, v16i8); 46056 v4i32 __builtin_msa_hadd_s_w (v8i16, v8i16); 46057 v2i64 __builtin_msa_hadd_s_d (v4i32, v4i32); 46058 46059 v8u16 __builtin_msa_hadd_u_h (v16u8, v16u8); 46060 v4u32 __builtin_msa_hadd_u_w (v8u16, v8u16); 46061 v2u64 __builtin_msa_hadd_u_d (v4u32, v4u32); 46062 46063 v8i16 __builtin_msa_hsub_s_h (v16i8, v16i8); 46064 v4i32 __builtin_msa_hsub_s_w (v8i16, v8i16); 46065 v2i64 __builtin_msa_hsub_s_d (v4i32, v4i32); 46066 46067 v8i16 __builtin_msa_hsub_u_h (v16u8, v16u8); 46068 v4i32 __builtin_msa_hsub_u_w (v8u16, v8u16); 46069 v2i64 __builtin_msa_hsub_u_d (v4u32, v4u32); 46070 46071 v16i8 __builtin_msa_ilvev_b (v16i8, v16i8); 46072 v8i16 __builtin_msa_ilvev_h (v8i16, v8i16); 46073 v4i32 __builtin_msa_ilvev_w (v4i32, v4i32); 46074 v2i64 __builtin_msa_ilvev_d (v2i64, v2i64); 46075 46076 v16i8 __builtin_msa_ilvl_b (v16i8, v16i8); 46077 v8i16 __builtin_msa_ilvl_h (v8i16, v8i16); 46078 v4i32 __builtin_msa_ilvl_w (v4i32, v4i32); 46079 v2i64 __builtin_msa_ilvl_d (v2i64, v2i64); 46080 46081 v16i8 __builtin_msa_ilvod_b (v16i8, v16i8); 46082 v8i16 __builtin_msa_ilvod_h (v8i16, v8i16); 46083 v4i32 __builtin_msa_ilvod_w (v4i32, v4i32); 46084 v2i64 __builtin_msa_ilvod_d (v2i64, v2i64); 46085 46086 v16i8 __builtin_msa_ilvr_b (v16i8, v16i8); 46087 v8i16 __builtin_msa_ilvr_h (v8i16, v8i16); 46088 v4i32 __builtin_msa_ilvr_w (v4i32, v4i32); 46089 v2i64 __builtin_msa_ilvr_d (v2i64, v2i64); 46090 46091 v16i8 __builtin_msa_insert_b (v16i8, imm0_15, i32); 46092 v8i16 __builtin_msa_insert_h (v8i16, imm0_7, i32); 46093 v4i32 __builtin_msa_insert_w (v4i32, imm0_3, i32); 46094 v2i64 __builtin_msa_insert_d (v2i64, imm0_1, i64); 46095 46096 v16i8 __builtin_msa_insve_b (v16i8, imm0_15, v16i8); 46097 v8i16 __builtin_msa_insve_h (v8i16, imm0_7, v8i16); 46098 v4i32 __builtin_msa_insve_w (v4i32, imm0_3, v4i32); 46099 v2i64 __builtin_msa_insve_d (v2i64, imm0_1, v2i64); 46100 46101 v16i8 __builtin_msa_ld_b (const void *, imm_n512_511); 46102 v8i16 __builtin_msa_ld_h (const void *, imm_n1024_1022); 46103 v4i32 __builtin_msa_ld_w (const void *, imm_n2048_2044); 46104 v2i64 __builtin_msa_ld_d (const void *, imm_n4096_4088); 46105 46106 v16i8 __builtin_msa_ldi_b (imm_n512_511); 46107 v8i16 __builtin_msa_ldi_h (imm_n512_511); 46108 v4i32 __builtin_msa_ldi_w (imm_n512_511); 46109 v2i64 __builtin_msa_ldi_d (imm_n512_511); 46110 46111 v8i16 __builtin_msa_madd_q_h (v8i16, v8i16, v8i16); 46112 v4i32 __builtin_msa_madd_q_w (v4i32, v4i32, v4i32); 46113 46114 v8i16 __builtin_msa_maddr_q_h (v8i16, v8i16, v8i16); 46115 v4i32 __builtin_msa_maddr_q_w (v4i32, v4i32, v4i32); 46116 46117 v16i8 __builtin_msa_maddv_b (v16i8, v16i8, v16i8); 46118 v8i16 __builtin_msa_maddv_h (v8i16, v8i16, v8i16); 46119 v4i32 __builtin_msa_maddv_w (v4i32, v4i32, v4i32); 46120 v2i64 __builtin_msa_maddv_d (v2i64, v2i64, v2i64); 46121 46122 v16i8 __builtin_msa_max_a_b (v16i8, v16i8); 46123 v8i16 __builtin_msa_max_a_h (v8i16, v8i16); 46124 v4i32 __builtin_msa_max_a_w (v4i32, v4i32); 46125 v2i64 __builtin_msa_max_a_d (v2i64, v2i64); 46126 46127 v16i8 __builtin_msa_max_s_b (v16i8, v16i8); 46128 v8i16 __builtin_msa_max_s_h (v8i16, v8i16); 46129 v4i32 __builtin_msa_max_s_w (v4i32, v4i32); 46130 v2i64 __builtin_msa_max_s_d (v2i64, v2i64); 46131 46132 v16u8 __builtin_msa_max_u_b (v16u8, v16u8); 46133 v8u16 __builtin_msa_max_u_h (v8u16, v8u16); 46134 v4u32 __builtin_msa_max_u_w (v4u32, v4u32); 46135 v2u64 __builtin_msa_max_u_d (v2u64, v2u64); 46136 46137 v16i8 __builtin_msa_maxi_s_b (v16i8, imm_n16_15); 46138 v8i16 __builtin_msa_maxi_s_h (v8i16, imm_n16_15); 46139 v4i32 __builtin_msa_maxi_s_w (v4i32, imm_n16_15); 46140 v2i64 __builtin_msa_maxi_s_d (v2i64, imm_n16_15); 46141 46142 v16u8 __builtin_msa_maxi_u_b (v16u8, imm0_31); 46143 v8u16 __builtin_msa_maxi_u_h (v8u16, imm0_31); 46144 v4u32 __builtin_msa_maxi_u_w (v4u32, imm0_31); 46145 v2u64 __builtin_msa_maxi_u_d (v2u64, imm0_31); 46146 46147 v16i8 __builtin_msa_min_a_b (v16i8, v16i8); 46148 v8i16 __builtin_msa_min_a_h (v8i16, v8i16); 46149 v4i32 __builtin_msa_min_a_w (v4i32, v4i32); 46150 v2i64 __builtin_msa_min_a_d (v2i64, v2i64); 46151 46152 v16i8 __builtin_msa_min_s_b (v16i8, v16i8); 46153 v8i16 __builtin_msa_min_s_h (v8i16, v8i16); 46154 v4i32 __builtin_msa_min_s_w (v4i32, v4i32); 46155 v2i64 __builtin_msa_min_s_d (v2i64, v2i64); 46156 46157 v16u8 __builtin_msa_min_u_b (v16u8, v16u8); 46158 v8u16 __builtin_msa_min_u_h (v8u16, v8u16); 46159 v4u32 __builtin_msa_min_u_w (v4u32, v4u32); 46160 v2u64 __builtin_msa_min_u_d (v2u64, v2u64); 46161 46162 v16i8 __builtin_msa_mini_s_b (v16i8, imm_n16_15); 46163 v8i16 __builtin_msa_mini_s_h (v8i16, imm_n16_15); 46164 v4i32 __builtin_msa_mini_s_w (v4i32, imm_n16_15); 46165 v2i64 __builtin_msa_mini_s_d (v2i64, imm_n16_15); 46166 46167 v16u8 __builtin_msa_mini_u_b (v16u8, imm0_31); 46168 v8u16 __builtin_msa_mini_u_h (v8u16, imm0_31); 46169 v4u32 __builtin_msa_mini_u_w (v4u32, imm0_31); 46170 v2u64 __builtin_msa_mini_u_d (v2u64, imm0_31); 46171 46172 v16i8 __builtin_msa_mod_s_b (v16i8, v16i8); 46173 v8i16 __builtin_msa_mod_s_h (v8i16, v8i16); 46174 v4i32 __builtin_msa_mod_s_w (v4i32, v4i32); 46175 v2i64 __builtin_msa_mod_s_d (v2i64, v2i64); 46176 46177 v16u8 __builtin_msa_mod_u_b (v16u8, v16u8); 46178 v8u16 __builtin_msa_mod_u_h (v8u16, v8u16); 46179 v4u32 __builtin_msa_mod_u_w (v4u32, v4u32); 46180 v2u64 __builtin_msa_mod_u_d (v2u64, v2u64); 46181 46182 v16i8 __builtin_msa_move_v (v16i8); 46183 46184 v8i16 __builtin_msa_msub_q_h (v8i16, v8i16, v8i16); 46185 v4i32 __builtin_msa_msub_q_w (v4i32, v4i32, v4i32); 46186 46187 v8i16 __builtin_msa_msubr_q_h (v8i16, v8i16, v8i16); 46188 v4i32 __builtin_msa_msubr_q_w (v4i32, v4i32, v4i32); 46189 46190 v16i8 __builtin_msa_msubv_b (v16i8, v16i8, v16i8); 46191 v8i16 __builtin_msa_msubv_h (v8i16, v8i16, v8i16); 46192 v4i32 __builtin_msa_msubv_w (v4i32, v4i32, v4i32); 46193 v2i64 __builtin_msa_msubv_d (v2i64, v2i64, v2i64); 46194 46195 v8i16 __builtin_msa_mul_q_h (v8i16, v8i16); 46196 v4i32 __builtin_msa_mul_q_w (v4i32, v4i32); 46197 46198 v8i16 __builtin_msa_mulr_q_h (v8i16, v8i16); 46199 v4i32 __builtin_msa_mulr_q_w (v4i32, v4i32); 46200 46201 v16i8 __builtin_msa_mulv_b (v16i8, v16i8); 46202 v8i16 __builtin_msa_mulv_h (v8i16, v8i16); 46203 v4i32 __builtin_msa_mulv_w (v4i32, v4i32); 46204 v2i64 __builtin_msa_mulv_d (v2i64, v2i64); 46205 46206 v16i8 __builtin_msa_nloc_b (v16i8); 46207 v8i16 __builtin_msa_nloc_h (v8i16); 46208 v4i32 __builtin_msa_nloc_w (v4i32); 46209 v2i64 __builtin_msa_nloc_d (v2i64); 46210 46211 v16i8 __builtin_msa_nlzc_b (v16i8); 46212 v8i16 __builtin_msa_nlzc_h (v8i16); 46213 v4i32 __builtin_msa_nlzc_w (v4i32); 46214 v2i64 __builtin_msa_nlzc_d (v2i64); 46215 46216 v16u8 __builtin_msa_nor_v (v16u8, v16u8); 46217 46218 v16u8 __builtin_msa_nori_b (v16u8, imm0_255); 46219 46220 v16u8 __builtin_msa_or_v (v16u8, v16u8); 46221 46222 v16u8 __builtin_msa_ori_b (v16u8, imm0_255); 46223 46224 v16i8 __builtin_msa_pckev_b (v16i8, v16i8); 46225 v8i16 __builtin_msa_pckev_h (v8i16, v8i16); 46226 v4i32 __builtin_msa_pckev_w (v4i32, v4i32); 46227 v2i64 __builtin_msa_pckev_d (v2i64, v2i64); 46228 46229 v16i8 __builtin_msa_pckod_b (v16i8, v16i8); 46230 v8i16 __builtin_msa_pckod_h (v8i16, v8i16); 46231 v4i32 __builtin_msa_pckod_w (v4i32, v4i32); 46232 v2i64 __builtin_msa_pckod_d (v2i64, v2i64); 46233 46234 v16i8 __builtin_msa_pcnt_b (v16i8); 46235 v8i16 __builtin_msa_pcnt_h (v8i16); 46236 v4i32 __builtin_msa_pcnt_w (v4i32); 46237 v2i64 __builtin_msa_pcnt_d (v2i64); 46238 46239 v16i8 __builtin_msa_sat_s_b (v16i8, imm0_7); 46240 v8i16 __builtin_msa_sat_s_h (v8i16, imm0_15); 46241 v4i32 __builtin_msa_sat_s_w (v4i32, imm0_31); 46242 v2i64 __builtin_msa_sat_s_d (v2i64, imm0_63); 46243 46244 v16u8 __builtin_msa_sat_u_b (v16u8, imm0_7); 46245 v8u16 __builtin_msa_sat_u_h (v8u16, imm0_15); 46246 v4u32 __builtin_msa_sat_u_w (v4u32, imm0_31); 46247 v2u64 __builtin_msa_sat_u_d (v2u64, imm0_63); 46248 46249 v16i8 __builtin_msa_shf_b (v16i8, imm0_255); 46250 v8i16 __builtin_msa_shf_h (v8i16, imm0_255); 46251 v4i32 __builtin_msa_shf_w (v4i32, imm0_255); 46252 46253 v16i8 __builtin_msa_sld_b (v16i8, v16i8, i32); 46254 v8i16 __builtin_msa_sld_h (v8i16, v8i16, i32); 46255 v4i32 __builtin_msa_sld_w (v4i32, v4i32, i32); 46256 v2i64 __builtin_msa_sld_d (v2i64, v2i64, i32); 46257 46258 v16i8 __builtin_msa_sldi_b (v16i8, v16i8, imm0_15); 46259 v8i16 __builtin_msa_sldi_h (v8i16, v8i16, imm0_7); 46260 v4i32 __builtin_msa_sldi_w (v4i32, v4i32, imm0_3); 46261 v2i64 __builtin_msa_sldi_d (v2i64, v2i64, imm0_1); 46262 46263 v16i8 __builtin_msa_sll_b (v16i8, v16i8); 46264 v8i16 __builtin_msa_sll_h (v8i16, v8i16); 46265 v4i32 __builtin_msa_sll_w (v4i32, v4i32); 46266 v2i64 __builtin_msa_sll_d (v2i64, v2i64); 46267 46268 v16i8 __builtin_msa_slli_b (v16i8, imm0_7); 46269 v8i16 __builtin_msa_slli_h (v8i16, imm0_15); 46270 v4i32 __builtin_msa_slli_w (v4i32, imm0_31); 46271 v2i64 __builtin_msa_slli_d (v2i64, imm0_63); 46272 46273 v16i8 __builtin_msa_splat_b (v16i8, i32); 46274 v8i16 __builtin_msa_splat_h (v8i16, i32); 46275 v4i32 __builtin_msa_splat_w (v4i32, i32); 46276 v2i64 __builtin_msa_splat_d (v2i64, i32); 46277 46278 v16i8 __builtin_msa_splati_b (v16i8, imm0_15); 46279 v8i16 __builtin_msa_splati_h (v8i16, imm0_7); 46280 v4i32 __builtin_msa_splati_w (v4i32, imm0_3); 46281 v2i64 __builtin_msa_splati_d (v2i64, imm0_1); 46282 46283 v16i8 __builtin_msa_sra_b (v16i8, v16i8); 46284 v8i16 __builtin_msa_sra_h (v8i16, v8i16); 46285 v4i32 __builtin_msa_sra_w (v4i32, v4i32); 46286 v2i64 __builtin_msa_sra_d (v2i64, v2i64); 46287 46288 v16i8 __builtin_msa_srai_b (v16i8, imm0_7); 46289 v8i16 __builtin_msa_srai_h (v8i16, imm0_15); 46290 v4i32 __builtin_msa_srai_w (v4i32, imm0_31); 46291 v2i64 __builtin_msa_srai_d (v2i64, imm0_63); 46292 46293 v16i8 __builtin_msa_srar_b (v16i8, v16i8); 46294 v8i16 __builtin_msa_srar_h (v8i16, v8i16); 46295 v4i32 __builtin_msa_srar_w (v4i32, v4i32); 46296 v2i64 __builtin_msa_srar_d (v2i64, v2i64); 46297 46298 v16i8 __builtin_msa_srari_b (v16i8, imm0_7); 46299 v8i16 __builtin_msa_srari_h (v8i16, imm0_15); 46300 v4i32 __builtin_msa_srari_w (v4i32, imm0_31); 46301 v2i64 __builtin_msa_srari_d (v2i64, imm0_63); 46302 46303 v16i8 __builtin_msa_srl_b (v16i8, v16i8); 46304 v8i16 __builtin_msa_srl_h (v8i16, v8i16); 46305 v4i32 __builtin_msa_srl_w (v4i32, v4i32); 46306 v2i64 __builtin_msa_srl_d (v2i64, v2i64); 46307 46308 v16i8 __builtin_msa_srli_b (v16i8, imm0_7); 46309 v8i16 __builtin_msa_srli_h (v8i16, imm0_15); 46310 v4i32 __builtin_msa_srli_w (v4i32, imm0_31); 46311 v2i64 __builtin_msa_srli_d (v2i64, imm0_63); 46312 46313 v16i8 __builtin_msa_srlr_b (v16i8, v16i8); 46314 v8i16 __builtin_msa_srlr_h (v8i16, v8i16); 46315 v4i32 __builtin_msa_srlr_w (v4i32, v4i32); 46316 v2i64 __builtin_msa_srlr_d (v2i64, v2i64); 46317 46318 v16i8 __builtin_msa_srlri_b (v16i8, imm0_7); 46319 v8i16 __builtin_msa_srlri_h (v8i16, imm0_15); 46320 v4i32 __builtin_msa_srlri_w (v4i32, imm0_31); 46321 v2i64 __builtin_msa_srlri_d (v2i64, imm0_63); 46322 46323 void __builtin_msa_st_b (v16i8, void *, imm_n512_511); 46324 void __builtin_msa_st_h (v8i16, void *, imm_n1024_1022); 46325 void __builtin_msa_st_w (v4i32, void *, imm_n2048_2044); 46326 void __builtin_msa_st_d (v2i64, void *, imm_n4096_4088); 46327 46328 v16i8 __builtin_msa_subs_s_b (v16i8, v16i8); 46329 v8i16 __builtin_msa_subs_s_h (v8i16, v8i16); 46330 v4i32 __builtin_msa_subs_s_w (v4i32, v4i32); 46331 v2i64 __builtin_msa_subs_s_d (v2i64, v2i64); 46332 46333 v16u8 __builtin_msa_subs_u_b (v16u8, v16u8); 46334 v8u16 __builtin_msa_subs_u_h (v8u16, v8u16); 46335 v4u32 __builtin_msa_subs_u_w (v4u32, v4u32); 46336 v2u64 __builtin_msa_subs_u_d (v2u64, v2u64); 46337 46338 v16u8 __builtin_msa_subsus_u_b (v16u8, v16i8); 46339 v8u16 __builtin_msa_subsus_u_h (v8u16, v8i16); 46340 v4u32 __builtin_msa_subsus_u_w (v4u32, v4i32); 46341 v2u64 __builtin_msa_subsus_u_d (v2u64, v2i64); 46342 46343 v16i8 __builtin_msa_subsuu_s_b (v16u8, v16u8); 46344 v8i16 __builtin_msa_subsuu_s_h (v8u16, v8u16); 46345 v4i32 __builtin_msa_subsuu_s_w (v4u32, v4u32); 46346 v2i64 __builtin_msa_subsuu_s_d (v2u64, v2u64); 46347 46348 v16i8 __builtin_msa_subv_b (v16i8, v16i8); 46349 v8i16 __builtin_msa_subv_h (v8i16, v8i16); 46350 v4i32 __builtin_msa_subv_w (v4i32, v4i32); 46351 v2i64 __builtin_msa_subv_d (v2i64, v2i64); 46352 46353 v16i8 __builtin_msa_subvi_b (v16i8, imm0_31); 46354 v8i16 __builtin_msa_subvi_h (v8i16, imm0_31); 46355 v4i32 __builtin_msa_subvi_w (v4i32, imm0_31); 46356 v2i64 __builtin_msa_subvi_d (v2i64, imm0_31); 46357 46358 v16i8 __builtin_msa_vshf_b (v16i8, v16i8, v16i8); 46359 v8i16 __builtin_msa_vshf_h (v8i16, v8i16, v8i16); 46360 v4i32 __builtin_msa_vshf_w (v4i32, v4i32, v4i32); 46361 v2i64 __builtin_msa_vshf_d (v2i64, v2i64, v2i64); 46362 46363 v16u8 __builtin_msa_xor_v (v16u8, v16u8); 46364 46365 v16u8 __builtin_msa_xori_b (v16u8, imm0_255); 46366 46367 46368File: gcc.info, Node: Other MIPS Built-in Functions, Next: MSP430 Built-in Functions, Prev: MIPS SIMD Architecture (MSA) Support, Up: Target Builtins 46369 463706.60.18 Other MIPS Built-in Functions 46371------------------------------------- 46372 46373GCC provides other MIPS-specific built-in functions: 46374 46375'void __builtin_mips_cache (int OP, const volatile void *ADDR)' 46376 Insert a 'cache' instruction with operands OP and ADDR. GCC 46377 defines the preprocessor macro '___GCC_HAVE_BUILTIN_MIPS_CACHE' 46378 when this function is available. 46379 46380'unsigned int __builtin_mips_get_fcsr (void)' 46381'void __builtin_mips_set_fcsr (unsigned int VALUE)' 46382 Get and set the contents of the floating-point control and status 46383 register (FPU control register 31). These functions are only 46384 available in hard-float code but can be called in both MIPS16 and 46385 non-MIPS16 contexts. 46386 46387 '__builtin_mips_set_fcsr' can be used to change any bit of the 46388 register except the condition codes, which GCC assumes are 46389 preserved. 46390 46391 46392File: gcc.info, Node: MSP430 Built-in Functions, Next: NDS32 Built-in Functions, Prev: Other MIPS Built-in Functions, Up: Target Builtins 46393 463946.60.19 MSP430 Built-in Functions 46395--------------------------------- 46396 46397GCC provides a couple of special builtin functions to aid in the writing 46398of interrupt handlers in C. 46399 46400'__bic_SR_register_on_exit (int MASK)' 46401 This clears the indicated bits in the saved copy of the status 46402 register currently residing on the stack. This only works inside 46403 interrupt handlers and the changes to the status register will only 46404 take affect once the handler returns. 46405 46406'__bis_SR_register_on_exit (int MASK)' 46407 This sets the indicated bits in the saved copy of the status 46408 register currently residing on the stack. This only works inside 46409 interrupt handlers and the changes to the status register will only 46410 take affect once the handler returns. 46411 46412'__delay_cycles (long long CYCLES)' 46413 This inserts an instruction sequence that takes exactly CYCLES 46414 cycles (between 0 and about 17E9) to complete. The inserted 46415 sequence may use jumps, loops, or no-ops, and does not interfere 46416 with any other instructions. Note that CYCLES must be a 46417 compile-time constant integer - that is, you must pass a number, 46418 not a variable that may be optimized to a constant later. The 46419 number of cycles delayed by this builtin is exact. 46420 46421 46422File: gcc.info, Node: NDS32 Built-in Functions, Next: picoChip Built-in Functions, Prev: MSP430 Built-in Functions, Up: Target Builtins 46423 464246.60.20 NDS32 Built-in Functions 46425-------------------------------- 46426 46427These built-in functions are available for the NDS32 target: 46428 46429 -- Built-in Function: void __builtin_nds32_isync (int *ADDR) 46430 Insert an ISYNC instruction into the instruction stream where ADDR 46431 is an instruction address for serialization. 46432 46433 -- Built-in Function: void __builtin_nds32_isb (void) 46434 Insert an ISB instruction into the instruction stream. 46435 46436 -- Built-in Function: int __builtin_nds32_mfsr (int SR) 46437 Return the content of a system register which is mapped by SR. 46438 46439 -- Built-in Function: int __builtin_nds32_mfusr (int USR) 46440 Return the content of a user space register which is mapped by USR. 46441 46442 -- Built-in Function: void __builtin_nds32_mtsr (int VALUE, int SR) 46443 Move the VALUE to a system register which is mapped by SR. 46444 46445 -- Built-in Function: void __builtin_nds32_mtusr (int VALUE, int USR) 46446 Move the VALUE to a user space register which is mapped by USR. 46447 46448 -- Built-in Function: void __builtin_nds32_setgie_en (void) 46449 Enable global interrupt. 46450 46451 -- Built-in Function: void __builtin_nds32_setgie_dis (void) 46452 Disable global interrupt. 46453 46454 46455File: gcc.info, Node: picoChip Built-in Functions, Next: Basic PowerPC Built-in Functions, Prev: NDS32 Built-in Functions, Up: Target Builtins 46456 464576.60.21 picoChip Built-in Functions 46458----------------------------------- 46459 46460GCC provides an interface to selected machine instructions from the 46461picoChip instruction set. 46462 46463'int __builtin_sbc (int VALUE)' 46464 Sign bit count. Return the number of consecutive bits in VALUE 46465 that have the same value as the sign bit. The result is the number 46466 of leading sign bits minus one, giving the number of redundant sign 46467 bits in VALUE. 46468 46469'int __builtin_byteswap (int VALUE)' 46470 Byte swap. Return the result of swapping the upper and lower bytes 46471 of VALUE. 46472 46473'int __builtin_brev (int VALUE)' 46474 Bit reversal. Return the result of reversing the bits in VALUE. 46475 Bit 15 is swapped with bit 0, bit 14 is swapped with bit 1, and so 46476 on. 46477 46478'int __builtin_adds (int X, int Y)' 46479 Saturating addition. Return the result of adding X and Y, storing 46480 the value 32767 if the result overflows. 46481 46482'int __builtin_subs (int X, int Y)' 46483 Saturating subtraction. Return the result of subtracting Y from X, 46484 storing the value -32768 if the result overflows. 46485 46486'void __builtin_halt (void)' 46487 Halt. The processor stops execution. This built-in is useful for 46488 implementing assertions. 46489 46490 46491File: gcc.info, Node: Basic PowerPC Built-in Functions, Next: PowerPC AltiVec/VSX Built-in Functions, Prev: picoChip Built-in Functions, Up: Target Builtins 46492 464936.60.22 Basic PowerPC Built-in Functions 46494---------------------------------------- 46495 46496* Menu: 46497 46498* Basic PowerPC Built-in Functions Available on all Configurations:: 46499* Basic PowerPC Built-in Functions Available on ISA 2.05:: 46500* Basic PowerPC Built-in Functions Available on ISA 2.06:: 46501* Basic PowerPC Built-in Functions Available on ISA 2.07:: 46502* Basic PowerPC Built-in Functions Available on ISA 3.0:: 46503 46504This section describes PowerPC built-in functions that do not require 46505the inclusion of any special header files to declare prototypes or 46506provide macro definitions. The sections that follow describe additional 46507PowerPC built-in functions. 46508 46509 46510File: gcc.info, Node: Basic PowerPC Built-in Functions Available on all Configurations, Next: Basic PowerPC Built-in Functions Available on ISA 2.05, Up: Basic PowerPC Built-in Functions 46511 465126.60.22.1 Basic PowerPC Built-in Functions Available on all Configurations 46513.......................................................................... 46514 46515 -- Built-in Function: void __builtin_cpu_init (void) 46516 This function is a 'nop' on the PowerPC platform and is included 46517 solely to maintain API compatibility with the x86 builtins. 46518 46519 -- Built-in Function: int __builtin_cpu_is (const char *CPUNAME) 46520 This function returns a value of '1' if the run-time CPU is of type 46521 CPUNAME and returns '0' otherwise 46522 46523 The '__builtin_cpu_is' function requires GLIBC 2.23 or newer which 46524 exports the hardware capability bits. GCC defines the macro 46525 '__BUILTIN_CPU_SUPPORTS__' if the '__builtin_cpu_supports' built-in 46526 function is fully supported. 46527 46528 If GCC was configured to use a GLIBC before 2.23, the built-in 46529 function '__builtin_cpu_is' always returns a 0 and the compiler 46530 issues a warning. 46531 46532 The following CPU names can be detected: 46533 46534 'power10' 46535 IBM POWER10 Server CPU. 46536 'power9' 46537 IBM POWER9 Server CPU. 46538 'power8' 46539 IBM POWER8 Server CPU. 46540 'power7' 46541 IBM POWER7 Server CPU. 46542 'power6x' 46543 IBM POWER6 Server CPU (RAW mode). 46544 'power6' 46545 IBM POWER6 Server CPU (Architected mode). 46546 'power5+' 46547 IBM POWER5+ Server CPU. 46548 'power5' 46549 IBM POWER5 Server CPU. 46550 'ppc970' 46551 IBM 970 Server CPU (ie, Apple G5). 46552 'power4' 46553 IBM POWER4 Server CPU. 46554 'ppca2' 46555 IBM A2 64-bit Embedded CPU 46556 'ppc476' 46557 IBM PowerPC 476FP 32-bit Embedded CPU. 46558 'ppc464' 46559 IBM PowerPC 464 32-bit Embedded CPU. 46560 'ppc440' 46561 PowerPC 440 32-bit Embedded CPU. 46562 'ppc405' 46563 PowerPC 405 32-bit Embedded CPU. 46564 'ppc-cell-be' 46565 IBM PowerPC Cell Broadband Engine Architecture CPU. 46566 46567 Here is an example: 46568 #ifdef __BUILTIN_CPU_SUPPORTS__ 46569 if (__builtin_cpu_is ("power8")) 46570 { 46571 do_power8 (); // POWER8 specific implementation. 46572 } 46573 else 46574 #endif 46575 { 46576 do_generic (); // Generic implementation. 46577 } 46578 46579 -- Built-in Function: int __builtin_cpu_supports (const char *FEATURE) 46580 This function returns a value of '1' if the run-time CPU supports 46581 the HWCAP feature FEATURE and returns '0' otherwise. 46582 46583 The '__builtin_cpu_supports' function requires GLIBC 2.23 or newer 46584 which exports the hardware capability bits. GCC defines the macro 46585 '__BUILTIN_CPU_SUPPORTS__' if the '__builtin_cpu_supports' built-in 46586 function is fully supported. 46587 46588 If GCC was configured to use a GLIBC before 2.23, the built-in 46589 function '__builtin_cpu_suports' always returns a 0 and the 46590 compiler issues a warning. 46591 46592 The following features can be detected: 46593 46594 '4xxmac' 46595 4xx CPU has a Multiply Accumulator. 46596 'altivec' 46597 CPU has a SIMD/Vector Unit. 46598 'arch_2_05' 46599 CPU supports ISA 2.05 (eg, POWER6) 46600 'arch_2_06' 46601 CPU supports ISA 2.06 (eg, POWER7) 46602 'arch_2_07' 46603 CPU supports ISA 2.07 (eg, POWER8) 46604 'arch_3_00' 46605 CPU supports ISA 3.0 (eg, POWER9) 46606 'arch_3_1' 46607 CPU supports ISA 3.1 (eg, POWER10) 46608 'archpmu' 46609 CPU supports the set of compatible performance monitoring 46610 events. 46611 'booke' 46612 CPU supports the Embedded ISA category. 46613 'cellbe' 46614 CPU has a CELL broadband engine. 46615 'darn' 46616 CPU supports the 'darn' (deliver a random number) instruction. 46617 'dfp' 46618 CPU has a decimal floating point unit. 46619 'dscr' 46620 CPU supports the data stream control register. 46621 'ebb' 46622 CPU supports event base branching. 46623 'efpdouble' 46624 CPU has a SPE double precision floating point unit. 46625 'efpsingle' 46626 CPU has a SPE single precision floating point unit. 46627 'fpu' 46628 CPU has a floating point unit. 46629 'htm' 46630 CPU has hardware transaction memory instructions. 46631 'htm-nosc' 46632 Kernel aborts hardware transactions when a syscall is made. 46633 'htm-no-suspend' 46634 CPU supports hardware transaction memory but does not support 46635 the 'tsuspend.' instruction. 46636 'ic_snoop' 46637 CPU supports icache snooping capabilities. 46638 'ieee128' 46639 CPU supports 128-bit IEEE binary floating point instructions. 46640 'isel' 46641 CPU supports the integer select instruction. 46642 'mma' 46643 CPU supports the matrix-multiply assist instructions. 46644 'mmu' 46645 CPU has a memory management unit. 46646 'notb' 46647 CPU does not have a timebase (eg, 601 and 403gx). 46648 'pa6t' 46649 CPU supports the PA Semi 6T CORE ISA. 46650 'power4' 46651 CPU supports ISA 2.00 (eg, POWER4) 46652 'power5' 46653 CPU supports ISA 2.02 (eg, POWER5) 46654 'power5+' 46655 CPU supports ISA 2.03 (eg, POWER5+) 46656 'power6x' 46657 CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and 46658 mftgpr. 46659 'ppc32' 46660 CPU supports 32-bit mode execution. 46661 'ppc601' 46662 CPU supports the old POWER ISA (eg, 601) 46663 'ppc64' 46664 CPU supports 64-bit mode execution. 46665 'ppcle' 46666 CPU supports a little-endian mode that uses address swizzling. 46667 'scv' 46668 Kernel supports system call vectored. 46669 'smt' 46670 CPU support simultaneous multi-threading. 46671 'spe' 46672 CPU has a signal processing extension unit. 46673 'tar' 46674 CPU supports the target address register. 46675 'true_le' 46676 CPU supports true little-endian mode. 46677 'ucache' 46678 CPU has unified I/D cache. 46679 'vcrypto' 46680 CPU supports the vector cryptography instructions. 46681 'vsx' 46682 CPU supports the vector-scalar extension. 46683 46684 Here is an example: 46685 #ifdef __BUILTIN_CPU_SUPPORTS__ 46686 if (__builtin_cpu_supports ("fpu")) 46687 { 46688 asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2)); 46689 } 46690 else 46691 #endif 46692 { 46693 dst = __fadd (src1, src2); // Software FP addition function. 46694 } 46695 46696 The following built-in functions are also available on all PowerPC 46697processors: 46698 uint64_t __builtin_ppc_get_timebase (); 46699 unsigned long __builtin_ppc_mftb (); 46700 double __builtin_unpack_ibm128 (__ibm128, int); 46701 __ibm128 __builtin_pack_ibm128 (double, double); 46702 double __builtin_mffs (void); 46703 void __builtin_mtfsf (const int, double); 46704 void __builtin_mtfsb0 (const int); 46705 void __builtin_mtfsb1 (const int); 46706 void __builtin_set_fpscr_rn (int); 46707 46708 The '__builtin_ppc_get_timebase' and '__builtin_ppc_mftb' functions 46709generate instructions to read the Time Base Register. The 46710'__builtin_ppc_get_timebase' function may generate multiple instructions 46711and always returns the 64 bits of the Time Base Register. The 46712'__builtin_ppc_mftb' function always generates one instruction and 46713returns the Time Base Register value as an unsigned long, throwing away 46714the most significant word on 32-bit environments. The '__builtin_mffs' 46715return the value of the FPSCR register. Note, ISA 3.0 supports the 46716'__builtin_mffsl()' which permits software to read the control and 46717non-sticky status bits in the FSPCR without the higher latency 46718associated with accessing the sticky status bits. The '__builtin_mtfsf' 46719takes a constant 8-bit integer field mask and a double precision 46720floating point argument and generates the 'mtfsf' (extended mnemonic) 46721instruction to write new values to selected fields of the FPSCR. The 46722'__builtin_mtfsb0' and '__builtin_mtfsb1' take the bit to change as an 46723argument. The valid bit range is between 0 and 31. The builtins map to 46724the 'mtfsb0' and 'mtfsb1' instructions which take the argument and add 4672532. Hence these instructions only modify the FPSCR[32:63] bits by 46726changing the specified bit to a zero or one respectively. The 46727'__builtin_set_fpscr_rn' builtin allows changing both of the floating 46728point rounding mode bits. The argument is a 2-bit value. The argument 46729can either be a 'const int' or stored in a variable. The builtin uses 46730the ISA 3.0 instruction 'mffscrn' if available, otherwise it reads the 46731FPSCR, masks the current rounding mode bits out and OR's in the new 46732value. 46733 46734 46735File: gcc.info, Node: Basic PowerPC Built-in Functions Available on ISA 2.05, Next: Basic PowerPC Built-in Functions Available on ISA 2.06, Prev: Basic PowerPC Built-in Functions Available on all Configurations, Up: Basic PowerPC Built-in Functions 46736 467376.60.22.2 Basic PowerPC Built-in Functions Available on ISA 2.05 46738................................................................ 46739 46740The basic built-in functions described in this section are available on 46741the PowerPC family of processors starting with ISA 2.05 or later. 46742Unless specific options are explicitly disabled on the command line, 46743specifying option '-mcpu=power6' has the effect of enabling the 46744'-mpowerpc64', '-mpowerpc-gpopt', '-mpowerpc-gfxopt', '-mmfcrf', 46745'-mpopcntb', '-mfprnd', '-mcmpb', '-mhard-dfp', and '-mrecip-precision' 46746options. Specify the '-maltivec' option explicitly in combination with 46747the above options if desired. 46748 46749 The following functions require option '-mcmpb'. 46750 unsigned long long __builtin_cmpb (unsigned long long int, unsigned long long int); 46751 unsigned int __builtin_cmpb (unsigned int, unsigned int); 46752 46753 The '__builtin_cmpb' function performs a byte-wise compare on the 46754contents of its two arguments, returning the result of the byte-wise 46755comparison as the returned value. For each byte comparison, the 46756corresponding byte of the return value holds 0xff if the input bytes are 46757equal and 0 if the input bytes are not equal. If either of the 46758arguments to this built-in function is wider than 32 bits, the function 46759call expands into the form that expects 'unsigned long long int' 46760arguments which is only available on 64-bit targets. 46761 46762 The following built-in functions are available when hardware decimal 46763floating point ('-mhard-dfp') is available: 46764 void __builtin_set_fpscr_drn(int); 46765 _Decimal64 __builtin_ddedpd (int, _Decimal64); 46766 _Decimal128 __builtin_ddedpdq (int, _Decimal128); 46767 _Decimal64 __builtin_denbcd (int, _Decimal64); 46768 _Decimal128 __builtin_denbcdq (int, _Decimal128); 46769 _Decimal64 __builtin_diex (long long, _Decimal64); 46770 _Decimal128 _builtin_diexq (long long, _Decimal128); 46771 _Decimal64 __builtin_dscli (_Decimal64, int); 46772 _Decimal128 __builtin_dscliq (_Decimal128, int); 46773 _Decimal64 __builtin_dscri (_Decimal64, int); 46774 _Decimal128 __builtin_dscriq (_Decimal128, int); 46775 long long __builtin_dxex (_Decimal64); 46776 long long __builtin_dxexq (_Decimal128); 46777 _Decimal128 __builtin_pack_dec128 (unsigned long long, unsigned long long); 46778 unsigned long long __builtin_unpack_dec128 (_Decimal128, int); 46779 46780 The __builtin_set_fpscr_drn builtin allows changing the three decimal 46781 floating point rounding mode bits. The argument is a 3-bit value. The 46782 argument can either be a const int or the value can be stored in 46783 a variable. 46784 The builtin uses the ISA 3.0 instruction mffscdrn if available. 46785 Otherwise the builtin reads the FPSCR, masks the current decimal rounding 46786 mode bits out and OR's in the new value. 46787 46788 46789 The following functions require '-mhard-float', '-mpowerpc-gfxopt', and 46790'-mpopcntb' options. 46791 46792 double __builtin_recipdiv (double, double); 46793 float __builtin_recipdivf (float, float); 46794 double __builtin_rsqrt (double); 46795 float __builtin_rsqrtf (float); 46796 46797 The 'vec_rsqrt', '__builtin_rsqrt', and '__builtin_rsqrtf' functions 46798generate multiple instructions to implement the reciprocal sqrt 46799functionality using reciprocal sqrt estimate instructions. 46800 46801 The '__builtin_recipdiv', and '__builtin_recipdivf' functions generate 46802multiple instructions to implement division using the reciprocal 46803estimate instructions. 46804 46805 The following functions require '-mhard-float' and '-mmultiple' 46806options. 46807 46808 The '__builtin_unpack_longdouble' function takes a 'long double' 46809argument and a compile time constant of 0 or 1. If the constant is 0, 46810the first 'double' within the 'long double' is returned, otherwise the 46811second 'double' is returned. The '__builtin_unpack_longdouble' function 46812is only available if 'long double' uses the IBM extended double 46813representation. 46814 46815 The '__builtin_pack_longdouble' function takes two 'double' arguments 46816and returns a 'long double' value that combines the two arguments. The 46817'__builtin_pack_longdouble' function is only available if 'long double' 46818uses the IBM extended double representation. 46819 46820 The '__builtin_unpack_ibm128' function takes a '__ibm128' argument and 46821a compile time constant of 0 or 1. If the constant is 0, the first 46822'double' within the '__ibm128' is returned, otherwise the second 46823'double' is returned. 46824 46825 The '__builtin_pack_ibm128' function takes two 'double' arguments and 46826returns a '__ibm128' value that combines the two arguments. 46827 46828 Additional built-in functions are available for the 64-bit PowerPC 46829family of processors, for efficient use of 128-bit floating point 46830('__float128') values. 46831 46832 46833File: gcc.info, Node: Basic PowerPC Built-in Functions Available on ISA 2.06, Next: Basic PowerPC Built-in Functions Available on ISA 2.07, Prev: Basic PowerPC Built-in Functions Available on ISA 2.05, Up: Basic PowerPC Built-in Functions 46834 468356.60.22.3 Basic PowerPC Built-in Functions Available on ISA 2.06 46836................................................................ 46837 46838The basic built-in functions described in this section are available on 46839the PowerPC family of processors starting with ISA 2.05 or later. 46840Unless specific options are explicitly disabled on the command line, 46841specifying option '-mcpu=power7' has the effect of enabling all the same 46842options as for '-mcpu=power6' in addition to the '-maltivec', 46843'-mpopcntd', and '-mvsx' options. 46844 46845 The following basic built-in functions require '-mpopcntd': 46846 unsigned int __builtin_addg6s (unsigned int, unsigned int); 46847 long long __builtin_bpermd (long long, long long); 46848 unsigned int __builtin_cbcdtd (unsigned int); 46849 unsigned int __builtin_cdtbcd (unsigned int); 46850 long long __builtin_divde (long long, long long); 46851 unsigned long long __builtin_divdeu (unsigned long long, unsigned long long); 46852 int __builtin_divwe (int, int); 46853 unsigned int __builtin_divweu (unsigned int, unsigned int); 46854 vector __int128 __builtin_pack_vector_int128 (long long, long long); 46855 void __builtin_rs6000_speculation_barrier (void); 46856 long long __builtin_unpack_vector_int128 (vector __int128, signed char); 46857 46858 Of these, the '__builtin_divde' and '__builtin_divdeu' functions 46859require a 64-bit environment. 46860 46861 The following basic built-in functions, which are also supported on x86 46862targets, require '-mfloat128'. 46863 __float128 __builtin_fabsq (__float128); 46864 __float128 __builtin_copysignq (__float128, __float128); 46865 __float128 __builtin_infq (void); 46866 __float128 __builtin_huge_valq (void); 46867 __float128 __builtin_nanq (void); 46868 __float128 __builtin_nansq (void); 46869 46870 __float128 __builtin_sqrtf128 (__float128); 46871 __float128 __builtin_fmaf128 (__float128, __float128, __float128); 46872 46873 46874File: gcc.info, Node: Basic PowerPC Built-in Functions Available on ISA 2.07, Next: Basic PowerPC Built-in Functions Available on ISA 3.0, Prev: Basic PowerPC Built-in Functions Available on ISA 2.06, Up: Basic PowerPC Built-in Functions 46875 468766.60.22.4 Basic PowerPC Built-in Functions Available on ISA 2.07 46877................................................................ 46878 46879The basic built-in functions described in this section are available on 46880the PowerPC family of processors starting with ISA 2.07 or later. 46881Unless specific options are explicitly disabled on the command line, 46882specifying option '-mcpu=power8' has the effect of enabling all the same 46883options as for '-mcpu=power7' in addition to the '-mpower8-fusion', 46884'-mpower8-vector', '-mcrypto', '-mhtm', '-mquad-memory', and 46885'-mquad-memory-atomic' options. 46886 46887 This section intentionally empty. 46888 46889 46890File: gcc.info, Node: Basic PowerPC Built-in Functions Available on ISA 3.0, Prev: Basic PowerPC Built-in Functions Available on ISA 2.07, Up: Basic PowerPC Built-in Functions 46891 468926.60.22.5 Basic PowerPC Built-in Functions Available on ISA 3.0 46893............................................................... 46894 46895The basic built-in functions described in this section are available on 46896the PowerPC family of processors starting with ISA 3.0 or later. Unless 46897specific options are explicitly disabled on the command line, specifying 46898option '-mcpu=power9' has the effect of enabling all the same options as 46899for '-mcpu=power8' in addition to the '-misel' option. 46900 46901 The following built-in functions are available on Linux 64-bit systems 46902that use the ISA 3.0 instruction set ('-mcpu=power9'): 46903 46904'__float128 __builtin_addf128_round_to_odd (__float128, __float128)' 46905 Perform a 128-bit IEEE floating point add using round to odd as the 46906 rounding mode. 46907 46908'__float128 __builtin_subf128_round_to_odd (__float128, __float128)' 46909 Perform a 128-bit IEEE floating point subtract using round to odd 46910 as the rounding mode. 46911 46912'__float128 __builtin_mulf128_round_to_odd (__float128, __float128)' 46913 Perform a 128-bit IEEE floating point multiply using round to odd 46914 as the rounding mode. 46915 46916'__float128 __builtin_divf128_round_to_odd (__float128, __float128)' 46917 Perform a 128-bit IEEE floating point divide using round to odd as 46918 the rounding mode. 46919 46920'__float128 __builtin_sqrtf128_round_to_odd (__float128)' 46921 Perform a 128-bit IEEE floating point square root using round to 46922 odd as the rounding mode. 46923 46924'__float128 __builtin_fmaf128_round_to_odd (__float128, __float128, __float128)' 46925 Perform a 128-bit IEEE floating point fused multiply and add 46926 operation using round to odd as the rounding mode. 46927 46928'double __builtin_truncf128_round_to_odd (__float128)' 46929 Convert a 128-bit IEEE floating point value to 'double' using round 46930 to odd as the rounding mode. 46931 46932 The following additional built-in functions are also available for the 46933PowerPC family of processors, starting with ISA 3.0 or later: 46934 long long __builtin_darn (void); 46935 long long __builtin_darn_raw (void); 46936 int __builtin_darn_32 (void); 46937 46938 The '__builtin_darn' and '__builtin_darn_raw' functions require a 4693964-bit environment supporting ISA 3.0 or later. The '__builtin_darn' 46940function provides a 64-bit conditioned random number. The 46941'__builtin_darn_raw' function provides a 64-bit raw random number. The 46942'__builtin_darn_32' function provides a 32-bit conditioned random 46943number. 46944 46945 The following additional built-in functions are also available for the 46946PowerPC family of processors, starting with ISA 3.0 or later: 46947 46948 int __builtin_byte_in_set (unsigned char u, unsigned long long set); 46949 int __builtin_byte_in_range (unsigned char u, unsigned int range); 46950 int __builtin_byte_in_either_range (unsigned char u, unsigned int ranges); 46951 46952 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal64 value); 46953 int __builtin_dfp_dtstsfi_lt (unsigned int comparison, _Decimal128 value); 46954 int __builtin_dfp_dtstsfi_lt_dd (unsigned int comparison, _Decimal64 value); 46955 int __builtin_dfp_dtstsfi_lt_td (unsigned int comparison, _Decimal128 value); 46956 46957 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal64 value); 46958 int __builtin_dfp_dtstsfi_gt (unsigned int comparison, _Decimal128 value); 46959 int __builtin_dfp_dtstsfi_gt_dd (unsigned int comparison, _Decimal64 value); 46960 int __builtin_dfp_dtstsfi_gt_td (unsigned int comparison, _Decimal128 value); 46961 46962 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal64 value); 46963 int __builtin_dfp_dtstsfi_eq (unsigned int comparison, _Decimal128 value); 46964 int __builtin_dfp_dtstsfi_eq_dd (unsigned int comparison, _Decimal64 value); 46965 int __builtin_dfp_dtstsfi_eq_td (unsigned int comparison, _Decimal128 value); 46966 46967 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal64 value); 46968 int __builtin_dfp_dtstsfi_ov (unsigned int comparison, _Decimal128 value); 46969 int __builtin_dfp_dtstsfi_ov_dd (unsigned int comparison, _Decimal64 value); 46970 int __builtin_dfp_dtstsfi_ov_td (unsigned int comparison, _Decimal128 value); 46971 46972 double __builtin_mffsl(void); 46973 46974 The '__builtin_byte_in_set' function requires a 64-bit environment 46975supporting ISA 3.0 or later. This function returns a non-zero value if 46976and only if its 'u' argument exactly equals one of the eight bytes 46977contained within its 64-bit 'set' argument. 46978 46979 The '__builtin_byte_in_range' and '__builtin_byte_in_either_range' 46980require an environment supporting ISA 3.0 or later. For these two 46981functions, the 'range' argument is encoded as 4 bytes, organized as 46982'hi_1:lo_1:hi_2:lo_2'. The '__builtin_byte_in_range' function returns a 46983non-zero value if and only if its 'u' argument is within the range 46984bounded between 'lo_2' and 'hi_2' inclusive. The 46985'__builtin_byte_in_either_range' function returns non-zero if and only 46986if its 'u' argument is within either the range bounded between 'lo_1' 46987and 'hi_1' inclusive or the range bounded between 'lo_2' and 'hi_2' 46988inclusive. 46989 46990 The '__builtin_dfp_dtstsfi_lt' function returns a non-zero value if and 46991only if the number of signficant digits of its 'value' argument is less 46992than its 'comparison' argument. The '__builtin_dfp_dtstsfi_lt_dd' and 46993'__builtin_dfp_dtstsfi_lt_td' functions behave similarly, but require 46994that the type of the 'value' argument be '__Decimal64' and 46995'__Decimal128' respectively. 46996 46997 The '__builtin_dfp_dtstsfi_gt' function returns a non-zero value if and 46998only if the number of signficant digits of its 'value' argument is 46999greater than its 'comparison' argument. The 47000'__builtin_dfp_dtstsfi_gt_dd' and '__builtin_dfp_dtstsfi_gt_td' 47001functions behave similarly, but require that the type of the 'value' 47002argument be '__Decimal64' and '__Decimal128' respectively. 47003 47004 The '__builtin_dfp_dtstsfi_eq' function returns a non-zero value if and 47005only if the number of signficant digits of its 'value' argument equals 47006its 'comparison' argument. The '__builtin_dfp_dtstsfi_eq_dd' and 47007'__builtin_dfp_dtstsfi_eq_td' functions behave similarly, but require 47008that the type of the 'value' argument be '__Decimal64' and 47009'__Decimal128' respectively. 47010 47011 The '__builtin_dfp_dtstsfi_ov' function returns a non-zero value if and 47012only if its 'value' argument has an undefined number of significant 47013digits, such as when 'value' is an encoding of 'NaN'. The 47014'__builtin_dfp_dtstsfi_ov_dd' and '__builtin_dfp_dtstsfi_ov_td' 47015functions behave similarly, but require that the type of the 'value' 47016argument be '__Decimal64' and '__Decimal128' respectively. 47017 47018 The '__builtin_mffsl' uses the ISA 3.0 'mffsl' instruction to read the 47019FPSCR. The instruction is a lower latency version of the 'mffs' 47020instruction. If the 'mffsl' instruction is not available, then the 47021builtin uses the older 'mffs' instruction to read the FPSCR. 47022 47023 47024File: gcc.info, Node: PowerPC AltiVec/VSX Built-in Functions, Next: PowerPC Hardware Transactional Memory Built-in Functions, Prev: Basic PowerPC Built-in Functions, Up: Target Builtins 47025 470266.60.23 PowerPC AltiVec/VSX Built-in Functions 47027---------------------------------------------- 47028 47029GCC provides an interface for the PowerPC family of processors to access 47030the AltiVec operations described in Motorola's AltiVec Programming 47031Interface Manual. The interface is made available by including 47032'<altivec.h>' and using '-maltivec' and '-mabi=altivec'. The interface 47033supports the following vector types. 47034 47035 vector unsigned char 47036 vector signed char 47037 vector bool char 47038 47039 vector unsigned short 47040 vector signed short 47041 vector bool short 47042 vector pixel 47043 47044 vector unsigned int 47045 vector signed int 47046 vector bool int 47047 vector float 47048 47049 GCC's implementation of the high-level language interface available 47050from C and C++ code differs from Motorola's documentation in several 47051ways. 47052 47053 * A vector constant is a list of constant expressions within curly 47054 braces. 47055 47056 * A vector initializer requires no cast if the vector constant is of 47057 the same type as the variable it is initializing. 47058 47059 * If 'signed' or 'unsigned' is omitted, the signedness of the vector 47060 type is the default signedness of the base type. The default 47061 varies depending on the operating system, so a portable program 47062 should always specify the signedness. 47063 47064 * Compiling with '-maltivec' adds keywords '__vector', 'vector', 47065 '__pixel', 'pixel', '__bool' and 'bool'. When compiling ISO C, the 47066 context-sensitive substitution of the keywords 'vector', 'pixel' 47067 and 'bool' is disabled. To use them, you must include 47068 '<altivec.h>' instead. 47069 47070 * GCC allows using a 'typedef' name as the type specifier for a 47071 vector type, but only under the following circumstances: 47072 47073 * When using '__vector' instead of 'vector'; for example, 47074 47075 typedef signed short int16; 47076 __vector int16 data; 47077 47078 * When using 'vector' in keyword-and-predefine mode; for 47079 example, 47080 47081 typedef signed short int16; 47082 vector int16 data; 47083 47084 Note that keyword-and-predefine mode is enabled by disabling 47085 GNU extensions (e.g., by using '-std=c11') and including 47086 '<altivec.h>'. 47087 47088 * For C, overloaded functions are implemented with macros so the 47089 following does not work: 47090 47091 vec_add ((vector signed int){1, 2, 3, 4}, foo); 47092 47093 Since 'vec_add' is a macro, the vector constant in the example is 47094 treated as four separate arguments. Wrap the entire argument in 47095 parentheses for this to work. 47096 47097 _Note:_ Only the '<altivec.h>' interface is supported. Internally, GCC 47098uses built-in functions to achieve the functionality in the 47099aforementioned header file, but they are not supported and are subject 47100to change without notice. 47101 47102 GCC complies with the OpenPOWER 64-Bit ELF V2 ABI Specification, which 47103may be found at 47104<https://openpowerfoundation.org/?resource_lib=64-bit-elf-v2-abi-specification-power-architecture>. 47105Appendix A of this document lists the vector API interfaces that must be 47106provided by compliant compilers. Programmers should preferentially use 47107the interfaces described therein. However, historically GCC has 47108provided additional interfaces for access to vector instructions. These 47109are briefly described below. 47110 47111* Menu: 47112 47113* PowerPC AltiVec Built-in Functions on ISA 2.05:: 47114* PowerPC AltiVec Built-in Functions Available on ISA 2.06:: 47115* PowerPC AltiVec Built-in Functions Available on ISA 2.07:: 47116* PowerPC AltiVec Built-in Functions Available on ISA 3.0:: 47117 47118 47119File: gcc.info, Node: PowerPC AltiVec Built-in Functions on ISA 2.05, Next: PowerPC AltiVec Built-in Functions Available on ISA 2.06, Up: PowerPC AltiVec/VSX Built-in Functions 47120 471216.60.23.1 PowerPC AltiVec Built-in Functions on ISA 2.05 47122........................................................ 47123 47124The following interfaces are supported for the generic and specific 47125AltiVec operations and the AltiVec predicates. In cases where there is 47126a direct mapping between generic and specific operations, only the 47127generic names are shown here, although the specific operations can also 47128be used. 47129 47130 Arguments that are documented as 'const int' require literal integral 47131values within the range required for that operation. 47132 47133 vector signed char vec_abs (vector signed char); 47134 vector signed short vec_abs (vector signed short); 47135 vector signed int vec_abs (vector signed int); 47136 vector float vec_abs (vector float); 47137 47138 vector signed char vec_abss (vector signed char); 47139 vector signed short vec_abss (vector signed short); 47140 vector signed int vec_abss (vector signed int); 47141 47142 vector signed char vec_add (vector bool char, vector signed char); 47143 vector signed char vec_add (vector signed char, vector bool char); 47144 vector signed char vec_add (vector signed char, vector signed char); 47145 vector unsigned char vec_add (vector bool char, vector unsigned char); 47146 vector unsigned char vec_add (vector unsigned char, vector bool char); 47147 vector unsigned char vec_add (vector unsigned char, vector unsigned char); 47148 vector signed short vec_add (vector bool short, vector signed short); 47149 vector signed short vec_add (vector signed short, vector bool short); 47150 vector signed short vec_add (vector signed short, vector signed short); 47151 vector unsigned short vec_add (vector bool short, vector unsigned short); 47152 vector unsigned short vec_add (vector unsigned short, vector bool short); 47153 vector unsigned short vec_add (vector unsigned short, vector unsigned short); 47154 vector signed int vec_add (vector bool int, vector signed int); 47155 vector signed int vec_add (vector signed int, vector bool int); 47156 vector signed int vec_add (vector signed int, vector signed int); 47157 vector unsigned int vec_add (vector bool int, vector unsigned int); 47158 vector unsigned int vec_add (vector unsigned int, vector bool int); 47159 vector unsigned int vec_add (vector unsigned int, vector unsigned int); 47160 vector float vec_add (vector float, vector float); 47161 47162 vector unsigned int vec_addc (vector unsigned int, vector unsigned int); 47163 47164 vector unsigned char vec_adds (vector bool char, vector unsigned char); 47165 vector unsigned char vec_adds (vector unsigned char, vector bool char); 47166 vector unsigned char vec_adds (vector unsigned char, vector unsigned char); 47167 vector signed char vec_adds (vector bool char, vector signed char); 47168 vector signed char vec_adds (vector signed char, vector bool char); 47169 vector signed char vec_adds (vector signed char, vector signed char); 47170 vector unsigned short vec_adds (vector bool short, vector unsigned short); 47171 vector unsigned short vec_adds (vector unsigned short, vector bool short); 47172 vector unsigned short vec_adds (vector unsigned short, vector unsigned short); 47173 vector signed short vec_adds (vector bool short, vector signed short); 47174 vector signed short vec_adds (vector signed short, vector bool short); 47175 vector signed short vec_adds (vector signed short, vector signed short); 47176 vector unsigned int vec_adds (vector bool int, vector unsigned int); 47177 vector unsigned int vec_adds (vector unsigned int, vector bool int); 47178 vector unsigned int vec_adds (vector unsigned int, vector unsigned int); 47179 vector signed int vec_adds (vector bool int, vector signed int); 47180 vector signed int vec_adds (vector signed int, vector bool int); 47181 vector signed int vec_adds (vector signed int, vector signed int); 47182 47183 int vec_all_eq (vector signed char, vector bool char); 47184 int vec_all_eq (vector signed char, vector signed char); 47185 int vec_all_eq (vector unsigned char, vector bool char); 47186 int vec_all_eq (vector unsigned char, vector unsigned char); 47187 int vec_all_eq (vector bool char, vector bool char); 47188 int vec_all_eq (vector bool char, vector unsigned char); 47189 int vec_all_eq (vector bool char, vector signed char); 47190 int vec_all_eq (vector signed short, vector bool short); 47191 int vec_all_eq (vector signed short, vector signed short); 47192 int vec_all_eq (vector unsigned short, vector bool short); 47193 int vec_all_eq (vector unsigned short, vector unsigned short); 47194 int vec_all_eq (vector bool short, vector bool short); 47195 int vec_all_eq (vector bool short, vector unsigned short); 47196 int vec_all_eq (vector bool short, vector signed short); 47197 int vec_all_eq (vector pixel, vector pixel); 47198 int vec_all_eq (vector signed int, vector bool int); 47199 int vec_all_eq (vector signed int, vector signed int); 47200 int vec_all_eq (vector unsigned int, vector bool int); 47201 int vec_all_eq (vector unsigned int, vector unsigned int); 47202 int vec_all_eq (vector bool int, vector bool int); 47203 int vec_all_eq (vector bool int, vector unsigned int); 47204 int vec_all_eq (vector bool int, vector signed int); 47205 int vec_all_eq (vector float, vector float); 47206 47207 int vec_all_ge (vector bool char, vector unsigned char); 47208 int vec_all_ge (vector unsigned char, vector bool char); 47209 int vec_all_ge (vector unsigned char, vector unsigned char); 47210 int vec_all_ge (vector bool char, vector signed char); 47211 int vec_all_ge (vector signed char, vector bool char); 47212 int vec_all_ge (vector signed char, vector signed char); 47213 int vec_all_ge (vector bool short, vector unsigned short); 47214 int vec_all_ge (vector unsigned short, vector bool short); 47215 int vec_all_ge (vector unsigned short, vector unsigned short); 47216 int vec_all_ge (vector signed short, vector signed short); 47217 int vec_all_ge (vector bool short, vector signed short); 47218 int vec_all_ge (vector signed short, vector bool short); 47219 int vec_all_ge (vector bool int, vector unsigned int); 47220 int vec_all_ge (vector unsigned int, vector bool int); 47221 int vec_all_ge (vector unsigned int, vector unsigned int); 47222 int vec_all_ge (vector bool int, vector signed int); 47223 int vec_all_ge (vector signed int, vector bool int); 47224 int vec_all_ge (vector signed int, vector signed int); 47225 int vec_all_ge (vector float, vector float); 47226 47227 int vec_all_gt (vector bool char, vector unsigned char); 47228 int vec_all_gt (vector unsigned char, vector bool char); 47229 int vec_all_gt (vector unsigned char, vector unsigned char); 47230 int vec_all_gt (vector bool char, vector signed char); 47231 int vec_all_gt (vector signed char, vector bool char); 47232 int vec_all_gt (vector signed char, vector signed char); 47233 int vec_all_gt (vector bool short, vector unsigned short); 47234 int vec_all_gt (vector unsigned short, vector bool short); 47235 int vec_all_gt (vector unsigned short, vector unsigned short); 47236 int vec_all_gt (vector bool short, vector signed short); 47237 int vec_all_gt (vector signed short, vector bool short); 47238 int vec_all_gt (vector signed short, vector signed short); 47239 int vec_all_gt (vector bool int, vector unsigned int); 47240 int vec_all_gt (vector unsigned int, vector bool int); 47241 int vec_all_gt (vector unsigned int, vector unsigned int); 47242 int vec_all_gt (vector bool int, vector signed int); 47243 int vec_all_gt (vector signed int, vector bool int); 47244 int vec_all_gt (vector signed int, vector signed int); 47245 int vec_all_gt (vector float, vector float); 47246 47247 int vec_all_in (vector float, vector float); 47248 47249 int vec_all_le (vector bool char, vector unsigned char); 47250 int vec_all_le (vector unsigned char, vector bool char); 47251 int vec_all_le (vector unsigned char, vector unsigned char); 47252 int vec_all_le (vector bool char, vector signed char); 47253 int vec_all_le (vector signed char, vector bool char); 47254 int vec_all_le (vector signed char, vector signed char); 47255 int vec_all_le (vector bool short, vector unsigned short); 47256 int vec_all_le (vector unsigned short, vector bool short); 47257 int vec_all_le (vector unsigned short, vector unsigned short); 47258 int vec_all_le (vector bool short, vector signed short); 47259 int vec_all_le (vector signed short, vector bool short); 47260 int vec_all_le (vector signed short, vector signed short); 47261 int vec_all_le (vector bool int, vector unsigned int); 47262 int vec_all_le (vector unsigned int, vector bool int); 47263 int vec_all_le (vector unsigned int, vector unsigned int); 47264 int vec_all_le (vector bool int, vector signed int); 47265 int vec_all_le (vector signed int, vector bool int); 47266 int vec_all_le (vector signed int, vector signed int); 47267 int vec_all_le (vector float, vector float); 47268 47269 int vec_all_lt (vector bool char, vector unsigned char); 47270 int vec_all_lt (vector unsigned char, vector bool char); 47271 int vec_all_lt (vector unsigned char, vector unsigned char); 47272 int vec_all_lt (vector bool char, vector signed char); 47273 int vec_all_lt (vector signed char, vector bool char); 47274 int vec_all_lt (vector signed char, vector signed char); 47275 int vec_all_lt (vector bool short, vector unsigned short); 47276 int vec_all_lt (vector unsigned short, vector bool short); 47277 int vec_all_lt (vector unsigned short, vector unsigned short); 47278 int vec_all_lt (vector bool short, vector signed short); 47279 int vec_all_lt (vector signed short, vector bool short); 47280 int vec_all_lt (vector signed short, vector signed short); 47281 int vec_all_lt (vector bool int, vector unsigned int); 47282 int vec_all_lt (vector unsigned int, vector bool int); 47283 int vec_all_lt (vector unsigned int, vector unsigned int); 47284 int vec_all_lt (vector bool int, vector signed int); 47285 int vec_all_lt (vector signed int, vector bool int); 47286 int vec_all_lt (vector signed int, vector signed int); 47287 int vec_all_lt (vector float, vector float); 47288 47289 int vec_all_nan (vector float); 47290 47291 int vec_all_ne (vector signed char, vector bool char); 47292 int vec_all_ne (vector signed char, vector signed char); 47293 int vec_all_ne (vector unsigned char, vector bool char); 47294 int vec_all_ne (vector unsigned char, vector unsigned char); 47295 int vec_all_ne (vector bool char, vector bool char); 47296 int vec_all_ne (vector bool char, vector unsigned char); 47297 int vec_all_ne (vector bool char, vector signed char); 47298 int vec_all_ne (vector signed short, vector bool short); 47299 int vec_all_ne (vector signed short, vector signed short); 47300 int vec_all_ne (vector unsigned short, vector bool short); 47301 int vec_all_ne (vector unsigned short, vector unsigned short); 47302 int vec_all_ne (vector bool short, vector bool short); 47303 int vec_all_ne (vector bool short, vector unsigned short); 47304 int vec_all_ne (vector bool short, vector signed short); 47305 int vec_all_ne (vector pixel, vector pixel); 47306 int vec_all_ne (vector signed int, vector bool int); 47307 int vec_all_ne (vector signed int, vector signed int); 47308 int vec_all_ne (vector unsigned int, vector bool int); 47309 int vec_all_ne (vector unsigned int, vector unsigned int); 47310 int vec_all_ne (vector bool int, vector bool int); 47311 int vec_all_ne (vector bool int, vector unsigned int); 47312 int vec_all_ne (vector bool int, vector signed int); 47313 int vec_all_ne (vector float, vector float); 47314 47315 int vec_all_nge (vector float, vector float); 47316 47317 int vec_all_ngt (vector float, vector float); 47318 47319 int vec_all_nle (vector float, vector float); 47320 47321 int vec_all_nlt (vector float, vector float); 47322 47323 int vec_all_numeric (vector float); 47324 47325 vector float vec_and (vector float, vector float); 47326 vector float vec_and (vector float, vector bool int); 47327 vector float vec_and (vector bool int, vector float); 47328 vector bool int vec_and (vector bool int, vector bool int); 47329 vector signed int vec_and (vector bool int, vector signed int); 47330 vector signed int vec_and (vector signed int, vector bool int); 47331 vector signed int vec_and (vector signed int, vector signed int); 47332 vector unsigned int vec_and (vector bool int, vector unsigned int); 47333 vector unsigned int vec_and (vector unsigned int, vector bool int); 47334 vector unsigned int vec_and (vector unsigned int, vector unsigned int); 47335 vector bool short vec_and (vector bool short, vector bool short); 47336 vector signed short vec_and (vector bool short, vector signed short); 47337 vector signed short vec_and (vector signed short, vector bool short); 47338 vector signed short vec_and (vector signed short, vector signed short); 47339 vector unsigned short vec_and (vector bool short, vector unsigned short); 47340 vector unsigned short vec_and (vector unsigned short, vector bool short); 47341 vector unsigned short vec_and (vector unsigned short, vector unsigned short); 47342 vector signed char vec_and (vector bool char, vector signed char); 47343 vector bool char vec_and (vector bool char, vector bool char); 47344 vector signed char vec_and (vector signed char, vector bool char); 47345 vector signed char vec_and (vector signed char, vector signed char); 47346 vector unsigned char vec_and (vector bool char, vector unsigned char); 47347 vector unsigned char vec_and (vector unsigned char, vector bool char); 47348 vector unsigned char vec_and (vector unsigned char, vector unsigned char); 47349 47350 vector float vec_andc (vector float, vector float); 47351 vector float vec_andc (vector float, vector bool int); 47352 vector float vec_andc (vector bool int, vector float); 47353 vector bool int vec_andc (vector bool int, vector bool int); 47354 vector signed int vec_andc (vector bool int, vector signed int); 47355 vector signed int vec_andc (vector signed int, vector bool int); 47356 vector signed int vec_andc (vector signed int, vector signed int); 47357 vector unsigned int vec_andc (vector bool int, vector unsigned int); 47358 vector unsigned int vec_andc (vector unsigned int, vector bool int); 47359 vector unsigned int vec_andc (vector unsigned int, vector unsigned int); 47360 vector bool short vec_andc (vector bool short, vector bool short); 47361 vector signed short vec_andc (vector bool short, vector signed short); 47362 vector signed short vec_andc (vector signed short, vector bool short); 47363 vector signed short vec_andc (vector signed short, vector signed short); 47364 vector unsigned short vec_andc (vector bool short, vector unsigned short); 47365 vector unsigned short vec_andc (vector unsigned short, vector bool short); 47366 vector unsigned short vec_andc (vector unsigned short, vector unsigned short); 47367 vector signed char vec_andc (vector bool char, vector signed char); 47368 vector bool char vec_andc (vector bool char, vector bool char); 47369 vector signed char vec_andc (vector signed char, vector bool char); 47370 vector signed char vec_andc (vector signed char, vector signed char); 47371 vector unsigned char vec_andc (vector bool char, vector unsigned char); 47372 vector unsigned char vec_andc (vector unsigned char, vector bool char); 47373 vector unsigned char vec_andc (vector unsigned char, vector unsigned char); 47374 47375 int vec_any_eq (vector signed char, vector bool char); 47376 int vec_any_eq (vector signed char, vector signed char); 47377 int vec_any_eq (vector unsigned char, vector bool char); 47378 int vec_any_eq (vector unsigned char, vector unsigned char); 47379 int vec_any_eq (vector bool char, vector bool char); 47380 int vec_any_eq (vector bool char, vector unsigned char); 47381 int vec_any_eq (vector bool char, vector signed char); 47382 int vec_any_eq (vector signed short, vector bool short); 47383 int vec_any_eq (vector signed short, vector signed short); 47384 int vec_any_eq (vector unsigned short, vector bool short); 47385 int vec_any_eq (vector unsigned short, vector unsigned short); 47386 int vec_any_eq (vector bool short, vector bool short); 47387 int vec_any_eq (vector bool short, vector unsigned short); 47388 int vec_any_eq (vector bool short, vector signed short); 47389 int vec_any_eq (vector pixel, vector pixel); 47390 int vec_any_eq (vector signed int, vector bool int); 47391 int vec_any_eq (vector signed int, vector signed int); 47392 int vec_any_eq (vector unsigned int, vector bool int); 47393 int vec_any_eq (vector unsigned int, vector unsigned int); 47394 int vec_any_eq (vector bool int, vector bool int); 47395 int vec_any_eq (vector bool int, vector unsigned int); 47396 int vec_any_eq (vector bool int, vector signed int); 47397 int vec_any_eq (vector float, vector float); 47398 47399 int vec_any_ge (vector signed char, vector bool char); 47400 int vec_any_ge (vector unsigned char, vector bool char); 47401 int vec_any_ge (vector unsigned char, vector unsigned char); 47402 int vec_any_ge (vector signed char, vector signed char); 47403 int vec_any_ge (vector bool char, vector unsigned char); 47404 int vec_any_ge (vector bool char, vector signed char); 47405 int vec_any_ge (vector unsigned short, vector bool short); 47406 int vec_any_ge (vector unsigned short, vector unsigned short); 47407 int vec_any_ge (vector signed short, vector signed short); 47408 int vec_any_ge (vector signed short, vector bool short); 47409 int vec_any_ge (vector bool short, vector unsigned short); 47410 int vec_any_ge (vector bool short, vector signed short); 47411 int vec_any_ge (vector signed int, vector bool int); 47412 int vec_any_ge (vector unsigned int, vector bool int); 47413 int vec_any_ge (vector unsigned int, vector unsigned int); 47414 int vec_any_ge (vector signed int, vector signed int); 47415 int vec_any_ge (vector bool int, vector unsigned int); 47416 int vec_any_ge (vector bool int, vector signed int); 47417 int vec_any_ge (vector float, vector float); 47418 47419 int vec_any_gt (vector bool char, vector unsigned char); 47420 int vec_any_gt (vector unsigned char, vector bool char); 47421 int vec_any_gt (vector unsigned char, vector unsigned char); 47422 int vec_any_gt (vector bool char, vector signed char); 47423 int vec_any_gt (vector signed char, vector bool char); 47424 int vec_any_gt (vector signed char, vector signed char); 47425 int vec_any_gt (vector bool short, vector unsigned short); 47426 int vec_any_gt (vector unsigned short, vector bool short); 47427 int vec_any_gt (vector unsigned short, vector unsigned short); 47428 int vec_any_gt (vector bool short, vector signed short); 47429 int vec_any_gt (vector signed short, vector bool short); 47430 int vec_any_gt (vector signed short, vector signed short); 47431 int vec_any_gt (vector bool int, vector unsigned int); 47432 int vec_any_gt (vector unsigned int, vector bool int); 47433 int vec_any_gt (vector unsigned int, vector unsigned int); 47434 int vec_any_gt (vector bool int, vector signed int); 47435 int vec_any_gt (vector signed int, vector bool int); 47436 int vec_any_gt (vector signed int, vector signed int); 47437 int vec_any_gt (vector float, vector float); 47438 47439 int vec_any_le (vector bool char, vector unsigned char); 47440 int vec_any_le (vector unsigned char, vector bool char); 47441 int vec_any_le (vector unsigned char, vector unsigned char); 47442 int vec_any_le (vector bool char, vector signed char); 47443 int vec_any_le (vector signed char, vector bool char); 47444 int vec_any_le (vector signed char, vector signed char); 47445 int vec_any_le (vector bool short, vector unsigned short); 47446 int vec_any_le (vector unsigned short, vector bool short); 47447 int vec_any_le (vector unsigned short, vector unsigned short); 47448 int vec_any_le (vector bool short, vector signed short); 47449 int vec_any_le (vector signed short, vector bool short); 47450 int vec_any_le (vector signed short, vector signed short); 47451 int vec_any_le (vector bool int, vector unsigned int); 47452 int vec_any_le (vector unsigned int, vector bool int); 47453 int vec_any_le (vector unsigned int, vector unsigned int); 47454 int vec_any_le (vector bool int, vector signed int); 47455 int vec_any_le (vector signed int, vector bool int); 47456 int vec_any_le (vector signed int, vector signed int); 47457 int vec_any_le (vector float, vector float); 47458 47459 int vec_any_lt (vector bool char, vector unsigned char); 47460 int vec_any_lt (vector unsigned char, vector bool char); 47461 int vec_any_lt (vector unsigned char, vector unsigned char); 47462 int vec_any_lt (vector bool char, vector signed char); 47463 int vec_any_lt (vector signed char, vector bool char); 47464 int vec_any_lt (vector signed char, vector signed char); 47465 int vec_any_lt (vector bool short, vector unsigned short); 47466 int vec_any_lt (vector unsigned short, vector bool short); 47467 int vec_any_lt (vector unsigned short, vector unsigned short); 47468 int vec_any_lt (vector bool short, vector signed short); 47469 int vec_any_lt (vector signed short, vector bool short); 47470 int vec_any_lt (vector signed short, vector signed short); 47471 int vec_any_lt (vector bool int, vector unsigned int); 47472 int vec_any_lt (vector unsigned int, vector bool int); 47473 int vec_any_lt (vector unsigned int, vector unsigned int); 47474 int vec_any_lt (vector bool int, vector signed int); 47475 int vec_any_lt (vector signed int, vector bool int); 47476 int vec_any_lt (vector signed int, vector signed int); 47477 int vec_any_lt (vector float, vector float); 47478 47479 int vec_any_nan (vector float); 47480 47481 int vec_any_ne (vector signed char, vector bool char); 47482 int vec_any_ne (vector signed char, vector signed char); 47483 int vec_any_ne (vector unsigned char, vector bool char); 47484 int vec_any_ne (vector unsigned char, vector unsigned char); 47485 int vec_any_ne (vector bool char, vector bool char); 47486 int vec_any_ne (vector bool char, vector unsigned char); 47487 int vec_any_ne (vector bool char, vector signed char); 47488 int vec_any_ne (vector signed short, vector bool short); 47489 int vec_any_ne (vector signed short, vector signed short); 47490 int vec_any_ne (vector unsigned short, vector bool short); 47491 int vec_any_ne (vector unsigned short, vector unsigned short); 47492 int vec_any_ne (vector bool short, vector bool short); 47493 int vec_any_ne (vector bool short, vector unsigned short); 47494 int vec_any_ne (vector bool short, vector signed short); 47495 int vec_any_ne (vector pixel, vector pixel); 47496 int vec_any_ne (vector signed int, vector bool int); 47497 int vec_any_ne (vector signed int, vector signed int); 47498 int vec_any_ne (vector unsigned int, vector bool int); 47499 int vec_any_ne (vector unsigned int, vector unsigned int); 47500 int vec_any_ne (vector bool int, vector bool int); 47501 int vec_any_ne (vector bool int, vector unsigned int); 47502 int vec_any_ne (vector bool int, vector signed int); 47503 int vec_any_ne (vector float, vector float); 47504 47505 int vec_any_nge (vector float, vector float); 47506 47507 int vec_any_ngt (vector float, vector float); 47508 47509 int vec_any_nle (vector float, vector float); 47510 47511 int vec_any_nlt (vector float, vector float); 47512 47513 int vec_any_numeric (vector float); 47514 47515 int vec_any_out (vector float, vector float); 47516 47517 vector unsigned char vec_avg (vector unsigned char, vector unsigned char); 47518 vector signed char vec_avg (vector signed char, vector signed char); 47519 vector unsigned short vec_avg (vector unsigned short, vector unsigned short); 47520 vector signed short vec_avg (vector signed short, vector signed short); 47521 vector unsigned int vec_avg (vector unsigned int, vector unsigned int); 47522 vector signed int vec_avg (vector signed int, vector signed int); 47523 47524 vector float vec_ceil (vector float); 47525 47526 vector signed int vec_cmpb (vector float, vector float); 47527 47528 vector bool char vec_cmpeq (vector bool char, vector bool char); 47529 vector bool short vec_cmpeq (vector bool short, vector bool short); 47530 vector bool int vec_cmpeq (vector bool int, vector bool int); 47531 vector bool char vec_cmpeq (vector signed char, vector signed char); 47532 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char); 47533 vector bool short vec_cmpeq (vector signed short, vector signed short); 47534 vector bool short vec_cmpeq (vector unsigned short, vector unsigned short); 47535 vector bool int vec_cmpeq (vector signed int, vector signed int); 47536 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int); 47537 vector bool int vec_cmpeq (vector float, vector float); 47538 47539 vector bool int vec_cmpge (vector float, vector float); 47540 47541 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char); 47542 vector bool char vec_cmpgt (vector signed char, vector signed char); 47543 vector bool short vec_cmpgt (vector unsigned short, vector unsigned short); 47544 vector bool short vec_cmpgt (vector signed short, vector signed short); 47545 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int); 47546 vector bool int vec_cmpgt (vector signed int, vector signed int); 47547 vector bool int vec_cmpgt (vector float, vector float); 47548 47549 vector bool int vec_cmple (vector float, vector float); 47550 47551 vector bool char vec_cmplt (vector unsigned char, vector unsigned char); 47552 vector bool char vec_cmplt (vector signed char, vector signed char); 47553 vector bool short vec_cmplt (vector unsigned short, vector unsigned short); 47554 vector bool short vec_cmplt (vector signed short, vector signed short); 47555 vector bool int vec_cmplt (vector unsigned int, vector unsigned int); 47556 vector bool int vec_cmplt (vector signed int, vector signed int); 47557 vector bool int vec_cmplt (vector float, vector float); 47558 47559 vector float vec_cpsgn (vector float, vector float); 47560 47561 vector float vec_ctf (vector unsigned int, const int); 47562 vector float vec_ctf (vector signed int, const int); 47563 47564 vector signed int vec_cts (vector float, const int); 47565 47566 vector unsigned int vec_ctu (vector float, const int); 47567 47568 void vec_dss (const int); 47569 47570 void vec_dssall (void); 47571 47572 void vec_dst (const vector unsigned char *, int, const int); 47573 void vec_dst (const vector signed char *, int, const int); 47574 void vec_dst (const vector bool char *, int, const int); 47575 void vec_dst (const vector unsigned short *, int, const int); 47576 void vec_dst (const vector signed short *, int, const int); 47577 void vec_dst (const vector bool short *, int, const int); 47578 void vec_dst (const vector pixel *, int, const int); 47579 void vec_dst (const vector unsigned int *, int, const int); 47580 void vec_dst (const vector signed int *, int, const int); 47581 void vec_dst (const vector bool int *, int, const int); 47582 void vec_dst (const vector float *, int, const int); 47583 void vec_dst (const unsigned char *, int, const int); 47584 void vec_dst (const signed char *, int, const int); 47585 void vec_dst (const unsigned short *, int, const int); 47586 void vec_dst (const short *, int, const int); 47587 void vec_dst (const unsigned int *, int, const int); 47588 void vec_dst (const int *, int, const int); 47589 void vec_dst (const float *, int, const int); 47590 47591 void vec_dstst (const vector unsigned char *, int, const int); 47592 void vec_dstst (const vector signed char *, int, const int); 47593 void vec_dstst (const vector bool char *, int, const int); 47594 void vec_dstst (const vector unsigned short *, int, const int); 47595 void vec_dstst (const vector signed short *, int, const int); 47596 void vec_dstst (const vector bool short *, int, const int); 47597 void vec_dstst (const vector pixel *, int, const int); 47598 void vec_dstst (const vector unsigned int *, int, const int); 47599 void vec_dstst (const vector signed int *, int, const int); 47600 void vec_dstst (const vector bool int *, int, const int); 47601 void vec_dstst (const vector float *, int, const int); 47602 void vec_dstst (const unsigned char *, int, const int); 47603 void vec_dstst (const signed char *, int, const int); 47604 void vec_dstst (const unsigned short *, int, const int); 47605 void vec_dstst (const short *, int, const int); 47606 void vec_dstst (const unsigned int *, int, const int); 47607 void vec_dstst (const int *, int, const int); 47608 void vec_dstst (const unsigned long *, int, const int); 47609 void vec_dstst (const long *, int, const int); 47610 void vec_dstst (const float *, int, const int); 47611 47612 void vec_dststt (const vector unsigned char *, int, const int); 47613 void vec_dststt (const vector signed char *, int, const int); 47614 void vec_dststt (const vector bool char *, int, const int); 47615 void vec_dststt (const vector unsigned short *, int, const int); 47616 void vec_dststt (const vector signed short *, int, const int); 47617 void vec_dststt (const vector bool short *, int, const int); 47618 void vec_dststt (const vector pixel *, int, const int); 47619 void vec_dststt (const vector unsigned int *, int, const int); 47620 void vec_dststt (const vector signed int *, int, const int); 47621 void vec_dststt (const vector bool int *, int, const int); 47622 void vec_dststt (const vector float *, int, const int); 47623 void vec_dststt (const unsigned char *, int, const int); 47624 void vec_dststt (const signed char *, int, const int); 47625 void vec_dststt (const unsigned short *, int, const int); 47626 void vec_dststt (const short *, int, const int); 47627 void vec_dststt (const unsigned int *, int, const int); 47628 void vec_dststt (const int *, int, const int); 47629 void vec_dststt (const float *, int, const int); 47630 47631 void vec_dstt (const vector unsigned char *, int, const int); 47632 void vec_dstt (const vector signed char *, int, const int); 47633 void vec_dstt (const vector bool char *, int, const int); 47634 void vec_dstt (const vector unsigned short *, int, const int); 47635 void vec_dstt (const vector signed short *, int, const int); 47636 void vec_dstt (const vector bool short *, int, const int); 47637 void vec_dstt (const vector pixel *, int, const int); 47638 void vec_dstt (const vector unsigned int *, int, const int); 47639 void vec_dstt (const vector signed int *, int, const int); 47640 void vec_dstt (const vector bool int *, int, const int); 47641 void vec_dstt (const vector float *, int, const int); 47642 void vec_dstt (const unsigned char *, int, const int); 47643 void vec_dstt (const signed char *, int, const int); 47644 void vec_dstt (const unsigned short *, int, const int); 47645 void vec_dstt (const short *, int, const int); 47646 void vec_dstt (const unsigned int *, int, const int); 47647 void vec_dstt (const int *, int, const int); 47648 void vec_dstt (const float *, int, const int); 47649 47650 vector float vec_expte (vector float); 47651 47652 vector float vec_floor (vector float); 47653 47654 vector float vec_ld (int, const vector float *); 47655 vector float vec_ld (int, const float *); 47656 vector bool int vec_ld (int, const vector bool int *); 47657 vector signed int vec_ld (int, const vector signed int *); 47658 vector signed int vec_ld (int, const int *); 47659 vector unsigned int vec_ld (int, const vector unsigned int *); 47660 vector unsigned int vec_ld (int, const unsigned int *); 47661 vector bool short vec_ld (int, const vector bool short *); 47662 vector pixel vec_ld (int, const vector pixel *); 47663 vector signed short vec_ld (int, const vector signed short *); 47664 vector signed short vec_ld (int, const short *); 47665 vector unsigned short vec_ld (int, const vector unsigned short *); 47666 vector unsigned short vec_ld (int, const unsigned short *); 47667 vector bool char vec_ld (int, const vector bool char *); 47668 vector signed char vec_ld (int, const vector signed char *); 47669 vector signed char vec_ld (int, const signed char *); 47670 vector unsigned char vec_ld (int, const vector unsigned char *); 47671 vector unsigned char vec_ld (int, const unsigned char *); 47672 47673 vector signed char vec_lde (int, const signed char *); 47674 vector unsigned char vec_lde (int, const unsigned char *); 47675 vector signed short vec_lde (int, const short *); 47676 vector unsigned short vec_lde (int, const unsigned short *); 47677 vector float vec_lde (int, const float *); 47678 vector signed int vec_lde (int, const int *); 47679 vector unsigned int vec_lde (int, const unsigned int *); 47680 47681 vector float vec_ldl (int, const vector float *); 47682 vector float vec_ldl (int, const float *); 47683 vector bool int vec_ldl (int, const vector bool int *); 47684 vector signed int vec_ldl (int, const vector signed int *); 47685 vector signed int vec_ldl (int, const int *); 47686 vector unsigned int vec_ldl (int, const vector unsigned int *); 47687 vector unsigned int vec_ldl (int, const unsigned int *); 47688 vector bool short vec_ldl (int, const vector bool short *); 47689 vector pixel vec_ldl (int, const vector pixel *); 47690 vector signed short vec_ldl (int, const vector signed short *); 47691 vector signed short vec_ldl (int, const short *); 47692 vector unsigned short vec_ldl (int, const vector unsigned short *); 47693 vector unsigned short vec_ldl (int, const unsigned short *); 47694 vector bool char vec_ldl (int, const vector bool char *); 47695 vector signed char vec_ldl (int, const vector signed char *); 47696 vector signed char vec_ldl (int, const signed char *); 47697 vector unsigned char vec_ldl (int, const vector unsigned char *); 47698 vector unsigned char vec_ldl (int, const unsigned char *); 47699 47700 vector float vec_loge (vector float); 47701 47702 vector signed char vec_lvebx (int, char *); 47703 vector unsigned char vec_lvebx (int, unsigned char *); 47704 47705 vector signed short vec_lvehx (int, short *); 47706 vector unsigned short vec_lvehx (int, unsigned short *); 47707 47708 vector float vec_lvewx (int, float *); 47709 vector signed int vec_lvewx (int, int *); 47710 vector unsigned int vec_lvewx (int, unsigned int *); 47711 47712 vector unsigned char vec_lvsl (int, const unsigned char *); 47713 vector unsigned char vec_lvsl (int, const signed char *); 47714 vector unsigned char vec_lvsl (int, const unsigned short *); 47715 vector unsigned char vec_lvsl (int, const short *); 47716 vector unsigned char vec_lvsl (int, const unsigned int *); 47717 vector unsigned char vec_lvsl (int, const int *); 47718 vector unsigned char vec_lvsl (int, const float *); 47719 47720 vector unsigned char vec_lvsr (int, const unsigned char *); 47721 vector unsigned char vec_lvsr (int, const signed char *); 47722 vector unsigned char vec_lvsr (int, const unsigned short *); 47723 vector unsigned char vec_lvsr (int, const short *); 47724 vector unsigned char vec_lvsr (int, const unsigned int *); 47725 vector unsigned char vec_lvsr (int, const int *); 47726 vector unsigned char vec_lvsr (int, const float *); 47727 47728 vector float vec_madd (vector float, vector float, vector float); 47729 47730 vector signed short vec_madds (vector signed short, vector signed short, 47731 vector signed short); 47732 47733 vector unsigned char vec_max (vector bool char, vector unsigned char); 47734 vector unsigned char vec_max (vector unsigned char, vector bool char); 47735 vector unsigned char vec_max (vector unsigned char, vector unsigned char); 47736 vector signed char vec_max (vector bool char, vector signed char); 47737 vector signed char vec_max (vector signed char, vector bool char); 47738 vector signed char vec_max (vector signed char, vector signed char); 47739 vector unsigned short vec_max (vector bool short, vector unsigned short); 47740 vector unsigned short vec_max (vector unsigned short, vector bool short); 47741 vector unsigned short vec_max (vector unsigned short, vector unsigned short); 47742 vector signed short vec_max (vector bool short, vector signed short); 47743 vector signed short vec_max (vector signed short, vector bool short); 47744 vector signed short vec_max (vector signed short, vector signed short); 47745 vector unsigned int vec_max (vector bool int, vector unsigned int); 47746 vector unsigned int vec_max (vector unsigned int, vector bool int); 47747 vector unsigned int vec_max (vector unsigned int, vector unsigned int); 47748 vector signed int vec_max (vector bool int, vector signed int); 47749 vector signed int vec_max (vector signed int, vector bool int); 47750 vector signed int vec_max (vector signed int, vector signed int); 47751 vector float vec_max (vector float, vector float); 47752 47753 vector bool char vec_mergeh (vector bool char, vector bool char); 47754 vector signed char vec_mergeh (vector signed char, vector signed char); 47755 vector unsigned char vec_mergeh (vector unsigned char, vector unsigned char); 47756 vector bool short vec_mergeh (vector bool short, vector bool short); 47757 vector pixel vec_mergeh (vector pixel, vector pixel); 47758 vector signed short vec_mergeh (vector signed short, vector signed short); 47759 vector unsigned short vec_mergeh (vector unsigned short, vector unsigned short); 47760 vector float vec_mergeh (vector float, vector float); 47761 vector bool int vec_mergeh (vector bool int, vector bool int); 47762 vector signed int vec_mergeh (vector signed int, vector signed int); 47763 vector unsigned int vec_mergeh (vector unsigned int, vector unsigned int); 47764 47765 vector bool char vec_mergel (vector bool char, vector bool char); 47766 vector signed char vec_mergel (vector signed char, vector signed char); 47767 vector unsigned char vec_mergel (vector unsigned char, vector unsigned char); 47768 vector bool short vec_mergel (vector bool short, vector bool short); 47769 vector pixel vec_mergel (vector pixel, vector pixel); 47770 vector signed short vec_mergel (vector signed short, vector signed short); 47771 vector unsigned short vec_mergel (vector unsigned short, vector unsigned short); 47772 vector float vec_mergel (vector float, vector float); 47773 vector bool int vec_mergel (vector bool int, vector bool int); 47774 vector signed int vec_mergel (vector signed int, vector signed int); 47775 vector unsigned int vec_mergel (vector unsigned int, vector unsigned int); 47776 47777 vector unsigned short vec_mfvscr (void); 47778 47779 vector unsigned char vec_min (vector bool char, vector unsigned char); 47780 vector unsigned char vec_min (vector unsigned char, vector bool char); 47781 vector unsigned char vec_min (vector unsigned char, vector unsigned char); 47782 vector signed char vec_min (vector bool char, vector signed char); 47783 vector signed char vec_min (vector signed char, vector bool char); 47784 vector signed char vec_min (vector signed char, vector signed char); 47785 vector unsigned short vec_min (vector bool short, vector unsigned short); 47786 vector unsigned short vec_min (vector unsigned short, vector bool short); 47787 vector unsigned short vec_min (vector unsigned short, vector unsigned short); 47788 vector signed short vec_min (vector bool short, vector signed short); 47789 vector signed short vec_min (vector signed short, vector bool short); 47790 vector signed short vec_min (vector signed short, vector signed short); 47791 vector unsigned int vec_min (vector bool int, vector unsigned int); 47792 vector unsigned int vec_min (vector unsigned int, vector bool int); 47793 vector unsigned int vec_min (vector unsigned int, vector unsigned int); 47794 vector signed int vec_min (vector bool int, vector signed int); 47795 vector signed int vec_min (vector signed int, vector bool int); 47796 vector signed int vec_min (vector signed int, vector signed int); 47797 vector float vec_min (vector float, vector float); 47798 47799 vector signed short vec_mladd (vector signed short, vector signed short, 47800 vector signed short); 47801 vector signed short vec_mladd (vector signed short, vector unsigned short, 47802 vector unsigned short); 47803 vector signed short vec_mladd (vector unsigned short, vector signed short, 47804 vector signed short); 47805 vector unsigned short vec_mladd (vector unsigned short, vector unsigned short, 47806 vector unsigned short); 47807 47808 vector signed short vec_mradds (vector signed short, vector signed short, 47809 vector signed short); 47810 47811 vector unsigned int vec_msum (vector unsigned char, vector unsigned char, 47812 vector unsigned int); 47813 vector signed int vec_msum (vector signed char, vector unsigned char, 47814 vector signed int); 47815 vector unsigned int vec_msum (vector unsigned short, vector unsigned short, 47816 vector unsigned int); 47817 vector signed int vec_msum (vector signed short, vector signed short, 47818 vector signed int); 47819 47820 vector unsigned int vec_msums (vector unsigned short, vector unsigned short, 47821 vector unsigned int); 47822 vector signed int vec_msums (vector signed short, vector signed short, 47823 vector signed int); 47824 47825 void vec_mtvscr (vector signed int); 47826 void vec_mtvscr (vector unsigned int); 47827 void vec_mtvscr (vector bool int); 47828 void vec_mtvscr (vector signed short); 47829 void vec_mtvscr (vector unsigned short); 47830 void vec_mtvscr (vector bool short); 47831 void vec_mtvscr (vector pixel); 47832 void vec_mtvscr (vector signed char); 47833 void vec_mtvscr (vector unsigned char); 47834 void vec_mtvscr (vector bool char); 47835 47836 vector float vec_mul (vector float, vector float); 47837 47838 vector unsigned short vec_mule (vector unsigned char, vector unsigned char); 47839 vector signed short vec_mule (vector signed char, vector signed char); 47840 vector unsigned int vec_mule (vector unsigned short, vector unsigned short); 47841 vector signed int vec_mule (vector signed short, vector signed short); 47842 47843 vector unsigned short vec_mulo (vector unsigned char, vector unsigned char); 47844 vector signed short vec_mulo (vector signed char, vector signed char); 47845 vector unsigned int vec_mulo (vector unsigned short, vector unsigned short); 47846 vector signed int vec_mulo (vector signed short, vector signed short); 47847 47848 vector signed char vec_nabs (vector signed char); 47849 vector signed short vec_nabs (vector signed short); 47850 vector signed int vec_nabs (vector signed int); 47851 vector float vec_nabs (vector float); 47852 47853 vector float vec_nmsub (vector float, vector float, vector float); 47854 47855 vector float vec_nor (vector float, vector float); 47856 vector signed int vec_nor (vector signed int, vector signed int); 47857 vector unsigned int vec_nor (vector unsigned int, vector unsigned int); 47858 vector bool int vec_nor (vector bool int, vector bool int); 47859 vector signed short vec_nor (vector signed short, vector signed short); 47860 vector unsigned short vec_nor (vector unsigned short, vector unsigned short); 47861 vector bool short vec_nor (vector bool short, vector bool short); 47862 vector signed char vec_nor (vector signed char, vector signed char); 47863 vector unsigned char vec_nor (vector unsigned char, vector unsigned char); 47864 vector bool char vec_nor (vector bool char, vector bool char); 47865 47866 vector float vec_or (vector float, vector float); 47867 vector float vec_or (vector float, vector bool int); 47868 vector float vec_or (vector bool int, vector float); 47869 vector bool int vec_or (vector bool int, vector bool int); 47870 vector signed int vec_or (vector bool int, vector signed int); 47871 vector signed int vec_or (vector signed int, vector bool int); 47872 vector signed int vec_or (vector signed int, vector signed int); 47873 vector unsigned int vec_or (vector bool int, vector unsigned int); 47874 vector unsigned int vec_or (vector unsigned int, vector bool int); 47875 vector unsigned int vec_or (vector unsigned int, vector unsigned int); 47876 vector bool short vec_or (vector bool short, vector bool short); 47877 vector signed short vec_or (vector bool short, vector signed short); 47878 vector signed short vec_or (vector signed short, vector bool short); 47879 vector signed short vec_or (vector signed short, vector signed short); 47880 vector unsigned short vec_or (vector bool short, vector unsigned short); 47881 vector unsigned short vec_or (vector unsigned short, vector bool short); 47882 vector unsigned short vec_or (vector unsigned short, vector unsigned short); 47883 vector signed char vec_or (vector bool char, vector signed char); 47884 vector bool char vec_or (vector bool char, vector bool char); 47885 vector signed char vec_or (vector signed char, vector bool char); 47886 vector signed char vec_or (vector signed char, vector signed char); 47887 vector unsigned char vec_or (vector bool char, vector unsigned char); 47888 vector unsigned char vec_or (vector unsigned char, vector bool char); 47889 vector unsigned char vec_or (vector unsigned char, vector unsigned char); 47890 47891 vector signed char vec_pack (vector signed short, vector signed short); 47892 vector unsigned char vec_pack (vector unsigned short, vector unsigned short); 47893 vector bool char vec_pack (vector bool short, vector bool short); 47894 vector signed short vec_pack (vector signed int, vector signed int); 47895 vector unsigned short vec_pack (vector unsigned int, vector unsigned int); 47896 vector bool short vec_pack (vector bool int, vector bool int); 47897 47898 vector pixel vec_packpx (vector unsigned int, vector unsigned int); 47899 47900 vector unsigned char vec_packs (vector unsigned short, vector unsigned short); 47901 vector signed char vec_packs (vector signed short, vector signed short); 47902 vector unsigned short vec_packs (vector unsigned int, vector unsigned int); 47903 vector signed short vec_packs (vector signed int, vector signed int); 47904 47905 vector unsigned char vec_packsu (vector unsigned short, vector unsigned short); 47906 vector unsigned char vec_packsu (vector signed short, vector signed short); 47907 vector unsigned short vec_packsu (vector unsigned int, vector unsigned int); 47908 vector unsigned short vec_packsu (vector signed int, vector signed int); 47909 47910 vector float vec_perm (vector float, vector float, vector unsigned char); 47911 vector signed int vec_perm (vector signed int, vector signed int, vector unsigned char); 47912 vector unsigned int vec_perm (vector unsigned int, vector unsigned int, 47913 vector unsigned char); 47914 vector bool int vec_perm (vector bool int, vector bool int, vector unsigned char); 47915 vector signed short vec_perm (vector signed short, vector signed short, 47916 vector unsigned char); 47917 vector unsigned short vec_perm (vector unsigned short, vector unsigned short, 47918 vector unsigned char); 47919 vector bool short vec_perm (vector bool short, vector bool short, vector unsigned char); 47920 vector pixel vec_perm (vector pixel, vector pixel, vector unsigned char); 47921 vector signed char vec_perm (vector signed char, vector signed char, 47922 vector unsigned char); 47923 vector unsigned char vec_perm (vector unsigned char, vector unsigned char, 47924 vector unsigned char); 47925 vector bool char vec_perm (vector bool char, vector bool char, vector unsigned char); 47926 47927 vector float vec_re (vector float); 47928 47929 vector bool char vec_reve (vector bool char); 47930 vector signed char vec_reve (vector signed char); 47931 vector unsigned char vec_reve (vector unsigned char); 47932 vector bool int vec_reve (vector bool int); 47933 vector signed int vec_reve (vector signed int); 47934 vector unsigned int vec_reve (vector unsigned int); 47935 vector bool short vec_reve (vector bool short); 47936 vector signed short vec_reve (vector signed short); 47937 vector unsigned short vec_reve (vector unsigned short); 47938 47939 vector signed char vec_rl (vector signed char, vector unsigned char); 47940 vector unsigned char vec_rl (vector unsigned char, vector unsigned char); 47941 vector signed short vec_rl (vector signed short, vector unsigned short); 47942 vector unsigned short vec_rl (vector unsigned short, vector unsigned short); 47943 vector signed int vec_rl (vector signed int, vector unsigned int); 47944 vector unsigned int vec_rl (vector unsigned int, vector unsigned int); 47945 47946 vector float vec_round (vector float); 47947 47948 vector float vec_rsqrt (vector float); 47949 47950 vector float vec_rsqrte (vector float); 47951 47952 vector float vec_sel (vector float, vector float, vector bool int); 47953 vector float vec_sel (vector float, vector float, vector unsigned int); 47954 vector signed int vec_sel (vector signed int, vector signed int, vector bool int); 47955 vector signed int vec_sel (vector signed int, vector signed int, vector unsigned int); 47956 vector unsigned int vec_sel (vector unsigned int, vector unsigned int, vector bool int); 47957 vector unsigned int vec_sel (vector unsigned int, vector unsigned int, 47958 vector unsigned int); 47959 vector bool int vec_sel (vector bool int, vector bool int, vector bool int); 47960 vector bool int vec_sel (vector bool int, vector bool int, vector unsigned int); 47961 vector signed short vec_sel (vector signed short, vector signed short, 47962 vector bool short); 47963 vector signed short vec_sel (vector signed short, vector signed short, 47964 vector unsigned short); 47965 vector unsigned short vec_sel (vector unsigned short, vector unsigned short, 47966 vector bool short); 47967 vector unsigned short vec_sel (vector unsigned short, vector unsigned short, 47968 vector unsigned short); 47969 vector bool short vec_sel (vector bool short, vector bool short, vector bool short); 47970 vector bool short vec_sel (vector bool short, vector bool short, vector unsigned short); 47971 vector signed char vec_sel (vector signed char, vector signed char, vector bool char); 47972 vector signed char vec_sel (vector signed char, vector signed char, 47973 vector unsigned char); 47974 vector unsigned char vec_sel (vector unsigned char, vector unsigned char, 47975 vector bool char); 47976 vector unsigned char vec_sel (vector unsigned char, vector unsigned char, 47977 vector unsigned char); 47978 vector bool char vec_sel (vector bool char, vector bool char, vector bool char); 47979 vector bool char vec_sel (vector bool char, vector bool char, vector unsigned char); 47980 47981 vector signed char vec_sl (vector signed char, vector unsigned char); 47982 vector unsigned char vec_sl (vector unsigned char, vector unsigned char); 47983 vector signed short vec_sl (vector signed short, vector unsigned short); 47984 vector unsigned short vec_sl (vector unsigned short, vector unsigned short); 47985 vector signed int vec_sl (vector signed int, vector unsigned int); 47986 vector unsigned int vec_sl (vector unsigned int, vector unsigned int); 47987 47988 vector float vec_sld (vector float, vector float, const int); 47989 vector signed int vec_sld (vector signed int, vector signed int, const int); 47990 vector unsigned int vec_sld (vector unsigned int, vector unsigned int, const int); 47991 vector bool int vec_sld (vector bool int, vector bool int, const int); 47992 vector signed short vec_sld (vector signed short, vector signed short, const int); 47993 vector unsigned short vec_sld (vector unsigned short, vector unsigned short, const int); 47994 vector bool short vec_sld (vector bool short, vector bool short, const int); 47995 vector pixel vec_sld (vector pixel, vector pixel, const int); 47996 vector signed char vec_sld (vector signed char, vector signed char, const int); 47997 vector unsigned char vec_sld (vector unsigned char, vector unsigned char, const int); 47998 vector bool char vec_sld (vector bool char, vector bool char, const int); 47999 48000 vector signed int vec_sll (vector signed int, vector unsigned int); 48001 vector signed int vec_sll (vector signed int, vector unsigned short); 48002 vector signed int vec_sll (vector signed int, vector unsigned char); 48003 vector unsigned int vec_sll (vector unsigned int, vector unsigned int); 48004 vector unsigned int vec_sll (vector unsigned int, vector unsigned short); 48005 vector unsigned int vec_sll (vector unsigned int, vector unsigned char); 48006 vector bool int vec_sll (vector bool int, vector unsigned int); 48007 vector bool int vec_sll (vector bool int, vector unsigned short); 48008 vector bool int vec_sll (vector bool int, vector unsigned char); 48009 vector signed short vec_sll (vector signed short, vector unsigned int); 48010 vector signed short vec_sll (vector signed short, vector unsigned short); 48011 vector signed short vec_sll (vector signed short, vector unsigned char); 48012 vector unsigned short vec_sll (vector unsigned short, vector unsigned int); 48013 vector unsigned short vec_sll (vector unsigned short, vector unsigned short); 48014 vector unsigned short vec_sll (vector unsigned short, vector unsigned char); 48015 vector bool short vec_sll (vector bool short, vector unsigned int); 48016 vector bool short vec_sll (vector bool short, vector unsigned short); 48017 vector bool short vec_sll (vector bool short, vector unsigned char); 48018 vector pixel vec_sll (vector pixel, vector unsigned int); 48019 vector pixel vec_sll (vector pixel, vector unsigned short); 48020 vector pixel vec_sll (vector pixel, vector unsigned char); 48021 vector signed char vec_sll (vector signed char, vector unsigned int); 48022 vector signed char vec_sll (vector signed char, vector unsigned short); 48023 vector signed char vec_sll (vector signed char, vector unsigned char); 48024 vector unsigned char vec_sll (vector unsigned char, vector unsigned int); 48025 vector unsigned char vec_sll (vector unsigned char, vector unsigned short); 48026 vector unsigned char vec_sll (vector unsigned char, vector unsigned char); 48027 vector bool char vec_sll (vector bool char, vector unsigned int); 48028 vector bool char vec_sll (vector bool char, vector unsigned short); 48029 vector bool char vec_sll (vector bool char, vector unsigned char); 48030 48031 vector float vec_slo (vector float, vector signed char); 48032 vector float vec_slo (vector float, vector unsigned char); 48033 vector signed int vec_slo (vector signed int, vector signed char); 48034 vector signed int vec_slo (vector signed int, vector unsigned char); 48035 vector unsigned int vec_slo (vector unsigned int, vector signed char); 48036 vector unsigned int vec_slo (vector unsigned int, vector unsigned char); 48037 vector signed short vec_slo (vector signed short, vector signed char); 48038 vector signed short vec_slo (vector signed short, vector unsigned char); 48039 vector unsigned short vec_slo (vector unsigned short, vector signed char); 48040 vector unsigned short vec_slo (vector unsigned short, vector unsigned char); 48041 vector pixel vec_slo (vector pixel, vector signed char); 48042 vector pixel vec_slo (vector pixel, vector unsigned char); 48043 vector signed char vec_slo (vector signed char, vector signed char); 48044 vector signed char vec_slo (vector signed char, vector unsigned char); 48045 vector unsigned char vec_slo (vector unsigned char, vector signed char); 48046 vector unsigned char vec_slo (vector unsigned char, vector unsigned char); 48047 48048 vector signed char vec_splat (vector signed char, const int); 48049 vector unsigned char vec_splat (vector unsigned char, const int); 48050 vector bool char vec_splat (vector bool char, const int); 48051 vector signed short vec_splat (vector signed short, const int); 48052 vector unsigned short vec_splat (vector unsigned short, const int); 48053 vector bool short vec_splat (vector bool short, const int); 48054 vector pixel vec_splat (vector pixel, const int); 48055 vector float vec_splat (vector float, const int); 48056 vector signed int vec_splat (vector signed int, const int); 48057 vector unsigned int vec_splat (vector unsigned int, const int); 48058 vector bool int vec_splat (vector bool int, const int); 48059 48060 vector signed short vec_splat_s16 (const int); 48061 48062 vector signed int vec_splat_s32 (const int); 48063 48064 vector signed char vec_splat_s8 (const int); 48065 48066 vector unsigned short vec_splat_u16 (const int); 48067 48068 vector unsigned int vec_splat_u32 (const int); 48069 48070 vector unsigned char vec_splat_u8 (const int); 48071 48072 vector signed char vec_splats (signed char); 48073 vector unsigned char vec_splats (unsigned char); 48074 vector signed short vec_splats (signed short); 48075 vector unsigned short vec_splats (unsigned short); 48076 vector signed int vec_splats (signed int); 48077 vector unsigned int vec_splats (unsigned int); 48078 vector float vec_splats (float); 48079 48080 vector signed char vec_sr (vector signed char, vector unsigned char); 48081 vector unsigned char vec_sr (vector unsigned char, vector unsigned char); 48082 vector signed short vec_sr (vector signed short, vector unsigned short); 48083 vector unsigned short vec_sr (vector unsigned short, vector unsigned short); 48084 vector signed int vec_sr (vector signed int, vector unsigned int); 48085 vector unsigned int vec_sr (vector unsigned int, vector unsigned int); 48086 48087 vector signed char vec_sra (vector signed char, vector unsigned char); 48088 vector unsigned char vec_sra (vector unsigned char, vector unsigned char); 48089 vector signed short vec_sra (vector signed short, vector unsigned short); 48090 vector unsigned short vec_sra (vector unsigned short, vector unsigned short); 48091 vector signed int vec_sra (vector signed int, vector unsigned int); 48092 vector unsigned int vec_sra (vector unsigned int, vector unsigned int); 48093 48094 vector signed int vec_srl (vector signed int, vector unsigned int); 48095 vector signed int vec_srl (vector signed int, vector unsigned short); 48096 vector signed int vec_srl (vector signed int, vector unsigned char); 48097 vector unsigned int vec_srl (vector unsigned int, vector unsigned int); 48098 vector unsigned int vec_srl (vector unsigned int, vector unsigned short); 48099 vector unsigned int vec_srl (vector unsigned int, vector unsigned char); 48100 vector bool int vec_srl (vector bool int, vector unsigned int); 48101 vector bool int vec_srl (vector bool int, vector unsigned short); 48102 vector bool int vec_srl (vector bool int, vector unsigned char); 48103 vector signed short vec_srl (vector signed short, vector unsigned int); 48104 vector signed short vec_srl (vector signed short, vector unsigned short); 48105 vector signed short vec_srl (vector signed short, vector unsigned char); 48106 vector unsigned short vec_srl (vector unsigned short, vector unsigned int); 48107 vector unsigned short vec_srl (vector unsigned short, vector unsigned short); 48108 vector unsigned short vec_srl (vector unsigned short, vector unsigned char); 48109 vector bool short vec_srl (vector bool short, vector unsigned int); 48110 vector bool short vec_srl (vector bool short, vector unsigned short); 48111 vector bool short vec_srl (vector bool short, vector unsigned char); 48112 vector pixel vec_srl (vector pixel, vector unsigned int); 48113 vector pixel vec_srl (vector pixel, vector unsigned short); 48114 vector pixel vec_srl (vector pixel, vector unsigned char); 48115 vector signed char vec_srl (vector signed char, vector unsigned int); 48116 vector signed char vec_srl (vector signed char, vector unsigned short); 48117 vector signed char vec_srl (vector signed char, vector unsigned char); 48118 vector unsigned char vec_srl (vector unsigned char, vector unsigned int); 48119 vector unsigned char vec_srl (vector unsigned char, vector unsigned short); 48120 vector unsigned char vec_srl (vector unsigned char, vector unsigned char); 48121 vector bool char vec_srl (vector bool char, vector unsigned int); 48122 vector bool char vec_srl (vector bool char, vector unsigned short); 48123 vector bool char vec_srl (vector bool char, vector unsigned char); 48124 48125 vector float vec_sro (vector float, vector signed char); 48126 vector float vec_sro (vector float, vector unsigned char); 48127 vector signed int vec_sro (vector signed int, vector signed char); 48128 vector signed int vec_sro (vector signed int, vector unsigned char); 48129 vector unsigned int vec_sro (vector unsigned int, vector signed char); 48130 vector unsigned int vec_sro (vector unsigned int, vector unsigned char); 48131 vector signed short vec_sro (vector signed short, vector signed char); 48132 vector signed short vec_sro (vector signed short, vector unsigned char); 48133 vector unsigned short vec_sro (vector unsigned short, vector signed char); 48134 vector unsigned short vec_sro (vector unsigned short, vector unsigned char); 48135 vector pixel vec_sro (vector pixel, vector signed char); 48136 vector pixel vec_sro (vector pixel, vector unsigned char); 48137 vector signed char vec_sro (vector signed char, vector signed char); 48138 vector signed char vec_sro (vector signed char, vector unsigned char); 48139 vector unsigned char vec_sro (vector unsigned char, vector signed char); 48140 vector unsigned char vec_sro (vector unsigned char, vector unsigned char); 48141 48142 void vec_st (vector float, int, vector float *); 48143 void vec_st (vector float, int, float *); 48144 void vec_st (vector signed int, int, vector signed int *); 48145 void vec_st (vector signed int, int, int *); 48146 void vec_st (vector unsigned int, int, vector unsigned int *); 48147 void vec_st (vector unsigned int, int, unsigned int *); 48148 void vec_st (vector bool int, int, vector bool int *); 48149 void vec_st (vector bool int, int, unsigned int *); 48150 void vec_st (vector bool int, int, int *); 48151 void vec_st (vector signed short, int, vector signed short *); 48152 void vec_st (vector signed short, int, short *); 48153 void vec_st (vector unsigned short, int, vector unsigned short *); 48154 void vec_st (vector unsigned short, int, unsigned short *); 48155 void vec_st (vector bool short, int, vector bool short *); 48156 void vec_st (vector bool short, int, unsigned short *); 48157 void vec_st (vector pixel, int, vector pixel *); 48158 void vec_st (vector bool short, int, short *); 48159 void vec_st (vector signed char, int, vector signed char *); 48160 void vec_st (vector signed char, int, signed char *); 48161 void vec_st (vector unsigned char, int, vector unsigned char *); 48162 void vec_st (vector unsigned char, int, unsigned char *); 48163 void vec_st (vector bool char, int, vector bool char *); 48164 void vec_st (vector bool char, int, unsigned char *); 48165 void vec_st (vector bool char, int, signed char *); 48166 48167 void vec_ste (vector signed char, int, signed char *); 48168 void vec_ste (vector unsigned char, int, unsigned char *); 48169 void vec_ste (vector bool char, int, signed char *); 48170 void vec_ste (vector bool char, int, unsigned char *); 48171 void vec_ste (vector signed short, int, short *); 48172 void vec_ste (vector unsigned short, int, unsigned short *); 48173 void vec_ste (vector bool short, int, short *); 48174 void vec_ste (vector bool short, int, unsigned short *); 48175 void vec_ste (vector pixel, int, short *); 48176 void vec_ste (vector pixel, int, unsigned short *); 48177 void vec_ste (vector float, int, float *); 48178 void vec_ste (vector signed int, int, int *); 48179 void vec_ste (vector unsigned int, int, unsigned int *); 48180 void vec_ste (vector bool int, int, int *); 48181 void vec_ste (vector bool int, int, unsigned int *); 48182 48183 void vec_stl (vector float, int, vector float *); 48184 void vec_stl (vector float, int, float *); 48185 void vec_stl (vector signed int, int, vector signed int *); 48186 void vec_stl (vector signed int, int, int *); 48187 void vec_stl (vector unsigned int, int, vector unsigned int *); 48188 void vec_stl (vector unsigned int, int, unsigned int *); 48189 void vec_stl (vector bool int, int, vector bool int *); 48190 void vec_stl (vector bool int, int, unsigned int *); 48191 void vec_stl (vector bool int, int, int *); 48192 void vec_stl (vector signed short, int, vector signed short *); 48193 void vec_stl (vector signed short, int, short *); 48194 void vec_stl (vector unsigned short, int, vector unsigned short *); 48195 void vec_stl (vector unsigned short, int, unsigned short *); 48196 void vec_stl (vector bool short, int, vector bool short *); 48197 void vec_stl (vector bool short, int, unsigned short *); 48198 void vec_stl (vector bool short, int, short *); 48199 void vec_stl (vector pixel, int, vector pixel *); 48200 void vec_stl (vector signed char, int, vector signed char *); 48201 void vec_stl (vector signed char, int, signed char *); 48202 void vec_stl (vector unsigned char, int, vector unsigned char *); 48203 void vec_stl (vector unsigned char, int, unsigned char *); 48204 void vec_stl (vector bool char, int, vector bool char *); 48205 void vec_stl (vector bool char, int, unsigned char *); 48206 void vec_stl (vector bool char, int, signed char *); 48207 48208 void vec_stvebx (vector signed char, int, signed char *); 48209 void vec_stvebx (vector unsigned char, int, unsigned char *); 48210 void vec_stvebx (vector bool char, int, signed char *); 48211 void vec_stvebx (vector bool char, int, unsigned char *); 48212 48213 void vec_stvehx (vector signed short, int, short *); 48214 void vec_stvehx (vector unsigned short, int, unsigned short *); 48215 void vec_stvehx (vector bool short, int, short *); 48216 void vec_stvehx (vector bool short, int, unsigned short *); 48217 48218 void vec_stvewx (vector float, int, float *); 48219 void vec_stvewx (vector signed int, int, int *); 48220 void vec_stvewx (vector unsigned int, int, unsigned int *); 48221 void vec_stvewx (vector bool int, int, int *); 48222 void vec_stvewx (vector bool int, int, unsigned int *); 48223 48224 vector signed char vec_sub (vector bool char, vector signed char); 48225 vector signed char vec_sub (vector signed char, vector bool char); 48226 vector signed char vec_sub (vector signed char, vector signed char); 48227 vector unsigned char vec_sub (vector bool char, vector unsigned char); 48228 vector unsigned char vec_sub (vector unsigned char, vector bool char); 48229 vector unsigned char vec_sub (vector unsigned char, vector unsigned char); 48230 vector signed short vec_sub (vector bool short, vector signed short); 48231 vector signed short vec_sub (vector signed short, vector bool short); 48232 vector signed short vec_sub (vector signed short, vector signed short); 48233 vector unsigned short vec_sub (vector bool short, vector unsigned short); 48234 vector unsigned short vec_sub (vector unsigned short, vector bool short); 48235 vector unsigned short vec_sub (vector unsigned short, vector unsigned short); 48236 vector signed int vec_sub (vector bool int, vector signed int); 48237 vector signed int vec_sub (vector signed int, vector bool int); 48238 vector signed int vec_sub (vector signed int, vector signed int); 48239 vector unsigned int vec_sub (vector bool int, vector unsigned int); 48240 vector unsigned int vec_sub (vector unsigned int, vector bool int); 48241 vector unsigned int vec_sub (vector unsigned int, vector unsigned int); 48242 vector float vec_sub (vector float, vector float); 48243 48244 vector signed int vec_subc (vector signed int, vector signed int); 48245 vector unsigned int vec_subc (vector unsigned int, vector unsigned int); 48246 48247 vector signed int vec_sube (vector signed int, vector signed int, 48248 vector signed int); 48249 vector unsigned int vec_sube (vector unsigned int, vector unsigned int, 48250 vector unsigned int); 48251 48252 vector signed int vec_subec (vector signed int, vector signed int, 48253 vector signed int); 48254 vector unsigned int vec_subec (vector unsigned int, vector unsigned int, 48255 vector unsigned int); 48256 48257 vector unsigned char vec_subs (vector bool char, vector unsigned char); 48258 vector unsigned char vec_subs (vector unsigned char, vector bool char); 48259 vector unsigned char vec_subs (vector unsigned char, vector unsigned char); 48260 vector signed char vec_subs (vector bool char, vector signed char); 48261 vector signed char vec_subs (vector signed char, vector bool char); 48262 vector signed char vec_subs (vector signed char, vector signed char); 48263 vector unsigned short vec_subs (vector bool short, vector unsigned short); 48264 vector unsigned short vec_subs (vector unsigned short, vector bool short); 48265 vector unsigned short vec_subs (vector unsigned short, vector unsigned short); 48266 vector signed short vec_subs (vector bool short, vector signed short); 48267 vector signed short vec_subs (vector signed short, vector bool short); 48268 vector signed short vec_subs (vector signed short, vector signed short); 48269 vector unsigned int vec_subs (vector bool int, vector unsigned int); 48270 vector unsigned int vec_subs (vector unsigned int, vector bool int); 48271 vector unsigned int vec_subs (vector unsigned int, vector unsigned int); 48272 vector signed int vec_subs (vector bool int, vector signed int); 48273 vector signed int vec_subs (vector signed int, vector bool int); 48274 vector signed int vec_subs (vector signed int, vector signed int); 48275 48276 vector signed int vec_sum2s (vector signed int, vector signed int); 48277 48278 vector unsigned int vec_sum4s (vector unsigned char, vector unsigned int); 48279 vector signed int vec_sum4s (vector signed char, vector signed int); 48280 vector signed int vec_sum4s (vector signed short, vector signed int); 48281 48282 vector signed int vec_sums (vector signed int, vector signed int); 48283 48284 vector float vec_trunc (vector float); 48285 48286 vector signed short vec_unpackh (vector signed char); 48287 vector bool short vec_unpackh (vector bool char); 48288 vector signed int vec_unpackh (vector signed short); 48289 vector bool int vec_unpackh (vector bool short); 48290 vector unsigned int vec_unpackh (vector pixel); 48291 48292 vector signed short vec_unpackl (vector signed char); 48293 vector bool short vec_unpackl (vector bool char); 48294 vector unsigned int vec_unpackl (vector pixel); 48295 vector signed int vec_unpackl (vector signed short); 48296 vector bool int vec_unpackl (vector bool short); 48297 48298 vector float vec_vaddfp (vector float, vector float); 48299 48300 vector signed char vec_vaddsbs (vector bool char, vector signed char); 48301 vector signed char vec_vaddsbs (vector signed char, vector bool char); 48302 vector signed char vec_vaddsbs (vector signed char, vector signed char); 48303 48304 vector signed short vec_vaddshs (vector bool short, vector signed short); 48305 vector signed short vec_vaddshs (vector signed short, vector bool short); 48306 vector signed short vec_vaddshs (vector signed short, vector signed short); 48307 48308 vector signed int vec_vaddsws (vector bool int, vector signed int); 48309 vector signed int vec_vaddsws (vector signed int, vector bool int); 48310 vector signed int vec_vaddsws (vector signed int, vector signed int); 48311 48312 vector signed char vec_vaddubm (vector bool char, vector signed char); 48313 vector signed char vec_vaddubm (vector signed char, vector bool char); 48314 vector signed char vec_vaddubm (vector signed char, vector signed char); 48315 vector unsigned char vec_vaddubm (vector bool char, vector unsigned char); 48316 vector unsigned char vec_vaddubm (vector unsigned char, vector bool char); 48317 vector unsigned char vec_vaddubm (vector unsigned char, vector unsigned char); 48318 48319 vector unsigned char vec_vaddubs (vector bool char, vector unsigned char); 48320 vector unsigned char vec_vaddubs (vector unsigned char, vector bool char); 48321 vector unsigned char vec_vaddubs (vector unsigned char, vector unsigned char); 48322 48323 vector signed short vec_vadduhm (vector bool short, vector signed short); 48324 vector signed short vec_vadduhm (vector signed short, vector bool short); 48325 vector signed short vec_vadduhm (vector signed short, vector signed short); 48326 vector unsigned short vec_vadduhm (vector bool short, vector unsigned short); 48327 vector unsigned short vec_vadduhm (vector unsigned short, vector bool short); 48328 vector unsigned short vec_vadduhm (vector unsigned short, vector unsigned short); 48329 48330 vector unsigned short vec_vadduhs (vector bool short, vector unsigned short); 48331 vector unsigned short vec_vadduhs (vector unsigned short, vector bool short); 48332 vector unsigned short vec_vadduhs (vector unsigned short, vector unsigned short); 48333 48334 vector signed int vec_vadduwm (vector bool int, vector signed int); 48335 vector signed int vec_vadduwm (vector signed int, vector bool int); 48336 vector signed int vec_vadduwm (vector signed int, vector signed int); 48337 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int); 48338 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int); 48339 vector unsigned int vec_vadduwm (vector unsigned int, vector unsigned int); 48340 48341 vector unsigned int vec_vadduws (vector bool int, vector unsigned int); 48342 vector unsigned int vec_vadduws (vector unsigned int, vector bool int); 48343 vector unsigned int vec_vadduws (vector unsigned int, vector unsigned int); 48344 48345 vector signed char vec_vavgsb (vector signed char, vector signed char); 48346 48347 vector signed short vec_vavgsh (vector signed short, vector signed short); 48348 48349 vector signed int vec_vavgsw (vector signed int, vector signed int); 48350 48351 vector unsigned char vec_vavgub (vector unsigned char, vector unsigned char); 48352 48353 vector unsigned short vec_vavguh (vector unsigned short, vector unsigned short); 48354 48355 vector unsigned int vec_vavguw (vector unsigned int, vector unsigned int); 48356 48357 vector float vec_vcfsx (vector signed int, const int); 48358 48359 vector float vec_vcfux (vector unsigned int, const int); 48360 48361 vector bool int vec_vcmpeqfp (vector float, vector float); 48362 48363 vector bool char vec_vcmpequb (vector signed char, vector signed char); 48364 vector bool char vec_vcmpequb (vector unsigned char, vector unsigned char); 48365 48366 vector bool short vec_vcmpequh (vector signed short, vector signed short); 48367 vector bool short vec_vcmpequh (vector unsigned short, vector unsigned short); 48368 48369 vector bool int vec_vcmpequw (vector signed int, vector signed int); 48370 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int); 48371 48372 vector bool int vec_vcmpgtfp (vector float, vector float); 48373 48374 vector bool char vec_vcmpgtsb (vector signed char, vector signed char); 48375 48376 vector bool short vec_vcmpgtsh (vector signed short, vector signed short); 48377 48378 vector bool int vec_vcmpgtsw (vector signed int, vector signed int); 48379 48380 vector bool char vec_vcmpgtub (vector unsigned char, vector unsigned char); 48381 48382 vector bool short vec_vcmpgtuh (vector unsigned short, vector unsigned short); 48383 48384 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int); 48385 48386 vector float vec_vmaxfp (vector float, vector float); 48387 48388 vector signed char vec_vmaxsb (vector bool char, vector signed char); 48389 vector signed char vec_vmaxsb (vector signed char, vector bool char); 48390 vector signed char vec_vmaxsb (vector signed char, vector signed char); 48391 48392 vector signed short vec_vmaxsh (vector bool short, vector signed short); 48393 vector signed short vec_vmaxsh (vector signed short, vector bool short); 48394 vector signed short vec_vmaxsh (vector signed short, vector signed short); 48395 48396 vector signed int vec_vmaxsw (vector bool int, vector signed int); 48397 vector signed int vec_vmaxsw (vector signed int, vector bool int); 48398 vector signed int vec_vmaxsw (vector signed int, vector signed int); 48399 48400 vector unsigned char vec_vmaxub (vector bool char, vector unsigned char); 48401 vector unsigned char vec_vmaxub (vector unsigned char, vector bool char); 48402 vector unsigned char vec_vmaxub (vector unsigned char, vector unsigned char); 48403 48404 vector unsigned short vec_vmaxuh (vector bool short, vector unsigned short); 48405 vector unsigned short vec_vmaxuh (vector unsigned short, vector bool short); 48406 vector unsigned short vec_vmaxuh (vector unsigned short, vector unsigned short); 48407 48408 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int); 48409 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int); 48410 vector unsigned int vec_vmaxuw (vector unsigned int, vector unsigned int); 48411 48412 vector float vec_vminfp (vector float, vector float); 48413 48414 vector signed char vec_vminsb (vector bool char, vector signed char); 48415 vector signed char vec_vminsb (vector signed char, vector bool char); 48416 vector signed char vec_vminsb (vector signed char, vector signed char); 48417 48418 vector signed short vec_vminsh (vector bool short, vector signed short); 48419 vector signed short vec_vminsh (vector signed short, vector bool short); 48420 vector signed short vec_vminsh (vector signed short, vector signed short); 48421 48422 vector signed int vec_vminsw (vector bool int, vector signed int); 48423 vector signed int vec_vminsw (vector signed int, vector bool int); 48424 vector signed int vec_vminsw (vector signed int, vector signed int); 48425 48426 vector unsigned char vec_vminub (vector bool char, vector unsigned char); 48427 vector unsigned char vec_vminub (vector unsigned char, vector bool char); 48428 vector unsigned char vec_vminub (vector unsigned char, vector unsigned char); 48429 48430 vector unsigned short vec_vminuh (vector bool short, vector unsigned short); 48431 vector unsigned short vec_vminuh (vector unsigned short, vector bool short); 48432 vector unsigned short vec_vminuh (vector unsigned short, vector unsigned short); 48433 48434 vector unsigned int vec_vminuw (vector bool int, vector unsigned int); 48435 vector unsigned int vec_vminuw (vector unsigned int, vector bool int); 48436 vector unsigned int vec_vminuw (vector unsigned int, vector unsigned int); 48437 48438 vector bool char vec_vmrghb (vector bool char, vector bool char); 48439 vector signed char vec_vmrghb (vector signed char, vector signed char); 48440 vector unsigned char vec_vmrghb (vector unsigned char, vector unsigned char); 48441 48442 vector bool short vec_vmrghh (vector bool short, vector bool short); 48443 vector signed short vec_vmrghh (vector signed short, vector signed short); 48444 vector unsigned short vec_vmrghh (vector unsigned short, vector unsigned short); 48445 vector pixel vec_vmrghh (vector pixel, vector pixel); 48446 48447 vector float vec_vmrghw (vector float, vector float); 48448 vector bool int vec_vmrghw (vector bool int, vector bool int); 48449 vector signed int vec_vmrghw (vector signed int, vector signed int); 48450 vector unsigned int vec_vmrghw (vector unsigned int, vector unsigned int); 48451 48452 vector bool char vec_vmrglb (vector bool char, vector bool char); 48453 vector signed char vec_vmrglb (vector signed char, vector signed char); 48454 vector unsigned char vec_vmrglb (vector unsigned char, vector unsigned char); 48455 48456 vector bool short vec_vmrglh (vector bool short, vector bool short); 48457 vector signed short vec_vmrglh (vector signed short, vector signed short); 48458 vector unsigned short vec_vmrglh (vector unsigned short, vector unsigned short); 48459 vector pixel vec_vmrglh (vector pixel, vector pixel); 48460 48461 vector float vec_vmrglw (vector float, vector float); 48462 vector signed int vec_vmrglw (vector signed int, vector signed int); 48463 vector unsigned int vec_vmrglw (vector unsigned int, vector unsigned int); 48464 vector bool int vec_vmrglw (vector bool int, vector bool int); 48465 48466 vector signed int vec_vmsummbm (vector signed char, vector unsigned char, 48467 vector signed int); 48468 48469 vector signed int vec_vmsumshm (vector signed short, vector signed short, 48470 vector signed int); 48471 48472 vector signed int vec_vmsumshs (vector signed short, vector signed short, 48473 vector signed int); 48474 48475 vector unsigned int vec_vmsumubm (vector unsigned char, vector unsigned char, 48476 vector unsigned int); 48477 48478 vector unsigned int vec_vmsumuhm (vector unsigned short, vector unsigned short, 48479 vector unsigned int); 48480 48481 vector unsigned int vec_vmsumuhs (vector unsigned short, vector unsigned short, 48482 vector unsigned int); 48483 48484 vector signed short vec_vmulesb (vector signed char, vector signed char); 48485 48486 vector signed int vec_vmulesh (vector signed short, vector signed short); 48487 48488 vector unsigned short vec_vmuleub (vector unsigned char, vector unsigned char); 48489 48490 vector unsigned int vec_vmuleuh (vector unsigned short, vector unsigned short); 48491 48492 vector signed short vec_vmulosb (vector signed char, vector signed char); 48493 48494 vector signed int vec_vmulosh (vector signed short, vector signed short); 48495 48496 vector unsigned short vec_vmuloub (vector unsigned char, vector unsigned char); 48497 48498 vector unsigned int vec_vmulouh (vector unsigned short, vector unsigned short); 48499 48500 vector signed char vec_vpkshss (vector signed short, vector signed short); 48501 48502 vector unsigned char vec_vpkshus (vector signed short, vector signed short); 48503 48504 vector signed short vec_vpkswss (vector signed int, vector signed int); 48505 48506 vector unsigned short vec_vpkswus (vector signed int, vector signed int); 48507 48508 vector bool char vec_vpkuhum (vector bool short, vector bool short); 48509 vector signed char vec_vpkuhum (vector signed short, vector signed short); 48510 vector unsigned char vec_vpkuhum (vector unsigned short, vector unsigned short); 48511 48512 vector unsigned char vec_vpkuhus (vector unsigned short, vector unsigned short); 48513 48514 vector bool short vec_vpkuwum (vector bool int, vector bool int); 48515 vector signed short vec_vpkuwum (vector signed int, vector signed int); 48516 vector unsigned short vec_vpkuwum (vector unsigned int, vector unsigned int); 48517 48518 vector unsigned short vec_vpkuwus (vector unsigned int, vector unsigned int); 48519 48520 vector signed char vec_vrlb (vector signed char, vector unsigned char); 48521 vector unsigned char vec_vrlb (vector unsigned char, vector unsigned char); 48522 48523 vector signed short vec_vrlh (vector signed short, vector unsigned short); 48524 vector unsigned short vec_vrlh (vector unsigned short, vector unsigned short); 48525 48526 vector signed int vec_vrlw (vector signed int, vector unsigned int); 48527 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int); 48528 48529 vector signed char vec_vslb (vector signed char, vector unsigned char); 48530 vector unsigned char vec_vslb (vector unsigned char, vector unsigned char); 48531 48532 vector signed short vec_vslh (vector signed short, vector unsigned short); 48533 vector unsigned short vec_vslh (vector unsigned short, vector unsigned short); 48534 48535 vector signed int vec_vslw (vector signed int, vector unsigned int); 48536 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int); 48537 48538 vector signed char vec_vspltb (vector signed char, const int); 48539 vector unsigned char vec_vspltb (vector unsigned char, const int); 48540 vector bool char vec_vspltb (vector bool char, const int); 48541 48542 vector bool short vec_vsplth (vector bool short, const int); 48543 vector signed short vec_vsplth (vector signed short, const int); 48544 vector unsigned short vec_vsplth (vector unsigned short, const int); 48545 vector pixel vec_vsplth (vector pixel, const int); 48546 48547 vector float vec_vspltw (vector float, const int); 48548 vector signed int vec_vspltw (vector signed int, const int); 48549 vector unsigned int vec_vspltw (vector unsigned int, const int); 48550 vector bool int vec_vspltw (vector bool int, const int); 48551 48552 vector signed char vec_vsrab (vector signed char, vector unsigned char); 48553 vector unsigned char vec_vsrab (vector unsigned char, vector unsigned char); 48554 48555 vector signed short vec_vsrah (vector signed short, vector unsigned short); 48556 vector unsigned short vec_vsrah (vector unsigned short, vector unsigned short); 48557 48558 vector signed int vec_vsraw (vector signed int, vector unsigned int); 48559 vector unsigned int vec_vsraw (vector unsigned int, vector unsigned int); 48560 48561 vector signed char vec_vsrb (vector signed char, vector unsigned char); 48562 vector unsigned char vec_vsrb (vector unsigned char, vector unsigned char); 48563 48564 vector signed short vec_vsrh (vector signed short, vector unsigned short); 48565 vector unsigned short vec_vsrh (vector unsigned short, vector unsigned short); 48566 48567 vector signed int vec_vsrw (vector signed int, vector unsigned int); 48568 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int); 48569 48570 vector float vec_vsubfp (vector float, vector float); 48571 48572 vector signed char vec_vsubsbs (vector bool char, vector signed char); 48573 vector signed char vec_vsubsbs (vector signed char, vector bool char); 48574 vector signed char vec_vsubsbs (vector signed char, vector signed char); 48575 48576 vector signed short vec_vsubshs (vector bool short, vector signed short); 48577 vector signed short vec_vsubshs (vector signed short, vector bool short); 48578 vector signed short vec_vsubshs (vector signed short, vector signed short); 48579 48580 vector signed int vec_vsubsws (vector bool int, vector signed int); 48581 vector signed int vec_vsubsws (vector signed int, vector bool int); 48582 vector signed int vec_vsubsws (vector signed int, vector signed int); 48583 48584 vector signed char vec_vsububm (vector bool char, vector signed char); 48585 vector signed char vec_vsububm (vector signed char, vector bool char); 48586 vector signed char vec_vsububm (vector signed char, vector signed char); 48587 vector unsigned char vec_vsububm (vector bool char, vector unsigned char); 48588 vector unsigned char vec_vsububm (vector unsigned char, vector bool char); 48589 vector unsigned char vec_vsububm (vector unsigned char, vector unsigned char); 48590 48591 vector unsigned char vec_vsububs (vector bool char, vector unsigned char); 48592 vector unsigned char vec_vsububs (vector unsigned char, vector bool char); 48593 vector unsigned char vec_vsububs (vector unsigned char, vector unsigned char); 48594 48595 vector signed short vec_vsubuhm (vector bool short, vector signed short); 48596 vector signed short vec_vsubuhm (vector signed short, vector bool short); 48597 vector signed short vec_vsubuhm (vector signed short, vector signed short); 48598 vector unsigned short vec_vsubuhm (vector bool short, vector unsigned short); 48599 vector unsigned short vec_vsubuhm (vector unsigned short, vector bool short); 48600 vector unsigned short vec_vsubuhm (vector unsigned short, vector unsigned short); 48601 48602 vector unsigned short vec_vsubuhs (vector bool short, vector unsigned short); 48603 vector unsigned short vec_vsubuhs (vector unsigned short, vector bool short); 48604 vector unsigned short vec_vsubuhs (vector unsigned short, vector unsigned short); 48605 48606 vector signed int vec_vsubuwm (vector bool int, vector signed int); 48607 vector signed int vec_vsubuwm (vector signed int, vector bool int); 48608 vector signed int vec_vsubuwm (vector signed int, vector signed int); 48609 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int); 48610 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int); 48611 vector unsigned int vec_vsubuwm (vector unsigned int, vector unsigned int); 48612 48613 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int); 48614 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int); 48615 vector unsigned int vec_vsubuws (vector unsigned int, vector unsigned int); 48616 48617 vector signed int vec_vsum4sbs (vector signed char, vector signed int); 48618 48619 vector signed int vec_vsum4shs (vector signed short, vector signed int); 48620 48621 vector unsigned int vec_vsum4ubs (vector unsigned char, vector unsigned int); 48622 48623 vector unsigned int vec_vupkhpx (vector pixel); 48624 48625 vector bool short vec_vupkhsb (vector bool char); 48626 vector signed short vec_vupkhsb (vector signed char); 48627 48628 vector bool int vec_vupkhsh (vector bool short); 48629 vector signed int vec_vupkhsh (vector signed short); 48630 48631 vector unsigned int vec_vupklpx (vector pixel); 48632 48633 vector bool short vec_vupklsb (vector bool char); 48634 vector signed short vec_vupklsb (vector signed char); 48635 48636 vector bool int vec_vupklsh (vector bool short); 48637 vector signed int vec_vupklsh (vector signed short); 48638 48639 vector float vec_xor (vector float, vector float); 48640 vector float vec_xor (vector float, vector bool int); 48641 vector float vec_xor (vector bool int, vector float); 48642 vector bool int vec_xor (vector bool int, vector bool int); 48643 vector signed int vec_xor (vector bool int, vector signed int); 48644 vector signed int vec_xor (vector signed int, vector bool int); 48645 vector signed int vec_xor (vector signed int, vector signed int); 48646 vector unsigned int vec_xor (vector bool int, vector unsigned int); 48647 vector unsigned int vec_xor (vector unsigned int, vector bool int); 48648 vector unsigned int vec_xor (vector unsigned int, vector unsigned int); 48649 vector bool short vec_xor (vector bool short, vector bool short); 48650 vector signed short vec_xor (vector bool short, vector signed short); 48651 vector signed short vec_xor (vector signed short, vector bool short); 48652 vector signed short vec_xor (vector signed short, vector signed short); 48653 vector unsigned short vec_xor (vector bool short, vector unsigned short); 48654 vector unsigned short vec_xor (vector unsigned short, vector bool short); 48655 vector unsigned short vec_xor (vector unsigned short, vector unsigned short); 48656 vector signed char vec_xor (vector bool char, vector signed char); 48657 vector bool char vec_xor (vector bool char, vector bool char); 48658 vector signed char vec_xor (vector signed char, vector bool char); 48659 vector signed char vec_xor (vector signed char, vector signed char); 48660 vector unsigned char vec_xor (vector bool char, vector unsigned char); 48661 vector unsigned char vec_xor (vector unsigned char, vector bool char); 48662 vector unsigned char vec_xor (vector unsigned char, vector unsigned char); 48663 48664 48665File: gcc.info, Node: PowerPC AltiVec Built-in Functions Available on ISA 2.06, Next: PowerPC AltiVec Built-in Functions Available on ISA 2.07, Prev: PowerPC AltiVec Built-in Functions on ISA 2.05, Up: PowerPC AltiVec/VSX Built-in Functions 48666 486676.60.23.2 PowerPC AltiVec Built-in Functions Available on ISA 2.06 48668.................................................................. 48669 48670The AltiVec built-in functions described in this section are available 48671on the PowerPC family of processors starting with ISA 2.06 or later. 48672These are normally enabled by adding '-mvsx' to the command line. 48673 48674 When '-mvsx' is used, the following additional vector types are 48675implemented. 48676 48677 vector unsigned __int128 48678 vector signed __int128 48679 vector unsigned long long int 48680 vector signed long long int 48681 vector double 48682 48683 The long long types are only implemented for 64-bit code generation. 48684 48685 48686 vector bool long long vec_and (vector bool long long int, vector bool long long); 48687 48688 vector double vec_ctf (vector unsigned long, const int); 48689 vector double vec_ctf (vector signed long, const int); 48690 48691 vector signed long vec_cts (vector double, const int); 48692 48693 vector unsigned long vec_ctu (vector double, const int); 48694 48695 void vec_dst (const unsigned long *, int, const int); 48696 void vec_dst (const long *, int, const int); 48697 48698 void vec_dststt (const unsigned long *, int, const int); 48699 void vec_dststt (const long *, int, const int); 48700 48701 void vec_dstt (const unsigned long *, int, const int); 48702 void vec_dstt (const long *, int, const int); 48703 48704 vector unsigned char vec_lvsl (int, const unsigned long *); 48705 vector unsigned char vec_lvsl (int, const long *); 48706 48707 vector unsigned char vec_lvsr (int, const unsigned long *); 48708 vector unsigned char vec_lvsr (int, const long *); 48709 48710 vector double vec_mul (vector double, vector double); 48711 vector long vec_mul (vector long, vector long); 48712 vector unsigned long vec_mul (vector unsigned long, vector unsigned long); 48713 48714 vector unsigned long long vec_mule (vector unsigned int, vector unsigned int); 48715 vector signed long long vec_mule (vector signed int, vector signed int); 48716 48717 vector unsigned long long vec_mulo (vector unsigned int, vector unsigned int); 48718 vector signed long long vec_mulo (vector signed int, vector signed int); 48719 48720 vector double vec_nabs (vector double); 48721 48722 vector bool long long vec_reve (vector bool long long); 48723 vector signed long long vec_reve (vector signed long long); 48724 vector unsigned long long vec_reve (vector unsigned long long); 48725 vector double vec_sld (vector double, vector double, const int); 48726 48727 vector bool long long int vec_sld (vector bool long long int, 48728 vector bool long long int, const int); 48729 vector long long int vec_sld (vector long long int, vector long long int, const int); 48730 vector unsigned long long int vec_sld (vector unsigned long long int, 48731 vector unsigned long long int, const int); 48732 48733 vector long long int vec_sll (vector long long int, vector unsigned char); 48734 vector unsigned long long int vec_sll (vector unsigned long long int, 48735 vector unsigned char); 48736 48737 vector signed long long vec_slo (vector signed long long, vector signed char); 48738 vector signed long long vec_slo (vector signed long long, vector unsigned char); 48739 vector unsigned long long vec_slo (vector unsigned long long, vector signed char); 48740 vector unsigned long long vec_slo (vector unsigned long long, vector unsigned char); 48741 48742 vector signed long vec_splat (vector signed long, const int); 48743 vector unsigned long vec_splat (vector unsigned long, const int); 48744 48745 vector long long int vec_srl (vector long long int, vector unsigned char); 48746 vector unsigned long long int vec_srl (vector unsigned long long int, 48747 vector unsigned char); 48748 48749 vector long long int vec_sro (vector long long int, vector char); 48750 vector long long int vec_sro (vector long long int, vector unsigned char); 48751 vector unsigned long long int vec_sro (vector unsigned long long int, vector char); 48752 vector unsigned long long int vec_sro (vector unsigned long long int, 48753 vector unsigned char); 48754 48755 vector signed __int128 vec_subc (vector signed __int128, vector signed __int128); 48756 vector unsigned __int128 vec_subc (vector unsigned __int128, vector unsigned __int128); 48757 48758 vector signed __int128 vec_sube (vector signed __int128, vector signed __int128, 48759 vector signed __int128); 48760 vector unsigned __int128 vec_sube (vector unsigned __int128, vector unsigned __int128, 48761 vector unsigned __int128); 48762 48763 vector signed __int128 vec_subec (vector signed __int128, vector signed __int128, 48764 vector signed __int128); 48765 vector unsigned __int128 vec_subec (vector unsigned __int128, vector unsigned __int128, 48766 vector unsigned __int128); 48767 48768 vector double vec_unpackh (vector float); 48769 48770 vector double vec_unpackl (vector float); 48771 48772 vector double vec_doublee (vector float); 48773 vector double vec_doublee (vector signed int); 48774 vector double vec_doublee (vector unsigned int); 48775 48776 vector double vec_doubleo (vector float); 48777 vector double vec_doubleo (vector signed int); 48778 vector double vec_doubleo (vector unsigned int); 48779 48780 vector double vec_doubleh (vector float); 48781 vector double vec_doubleh (vector signed int); 48782 vector double vec_doubleh (vector unsigned int); 48783 48784 vector double vec_doublel (vector float); 48785 vector double vec_doublel (vector signed int); 48786 vector double vec_doublel (vector unsigned int); 48787 48788 vector float vec_float (vector signed int); 48789 vector float vec_float (vector unsigned int); 48790 48791 vector float vec_float2 (vector signed long long, vector signed long long); 48792 vector float vec_float2 (vector unsigned long long, vector signed long long); 48793 48794 vector float vec_floate (vector double); 48795 vector float vec_floate (vector signed long long); 48796 vector float vec_floate (vector unsigned long long); 48797 48798 vector float vec_floato (vector double); 48799 vector float vec_floato (vector signed long long); 48800 vector float vec_floato (vector unsigned long long); 48801 48802 vector signed long long vec_signed (vector double); 48803 vector signed int vec_signed (vector float); 48804 48805 vector signed int vec_signede (vector double); 48806 48807 vector signed int vec_signedo (vector double); 48808 48809 vector signed char vec_sldw (vector signed char, vector signed char, const int); 48810 vector unsigned char vec_sldw (vector unsigned char, vector unsigned char, const int); 48811 vector signed short vec_sldw (vector signed short, vector signed short, const int); 48812 vector unsigned short vec_sldw (vector unsigned short, 48813 vector unsigned short, const int); 48814 vector signed int vec_sldw (vector signed int, vector signed int, const int); 48815 vector unsigned int vec_sldw (vector unsigned int, vector unsigned int, const int); 48816 vector signed long long vec_sldw (vector signed long long, 48817 vector signed long long, const int); 48818 vector unsigned long long vec_sldw (vector unsigned long long, 48819 vector unsigned long long, const int); 48820 48821 vector signed long long vec_unsigned (vector double); 48822 vector signed int vec_unsigned (vector float); 48823 48824 vector signed int vec_unsignede (vector double); 48825 48826 vector signed int vec_unsignedo (vector double); 48827 48828 vector double vec_abs (vector double); 48829 vector double vec_add (vector double, vector double); 48830 vector double vec_and (vector double, vector double); 48831 vector double vec_and (vector double, vector bool long); 48832 vector double vec_and (vector bool long, vector double); 48833 vector long vec_and (vector long, vector long); 48834 vector long vec_and (vector long, vector bool long); 48835 vector long vec_and (vector bool long, vector long); 48836 vector unsigned long vec_and (vector unsigned long, vector unsigned long); 48837 vector unsigned long vec_and (vector unsigned long, vector bool long); 48838 vector unsigned long vec_and (vector bool long, vector unsigned long); 48839 vector double vec_andc (vector double, vector double); 48840 vector double vec_andc (vector double, vector bool long); 48841 vector double vec_andc (vector bool long, vector double); 48842 vector long vec_andc (vector long, vector long); 48843 vector long vec_andc (vector long, vector bool long); 48844 vector long vec_andc (vector bool long, vector long); 48845 vector unsigned long vec_andc (vector unsigned long, vector unsigned long); 48846 vector unsigned long vec_andc (vector unsigned long, vector bool long); 48847 vector unsigned long vec_andc (vector bool long, vector unsigned long); 48848 vector double vec_ceil (vector double); 48849 vector bool long vec_cmpeq (vector double, vector double); 48850 vector bool long vec_cmpge (vector double, vector double); 48851 vector bool long vec_cmpgt (vector double, vector double); 48852 vector bool long vec_cmple (vector double, vector double); 48853 vector bool long vec_cmplt (vector double, vector double); 48854 vector double vec_cpsgn (vector double, vector double); 48855 vector float vec_div (vector float, vector float); 48856 vector double vec_div (vector double, vector double); 48857 vector long vec_div (vector long, vector long); 48858 vector unsigned long vec_div (vector unsigned long, vector unsigned long); 48859 vector double vec_floor (vector double); 48860 vector signed long long vec_ld (int, const vector signed long long *); 48861 vector signed long long vec_ld (int, const signed long long *); 48862 vector unsigned long long vec_ld (int, const vector unsigned long long *); 48863 vector unsigned long long vec_ld (int, const unsigned long long *); 48864 vector __int128 vec_ld (int, const vector __int128 *); 48865 vector unsigned __int128 vec_ld (int, const vector unsigned __int128 *); 48866 vector __int128 vec_ld (int, const __int128 *); 48867 vector unsigned __int128 vec_ld (int, const unsigned __int128 *); 48868 vector double vec_ld (int, const vector double *); 48869 vector double vec_ld (int, const double *); 48870 vector double vec_ldl (int, const vector double *); 48871 vector double vec_ldl (int, const double *); 48872 vector unsigned char vec_lvsl (int, const double *); 48873 vector unsigned char vec_lvsr (int, const double *); 48874 vector double vec_madd (vector double, vector double, vector double); 48875 vector double vec_max (vector double, vector double); 48876 vector signed long vec_mergeh (vector signed long, vector signed long); 48877 vector signed long vec_mergeh (vector signed long, vector bool long); 48878 vector signed long vec_mergeh (vector bool long, vector signed long); 48879 vector unsigned long vec_mergeh (vector unsigned long, vector unsigned long); 48880 vector unsigned long vec_mergeh (vector unsigned long, vector bool long); 48881 vector unsigned long vec_mergeh (vector bool long, vector unsigned long); 48882 vector signed long vec_mergel (vector signed long, vector signed long); 48883 vector signed long vec_mergel (vector signed long, vector bool long); 48884 vector signed long vec_mergel (vector bool long, vector signed long); 48885 vector unsigned long vec_mergel (vector unsigned long, vector unsigned long); 48886 vector unsigned long vec_mergel (vector unsigned long, vector bool long); 48887 vector unsigned long vec_mergel (vector bool long, vector unsigned long); 48888 vector double vec_min (vector double, vector double); 48889 vector float vec_msub (vector float, vector float, vector float); 48890 vector double vec_msub (vector double, vector double, vector double); 48891 vector float vec_nearbyint (vector float); 48892 vector double vec_nearbyint (vector double); 48893 vector float vec_nmadd (vector float, vector float, vector float); 48894 vector double vec_nmadd (vector double, vector double, vector double); 48895 vector double vec_nmsub (vector double, vector double, vector double); 48896 vector double vec_nor (vector double, vector double); 48897 vector long vec_nor (vector long, vector long); 48898 vector long vec_nor (vector long, vector bool long); 48899 vector long vec_nor (vector bool long, vector long); 48900 vector unsigned long vec_nor (vector unsigned long, vector unsigned long); 48901 vector unsigned long vec_nor (vector unsigned long, vector bool long); 48902 vector unsigned long vec_nor (vector bool long, vector unsigned long); 48903 vector double vec_or (vector double, vector double); 48904 vector double vec_or (vector double, vector bool long); 48905 vector double vec_or (vector bool long, vector double); 48906 vector long vec_or (vector long, vector long); 48907 vector long vec_or (vector long, vector bool long); 48908 vector long vec_or (vector bool long, vector long); 48909 vector unsigned long vec_or (vector unsigned long, vector unsigned long); 48910 vector unsigned long vec_or (vector unsigned long, vector bool long); 48911 vector unsigned long vec_or (vector bool long, vector unsigned long); 48912 vector double vec_perm (vector double, vector double, vector unsigned char); 48913 vector long vec_perm (vector long, vector long, vector unsigned char); 48914 vector unsigned long vec_perm (vector unsigned long, vector unsigned long, 48915 vector unsigned char); 48916 vector bool char vec_permxor (vector bool char, vector bool char, 48917 vector bool char); 48918 vector unsigned char vec_permxor (vector signed char, vector signed char, 48919 vector signed char); 48920 vector unsigned char vec_permxor (vector unsigned char, vector unsigned char, 48921 vector unsigned char); 48922 vector double vec_rint (vector double); 48923 vector double vec_recip (vector double, vector double); 48924 vector double vec_rsqrt (vector double); 48925 vector double vec_rsqrte (vector double); 48926 vector double vec_sel (vector double, vector double, vector bool long); 48927 vector double vec_sel (vector double, vector double, vector unsigned long); 48928 vector long vec_sel (vector long, vector long, vector long); 48929 vector long vec_sel (vector long, vector long, vector unsigned long); 48930 vector long vec_sel (vector long, vector long, vector bool long); 48931 vector unsigned long vec_sel (vector unsigned long, vector unsigned long, 48932 vector long); 48933 vector unsigned long vec_sel (vector unsigned long, vector unsigned long, 48934 vector unsigned long); 48935 vector unsigned long vec_sel (vector unsigned long, vector unsigned long, 48936 vector bool long); 48937 vector double vec_splats (double); 48938 vector signed long vec_splats (signed long); 48939 vector unsigned long vec_splats (unsigned long); 48940 vector float vec_sqrt (vector float); 48941 vector double vec_sqrt (vector double); 48942 void vec_st (vector signed long long, int, vector signed long long *); 48943 void vec_st (vector signed long long, int, signed long long *); 48944 void vec_st (vector unsigned long long, int, vector unsigned long long *); 48945 void vec_st (vector unsigned long long, int, unsigned long long *); 48946 void vec_st (vector bool long long, int, vector bool long long *); 48947 void vec_st (vector bool long long, int, signed long long *); 48948 void vec_st (vector bool long long, int, unsigned long long *); 48949 void vec_st (vector double, int, vector double *); 48950 void vec_st (vector double, int, double *); 48951 vector double vec_sub (vector double, vector double); 48952 vector double vec_trunc (vector double); 48953 vector double vec_xl (int, vector double *); 48954 vector double vec_xl (int, double *); 48955 vector long long vec_xl (int, vector long long *); 48956 vector long long vec_xl (int, long long *); 48957 vector unsigned long long vec_xl (int, vector unsigned long long *); 48958 vector unsigned long long vec_xl (int, unsigned long long *); 48959 vector float vec_xl (int, vector float *); 48960 vector float vec_xl (int, float *); 48961 vector int vec_xl (int, vector int *); 48962 vector int vec_xl (int, int *); 48963 vector unsigned int vec_xl (int, vector unsigned int *); 48964 vector unsigned int vec_xl (int, unsigned int *); 48965 vector double vec_xor (vector double, vector double); 48966 vector double vec_xor (vector double, vector bool long); 48967 vector double vec_xor (vector bool long, vector double); 48968 vector long vec_xor (vector long, vector long); 48969 vector long vec_xor (vector long, vector bool long); 48970 vector long vec_xor (vector bool long, vector long); 48971 vector unsigned long vec_xor (vector unsigned long, vector unsigned long); 48972 vector unsigned long vec_xor (vector unsigned long, vector bool long); 48973 vector unsigned long vec_xor (vector bool long, vector unsigned long); 48974 void vec_xst (vector double, int, vector double *); 48975 void vec_xst (vector double, int, double *); 48976 void vec_xst (vector long long, int, vector long long *); 48977 void vec_xst (vector long long, int, long long *); 48978 void vec_xst (vector unsigned long long, int, vector unsigned long long *); 48979 void vec_xst (vector unsigned long long, int, unsigned long long *); 48980 void vec_xst (vector float, int, vector float *); 48981 void vec_xst (vector float, int, float *); 48982 void vec_xst (vector int, int, vector int *); 48983 void vec_xst (vector int, int, int *); 48984 void vec_xst (vector unsigned int, int, vector unsigned int *); 48985 void vec_xst (vector unsigned int, int, unsigned int *); 48986 int vec_all_eq (vector double, vector double); 48987 int vec_all_ge (vector double, vector double); 48988 int vec_all_gt (vector double, vector double); 48989 int vec_all_le (vector double, vector double); 48990 int vec_all_lt (vector double, vector double); 48991 int vec_all_nan (vector double); 48992 int vec_all_ne (vector double, vector double); 48993 int vec_all_nge (vector double, vector double); 48994 int vec_all_ngt (vector double, vector double); 48995 int vec_all_nle (vector double, vector double); 48996 int vec_all_nlt (vector double, vector double); 48997 int vec_all_numeric (vector double); 48998 int vec_any_eq (vector double, vector double); 48999 int vec_any_ge (vector double, vector double); 49000 int vec_any_gt (vector double, vector double); 49001 int vec_any_le (vector double, vector double); 49002 int vec_any_lt (vector double, vector double); 49003 int vec_any_nan (vector double); 49004 int vec_any_ne (vector double, vector double); 49005 int vec_any_nge (vector double, vector double); 49006 int vec_any_ngt (vector double, vector double); 49007 int vec_any_nle (vector double, vector double); 49008 int vec_any_nlt (vector double, vector double); 49009 int vec_any_numeric (vector double); 49010 49011 vector double vec_vsx_ld (int, const vector double *); 49012 vector double vec_vsx_ld (int, const double *); 49013 vector float vec_vsx_ld (int, const vector float *); 49014 vector float vec_vsx_ld (int, const float *); 49015 vector bool int vec_vsx_ld (int, const vector bool int *); 49016 vector signed int vec_vsx_ld (int, const vector signed int *); 49017 vector signed int vec_vsx_ld (int, const int *); 49018 vector signed int vec_vsx_ld (int, const long *); 49019 vector unsigned int vec_vsx_ld (int, const vector unsigned int *); 49020 vector unsigned int vec_vsx_ld (int, const unsigned int *); 49021 vector unsigned int vec_vsx_ld (int, const unsigned long *); 49022 vector bool short vec_vsx_ld (int, const vector bool short *); 49023 vector pixel vec_vsx_ld (int, const vector pixel *); 49024 vector signed short vec_vsx_ld (int, const vector signed short *); 49025 vector signed short vec_vsx_ld (int, const short *); 49026 vector unsigned short vec_vsx_ld (int, const vector unsigned short *); 49027 vector unsigned short vec_vsx_ld (int, const unsigned short *); 49028 vector bool char vec_vsx_ld (int, const vector bool char *); 49029 vector signed char vec_vsx_ld (int, const vector signed char *); 49030 vector signed char vec_vsx_ld (int, const signed char *); 49031 vector unsigned char vec_vsx_ld (int, const vector unsigned char *); 49032 vector unsigned char vec_vsx_ld (int, const unsigned char *); 49033 49034 void vec_vsx_st (vector double, int, vector double *); 49035 void vec_vsx_st (vector double, int, double *); 49036 void vec_vsx_st (vector float, int, vector float *); 49037 void vec_vsx_st (vector float, int, float *); 49038 void vec_vsx_st (vector signed int, int, vector signed int *); 49039 void vec_vsx_st (vector signed int, int, int *); 49040 void vec_vsx_st (vector unsigned int, int, vector unsigned int *); 49041 void vec_vsx_st (vector unsigned int, int, unsigned int *); 49042 void vec_vsx_st (vector bool int, int, vector bool int *); 49043 void vec_vsx_st (vector bool int, int, unsigned int *); 49044 void vec_vsx_st (vector bool int, int, int *); 49045 void vec_vsx_st (vector signed short, int, vector signed short *); 49046 void vec_vsx_st (vector signed short, int, short *); 49047 void vec_vsx_st (vector unsigned short, int, vector unsigned short *); 49048 void vec_vsx_st (vector unsigned short, int, unsigned short *); 49049 void vec_vsx_st (vector bool short, int, vector bool short *); 49050 void vec_vsx_st (vector bool short, int, unsigned short *); 49051 void vec_vsx_st (vector pixel, int, vector pixel *); 49052 void vec_vsx_st (vector pixel, int, unsigned short *); 49053 void vec_vsx_st (vector pixel, int, short *); 49054 void vec_vsx_st (vector bool short, int, short *); 49055 void vec_vsx_st (vector signed char, int, vector signed char *); 49056 void vec_vsx_st (vector signed char, int, signed char *); 49057 void vec_vsx_st (vector unsigned char, int, vector unsigned char *); 49058 void vec_vsx_st (vector unsigned char, int, unsigned char *); 49059 void vec_vsx_st (vector bool char, int, vector bool char *); 49060 void vec_vsx_st (vector bool char, int, unsigned char *); 49061 void vec_vsx_st (vector bool char, int, signed char *); 49062 49063 vector double vec_xxpermdi (vector double, vector double, const int); 49064 vector float vec_xxpermdi (vector float, vector float, const int); 49065 vector long long vec_xxpermdi (vector long long, vector long long, const int); 49066 vector unsigned long long vec_xxpermdi (vector unsigned long long, 49067 vector unsigned long long, const int); 49068 vector int vec_xxpermdi (vector int, vector int, const int); 49069 vector unsigned int vec_xxpermdi (vector unsigned int, 49070 vector unsigned int, const int); 49071 vector short vec_xxpermdi (vector short, vector short, const int); 49072 vector unsigned short vec_xxpermdi (vector unsigned short, 49073 vector unsigned short, const int); 49074 vector signed char vec_xxpermdi (vector signed char, vector signed char, 49075 const int); 49076 vector unsigned char vec_xxpermdi (vector unsigned char, 49077 vector unsigned char, const int); 49078 49079 vector double vec_xxsldi (vector double, vector double, int); 49080 vector float vec_xxsldi (vector float, vector float, int); 49081 vector long long vec_xxsldi (vector long long, vector long long, int); 49082 vector unsigned long long vec_xxsldi (vector unsigned long long, 49083 vector unsigned long long, int); 49084 vector int vec_xxsldi (vector int, vector int, int); 49085 vector unsigned int vec_xxsldi (vector unsigned int, vector unsigned int, int); 49086 vector short vec_xxsldi (vector short, vector short, int); 49087 vector unsigned short vec_xxsldi (vector unsigned short, 49088 vector unsigned short, int); 49089 vector signed char vec_xxsldi (vector signed char, vector signed char, int); 49090 vector unsigned char vec_xxsldi (vector unsigned char, 49091 vector unsigned char, int); 49092 49093 Note that the 'vec_ld' and 'vec_st' built-in functions always generate 49094the AltiVec 'LVX' and 'STVX' instructions even if the VSX instruction 49095set is available. The 'vec_vsx_ld' and 'vec_vsx_st' built-in functions 49096always generate the VSX 'LXVD2X', 'LXVW4X', 'STXVD2X', and 'STXVW4X' 49097instructions. 49098 49099 49100File: gcc.info, Node: PowerPC AltiVec Built-in Functions Available on ISA 2.07, Next: PowerPC AltiVec Built-in Functions Available on ISA 3.0, Prev: PowerPC AltiVec Built-in Functions Available on ISA 2.06, Up: PowerPC AltiVec/VSX Built-in Functions 49101 491026.60.23.3 PowerPC AltiVec Built-in Functions Available on ISA 2.07 49103.................................................................. 49104 49105If the ISA 2.07 additions to the vector/scalar (power8-vector) 49106instruction set are available, the following additional functions are 49107available for both 32-bit and 64-bit targets. For 64-bit targets, you 49108can use VECTOR LONG instead of VECTOR LONG LONG, VECTOR BOOL LONG 49109instead of VECTOR BOOL LONG LONG, and VECTOR UNSIGNED LONG instead of 49110VECTOR UNSIGNED LONG LONG. 49111 49112 vector signed char vec_neg (vector signed char); 49113 vector signed short vec_neg (vector signed short); 49114 vector signed int vec_neg (vector signed int); 49115 vector signed long long vec_neg (vector signed long long); 49116 vector float char vec_neg (vector float); 49117 vector double vec_neg (vector double); 49118 49119 vector signed int vec_signed2 (vector double, vector double); 49120 49121 vector signed int vec_unsigned2 (vector double, vector double); 49122 49123 vector long long vec_abs (vector long long); 49124 49125 vector long long vec_add (vector long long, vector long long); 49126 vector unsigned long long vec_add (vector unsigned long long, 49127 vector unsigned long long); 49128 49129 int vec_all_eq (vector long long, vector long long); 49130 int vec_all_eq (vector unsigned long long, vector unsigned long long); 49131 int vec_all_ge (vector long long, vector long long); 49132 int vec_all_ge (vector unsigned long long, vector unsigned long long); 49133 int vec_all_gt (vector long long, vector long long); 49134 int vec_all_gt (vector unsigned long long, vector unsigned long long); 49135 int vec_all_le (vector long long, vector long long); 49136 int vec_all_le (vector unsigned long long, vector unsigned long long); 49137 int vec_all_lt (vector long long, vector long long); 49138 int vec_all_lt (vector unsigned long long, vector unsigned long long); 49139 int vec_all_ne (vector long long, vector long long); 49140 int vec_all_ne (vector unsigned long long, vector unsigned long long); 49141 49142 int vec_any_eq (vector long long, vector long long); 49143 int vec_any_eq (vector unsigned long long, vector unsigned long long); 49144 int vec_any_ge (vector long long, vector long long); 49145 int vec_any_ge (vector unsigned long long, vector unsigned long long); 49146 int vec_any_gt (vector long long, vector long long); 49147 int vec_any_gt (vector unsigned long long, vector unsigned long long); 49148 int vec_any_le (vector long long, vector long long); 49149 int vec_any_le (vector unsigned long long, vector unsigned long long); 49150 int vec_any_lt (vector long long, vector long long); 49151 int vec_any_lt (vector unsigned long long, vector unsigned long long); 49152 int vec_any_ne (vector long long, vector long long); 49153 int vec_any_ne (vector unsigned long long, vector unsigned long long); 49154 49155 vector bool long long vec_cmpeq (vector bool long long, vector bool long long); 49156 49157 vector long long vec_eqv (vector long long, vector long long); 49158 vector long long vec_eqv (vector bool long long, vector long long); 49159 vector long long vec_eqv (vector long long, vector bool long long); 49160 vector unsigned long long vec_eqv (vector unsigned long long, vector unsigned long long); 49161 vector unsigned long long vec_eqv (vector bool long long, vector unsigned long long); 49162 vector unsigned long long vec_eqv (vector unsigned long long, 49163 vector bool long long); 49164 vector int vec_eqv (vector int, vector int); 49165 vector int vec_eqv (vector bool int, vector int); 49166 vector int vec_eqv (vector int, vector bool int); 49167 vector unsigned int vec_eqv (vector unsigned int, vector unsigned int); 49168 vector unsigned int vec_eqv (vector bool unsigned int, vector unsigned int); 49169 vector unsigned int vec_eqv (vector unsigned int, vector bool unsigned int); 49170 vector short vec_eqv (vector short, vector short); 49171 vector short vec_eqv (vector bool short, vector short); 49172 vector short vec_eqv (vector short, vector bool short); 49173 vector unsigned short vec_eqv (vector unsigned short, vector unsigned short); 49174 vector unsigned short vec_eqv (vector bool unsigned short, vector unsigned short); 49175 vector unsigned short vec_eqv (vector unsigned short, vector bool unsigned short); 49176 vector signed char vec_eqv (vector signed char, vector signed char); 49177 vector signed char vec_eqv (vector bool signed char, vector signed char); 49178 vector signed char vec_eqv (vector signed char, vector bool signed char); 49179 vector unsigned char vec_eqv (vector unsigned char, vector unsigned char); 49180 vector unsigned char vec_eqv (vector bool unsigned char, vector unsigned char); 49181 vector unsigned char vec_eqv (vector unsigned char, vector bool unsigned char); 49182 49183 vector long long vec_max (vector long long, vector long long); 49184 vector unsigned long long vec_max (vector unsigned long long, 49185 vector unsigned long long); 49186 49187 vector signed int vec_mergee (vector signed int, vector signed int); 49188 vector unsigned int vec_mergee (vector unsigned int, vector unsigned int); 49189 vector bool int vec_mergee (vector bool int, vector bool int); 49190 49191 vector signed int vec_mergeo (vector signed int, vector signed int); 49192 vector unsigned int vec_mergeo (vector unsigned int, vector unsigned int); 49193 vector bool int vec_mergeo (vector bool int, vector bool int); 49194 49195 vector long long vec_min (vector long long, vector long long); 49196 vector unsigned long long vec_min (vector unsigned long long, 49197 vector unsigned long long); 49198 49199 vector signed long long vec_nabs (vector signed long long); 49200 49201 vector long long vec_nand (vector long long, vector long long); 49202 vector long long vec_nand (vector bool long long, vector long long); 49203 vector long long vec_nand (vector long long, vector bool long long); 49204 vector unsigned long long vec_nand (vector unsigned long long, 49205 vector unsigned long long); 49206 vector unsigned long long vec_nand (vector bool long long, vector unsigned long long); 49207 vector unsigned long long vec_nand (vector unsigned long long, vector bool long long); 49208 vector int vec_nand (vector int, vector int); 49209 vector int vec_nand (vector bool int, vector int); 49210 vector int vec_nand (vector int, vector bool int); 49211 vector unsigned int vec_nand (vector unsigned int, vector unsigned int); 49212 vector unsigned int vec_nand (vector bool unsigned int, vector unsigned int); 49213 vector unsigned int vec_nand (vector unsigned int, vector bool unsigned int); 49214 vector short vec_nand (vector short, vector short); 49215 vector short vec_nand (vector bool short, vector short); 49216 vector short vec_nand (vector short, vector bool short); 49217 vector unsigned short vec_nand (vector unsigned short, vector unsigned short); 49218 vector unsigned short vec_nand (vector bool unsigned short, vector unsigned short); 49219 vector unsigned short vec_nand (vector unsigned short, vector bool unsigned short); 49220 vector signed char vec_nand (vector signed char, vector signed char); 49221 vector signed char vec_nand (vector bool signed char, vector signed char); 49222 vector signed char vec_nand (vector signed char, vector bool signed char); 49223 vector unsigned char vec_nand (vector unsigned char, vector unsigned char); 49224 vector unsigned char vec_nand (vector bool unsigned char, vector unsigned char); 49225 vector unsigned char vec_nand (vector unsigned char, vector bool unsigned char); 49226 49227 vector long long vec_orc (vector long long, vector long long); 49228 vector long long vec_orc (vector bool long long, vector long long); 49229 vector long long vec_orc (vector long long, vector bool long long); 49230 vector unsigned long long vec_orc (vector unsigned long long, 49231 vector unsigned long long); 49232 vector unsigned long long vec_orc (vector bool long long, vector unsigned long long); 49233 vector unsigned long long vec_orc (vector unsigned long long, vector bool long long); 49234 vector int vec_orc (vector int, vector int); 49235 vector int vec_orc (vector bool int, vector int); 49236 vector int vec_orc (vector int, vector bool int); 49237 vector unsigned int vec_orc (vector unsigned int, vector unsigned int); 49238 vector unsigned int vec_orc (vector bool unsigned int, vector unsigned int); 49239 vector unsigned int vec_orc (vector unsigned int, vector bool unsigned int); 49240 vector short vec_orc (vector short, vector short); 49241 vector short vec_orc (vector bool short, vector short); 49242 vector short vec_orc (vector short, vector bool short); 49243 vector unsigned short vec_orc (vector unsigned short, vector unsigned short); 49244 vector unsigned short vec_orc (vector bool unsigned short, vector unsigned short); 49245 vector unsigned short vec_orc (vector unsigned short, vector bool unsigned short); 49246 vector signed char vec_orc (vector signed char, vector signed char); 49247 vector signed char vec_orc (vector bool signed char, vector signed char); 49248 vector signed char vec_orc (vector signed char, vector bool signed char); 49249 vector unsigned char vec_orc (vector unsigned char, vector unsigned char); 49250 vector unsigned char vec_orc (vector bool unsigned char, vector unsigned char); 49251 vector unsigned char vec_orc (vector unsigned char, vector bool unsigned char); 49252 49253 vector int vec_pack (vector long long, vector long long); 49254 vector unsigned int vec_pack (vector unsigned long long, vector unsigned long long); 49255 vector bool int vec_pack (vector bool long long, vector bool long long); 49256 vector float vec_pack (vector double, vector double); 49257 49258 vector int vec_packs (vector long long, vector long long); 49259 vector unsigned int vec_packs (vector unsigned long long, vector unsigned long long); 49260 49261 vector unsigned char vec_packsu (vector signed short, vector signed short) 49262 vector unsigned char vec_packsu (vector unsigned short, vector unsigned short) 49263 vector unsigned short int vec_packsu (vector signed int, vector signed int); 49264 vector unsigned short int vec_packsu (vector unsigned int, vector unsigned int); 49265 vector unsigned int vec_packsu (vector long long, vector long long); 49266 vector unsigned int vec_packsu (vector unsigned long long, vector unsigned long long); 49267 vector unsigned int vec_packsu (vector signed long long, vector signed long long); 49268 49269 vector unsigned char vec_popcnt (vector signed char); 49270 vector unsigned char vec_popcnt (vector unsigned char); 49271 vector unsigned short vec_popcnt (vector signed short); 49272 vector unsigned short vec_popcnt (vector unsigned short); 49273 vector unsigned int vec_popcnt (vector signed int); 49274 vector unsigned int vec_popcnt (vector unsigned int); 49275 vector unsigned long long vec_popcnt (vector signed long long); 49276 vector unsigned long long vec_popcnt (vector unsigned long long); 49277 49278 vector long long vec_rl (vector long long, vector unsigned long long); 49279 vector long long vec_rl (vector unsigned long long, vector unsigned long long); 49280 49281 vector long long vec_sl (vector long long, vector unsigned long long); 49282 vector long long vec_sl (vector unsigned long long, vector unsigned long long); 49283 49284 vector long long vec_sr (vector long long, vector unsigned long long); 49285 vector unsigned long long char vec_sr (vector unsigned long long, 49286 vector unsigned long long); 49287 49288 vector long long vec_sra (vector long long, vector unsigned long long); 49289 vector unsigned long long vec_sra (vector unsigned long long, 49290 vector unsigned long long); 49291 49292 vector long long vec_sub (vector long long, vector long long); 49293 vector unsigned long long vec_sub (vector unsigned long long, 49294 vector unsigned long long); 49295 49296 vector long long vec_unpackh (vector int); 49297 vector unsigned long long vec_unpackh (vector unsigned int); 49298 49299 vector long long vec_unpackl (vector int); 49300 vector unsigned long long vec_unpackl (vector unsigned int); 49301 49302 vector long long vec_vaddudm (vector long long, vector long long); 49303 vector long long vec_vaddudm (vector bool long long, vector long long); 49304 vector long long vec_vaddudm (vector long long, vector bool long long); 49305 vector unsigned long long vec_vaddudm (vector unsigned long long, 49306 vector unsigned long long); 49307 vector unsigned long long vec_vaddudm (vector bool unsigned long long, 49308 vector unsigned long long); 49309 vector unsigned long long vec_vaddudm (vector unsigned long long, 49310 vector bool unsigned long long); 49311 49312 vector long long vec_vbpermq (vector signed char, vector signed char); 49313 vector long long vec_vbpermq (vector unsigned char, vector unsigned char); 49314 49315 vector unsigned char vec_bperm (vector unsigned char, vector unsigned char); 49316 vector unsigned char vec_bperm (vector unsigned long long, vector unsigned char); 49317 vector unsigned long long vec_bperm (vector unsigned __int128, vector unsigned char); 49318 49319 vector long long vec_cntlz (vector long long); 49320 vector unsigned long long vec_cntlz (vector unsigned long long); 49321 vector int vec_cntlz (vector int); 49322 vector unsigned int vec_cntlz (vector int); 49323 vector short vec_cntlz (vector short); 49324 vector unsigned short vec_cntlz (vector unsigned short); 49325 vector signed char vec_cntlz (vector signed char); 49326 vector unsigned char vec_cntlz (vector unsigned char); 49327 49328 vector long long vec_vclz (vector long long); 49329 vector unsigned long long vec_vclz (vector unsigned long long); 49330 vector int vec_vclz (vector int); 49331 vector unsigned int vec_vclz (vector int); 49332 vector short vec_vclz (vector short); 49333 vector unsigned short vec_vclz (vector unsigned short); 49334 vector signed char vec_vclz (vector signed char); 49335 vector unsigned char vec_vclz (vector unsigned char); 49336 49337 vector signed char vec_vclzb (vector signed char); 49338 vector unsigned char vec_vclzb (vector unsigned char); 49339 49340 vector long long vec_vclzd (vector long long); 49341 vector unsigned long long vec_vclzd (vector unsigned long long); 49342 49343 vector short vec_vclzh (vector short); 49344 vector unsigned short vec_vclzh (vector unsigned short); 49345 49346 vector int vec_vclzw (vector int); 49347 vector unsigned int vec_vclzw (vector int); 49348 49349 vector signed char vec_vgbbd (vector signed char); 49350 vector unsigned char vec_vgbbd (vector unsigned char); 49351 49352 vector long long vec_vmaxsd (vector long long, vector long long); 49353 49354 vector unsigned long long vec_vmaxud (vector unsigned long long, 49355 unsigned vector long long); 49356 49357 vector long long vec_vminsd (vector long long, vector long long); 49358 49359 vector unsigned long long vec_vminud (vector long long, vector long long); 49360 49361 vector int vec_vpksdss (vector long long, vector long long); 49362 vector unsigned int vec_vpksdss (vector long long, vector long long); 49363 49364 vector unsigned int vec_vpkudus (vector unsigned long long, 49365 vector unsigned long long); 49366 49367 vector int vec_vpkudum (vector long long, vector long long); 49368 vector unsigned int vec_vpkudum (vector unsigned long long, 49369 vector unsigned long long); 49370 vector bool int vec_vpkudum (vector bool long long, vector bool long long); 49371 49372 vector long long vec_vpopcnt (vector long long); 49373 vector unsigned long long vec_vpopcnt (vector unsigned long long); 49374 vector int vec_vpopcnt (vector int); 49375 vector unsigned int vec_vpopcnt (vector int); 49376 vector short vec_vpopcnt (vector short); 49377 vector unsigned short vec_vpopcnt (vector unsigned short); 49378 vector signed char vec_vpopcnt (vector signed char); 49379 vector unsigned char vec_vpopcnt (vector unsigned char); 49380 49381 vector signed char vec_vpopcntb (vector signed char); 49382 vector unsigned char vec_vpopcntb (vector unsigned char); 49383 49384 vector long long vec_vpopcntd (vector long long); 49385 vector unsigned long long vec_vpopcntd (vector unsigned long long); 49386 49387 vector short vec_vpopcnth (vector short); 49388 vector unsigned short vec_vpopcnth (vector unsigned short); 49389 49390 vector int vec_vpopcntw (vector int); 49391 vector unsigned int vec_vpopcntw (vector int); 49392 49393 vector long long vec_vrld (vector long long, vector unsigned long long); 49394 vector unsigned long long vec_vrld (vector unsigned long long, 49395 vector unsigned long long); 49396 49397 vector long long vec_vsld (vector long long, vector unsigned long long); 49398 vector long long vec_vsld (vector unsigned long long, 49399 vector unsigned long long); 49400 49401 vector long long vec_vsrad (vector long long, vector unsigned long long); 49402 vector unsigned long long vec_vsrad (vector unsigned long long, 49403 vector unsigned long long); 49404 49405 vector long long vec_vsrd (vector long long, vector unsigned long long); 49406 vector unsigned long long char vec_vsrd (vector unsigned long long, 49407 vector unsigned long long); 49408 49409 vector long long vec_vsubudm (vector long long, vector long long); 49410 vector long long vec_vsubudm (vector bool long long, vector long long); 49411 vector long long vec_vsubudm (vector long long, vector bool long long); 49412 vector unsigned long long vec_vsubudm (vector unsigned long long, 49413 vector unsigned long long); 49414 vector unsigned long long vec_vsubudm (vector bool long long, 49415 vector unsigned long long); 49416 vector unsigned long long vec_vsubudm (vector unsigned long long, 49417 vector bool long long); 49418 49419 vector long long vec_vupkhsw (vector int); 49420 vector unsigned long long vec_vupkhsw (vector unsigned int); 49421 49422 vector long long vec_vupklsw (vector int); 49423 vector unsigned long long vec_vupklsw (vector int); 49424 49425 If the ISA 2.07 additions to the vector/scalar (power8-vector) 49426instruction set are available, the following additional functions are 49427available for 64-bit targets. New vector types (VECTOR __INT128 and 49428VECTOR __UINT128) are available to hold the __INT128 and __UINT128 types 49429to use these builtins. 49430 49431 The normal vector extract, and set operations work on VECTOR __INT128 49432and VECTOR __UINT128 types, but the index value must be 0. 49433 49434 vector __int128 vec_vaddcuq (vector __int128, vector __int128); 49435 vector __uint128 vec_vaddcuq (vector __uint128, vector __uint128); 49436 49437 vector __int128 vec_vadduqm (vector __int128, vector __int128); 49438 vector __uint128 vec_vadduqm (vector __uint128, vector __uint128); 49439 49440 vector __int128 vec_vaddecuq (vector __int128, vector __int128, 49441 vector __int128); 49442 vector __uint128 vec_vaddecuq (vector __uint128, vector __uint128, 49443 vector __uint128); 49444 49445 vector __int128 vec_vaddeuqm (vector __int128, vector __int128, 49446 vector __int128); 49447 vector __uint128 vec_vaddeuqm (vector __uint128, vector __uint128, 49448 vector __uint128); 49449 49450 vector __int128 vec_vsubecuq (vector __int128, vector __int128, 49451 vector __int128); 49452 vector __uint128 vec_vsubecuq (vector __uint128, vector __uint128, 49453 vector __uint128); 49454 49455 vector __int128 vec_vsubeuqm (vector __int128, vector __int128, 49456 vector __int128); 49457 vector __uint128 vec_vsubeuqm (vector __uint128, vector __uint128, 49458 vector __uint128); 49459 49460 vector __int128 vec_vsubcuq (vector __int128, vector __int128); 49461 vector __uint128 vec_vsubcuq (vector __uint128, vector __uint128); 49462 49463 __int128 vec_vsubuqm (__int128, __int128); 49464 __uint128 vec_vsubuqm (__uint128, __uint128); 49465 49466 vector __int128 __builtin_bcdadd (vector __int128, vector __int128, const int); 49467 int __builtin_bcdadd_lt (vector __int128, vector __int128, const int); 49468 int __builtin_bcdadd_eq (vector __int128, vector __int128, const int); 49469 int __builtin_bcdadd_gt (vector __int128, vector __int128, const int); 49470 int __builtin_bcdadd_ov (vector __int128, vector __int128, const int); 49471 vector __int128 __builtin_bcdsub (vector __int128, vector __int128, const int); 49472 int __builtin_bcdsub_lt (vector __int128, vector __int128, const int); 49473 int __builtin_bcdsub_eq (vector __int128, vector __int128, const int); 49474 int __builtin_bcdsub_gt (vector __int128, vector __int128, const int); 49475 int __builtin_bcdsub_ov (vector __int128, vector __int128, const int); 49476 49477 49478File: gcc.info, Node: PowerPC AltiVec Built-in Functions Available on ISA 3.0, Prev: PowerPC AltiVec Built-in Functions Available on ISA 2.07, Up: PowerPC AltiVec/VSX Built-in Functions 49479 494806.60.23.4 PowerPC AltiVec Built-in Functions Available on ISA 3.0 49481................................................................. 49482 49483The following additional built-in functions are also available for the 49484PowerPC family of processors, starting with ISA 3.0 ('-mcpu=power9') or 49485later: 49486 unsigned int scalar_extract_exp (double source); 49487 unsigned long long int scalar_extract_exp (__ieee128 source); 49488 49489 unsigned long long int scalar_extract_sig (double source); 49490 unsigned __int128 scalar_extract_sig (__ieee128 source); 49491 49492 double scalar_insert_exp (unsigned long long int significand, 49493 unsigned long long int exponent); 49494 double scalar_insert_exp (double significand, unsigned long long int exponent); 49495 49496 ieee_128 scalar_insert_exp (unsigned __int128 significand, 49497 unsigned long long int exponent); 49498 ieee_128 scalar_insert_exp (ieee_128 significand, unsigned long long int exponent); 49499 49500 int scalar_cmp_exp_gt (double arg1, double arg2); 49501 int scalar_cmp_exp_lt (double arg1, double arg2); 49502 int scalar_cmp_exp_eq (double arg1, double arg2); 49503 int scalar_cmp_exp_unordered (double arg1, double arg2); 49504 49505 bool scalar_test_data_class (float source, const int condition); 49506 bool scalar_test_data_class (double source, const int condition); 49507 bool scalar_test_data_class (__ieee128 source, const int condition); 49508 49509 bool scalar_test_neg (float source); 49510 bool scalar_test_neg (double source); 49511 bool scalar_test_neg (__ieee128 source); 49512 49513 vector _uint128_t vec_msum (vector unsigned long long, 49514 vector unsigned long long, 49515 vector _uint128_t); 49516 vector _int128_t vec_msum (vector signed long long, 49517 vector signed long long, 49518 vector _int128_t); 49519 49520 The 'scalar_extract_exp' and 'scalar_extract_sig' functions require a 4952164-bit environment supporting ISA 3.0 or later. The 49522'scalar_extract_exp' and 'scalar_extract_sig' built-in functions return 49523the significand and the biased exponent value respectively of their 49524'source' arguments. When supplied with a 64-bit 'source' argument, the 49525result returned by 'scalar_extract_sig' has the '0x0010000000000000' bit 49526set if the function's 'source' argument is in normalized form. 49527Otherwise, this bit is set to 0. When supplied with a 128-bit 'source' 49528argument, the '0x00010000000000000000000000000000' bit of the result is 49529treated similarly. Note that the sign of the significand is not 49530represented in the result returned from the 'scalar_extract_sig' 49531function. Use the 'scalar_test_neg' function to test the sign of its 49532'double' argument. The 'vec_msum' functions perform a vector 49533multiply-sum, returning the result of arg1*arg2+arg3. ISA 3.0 adds 49534support for vec_msum returning a vector int128 result. 49535 49536 The 'scalar_insert_exp' functions require a 64-bit environment 49537supporting ISA 3.0 or later. When supplied with a 64-bit first 49538argument, the 'scalar_insert_exp' built-in function returns a 49539double-precision floating point value that is constructed by assembling 49540the values of its 'significand' and 'exponent' arguments. The sign of 49541the result is copied from the most significant bit of the 'significand' 49542argument. The significand and exponent components of the result are 49543composed of the least significant 11 bits of the 'exponent' argument and 49544the least significant 52 bits of the 'significand' argument 49545respectively. 49546 49547 When supplied with a 128-bit first argument, the 'scalar_insert_exp' 49548built-in function returns a quad-precision ieee floating point value. 49549The sign bit of the result is copied from the most significant bit of 49550the 'significand' argument. The significand and exponent components of 49551the result are composed of the least significant 15 bits of the 49552'exponent' argument and the least significant 112 bits of the 49553'significand' argument respectively. 49554 49555 The 'scalar_cmp_exp_gt', 'scalar_cmp_exp_lt', 'scalar_cmp_exp_eq', and 49556'scalar_cmp_exp_unordered' built-in functions return a non-zero value if 49557'arg1' is greater than, less than, equal to, or not comparable to 'arg2' 49558respectively. The arguments are not comparable if one or the other 49559equals NaN (not a number). 49560 49561 The 'scalar_test_data_class' built-in function returns 1 if any of the 49562condition tests enabled by the value of the 'condition' variable are 49563true, and 0 otherwise. The 'condition' argument must be a compile-time 49564constant integer with value not exceeding 127. The 'condition' argument 49565is encoded as a bitmask with each bit enabling the testing of a 49566different condition, as characterized by the following: 49567 0x40 Test for NaN 49568 0x20 Test for +Infinity 49569 0x10 Test for -Infinity 49570 0x08 Test for +Zero 49571 0x04 Test for -Zero 49572 0x02 Test for +Denormal 49573 0x01 Test for -Denormal 49574 49575 The 'scalar_test_neg' built-in function returns 1 if its 'source' 49576argument holds a negative value, 0 otherwise. 49577 49578 The following built-in functions are also available for the PowerPC 49579family of processors, starting with ISA 3.0 or later ('-mcpu=power9'). 49580These string functions are described separately in order to group the 49581descriptions closer to the function prototypes: 49582 int vec_all_nez (vector signed char, vector signed char); 49583 int vec_all_nez (vector unsigned char, vector unsigned char); 49584 int vec_all_nez (vector signed short, vector signed short); 49585 int vec_all_nez (vector unsigned short, vector unsigned short); 49586 int vec_all_nez (vector signed int, vector signed int); 49587 int vec_all_nez (vector unsigned int, vector unsigned int); 49588 49589 int vec_any_eqz (vector signed char, vector signed char); 49590 int vec_any_eqz (vector unsigned char, vector unsigned char); 49591 int vec_any_eqz (vector signed short, vector signed short); 49592 int vec_any_eqz (vector unsigned short, vector unsigned short); 49593 int vec_any_eqz (vector signed int, vector signed int); 49594 int vec_any_eqz (vector unsigned int, vector unsigned int); 49595 49596 vector bool char vec_cmpnez (vector signed char arg1, vector signed char arg2); 49597 vector bool char vec_cmpnez (vector unsigned char arg1, vector unsigned char arg2); 49598 vector bool short vec_cmpnez (vector signed short arg1, vector signed short arg2); 49599 vector bool short vec_cmpnez (vector unsigned short arg1, vector unsigned short arg2); 49600 vector bool int vec_cmpnez (vector signed int arg1, vector signed int arg2); 49601 vector bool int vec_cmpnez (vector unsigned int, vector unsigned int); 49602 49603 vector signed char vec_cnttz (vector signed char); 49604 vector unsigned char vec_cnttz (vector unsigned char); 49605 vector signed short vec_cnttz (vector signed short); 49606 vector unsigned short vec_cnttz (vector unsigned short); 49607 vector signed int vec_cnttz (vector signed int); 49608 vector unsigned int vec_cnttz (vector unsigned int); 49609 vector signed long long vec_cnttz (vector signed long long); 49610 vector unsigned long long vec_cnttz (vector unsigned long long); 49611 49612 signed int vec_cntlz_lsbb (vector signed char); 49613 signed int vec_cntlz_lsbb (vector unsigned char); 49614 49615 signed int vec_cnttz_lsbb (vector signed char); 49616 signed int vec_cnttz_lsbb (vector unsigned char); 49617 49618 unsigned int vec_first_match_index (vector signed char, vector signed char); 49619 unsigned int vec_first_match_index (vector unsigned char, vector unsigned char); 49620 unsigned int vec_first_match_index (vector signed int, vector signed int); 49621 unsigned int vec_first_match_index (vector unsigned int, vector unsigned int); 49622 unsigned int vec_first_match_index (vector signed short, vector signed short); 49623 unsigned int vec_first_match_index (vector unsigned short, vector unsigned short); 49624 unsigned int vec_first_match_or_eos_index (vector signed char, vector signed char); 49625 unsigned int vec_first_match_or_eos_index (vector unsigned char, vector unsigned char); 49626 unsigned int vec_first_match_or_eos_index (vector signed int, vector signed int); 49627 unsigned int vec_first_match_or_eos_index (vector unsigned int, vector unsigned int); 49628 unsigned int vec_first_match_or_eos_index (vector signed short, vector signed short); 49629 unsigned int vec_first_match_or_eos_index (vector unsigned short, 49630 vector unsigned short); 49631 unsigned int vec_first_mismatch_index (vector signed char, vector signed char); 49632 unsigned int vec_first_mismatch_index (vector unsigned char, vector unsigned char); 49633 unsigned int vec_first_mismatch_index (vector signed int, vector signed int); 49634 unsigned int vec_first_mismatch_index (vector unsigned int, vector unsigned int); 49635 unsigned int vec_first_mismatch_index (vector signed short, vector signed short); 49636 unsigned int vec_first_mismatch_index (vector unsigned short, vector unsigned short); 49637 unsigned int vec_first_mismatch_or_eos_index (vector signed char, vector signed char); 49638 unsigned int vec_first_mismatch_or_eos_index (vector unsigned char, 49639 vector unsigned char); 49640 unsigned int vec_first_mismatch_or_eos_index (vector signed int, vector signed int); 49641 unsigned int vec_first_mismatch_or_eos_index (vector unsigned int, vector unsigned int); 49642 unsigned int vec_first_mismatch_or_eos_index (vector signed short, vector signed short); 49643 unsigned int vec_first_mismatch_or_eos_index (vector unsigned short, 49644 vector unsigned short); 49645 49646 vector unsigned short vec_pack_to_short_fp32 (vector float, vector float); 49647 49648 vector signed char vec_xl_be (signed long long, signed char *); 49649 vector unsigned char vec_xl_be (signed long long, unsigned char *); 49650 vector signed int vec_xl_be (signed long long, signed int *); 49651 vector unsigned int vec_xl_be (signed long long, unsigned int *); 49652 vector signed __int128 vec_xl_be (signed long long, signed __int128 *); 49653 vector unsigned __int128 vec_xl_be (signed long long, unsigned __int128 *); 49654 vector signed long long vec_xl_be (signed long long, signed long long *); 49655 vector unsigned long long vec_xl_be (signed long long, unsigned long long *); 49656 vector signed short vec_xl_be (signed long long, signed short *); 49657 vector unsigned short vec_xl_be (signed long long, unsigned short *); 49658 vector double vec_xl_be (signed long long, double *); 49659 vector float vec_xl_be (signed long long, float *); 49660 49661 vector signed char vec_xl_len (signed char *addr, size_t len); 49662 vector unsigned char vec_xl_len (unsigned char *addr, size_t len); 49663 vector signed int vec_xl_len (signed int *addr, size_t len); 49664 vector unsigned int vec_xl_len (unsigned int *addr, size_t len); 49665 vector signed __int128 vec_xl_len (signed __int128 *addr, size_t len); 49666 vector unsigned __int128 vec_xl_len (unsigned __int128 *addr, size_t len); 49667 vector signed long long vec_xl_len (signed long long *addr, size_t len); 49668 vector unsigned long long vec_xl_len (unsigned long long *addr, size_t len); 49669 vector signed short vec_xl_len (signed short *addr, size_t len); 49670 vector unsigned short vec_xl_len (unsigned short *addr, size_t len); 49671 vector double vec_xl_len (double *addr, size_t len); 49672 vector float vec_xl_len (float *addr, size_t len); 49673 49674 vector unsigned char vec_xl_len_r (unsigned char *addr, size_t len); 49675 49676 void vec_xst_len (vector signed char data, signed char *addr, size_t len); 49677 void vec_xst_len (vector unsigned char data, unsigned char *addr, size_t len); 49678 void vec_xst_len (vector signed int data, signed int *addr, size_t len); 49679 void vec_xst_len (vector unsigned int data, unsigned int *addr, size_t len); 49680 void vec_xst_len (vector unsigned __int128 data, unsigned __int128 *addr, size_t len); 49681 void vec_xst_len (vector signed long long data, signed long long *addr, size_t len); 49682 void vec_xst_len (vector unsigned long long data, unsigned long long *addr, size_t len); 49683 void vec_xst_len (vector signed short data, signed short *addr, size_t len); 49684 void vec_xst_len (vector unsigned short data, unsigned short *addr, size_t len); 49685 void vec_xst_len (vector signed __int128 data, signed __int128 *addr, size_t len); 49686 void vec_xst_len (vector double data, double *addr, size_t len); 49687 void vec_xst_len (vector float data, float *addr, size_t len); 49688 49689 void vec_xst_len_r (vector unsigned char data, unsigned char *addr, size_t len); 49690 49691 signed char vec_xlx (unsigned int index, vector signed char data); 49692 unsigned char vec_xlx (unsigned int index, vector unsigned char data); 49693 signed short vec_xlx (unsigned int index, vector signed short data); 49694 unsigned short vec_xlx (unsigned int index, vector unsigned short data); 49695 signed int vec_xlx (unsigned int index, vector signed int data); 49696 unsigned int vec_xlx (unsigned int index, vector unsigned int data); 49697 float vec_xlx (unsigned int index, vector float data); 49698 49699 signed char vec_xrx (unsigned int index, vector signed char data); 49700 unsigned char vec_xrx (unsigned int index, vector unsigned char data); 49701 signed short vec_xrx (unsigned int index, vector signed short data); 49702 unsigned short vec_xrx (unsigned int index, vector unsigned short data); 49703 signed int vec_xrx (unsigned int index, vector signed int data); 49704 unsigned int vec_xrx (unsigned int index, vector unsigned int data); 49705 float vec_xrx (unsigned int index, vector float data); 49706 49707 The 'vec_all_nez', 'vec_any_eqz', and 'vec_cmpnez' perform pairwise 49708comparisons between the elements at the same positions within their two 49709vector arguments. The 'vec_all_nez' function returns a non-zero value 49710if and only if all pairwise comparisons are not equal and no element of 49711either vector argument contains a zero. The 'vec_any_eqz' function 49712returns a non-zero value if and only if at least one pairwise comparison 49713is equal or if at least one element of either vector argument contains a 49714zero. The 'vec_cmpnez' function returns a vector of the same type as 49715its two arguments, within which each element consists of all ones to 49716denote that either the corresponding elements of the incoming arguments 49717are not equal or that at least one of the corresponding elements 49718contains zero. Otherwise, the element of the returned vector contains 49719all zeros. 49720 49721 The 'vec_cntlz_lsbb' function returns the count of the number of 49722consecutive leading byte elements (starting from position 0 within the 49723supplied vector argument) for which the least-significant bit equals 49724zero. The 'vec_cnttz_lsbb' function returns the count of the number of 49725consecutive trailing byte elements (starting from position 15 and 49726counting backwards within the supplied vector argument) for which the 49727least-significant bit equals zero. 49728 49729 The 'vec_xl_len' and 'vec_xst_len' functions require a 64-bit 49730environment supporting ISA 3.0 or later. The 'vec_xl_len' function 49731loads a variable length vector from memory. The 'vec_xst_len' function 49732stores a variable length vector to memory. With both the 'vec_xl_len' 49733and 'vec_xst_len' functions, the 'addr' argument represents the memory 49734address to or from which data will be transferred, and the 'len' 49735argument represents the number of bytes to be transferred, as computed 49736by the C expression 'min((len & 0xff), 16)'. If this expression's value 49737is not a multiple of the vector element's size, the behavior of this 49738function is undefined. In the case that the underlying computer is 49739configured to run in big-endian mode, the data transfer moves bytes 0 to 49740'(len - 1)' of the corresponding vector. In little-endian mode, the 49741data transfer moves bytes '(16 - len)' to '15' of the corresponding 49742vector. For the load function, any bytes of the result vector that are 49743not loaded from memory are set to zero. The value of the 'addr' 49744argument need not be aligned on a multiple of the vector's element size. 49745 49746 The 'vec_xlx' and 'vec_xrx' functions extract the single element 49747selected by the 'index' argument from the vector represented by the 49748'data' argument. The 'index' argument always specifies a byte offset, 49749regardless of the size of the vector element. With 'vec_xlx', 'index' 49750is the offset of the first byte of the element to be extracted. With 49751'vec_xrx', 'index' represents the last byte of the element to be 49752extracted, measured from the right end of the vector. In other words, 49753the last byte of the element to be extracted is found at position '(15 - 49754index)'. There is no requirement that 'index' be a multiple of the 49755vector element size. However, if the size of the vector element added 49756to 'index' is greater than 15, the content of the returned value is 49757undefined. 49758 49759 If the ISA 3.0 instruction set additions ('-mcpu=power9') are 49760available: 49761 49762 vector unsigned long long vec_bperm (vector unsigned long long, vector unsigned char); 49763 49764 vector bool char vec_cmpne (vector bool char, vector bool char); 49765 vector bool char vec_cmpne (vector signed char, vector signed char); 49766 vector bool char vec_cmpne (vector unsigned char, vector unsigned char); 49767 vector bool int vec_cmpne (vector bool int, vector bool int); 49768 vector bool int vec_cmpne (vector signed int, vector signed int); 49769 vector bool int vec_cmpne (vector unsigned int, vector unsigned int); 49770 vector bool long long vec_cmpne (vector bool long long, vector bool long long); 49771 vector bool long long vec_cmpne (vector signed long long, vector signed long long); 49772 vector bool long long vec_cmpne (vector unsigned long long, vector unsigned long long); 49773 vector bool short vec_cmpne (vector bool short, vector bool short); 49774 vector bool short vec_cmpne (vector signed short, vector signed short); 49775 vector bool short vec_cmpne (vector unsigned short, vector unsigned short); 49776 vector bool long long vec_cmpne (vector double, vector double); 49777 vector bool int vec_cmpne (vector float, vector float); 49778 49779 vector float vec_extract_fp32_from_shorth (vector unsigned short); 49780 vector float vec_extract_fp32_from_shortl (vector unsigned short); 49781 49782 vector long long vec_vctz (vector long long); 49783 vector unsigned long long vec_vctz (vector unsigned long long); 49784 vector int vec_vctz (vector int); 49785 vector unsigned int vec_vctz (vector int); 49786 vector short vec_vctz (vector short); 49787 vector unsigned short vec_vctz (vector unsigned short); 49788 vector signed char vec_vctz (vector signed char); 49789 vector unsigned char vec_vctz (vector unsigned char); 49790 49791 vector signed char vec_vctzb (vector signed char); 49792 vector unsigned char vec_vctzb (vector unsigned char); 49793 49794 vector long long vec_vctzd (vector long long); 49795 vector unsigned long long vec_vctzd (vector unsigned long long); 49796 49797 vector short vec_vctzh (vector short); 49798 vector unsigned short vec_vctzh (vector unsigned short); 49799 49800 vector int vec_vctzw (vector int); 49801 vector unsigned int vec_vctzw (vector int); 49802 49803 vector unsigned long long vec_extract4b (vector unsigned char, const int); 49804 49805 vector unsigned char vec_insert4b (vector signed int, vector unsigned char, 49806 const int); 49807 vector unsigned char vec_insert4b (vector unsigned int, vector unsigned char, 49808 const int); 49809 49810 vector unsigned int vec_parity_lsbb (vector signed int); 49811 vector unsigned int vec_parity_lsbb (vector unsigned int); 49812 vector unsigned __int128 vec_parity_lsbb (vector signed __int128); 49813 vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128); 49814 vector unsigned long long vec_parity_lsbb (vector signed long long); 49815 vector unsigned long long vec_parity_lsbb (vector unsigned long long); 49816 49817 vector int vec_vprtyb (vector int); 49818 vector unsigned int vec_vprtyb (vector unsigned int); 49819 vector long long vec_vprtyb (vector long long); 49820 vector unsigned long long vec_vprtyb (vector unsigned long long); 49821 49822 vector int vec_vprtybw (vector int); 49823 vector unsigned int vec_vprtybw (vector unsigned int); 49824 49825 vector long long vec_vprtybd (vector long long); 49826 vector unsigned long long vec_vprtybd (vector unsigned long long); 49827 49828 On 64-bit targets, if the ISA 3.0 additions ('-mcpu=power9') are 49829available: 49830 49831 vector long vec_vprtyb (vector long); 49832 vector unsigned long vec_vprtyb (vector unsigned long); 49833 vector __int128 vec_vprtyb (vector __int128); 49834 vector __uint128 vec_vprtyb (vector __uint128); 49835 49836 vector long vec_vprtybd (vector long); 49837 vector unsigned long vec_vprtybd (vector unsigned long); 49838 49839 vector __int128 vec_vprtybq (vector __int128); 49840 vector __uint128 vec_vprtybd (vector __uint128); 49841 49842 The following built-in vector functions are available for the PowerPC 49843family of processors, starting with ISA 3.0 or later ('-mcpu=power9'): 49844 __vector unsigned char 49845 vec_slv (__vector unsigned char src, __vector unsigned char shift_distance); 49846 __vector unsigned char 49847 vec_srv (__vector unsigned char src, __vector unsigned char shift_distance); 49848 49849 The 'vec_slv' and 'vec_srv' functions operate on all of the bytes of 49850their 'src' and 'shift_distance' arguments in parallel. The behavior of 49851the 'vec_slv' is as if there existed a temporary array of 17 unsigned 49852characters 'slv_array' within which elements 0 through 15 are the same 49853as the entries in the 'src' array and element 16 equals 0. The result 49854returned from the 'vec_slv' function is a '__vector' of 16 unsigned 49855characters within which element 'i' is computed using the C expression 49856'0xff & (*((unsigned short *)(slv_array + i)) << (0x07 & 49857shift_distance[i]))', with this resulting value coerced to the 'unsigned 49858char' type. The behavior of the 'vec_srv' is as if there existed a 49859temporary array of 17 unsigned characters 'srv_array' within which 49860element 0 equals zero and elements 1 through 16 equal the elements 0 49861through 15 of the 'src' array. The result returned from the 'vec_srv' 49862function is a '__vector' of 16 unsigned characters within which element 49863'i' is computed using the C expression '0xff & (*((unsigned short 49864*)(srv_array + i)) >> (0x07 & shift_distance[i]))', with this resulting 49865value coerced to the 'unsigned char' type. 49866 49867 The following built-in functions are available for the PowerPC family 49868of processors, starting with ISA 3.0 or later ('-mcpu=power9'): 49869 __vector unsigned char 49870 vec_absd (__vector unsigned char arg1, __vector unsigned char arg2); 49871 __vector unsigned short 49872 vec_absd (__vector unsigned short arg1, __vector unsigned short arg2); 49873 __vector unsigned int 49874 vec_absd (__vector unsigned int arg1, __vector unsigned int arg2); 49875 49876 __vector unsigned char 49877 vec_absdb (__vector unsigned char arg1, __vector unsigned char arg2); 49878 __vector unsigned short 49879 vec_absdh (__vector unsigned short arg1, __vector unsigned short arg2); 49880 __vector unsigned int 49881 vec_absdw (__vector unsigned int arg1, __vector unsigned int arg2); 49882 49883 The 'vec_absd', 'vec_absdb', 'vec_absdh', and 'vec_absdw' built-in 49884functions each computes the absolute differences of the pairs of vector 49885elements supplied in its two vector arguments, placing the absolute 49886differences into the corresponding elements of the vector result. 49887 49888 The following built-in functions are available for the PowerPC family 49889of processors, starting with ISA 3.0 or later ('-mcpu=power9'): 49890 __vector unsigned int vec_extract_exp (__vector float source); 49891 __vector unsigned long long int vec_extract_exp (__vector double source); 49892 49893 __vector unsigned int vec_extract_sig (__vector float source); 49894 __vector unsigned long long int vec_extract_sig (__vector double source); 49895 49896 __vector float vec_insert_exp (__vector unsigned int significands, 49897 __vector unsigned int exponents); 49898 __vector float vec_insert_exp (__vector unsigned float significands, 49899 __vector unsigned int exponents); 49900 __vector double vec_insert_exp (__vector unsigned long long int significands, 49901 __vector unsigned long long int exponents); 49902 __vector double vec_insert_exp (__vector unsigned double significands, 49903 __vector unsigned long long int exponents); 49904 49905 __vector bool int vec_test_data_class (__vector float source, const int condition); 49906 __vector bool long long int vec_test_data_class (__vector double source, 49907 const int condition); 49908 49909 The 'vec_extract_sig' and 'vec_extract_exp' built-in functions return 49910vectors representing the significands and biased exponent values of 49911their 'source' arguments respectively. Within the result vector 49912returned by 'vec_extract_sig', the '0x800000' bit of each vector element 49913returned when the function's 'source' argument is of type 'float' is set 49914to 1 if the corresponding floating point value is in normalized form. 49915Otherwise, this bit is set to 0. When the 'source' argument is of type 49916'double', the '0x10000000000000' bit within each of the result vector's 49917elements is set according to the same rules. Note that the sign of the 49918significand is not represented in the result returned from the 49919'vec_extract_sig' function. To extract the sign bits, use the 49920'vec_cpsgn' function, which returns a new vector within which all of the 49921sign bits of its second argument vector are overwritten with the sign 49922bits copied from the coresponding elements of its first argument vector, 49923and all other (non-sign) bits of the second argument vector are copied 49924unchanged into the result vector. 49925 49926 The 'vec_insert_exp' built-in functions return a vector of single- or 49927double-precision floating point values constructed by assembling the 49928values of their 'significands' and 'exponents' arguments into the 49929corresponding elements of the returned vector. The sign of each element 49930of the result is copied from the most significant bit of the 49931corresponding entry within the 'significands' argument. Note that the 49932relevant bits of the 'significands' argument are the same, for both 49933integer and floating point types. The significand and exponent 49934components of each element of the result are composed of the least 49935significant bits of the corresponding 'significands' element and the 49936least significant bits of the corresponding 'exponents' element. 49937 49938 The 'vec_test_data_class' built-in function returns a vector 49939representing the results of testing the 'source' vector for the 49940condition selected by the 'condition' argument. The 'condition' 49941argument must be a compile-time constant integer with value not 49942exceeding 127. The 'condition' argument is encoded as a bitmask with 49943each bit enabling the testing of a different condition, as characterized 49944by the following: 49945 0x40 Test for NaN 49946 0x20 Test for +Infinity 49947 0x10 Test for -Infinity 49948 0x08 Test for +Zero 49949 0x04 Test for -Zero 49950 0x02 Test for +Denormal 49951 0x01 Test for -Denormal 49952 49953 If any of the enabled test conditions is true, the corresponding entry 49954in the result vector is -1. Otherwise (all of the enabled test 49955conditions are false), the corresponding entry of the result vector is 499560. 49957 49958 The following built-in functions are available for the PowerPC family 49959of processors, starting with ISA 3.0 or later ('-mcpu=power9'): 49960 vector unsigned int vec_rlmi (vector unsigned int, vector unsigned int, 49961 vector unsigned int); 49962 vector unsigned long long vec_rlmi (vector unsigned long long, 49963 vector unsigned long long, 49964 vector unsigned long long); 49965 vector unsigned int vec_rlnm (vector unsigned int, vector unsigned int, 49966 vector unsigned int); 49967 vector unsigned long long vec_rlnm (vector unsigned long long, 49968 vector unsigned long long, 49969 vector unsigned long long); 49970 vector unsigned int vec_vrlnm (vector unsigned int, vector unsigned int); 49971 vector unsigned long long vec_vrlnm (vector unsigned long long, 49972 vector unsigned long long); 49973 49974 The result of 'vec_rlmi' is obtained by rotating each element of the 49975first argument vector left and inserting it under mask into the second 49976argument vector. The third argument vector contains the mask beginning 49977in bits 11:15, the mask end in bits 19:23, and the shift count in bits 4997827:31, of each element. 49979 49980 The result of 'vec_rlnm' is obtained by rotating each element of the 49981first argument vector left and ANDing it with a mask specified by the 49982second and third argument vectors. The second argument vector contains 49983the shift count for each element in the low-order byte. The third 49984argument vector contains the mask end for each element in the low-order 49985byte, with the mask begin in the next higher byte. 49986 49987 The result of 'vec_vrlnm' is obtained by rotating each element of the 49988first argument vector left and ANDing it with a mask. The second 49989argument vector contains the mask beginning in bits 11:15, the mask end 49990in bits 19:23, and the shift count in bits 27:31, of each element. 49991 49992 If the ISA 3.0 instruction set additions ('-mcpu=power9') are 49993available: 49994 vector signed bool char vec_revb (vector signed char); 49995 vector signed char vec_revb (vector signed char); 49996 vector unsigned char vec_revb (vector unsigned char); 49997 vector bool short vec_revb (vector bool short); 49998 vector short vec_revb (vector short); 49999 vector unsigned short vec_revb (vector unsigned short); 50000 vector bool int vec_revb (vector bool int); 50001 vector int vec_revb (vector int); 50002 vector unsigned int vec_revb (vector unsigned int); 50003 vector float vec_revb (vector float); 50004 vector bool long long vec_revb (vector bool long long); 50005 vector long long vec_revb (vector long long); 50006 vector unsigned long long vec_revb (vector unsigned long long); 50007 vector double vec_revb (vector double); 50008 50009 On 64-bit targets, if the ISA 3.0 additions ('-mcpu=power9') are 50010available: 50011 vector long vec_revb (vector long); 50012 vector unsigned long vec_revb (vector unsigned long); 50013 vector __int128 vec_revb (vector __int128); 50014 vector __uint128 vec_revb (vector __uint128); 50015 50016 The 'vec_revb' built-in function reverses the bytes on an element by 50017element basis. A vector of 'vector unsigned char' or 'vector signed 50018char' reverses the bytes in the whole word. 50019 50020 If the cryptographic instructions are enabled ('-mcrypto' or 50021'-mcpu=power8'), the following builtins are enabled. 50022 50023 vector unsigned long long __builtin_crypto_vsbox (vector unsigned long long); 50024 50025 vector unsigned char vec_sbox_be (vector unsigned char); 50026 50027 vector unsigned long long __builtin_crypto_vcipher (vector unsigned long long, 50028 vector unsigned long long); 50029 50030 vector unsigned char vec_cipher_be (vector unsigned char, vector unsigned char); 50031 50032 vector unsigned long long __builtin_crypto_vcipherlast 50033 (vector unsigned long long, 50034 vector unsigned long long); 50035 50036 vector unsigned char vec_cipherlast_be (vector unsigned char, 50037 vector unsigned char); 50038 50039 vector unsigned long long __builtin_crypto_vncipher (vector unsigned long long, 50040 vector unsigned long long); 50041 50042 vector unsigned char vec_ncipher_be (vector unsigned char, 50043 vector unsigned char); 50044 50045 vector unsigned long long __builtin_crypto_vncipherlast (vector unsigned long long, 50046 vector unsigned long long); 50047 50048 vector unsigned char vec_ncipherlast_be (vector unsigned char, 50049 vector unsigned char); 50050 50051 vector unsigned char __builtin_crypto_vpermxor (vector unsigned char, 50052 vector unsigned char, 50053 vector unsigned char); 50054 50055 vector unsigned short __builtin_crypto_vpermxor (vector unsigned short, 50056 vector unsigned short, 50057 vector unsigned short); 50058 50059 vector unsigned int __builtin_crypto_vpermxor (vector unsigned int, 50060 vector unsigned int, 50061 vector unsigned int); 50062 50063 vector unsigned long long __builtin_crypto_vpermxor (vector unsigned long long, 50064 vector unsigned long long, 50065 vector unsigned long long); 50066 50067 vector unsigned char __builtin_crypto_vpmsumb (vector unsigned char, 50068 vector unsigned char); 50069 50070 vector unsigned short __builtin_crypto_vpmsumh (vector unsigned short, 50071 vector unsigned short); 50072 50073 vector unsigned int __builtin_crypto_vpmsumw (vector unsigned int, 50074 vector unsigned int); 50075 50076 vector unsigned long long __builtin_crypto_vpmsumd (vector unsigned long long, 50077 vector unsigned long long); 50078 50079 vector unsigned long long __builtin_crypto_vshasigmad (vector unsigned long long, 50080 int, int); 50081 50082 vector unsigned int __builtin_crypto_vshasigmaw (vector unsigned int, int, int); 50083 50084 The second argument to __BUILTIN_CRYPTO_VSHASIGMAD and 50085__BUILTIN_CRYPTO_VSHASIGMAW must be a constant integer that is 0 or 1. 50086The third argument to these built-in functions must be a constant 50087integer in the range of 0 to 15. 50088 50089 If the ISA 3.0 instruction set additions are enabled ('-mcpu=power9'), 50090the following additional functions are available for both 32-bit and 5009164-bit targets. 50092 vector short vec_xl (int, vector short *); 50093 vector short vec_xl (int, short *); 50094 vector unsigned short vec_xl (int, vector unsigned short *); 50095 vector unsigned short vec_xl (int, unsigned short *); 50096 vector char vec_xl (int, vector char *); 50097 vector char vec_xl (int, char *); 50098 vector unsigned char vec_xl (int, vector unsigned char *); 50099 vector unsigned char vec_xl (int, unsigned char *); 50100 50101 void vec_xst (vector short, int, vector short *); 50102 void vec_xst (vector short, int, short *); 50103 void vec_xst (vector unsigned short, int, vector unsigned short *); 50104 void vec_xst (vector unsigned short, int, unsigned short *); 50105 void vec_xst (vector char, int, vector char *); 50106 void vec_xst (vector char, int, char *); 50107 void vec_xst (vector unsigned char, int, vector unsigned char *); 50108 void vec_xst (vector unsigned char, int, unsigned char *); 50109 50110 50111File: gcc.info, Node: PowerPC Hardware Transactional Memory Built-in Functions, Next: PowerPC Atomic Memory Operation Functions, Prev: PowerPC AltiVec/VSX Built-in Functions, Up: Target Builtins 50112 501136.60.24 PowerPC Hardware Transactional Memory Built-in Functions 50114---------------------------------------------------------------- 50115 50116GCC provides two interfaces for accessing the Hardware Transactional 50117Memory (HTM) instructions available on some of the PowerPC family of 50118processors (eg, POWER8). The two interfaces come in a low level 50119interface, consisting of built-in functions specific to PowerPC and a 50120higher level interface consisting of inline functions that are common 50121between PowerPC and S/390. 50122 501236.60.24.1 PowerPC HTM Low Level Built-in Functions 50124.................................................. 50125 50126The following low level built-in functions are available with '-mhtm' or 50127'-mcpu=CPU' where CPU is 'power8' or later. They all generate the 50128machine instruction that is part of the name. 50129 50130 The HTM builtins (with the exception of '__builtin_tbegin') return the 50131full 4-bit condition register value set by their associated hardware 50132instruction. The header file 'htmintrin.h' defines some macros that can 50133be used to decipher the return value. The '__builtin_tbegin' builtin 50134returns a simple 'true' or 'false' value depending on whether a 50135transaction was successfully started or not. The arguments of the 50136builtins match exactly the type and order of the associated hardware 50137instruction's operands, except for the '__builtin_tcheck' builtin, which 50138does not take any input arguments. Refer to the ISA manual for a 50139description of each instruction's operands. 50140 50141 unsigned int __builtin_tbegin (unsigned int) 50142 unsigned int __builtin_tend (unsigned int) 50143 50144 unsigned int __builtin_tabort (unsigned int) 50145 unsigned int __builtin_tabortdc (unsigned int, unsigned int, unsigned int) 50146 unsigned int __builtin_tabortdci (unsigned int, unsigned int, int) 50147 unsigned int __builtin_tabortwc (unsigned int, unsigned int, unsigned int) 50148 unsigned int __builtin_tabortwci (unsigned int, unsigned int, int) 50149 50150 unsigned int __builtin_tcheck (void) 50151 unsigned int __builtin_treclaim (unsigned int) 50152 unsigned int __builtin_trechkpt (void) 50153 unsigned int __builtin_tsr (unsigned int) 50154 50155 In addition to the above HTM built-ins, we have added built-ins for 50156some common extended mnemonics of the HTM instructions: 50157 50158 unsigned int __builtin_tendall (void) 50159 unsigned int __builtin_tresume (void) 50160 unsigned int __builtin_tsuspend (void) 50161 50162 Note that the semantics of the above HTM builtins are required to mimic 50163the locking semantics used for critical sections. Builtins that are 50164used to create a new transaction or restart a suspended transaction must 50165have lock acquisition like semantics while those builtins that end or 50166suspend a transaction must have lock release like semantics. 50167Specifically, this must mimic lock semantics as specified by C++11, for 50168example: Lock acquisition is as-if an execution of 50169__atomic_exchange_n(&globallock,1,__ATOMIC_ACQUIRE) that returns 0, and 50170lock release is as-if an execution of 50171__atomic_store(&globallock,0,__ATOMIC_RELEASE), with globallock being an 50172implicit implementation-defined lock used for all transactions. The HTM 50173instructions associated with with the builtins inherently provide the 50174correct acquisition and release hardware barriers required. However, 50175the compiler must also be prohibited from moving loads and stores across 50176the builtins in a way that would violate their semantics. This has been 50177accomplished by adding memory barriers to the associated HTM 50178instructions (which is a conservative approach to provide acquire and 50179release semantics). Earlier versions of the compiler did not treat the 50180HTM instructions as memory barriers. A '__TM_FENCE__' macro has been 50181added, which can be used to determine whether the current compiler 50182treats HTM instructions as memory barriers or not. This allows the user 50183to explicitly add memory barriers to their code when using an older 50184version of the compiler. 50185 50186 The following set of built-in functions are available to gain access to 50187the HTM specific special purpose registers. 50188 50189 unsigned long __builtin_get_texasr (void) 50190 unsigned long __builtin_get_texasru (void) 50191 unsigned long __builtin_get_tfhar (void) 50192 unsigned long __builtin_get_tfiar (void) 50193 50194 void __builtin_set_texasr (unsigned long); 50195 void __builtin_set_texasru (unsigned long); 50196 void __builtin_set_tfhar (unsigned long); 50197 void __builtin_set_tfiar (unsigned long); 50198 50199 Example usage of these low level built-in functions may look like: 50200 50201 #include <htmintrin.h> 50202 50203 int num_retries = 10; 50204 50205 while (1) 50206 { 50207 if (__builtin_tbegin (0)) 50208 { 50209 /* Transaction State Initiated. */ 50210 if (is_locked (lock)) 50211 __builtin_tabort (0); 50212 ... transaction code... 50213 __builtin_tend (0); 50214 break; 50215 } 50216 else 50217 { 50218 /* Transaction State Failed. Use locks if the transaction 50219 failure is "persistent" or we've tried too many times. */ 50220 if (num_retries-- <= 0 50221 || _TEXASRU_FAILURE_PERSISTENT (__builtin_get_texasru ())) 50222 { 50223 acquire_lock (lock); 50224 ... non transactional fallback path... 50225 release_lock (lock); 50226 break; 50227 } 50228 } 50229 } 50230 50231 One final built-in function has been added that returns the value of 50232the 2-bit Transaction State field of the Machine Status Register (MSR) 50233as stored in 'CR0'. 50234 50235 unsigned long __builtin_ttest (void) 50236 50237 This built-in can be used to determine the current transaction state 50238using the following code example: 50239 50240 #include <htmintrin.h> 50241 50242 unsigned char tx_state = _HTM_STATE (__builtin_ttest ()); 50243 50244 if (tx_state == _HTM_TRANSACTIONAL) 50245 { 50246 /* Code to use in transactional state. */ 50247 } 50248 else if (tx_state == _HTM_NONTRANSACTIONAL) 50249 { 50250 /* Code to use in non-transactional state. */ 50251 } 50252 else if (tx_state == _HTM_SUSPENDED) 50253 { 50254 /* Code to use in transaction suspended state. */ 50255 } 50256 502576.60.24.2 PowerPC HTM High Level Inline Functions 50258................................................. 50259 50260The following high level HTM interface is made available by including 50261'<htmxlintrin.h>' and using '-mhtm' or '-mcpu=CPU' where CPU is 'power8' 50262or later. This interface is common between PowerPC and S/390, allowing 50263users to write one HTM source implementation that can be compiled and 50264executed on either system. 50265 50266 long __TM_simple_begin (void) 50267 long __TM_begin (void* const TM_buff) 50268 long __TM_end (void) 50269 void __TM_abort (void) 50270 void __TM_named_abort (unsigned char const code) 50271 void __TM_resume (void) 50272 void __TM_suspend (void) 50273 50274 long __TM_is_user_abort (void* const TM_buff) 50275 long __TM_is_named_user_abort (void* const TM_buff, unsigned char *code) 50276 long __TM_is_illegal (void* const TM_buff) 50277 long __TM_is_footprint_exceeded (void* const TM_buff) 50278 long __TM_nesting_depth (void* const TM_buff) 50279 long __TM_is_nested_too_deep(void* const TM_buff) 50280 long __TM_is_conflict(void* const TM_buff) 50281 long __TM_is_failure_persistent(void* const TM_buff) 50282 long __TM_failure_address(void* const TM_buff) 50283 long long __TM_failure_code(void* const TM_buff) 50284 50285 Using these common set of HTM inline functions, we can create a more 50286portable version of the HTM example in the previous section that will 50287work on either PowerPC or S/390: 50288 50289 #include <htmxlintrin.h> 50290 50291 int num_retries = 10; 50292 TM_buff_type TM_buff; 50293 50294 while (1) 50295 { 50296 if (__TM_begin (TM_buff) == _HTM_TBEGIN_STARTED) 50297 { 50298 /* Transaction State Initiated. */ 50299 if (is_locked (lock)) 50300 __TM_abort (); 50301 ... transaction code... 50302 __TM_end (); 50303 break; 50304 } 50305 else 50306 { 50307 /* Transaction State Failed. Use locks if the transaction 50308 failure is "persistent" or we've tried too many times. */ 50309 if (num_retries-- <= 0 50310 || __TM_is_failure_persistent (TM_buff)) 50311 { 50312 acquire_lock (lock); 50313 ... non transactional fallback path... 50314 release_lock (lock); 50315 break; 50316 } 50317 } 50318 } 50319 50320 50321File: gcc.info, Node: PowerPC Atomic Memory Operation Functions, Next: PowerPC Matrix-Multiply Assist Built-in Functions, Prev: PowerPC Hardware Transactional Memory Built-in Functions, Up: Target Builtins 50322 503236.60.25 PowerPC Atomic Memory Operation Functions 50324------------------------------------------------- 50325 50326ISA 3.0 of the PowerPC added new atomic memory operation (amo) 50327instructions. GCC provides support for these instructions in 64-bit 50328environments. All of the functions are declared in the include file 50329'amo.h'. 50330 50331 The functions supported are: 50332 50333 #include <amo.h> 50334 50335 uint32_t amo_lwat_add (uint32_t *, uint32_t); 50336 uint32_t amo_lwat_xor (uint32_t *, uint32_t); 50337 uint32_t amo_lwat_ior (uint32_t *, uint32_t); 50338 uint32_t amo_lwat_and (uint32_t *, uint32_t); 50339 uint32_t amo_lwat_umax (uint32_t *, uint32_t); 50340 uint32_t amo_lwat_umin (uint32_t *, uint32_t); 50341 uint32_t amo_lwat_swap (uint32_t *, uint32_t); 50342 50343 int32_t amo_lwat_sadd (int32_t *, int32_t); 50344 int32_t amo_lwat_smax (int32_t *, int32_t); 50345 int32_t amo_lwat_smin (int32_t *, int32_t); 50346 int32_t amo_lwat_sswap (int32_t *, int32_t); 50347 50348 uint64_t amo_ldat_add (uint64_t *, uint64_t); 50349 uint64_t amo_ldat_xor (uint64_t *, uint64_t); 50350 uint64_t amo_ldat_ior (uint64_t *, uint64_t); 50351 uint64_t amo_ldat_and (uint64_t *, uint64_t); 50352 uint64_t amo_ldat_umax (uint64_t *, uint64_t); 50353 uint64_t amo_ldat_umin (uint64_t *, uint64_t); 50354 uint64_t amo_ldat_swap (uint64_t *, uint64_t); 50355 50356 int64_t amo_ldat_sadd (int64_t *, int64_t); 50357 int64_t amo_ldat_smax (int64_t *, int64_t); 50358 int64_t amo_ldat_smin (int64_t *, int64_t); 50359 int64_t amo_ldat_sswap (int64_t *, int64_t); 50360 50361 void amo_stwat_add (uint32_t *, uint32_t); 50362 void amo_stwat_xor (uint32_t *, uint32_t); 50363 void amo_stwat_ior (uint32_t *, uint32_t); 50364 void amo_stwat_and (uint32_t *, uint32_t); 50365 void amo_stwat_umax (uint32_t *, uint32_t); 50366 void amo_stwat_umin (uint32_t *, uint32_t); 50367 50368 void amo_stwat_sadd (int32_t *, int32_t); 50369 void amo_stwat_smax (int32_t *, int32_t); 50370 void amo_stwat_smin (int32_t *, int32_t); 50371 50372 void amo_stdat_add (uint64_t *, uint64_t); 50373 void amo_stdat_xor (uint64_t *, uint64_t); 50374 void amo_stdat_ior (uint64_t *, uint64_t); 50375 void amo_stdat_and (uint64_t *, uint64_t); 50376 void amo_stdat_umax (uint64_t *, uint64_t); 50377 void amo_stdat_umin (uint64_t *, uint64_t); 50378 50379 void amo_stdat_sadd (int64_t *, int64_t); 50380 void amo_stdat_smax (int64_t *, int64_t); 50381 void amo_stdat_smin (int64_t *, int64_t); 50382 50383 50384File: gcc.info, Node: PowerPC Matrix-Multiply Assist Built-in Functions, Next: RISC-V Built-in Functions, Prev: PowerPC Atomic Memory Operation Functions, Up: Target Builtins 50385 503866.60.26 PowerPC Matrix-Multiply Assist Built-in Functions 50387--------------------------------------------------------- 50388 50389ISA 3.1 of the PowerPC added new Matrix-Multiply Assist (MMA) 50390instructions. GCC provides support for these instructions through the 50391following built-in functions which are enabled with the '-mmma' option. 50392The vec_t type below is defined to be a normal vector unsigned char 50393type.�� The uint2, uint4 and uint8 parameters are 2-bit, 4-bit and 8-bit 50394unsigned integer constants respectively.�� The compiler will verify that 50395they are constants and that their values are within range.�� 50396 50397 The built-in functions supported are: 50398 50399 void __builtin_mma_xvi4ger8 (__vector_quad *, vec_t, vec_t); 50400 void __builtin_mma_xvi8ger4 (__vector_quad *, vec_t, vec_t); 50401 void __builtin_mma_xvi16ger2 (__vector_quad *, vec_t, vec_t); 50402 void __builtin_mma_xvi16ger2s (__vector_quad *, vec_t, vec_t); 50403 void __builtin_mma_xvf16ger2 (__vector_quad *, vec_t, vec_t); 50404 void __builtin_mma_xvbf16ger2 (__vector_quad *, vec_t, vec_t); 50405 void __builtin_mma_xvf32ger (__vector_quad *, vec_t, vec_t); 50406 50407 void __builtin_mma_xvi4ger8pp (__vector_quad *, vec_t, vec_t); 50408 void __builtin_mma_xvi8ger4pp (__vector_quad *, vec_t, vec_t); 50409 void __builtin_mma_xvi8ger4spp(__vector_quad *, vec_t, vec_t); 50410 void __builtin_mma_xvi16ger2pp (__vector_quad *, vec_t, vec_t); 50411 void __builtin_mma_xvi16ger2spp (__vector_quad *, vec_t, vec_t); 50412 void __builtin_mma_xvf16ger2pp (__vector_quad *, vec_t, vec_t); 50413 void __builtin_mma_xvf16ger2pn (__vector_quad *, vec_t, vec_t); 50414 void __builtin_mma_xvf16ger2np (__vector_quad *, vec_t, vec_t); 50415 void __builtin_mma_xvf16ger2nn (__vector_quad *, vec_t, vec_t); 50416 void __builtin_mma_xvbf16ger2pp (__vector_quad *, vec_t, vec_t); 50417 void __builtin_mma_xvbf16ger2pn (__vector_quad *, vec_t, vec_t); 50418 void __builtin_mma_xvbf16ger2np (__vector_quad *, vec_t, vec_t); 50419 void __builtin_mma_xvbf16ger2nn (__vector_quad *, vec_t, vec_t); 50420 void __builtin_mma_xvf32gerpp (__vector_quad *, vec_t, vec_t); 50421 void __builtin_mma_xvf32gerpn (__vector_quad *, vec_t, vec_t); 50422 void __builtin_mma_xvf32gernp (__vector_quad *, vec_t, vec_t); 50423 void __builtin_mma_xvf32gernn (__vector_quad *, vec_t, vec_t); 50424 50425 void __builtin_mma_pmxvi4ger8 (__vector_quad *, vec_t, vec_t, uint4, uint4, uint8); 50426 void __builtin_mma_pmxvi4ger8pp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint8); 50427 50428 void __builtin_mma_pmxvi8ger4 (__vector_quad *, vec_t, vec_t, uint4, uint4, uint4); 50429 void __builtin_mma_pmxvi8ger4pp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint4); 50430 void __builtin_mma_pmxvi8ger4spp(__vector_quad *, vec_t, vec_t, uint4, uint4, uint4); 50431 50432 void __builtin_mma_pmxvi16ger2 (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50433 void __builtin_mma_pmxvi16ger2s (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50434 void __builtin_mma_pmxvf16ger2 (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50435 void __builtin_mma_pmxvbf16ger2 (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50436 50437 void __builtin_mma_pmxvi16ger2pp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50438 void __builtin_mma_pmxvi16ger2spp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50439 void __builtin_mma_pmxvf16ger2pp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50440 void __builtin_mma_pmxvf16ger2pn (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50441 void __builtin_mma_pmxvf16ger2np (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50442 void __builtin_mma_pmxvf16ger2nn (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50443 void __builtin_mma_pmxvbf16ger2pp (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50444 void __builtin_mma_pmxvbf16ger2pn (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50445 void __builtin_mma_pmxvbf16ger2np (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50446 void __builtin_mma_pmxvbf16ger2nn (__vector_quad *, vec_t, vec_t, uint4, uint4, uint2); 50447 50448 void __builtin_mma_pmxvf32ger (__vector_quad *, vec_t, vec_t, uint4, uint4); 50449 void __builtin_mma_pmxvf32gerpp (__vector_quad *, vec_t, vec_t, uint4, uint4); 50450 void __builtin_mma_pmxvf32gerpn (__vector_quad *, vec_t, vec_t, uint4, uint4); 50451 void __builtin_mma_pmxvf32gernp (__vector_quad *, vec_t, vec_t, uint4, uint4); 50452 void __builtin_mma_pmxvf32gernn (__vector_quad *, vec_t, vec_t, uint4, uint4); 50453 50454 void __builtin_mma_xvf64ger (__vector_quad *, __vector_pair, vec_t); 50455 void __builtin_mma_xvf64gerpp (__vector_quad *, __vector_pair, vec_t); 50456 void __builtin_mma_xvf64gerpn (__vector_quad *, __vector_pair, vec_t); 50457 void __builtin_mma_xvf64gernp (__vector_quad *, __vector_pair, vec_t); 50458 void __builtin_mma_xvf64gernn (__vector_quad *, __vector_pair, vec_t); 50459 50460 void __builtin_mma_pmxvf64ger (__vector_quad *, __vector_pair, vec_t, uint4, uint2); 50461 void __builtin_mma_pmxvf64gerpp (__vector_quad *, __vector_pair, vec_t, uint4, uint2); 50462 void __builtin_mma_pmxvf64gerpn (__vector_quad *, __vector_pair, vec_t, uint4, uint2); 50463 void __builtin_mma_pmxvf64gernp (__vector_quad *, __vector_pair, vec_t, uint4, uint2); 50464 void __builtin_mma_pmxvf64gernn (__vector_quad *, __vector_pair, vec_t, uint4, uint2); 50465 50466 void __builtin_mma_xxmtacc (__vector_quad *); 50467 void __builtin_mma_xxmfacc (__vector_quad *); 50468 void __builtin_mma_xxsetaccz (__vector_quad *); 50469 50470 void __builtin_mma_build_acc (__vector_quad *, vec_t, vec_t, vec_t, vec_t); 50471 void __builtin_mma_disassemble_acc (void *, __vector_quad *); 50472 50473 void __builtin_vsx_build_pair (__vector_pair *, vec_t, vec_t); 50474 void __builtin_vsx_disassemble_pair (void *, __vector_pair *); 50475 50476 vec_t __builtin_vsx_xvcvspbf16 (vec_t); 50477 vec_t __builtin_vsx_xvcvbf16spn (vec_t); 50478 50479 __vector_pair __builtin_vsx_lxvp (size_t, __vector_pair *); 50480 void __builtin_vsx_stxvp (__vector_pair, size_t, __vector_pair *); 50481 50482 50483File: gcc.info, Node: RISC-V Built-in Functions, Next: RX Built-in Functions, Prev: PowerPC Matrix-Multiply Assist Built-in Functions, Up: Target Builtins 50484 504856.60.27 RISC-V Built-in Functions 50486--------------------------------- 50487 50488These built-in functions are available for the RISC-V family of 50489processors. 50490 50491 -- Built-in Function: void * __builtin_thread_pointer (void) 50492 Returns the value that is currently set in the 'tp' register. 50493 50494 50495File: gcc.info, Node: RX Built-in Functions, Next: S/390 System z Built-in Functions, Prev: RISC-V Built-in Functions, Up: Target Builtins 50496 504976.60.28 RX Built-in Functions 50498----------------------------- 50499 50500GCC supports some of the RX instructions which cannot be expressed in 50501the C programming language via the use of built-in functions. The 50502following functions are supported: 50503 50504 -- Built-in Function: void __builtin_rx_brk (void) 50505 Generates the 'brk' machine instruction. 50506 50507 -- Built-in Function: void __builtin_rx_clrpsw (int) 50508 Generates the 'clrpsw' machine instruction to clear the specified 50509 bit in the processor status word. 50510 50511 -- Built-in Function: void __builtin_rx_int (int) 50512 Generates the 'int' machine instruction to generate an interrupt 50513 with the specified value. 50514 50515 -- Built-in Function: void __builtin_rx_machi (int, int) 50516 Generates the 'machi' machine instruction to add the result of 50517 multiplying the top 16 bits of the two arguments into the 50518 accumulator. 50519 50520 -- Built-in Function: void __builtin_rx_maclo (int, int) 50521 Generates the 'maclo' machine instruction to add the result of 50522 multiplying the bottom 16 bits of the two arguments into the 50523 accumulator. 50524 50525 -- Built-in Function: void __builtin_rx_mulhi (int, int) 50526 Generates the 'mulhi' machine instruction to place the result of 50527 multiplying the top 16 bits of the two arguments into the 50528 accumulator. 50529 50530 -- Built-in Function: void __builtin_rx_mullo (int, int) 50531 Generates the 'mullo' machine instruction to place the result of 50532 multiplying the bottom 16 bits of the two arguments into the 50533 accumulator. 50534 50535 -- Built-in Function: int __builtin_rx_mvfachi (void) 50536 Generates the 'mvfachi' machine instruction to read the top 32 bits 50537 of the accumulator. 50538 50539 -- Built-in Function: int __builtin_rx_mvfacmi (void) 50540 Generates the 'mvfacmi' machine instruction to read the middle 32 50541 bits of the accumulator. 50542 50543 -- Built-in Function: int __builtin_rx_mvfc (int) 50544 Generates the 'mvfc' machine instruction which reads the control 50545 register specified in its argument and returns its value. 50546 50547 -- Built-in Function: void __builtin_rx_mvtachi (int) 50548 Generates the 'mvtachi' machine instruction to set the top 32 bits 50549 of the accumulator. 50550 50551 -- Built-in Function: void __builtin_rx_mvtaclo (int) 50552 Generates the 'mvtaclo' machine instruction to set the bottom 32 50553 bits of the accumulator. 50554 50555 -- Built-in Function: void __builtin_rx_mvtc (int reg, int val) 50556 Generates the 'mvtc' machine instruction which sets control 50557 register number 'reg' to 'val'. 50558 50559 -- Built-in Function: void __builtin_rx_mvtipl (int) 50560 Generates the 'mvtipl' machine instruction set the interrupt 50561 priority level. 50562 50563 -- Built-in Function: void __builtin_rx_racw (int) 50564 Generates the 'racw' machine instruction to round the accumulator 50565 according to the specified mode. 50566 50567 -- Built-in Function: int __builtin_rx_revw (int) 50568 Generates the 'revw' machine instruction which swaps the bytes in 50569 the argument so that bits 0-7 now occupy bits 8-15 and vice versa, 50570 and also bits 16-23 occupy bits 24-31 and vice versa. 50571 50572 -- Built-in Function: void __builtin_rx_rmpa (void) 50573 Generates the 'rmpa' machine instruction which initiates a repeated 50574 multiply and accumulate sequence. 50575 50576 -- Built-in Function: void __builtin_rx_round (float) 50577 Generates the 'round' machine instruction which returns the 50578 floating-point argument rounded according to the current rounding 50579 mode set in the floating-point status word register. 50580 50581 -- Built-in Function: int __builtin_rx_sat (int) 50582 Generates the 'sat' machine instruction which returns the saturated 50583 value of the argument. 50584 50585 -- Built-in Function: void __builtin_rx_setpsw (int) 50586 Generates the 'setpsw' machine instruction to set the specified bit 50587 in the processor status word. 50588 50589 -- Built-in Function: void __builtin_rx_wait (void) 50590 Generates the 'wait' machine instruction. 50591 50592 50593File: gcc.info, Node: S/390 System z Built-in Functions, Next: SH Built-in Functions, Prev: RX Built-in Functions, Up: Target Builtins 50594 505956.60.29 S/390 System z Built-in Functions 50596----------------------------------------- 50597 50598 -- Built-in Function: int __builtin_tbegin (void*) 50599 Generates the 'tbegin' machine instruction starting a 50600 non-constrained hardware transaction. If the parameter is non-NULL 50601 the memory area is used to store the transaction diagnostic buffer 50602 and will be passed as first operand to 'tbegin'. This buffer can 50603 be defined using the 'struct __htm_tdb' C struct defined in 50604 'htmintrin.h' and must reside on a double-word boundary. The 50605 second tbegin operand is set to '0xff0c'. This enables 50606 save/restore of all GPRs and disables aborts for FPR and AR 50607 manipulations inside the transaction body. The condition code set 50608 by the tbegin instruction is returned as integer value. The tbegin 50609 instruction by definition overwrites the content of all FPRs. The 50610 compiler will generate code which saves and restores the FPRs. For 50611 soft-float code it is recommended to used the '*_nofloat' variant. 50612 In order to prevent a TDB from being written it is required to pass 50613 a constant zero value as parameter. Passing a zero value through a 50614 variable is not sufficient. Although modifications of access 50615 registers inside the transaction will not trigger an transaction 50616 abort it is not supported to actually modify them. Access 50617 registers do not get saved when entering a transaction. They will 50618 have undefined state when reaching the abort code. 50619 50620 Macros for the possible return codes of tbegin are defined in the 50621'htmintrin.h' header file: 50622 50623'_HTM_TBEGIN_STARTED' 50624 'tbegin' has been executed as part of normal processing. The 50625 transaction body is supposed to be executed. 50626'_HTM_TBEGIN_INDETERMINATE' 50627 The transaction was aborted due to an indeterminate condition which 50628 might be persistent. 50629'_HTM_TBEGIN_TRANSIENT' 50630 The transaction aborted due to a transient failure. The 50631 transaction should be re-executed in that case. 50632'_HTM_TBEGIN_PERSISTENT' 50633 The transaction aborted due to a persistent failure. Re-execution 50634 under same circumstances will not be productive. 50635 50636 -- Macro: _HTM_FIRST_USER_ABORT_CODE 50637 The '_HTM_FIRST_USER_ABORT_CODE' defined in 'htmintrin.h' specifies 50638 the first abort code which can be used for '__builtin_tabort'. 50639 Values below this threshold are reserved for machine use. 50640 50641 -- Data type: struct __htm_tdb 50642 The 'struct __htm_tdb' defined in 'htmintrin.h' describes the 50643 structure of the transaction diagnostic block as specified in the 50644 Principles of Operation manual chapter 5-91. 50645 50646 -- Built-in Function: int __builtin_tbegin_nofloat (void*) 50647 Same as '__builtin_tbegin' but without FPR saves and restores. 50648 Using this variant in code making use of FPRs will leave the FPRs 50649 in undefined state when entering the transaction abort handler 50650 code. 50651 50652 -- Built-in Function: int __builtin_tbegin_retry (void*, int) 50653 In addition to '__builtin_tbegin' a loop for transient failures is 50654 generated. If tbegin returns a condition code of 2 the transaction 50655 will be retried as often as specified in the second argument. The 50656 perform processor assist instruction is used to tell the CPU about 50657 the number of fails so far. 50658 50659 -- Built-in Function: int __builtin_tbegin_retry_nofloat (void*, int) 50660 Same as '__builtin_tbegin_retry' but without FPR saves and 50661 restores. Using this variant in code making use of FPRs will leave 50662 the FPRs in undefined state when entering the transaction abort 50663 handler code. 50664 50665 -- Built-in Function: void __builtin_tbeginc (void) 50666 Generates the 'tbeginc' machine instruction starting a constrained 50667 hardware transaction. The second operand is set to '0xff08'. 50668 50669 -- Built-in Function: int __builtin_tend (void) 50670 Generates the 'tend' machine instruction finishing a transaction 50671 and making the changes visible to other threads. The condition 50672 code generated by tend is returned as integer value. 50673 50674 -- Built-in Function: void __builtin_tabort (int) 50675 Generates the 'tabort' machine instruction with the specified abort 50676 code. Abort codes from 0 through 255 are reserved and will result 50677 in an error message. 50678 50679 -- Built-in Function: void __builtin_tx_assist (int) 50680 Generates the 'ppa rX,rY,1' machine instruction. Where the integer 50681 parameter is loaded into rX and a value of zero is loaded into rY. 50682 The integer parameter specifies the number of times the transaction 50683 repeatedly aborted. 50684 50685 -- Built-in Function: int __builtin_tx_nesting_depth (void) 50686 Generates the 'etnd' machine instruction. The current nesting 50687 depth is returned as integer value. For a nesting depth of 0 the 50688 code is not executed as part of an transaction. 50689 50690 -- Built-in Function: void __builtin_non_tx_store (uint64_t *, 50691 uint64_t) 50692 50693 Generates the 'ntstg' machine instruction. The second argument is 50694 written to the first arguments location. The store operation will 50695 not be rolled-back in case of an transaction abort. 50696 50697 50698File: gcc.info, Node: SH Built-in Functions, Next: SPARC VIS Built-in Functions, Prev: S/390 System z Built-in Functions, Up: Target Builtins 50699 507006.60.30 SH Built-in Functions 50701----------------------------- 50702 50703The following built-in functions are supported on the SH1, SH2, SH3 and 50704SH4 families of processors: 50705 50706 -- Built-in Function: void __builtin_set_thread_pointer (void *PTR) 50707 Sets the 'GBR' register to the specified value PTR. This is 50708 usually used by system code that manages threads and execution 50709 contexts. The compiler normally does not generate code that 50710 modifies the contents of 'GBR' and thus the value is preserved 50711 across function calls. Changing the 'GBR' value in user code must 50712 be done with caution, since the compiler might use 'GBR' in order 50713 to access thread local variables. 50714 50715 -- Built-in Function: void * __builtin_thread_pointer (void) 50716 Returns the value that is currently set in the 'GBR' register. 50717 Memory loads and stores that use the thread pointer as a base 50718 address are turned into 'GBR' based displacement loads and stores, 50719 if possible. For example: 50720 struct my_tcb 50721 { 50722 int a, b, c, d, e; 50723 }; 50724 50725 int get_tcb_value (void) 50726 { 50727 // Generate 'mov.l @(8,gbr),r0' instruction 50728 return ((my_tcb*)__builtin_thread_pointer ())->c; 50729 } 50730 50731 50732 -- Built-in Function: unsigned int __builtin_sh_get_fpscr (void) 50733 Returns the value that is currently set in the 'FPSCR' register. 50734 50735 -- Built-in Function: void __builtin_sh_set_fpscr (unsigned int VAL) 50736 Sets the 'FPSCR' register to the specified value VAL, while 50737 preserving the current values of the FR, SZ and PR bits. 50738 50739 50740File: gcc.info, Node: SPARC VIS Built-in Functions, Next: TI C6X Built-in Functions, Prev: SH Built-in Functions, Up: Target Builtins 50741 507426.60.31 SPARC VIS Built-in Functions 50743------------------------------------ 50744 50745GCC supports SIMD operations on the SPARC using both the generic vector 50746extensions (*note Vector Extensions::) as well as built-in functions for 50747the SPARC Visual Instruction Set (VIS). When you use the '-mvis' switch, 50748the VIS extension is exposed as the following built-in functions: 50749 50750 typedef int v1si __attribute__ ((vector_size (4))); 50751 typedef int v2si __attribute__ ((vector_size (8))); 50752 typedef short v4hi __attribute__ ((vector_size (8))); 50753 typedef short v2hi __attribute__ ((vector_size (4))); 50754 typedef unsigned char v8qi __attribute__ ((vector_size (8))); 50755 typedef unsigned char v4qi __attribute__ ((vector_size (4))); 50756 50757 void __builtin_vis_write_gsr (int64_t); 50758 int64_t __builtin_vis_read_gsr (void); 50759 50760 void * __builtin_vis_alignaddr (void *, long); 50761 void * __builtin_vis_alignaddrl (void *, long); 50762 int64_t __builtin_vis_faligndatadi (int64_t, int64_t); 50763 v2si __builtin_vis_faligndatav2si (v2si, v2si); 50764 v4hi __builtin_vis_faligndatav4hi (v4si, v4si); 50765 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi); 50766 50767 v4hi __builtin_vis_fexpand (v4qi); 50768 50769 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi); 50770 v4hi __builtin_vis_fmul8x16au (v4qi, v2hi); 50771 v4hi __builtin_vis_fmul8x16al (v4qi, v2hi); 50772 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi); 50773 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi); 50774 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi); 50775 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi); 50776 50777 v4qi __builtin_vis_fpack16 (v4hi); 50778 v8qi __builtin_vis_fpack32 (v2si, v8qi); 50779 v2hi __builtin_vis_fpackfix (v2si); 50780 v8qi __builtin_vis_fpmerge (v4qi, v4qi); 50781 50782 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t); 50783 50784 long __builtin_vis_edge8 (void *, void *); 50785 long __builtin_vis_edge8l (void *, void *); 50786 long __builtin_vis_edge16 (void *, void *); 50787 long __builtin_vis_edge16l (void *, void *); 50788 long __builtin_vis_edge32 (void *, void *); 50789 long __builtin_vis_edge32l (void *, void *); 50790 50791 long __builtin_vis_fcmple16 (v4hi, v4hi); 50792 long __builtin_vis_fcmple32 (v2si, v2si); 50793 long __builtin_vis_fcmpne16 (v4hi, v4hi); 50794 long __builtin_vis_fcmpne32 (v2si, v2si); 50795 long __builtin_vis_fcmpgt16 (v4hi, v4hi); 50796 long __builtin_vis_fcmpgt32 (v2si, v2si); 50797 long __builtin_vis_fcmpeq16 (v4hi, v4hi); 50798 long __builtin_vis_fcmpeq32 (v2si, v2si); 50799 50800 v4hi __builtin_vis_fpadd16 (v4hi, v4hi); 50801 v2hi __builtin_vis_fpadd16s (v2hi, v2hi); 50802 v2si __builtin_vis_fpadd32 (v2si, v2si); 50803 v1si __builtin_vis_fpadd32s (v1si, v1si); 50804 v4hi __builtin_vis_fpsub16 (v4hi, v4hi); 50805 v2hi __builtin_vis_fpsub16s (v2hi, v2hi); 50806 v2si __builtin_vis_fpsub32 (v2si, v2si); 50807 v1si __builtin_vis_fpsub32s (v1si, v1si); 50808 50809 long __builtin_vis_array8 (long, long); 50810 long __builtin_vis_array16 (long, long); 50811 long __builtin_vis_array32 (long, long); 50812 50813 When you use the '-mvis2' switch, the VIS version 2.0 built-in 50814functions also become available: 50815 50816 long __builtin_vis_bmask (long, long); 50817 int64_t __builtin_vis_bshuffledi (int64_t, int64_t); 50818 v2si __builtin_vis_bshufflev2si (v2si, v2si); 50819 v4hi __builtin_vis_bshufflev2si (v4hi, v4hi); 50820 v8qi __builtin_vis_bshufflev2si (v8qi, v8qi); 50821 50822 long __builtin_vis_edge8n (void *, void *); 50823 long __builtin_vis_edge8ln (void *, void *); 50824 long __builtin_vis_edge16n (void *, void *); 50825 long __builtin_vis_edge16ln (void *, void *); 50826 long __builtin_vis_edge32n (void *, void *); 50827 long __builtin_vis_edge32ln (void *, void *); 50828 50829 When you use the '-mvis3' switch, the VIS version 3.0 built-in 50830functions also become available: 50831 50832 void __builtin_vis_cmask8 (long); 50833 void __builtin_vis_cmask16 (long); 50834 void __builtin_vis_cmask32 (long); 50835 50836 v4hi __builtin_vis_fchksm16 (v4hi, v4hi); 50837 50838 v4hi __builtin_vis_fsll16 (v4hi, v4hi); 50839 v4hi __builtin_vis_fslas16 (v4hi, v4hi); 50840 v4hi __builtin_vis_fsrl16 (v4hi, v4hi); 50841 v4hi __builtin_vis_fsra16 (v4hi, v4hi); 50842 v2si __builtin_vis_fsll16 (v2si, v2si); 50843 v2si __builtin_vis_fslas16 (v2si, v2si); 50844 v2si __builtin_vis_fsrl16 (v2si, v2si); 50845 v2si __builtin_vis_fsra16 (v2si, v2si); 50846 50847 long __builtin_vis_pdistn (v8qi, v8qi); 50848 50849 v4hi __builtin_vis_fmean16 (v4hi, v4hi); 50850 50851 int64_t __builtin_vis_fpadd64 (int64_t, int64_t); 50852 int64_t __builtin_vis_fpsub64 (int64_t, int64_t); 50853 50854 v4hi __builtin_vis_fpadds16 (v4hi, v4hi); 50855 v2hi __builtin_vis_fpadds16s (v2hi, v2hi); 50856 v4hi __builtin_vis_fpsubs16 (v4hi, v4hi); 50857 v2hi __builtin_vis_fpsubs16s (v2hi, v2hi); 50858 v2si __builtin_vis_fpadds32 (v2si, v2si); 50859 v1si __builtin_vis_fpadds32s (v1si, v1si); 50860 v2si __builtin_vis_fpsubs32 (v2si, v2si); 50861 v1si __builtin_vis_fpsubs32s (v1si, v1si); 50862 50863 long __builtin_vis_fucmple8 (v8qi, v8qi); 50864 long __builtin_vis_fucmpne8 (v8qi, v8qi); 50865 long __builtin_vis_fucmpgt8 (v8qi, v8qi); 50866 long __builtin_vis_fucmpeq8 (v8qi, v8qi); 50867 50868 float __builtin_vis_fhadds (float, float); 50869 double __builtin_vis_fhaddd (double, double); 50870 float __builtin_vis_fhsubs (float, float); 50871 double __builtin_vis_fhsubd (double, double); 50872 float __builtin_vis_fnhadds (float, float); 50873 double __builtin_vis_fnhaddd (double, double); 50874 50875 int64_t __builtin_vis_umulxhi (int64_t, int64_t); 50876 int64_t __builtin_vis_xmulx (int64_t, int64_t); 50877 int64_t __builtin_vis_xmulxhi (int64_t, int64_t); 50878 50879 When you use the '-mvis4' switch, the VIS version 4.0 built-in 50880functions also become available: 50881 50882 v8qi __builtin_vis_fpadd8 (v8qi, v8qi); 50883 v8qi __builtin_vis_fpadds8 (v8qi, v8qi); 50884 v8qi __builtin_vis_fpaddus8 (v8qi, v8qi); 50885 v4hi __builtin_vis_fpaddus16 (v4hi, v4hi); 50886 50887 v8qi __builtin_vis_fpsub8 (v8qi, v8qi); 50888 v8qi __builtin_vis_fpsubs8 (v8qi, v8qi); 50889 v8qi __builtin_vis_fpsubus8 (v8qi, v8qi); 50890 v4hi __builtin_vis_fpsubus16 (v4hi, v4hi); 50891 50892 long __builtin_vis_fpcmple8 (v8qi, v8qi); 50893 long __builtin_vis_fpcmpgt8 (v8qi, v8qi); 50894 long __builtin_vis_fpcmpule16 (v4hi, v4hi); 50895 long __builtin_vis_fpcmpugt16 (v4hi, v4hi); 50896 long __builtin_vis_fpcmpule32 (v2si, v2si); 50897 long __builtin_vis_fpcmpugt32 (v2si, v2si); 50898 50899 v8qi __builtin_vis_fpmax8 (v8qi, v8qi); 50900 v4hi __builtin_vis_fpmax16 (v4hi, v4hi); 50901 v2si __builtin_vis_fpmax32 (v2si, v2si); 50902 50903 v8qi __builtin_vis_fpmaxu8 (v8qi, v8qi); 50904 v4hi __builtin_vis_fpmaxu16 (v4hi, v4hi); 50905 v2si __builtin_vis_fpmaxu32 (v2si, v2si); 50906 50907 50908 v8qi __builtin_vis_fpmin8 (v8qi, v8qi); 50909 v4hi __builtin_vis_fpmin16 (v4hi, v4hi); 50910 v2si __builtin_vis_fpmin32 (v2si, v2si); 50911 50912 v8qi __builtin_vis_fpminu8 (v8qi, v8qi); 50913 v4hi __builtin_vis_fpminu16 (v4hi, v4hi); 50914 v2si __builtin_vis_fpminu32 (v2si, v2si); 50915 50916 When you use the '-mvis4b' switch, the VIS version 4.0B built-in 50917functions also become available: 50918 50919 v8qi __builtin_vis_dictunpack8 (double, int); 50920 v4hi __builtin_vis_dictunpack16 (double, int); 50921 v2si __builtin_vis_dictunpack32 (double, int); 50922 50923 long __builtin_vis_fpcmple8shl (v8qi, v8qi, int); 50924 long __builtin_vis_fpcmpgt8shl (v8qi, v8qi, int); 50925 long __builtin_vis_fpcmpeq8shl (v8qi, v8qi, int); 50926 long __builtin_vis_fpcmpne8shl (v8qi, v8qi, int); 50927 50928 long __builtin_vis_fpcmple16shl (v4hi, v4hi, int); 50929 long __builtin_vis_fpcmpgt16shl (v4hi, v4hi, int); 50930 long __builtin_vis_fpcmpeq16shl (v4hi, v4hi, int); 50931 long __builtin_vis_fpcmpne16shl (v4hi, v4hi, int); 50932 50933 long __builtin_vis_fpcmple32shl (v2si, v2si, int); 50934 long __builtin_vis_fpcmpgt32shl (v2si, v2si, int); 50935 long __builtin_vis_fpcmpeq32shl (v2si, v2si, int); 50936 long __builtin_vis_fpcmpne32shl (v2si, v2si, int); 50937 50938 long __builtin_vis_fpcmpule8shl (v8qi, v8qi, int); 50939 long __builtin_vis_fpcmpugt8shl (v8qi, v8qi, int); 50940 long __builtin_vis_fpcmpule16shl (v4hi, v4hi, int); 50941 long __builtin_vis_fpcmpugt16shl (v4hi, v4hi, int); 50942 long __builtin_vis_fpcmpule32shl (v2si, v2si, int); 50943 long __builtin_vis_fpcmpugt32shl (v2si, v2si, int); 50944 50945 long __builtin_vis_fpcmpde8shl (v8qi, v8qi, int); 50946 long __builtin_vis_fpcmpde16shl (v4hi, v4hi, int); 50947 long __builtin_vis_fpcmpde32shl (v2si, v2si, int); 50948 50949 long __builtin_vis_fpcmpur8shl (v8qi, v8qi, int); 50950 long __builtin_vis_fpcmpur16shl (v4hi, v4hi, int); 50951 long __builtin_vis_fpcmpur32shl (v2si, v2si, int); 50952 50953 50954File: gcc.info, Node: TI C6X Built-in Functions, Next: TILE-Gx Built-in Functions, Prev: SPARC VIS Built-in Functions, Up: Target Builtins 50955 509566.60.32 TI C6X Built-in Functions 50957--------------------------------- 50958 50959GCC provides intrinsics to access certain instructions of the TI C6X 50960processors. These intrinsics, listed below, are available after 50961inclusion of the 'c6x_intrinsics.h' header file. They map directly to 50962C6X instructions. 50963 50964 50965 int _sadd (int, int) 50966 int _ssub (int, int) 50967 int _sadd2 (int, int) 50968 int _ssub2 (int, int) 50969 long long _mpy2 (int, int) 50970 long long _smpy2 (int, int) 50971 int _add4 (int, int) 50972 int _sub4 (int, int) 50973 int _saddu4 (int, int) 50974 50975 int _smpy (int, int) 50976 int _smpyh (int, int) 50977 int _smpyhl (int, int) 50978 int _smpylh (int, int) 50979 50980 int _sshl (int, int) 50981 int _subc (int, int) 50982 50983 int _avg2 (int, int) 50984 int _avgu4 (int, int) 50985 50986 int _clrr (int, int) 50987 int _extr (int, int) 50988 int _extru (int, int) 50989 int _abs (int) 50990 int _abs2 (int) 50991 50992 50993 50994File: gcc.info, Node: TILE-Gx Built-in Functions, Next: TILEPro Built-in Functions, Prev: TI C6X Built-in Functions, Up: Target Builtins 50995 509966.60.33 TILE-Gx Built-in Functions 50997---------------------------------- 50998 50999GCC provides intrinsics to access every instruction of the TILE-Gx 51000processor. The intrinsics are of the form: 51001 51002 51003 unsigned long long __insn_OP (...) 51004 51005 51006 Where OP is the name of the instruction. Refer to the ISA manual for 51007the complete list of instructions. 51008 51009 GCC also provides intrinsics to directly access the network registers. 51010The intrinsics are: 51011 51012 51013 unsigned long long __tile_idn0_receive (void) 51014 unsigned long long __tile_idn1_receive (void) 51015 unsigned long long __tile_udn0_receive (void) 51016 unsigned long long __tile_udn1_receive (void) 51017 unsigned long long __tile_udn2_receive (void) 51018 unsigned long long __tile_udn3_receive (void) 51019 void __tile_idn_send (unsigned long long) 51020 void __tile_udn_send (unsigned long long) 51021 51022 51023 The intrinsic 'void __tile_network_barrier (void)' is used to guarantee 51024that no network operations before it are reordered with those after it. 51025 51026 51027File: gcc.info, Node: TILEPro Built-in Functions, Next: x86 Built-in Functions, Prev: TILE-Gx Built-in Functions, Up: Target Builtins 51028 510296.60.34 TILEPro Built-in Functions 51030---------------------------------- 51031 51032GCC provides intrinsics to access every instruction of the TILEPro 51033processor. The intrinsics are of the form: 51034 51035 51036 unsigned __insn_OP (...) 51037 51038 51039where OP is the name of the instruction. Refer to the ISA manual for 51040the complete list of instructions. 51041 51042 GCC also provides intrinsics to directly access the network registers. 51043The intrinsics are: 51044 51045 51046 unsigned __tile_idn0_receive (void) 51047 unsigned __tile_idn1_receive (void) 51048 unsigned __tile_sn_receive (void) 51049 unsigned __tile_udn0_receive (void) 51050 unsigned __tile_udn1_receive (void) 51051 unsigned __tile_udn2_receive (void) 51052 unsigned __tile_udn3_receive (void) 51053 void __tile_idn_send (unsigned) 51054 void __tile_sn_send (unsigned) 51055 void __tile_udn_send (unsigned) 51056 51057 51058 The intrinsic 'void __tile_network_barrier (void)' is used to guarantee 51059that no network operations before it are reordered with those after it. 51060 51061 51062File: gcc.info, Node: x86 Built-in Functions, Next: x86 transactional memory intrinsics, Prev: TILEPro Built-in Functions, Up: Target Builtins 51063 510646.60.35 x86 Built-in Functions 51065------------------------------ 51066 51067These built-in functions are available for the x86-32 and x86-64 family 51068of computers, depending on the command-line switches used. 51069 51070 If you specify command-line switches such as '-msse', the compiler 51071could use the extended instruction sets even if the built-ins are not 51072used explicitly in the program. For this reason, applications that 51073perform run-time CPU detection must compile separate files for each 51074supported architecture, using the appropriate flags. In particular, the 51075file containing the CPU detection code should be compiled without these 51076options. 51077 51078 The following machine modes are available for use with MMX built-in 51079functions (*note Vector Extensions::): 'V2SI' for a vector of two 32-bit 51080integers, 'V4HI' for a vector of four 16-bit integers, and 'V8QI' for a 51081vector of eight 8-bit integers. Some of the built-in functions operate 51082on MMX registers as a whole 64-bit entity, these use 'V1DI' as their 51083mode. 51084 51085 If 3DNow! extensions are enabled, 'V2SF' is used as a mode for a vector 51086of two 32-bit floating-point values. 51087 51088 If SSE extensions are enabled, 'V4SF' is used for a vector of four 5108932-bit floating-point values. Some instructions use a vector of four 5109032-bit integers, these use 'V4SI'. Finally, some instructions operate 51091on an entire vector register, interpreting it as a 128-bit integer, 51092these use mode 'TI'. 51093 51094 The x86-32 and x86-64 family of processors use additional built-in 51095functions for efficient use of 'TF' ('__float128') 128-bit floating 51096point and 'TC' 128-bit complex floating-point values. 51097 51098 The following floating-point built-in functions are always available. 51099All of them implement the function that is part of the name. 51100 51101 __float128 __builtin_fabsq (__float128) 51102 __float128 __builtin_copysignq (__float128, __float128) 51103 51104 The following built-in functions are always available. 51105 51106'__float128 __builtin_infq (void)' 51107 Similar to '__builtin_inf', except the return type is '__float128'. 51108 51109'__float128 __builtin_huge_valq (void)' 51110 Similar to '__builtin_huge_val', except the return type is 51111 '__float128'. 51112 51113'__float128 __builtin_nanq (void)' 51114 Similar to '__builtin_nan', except the return type is '__float128'. 51115 51116'__float128 __builtin_nansq (void)' 51117 Similar to '__builtin_nans', except the return type is 51118 '__float128'. 51119 51120 The following built-in function is always available. 51121 51122'void __builtin_ia32_pause (void)' 51123 Generates the 'pause' machine instruction with a compiler memory 51124 barrier. 51125 51126 The following built-in functions are always available and can be used 51127to check the target platform type. 51128 51129 -- Built-in Function: void __builtin_cpu_init (void) 51130 This function runs the CPU detection code to check the type of CPU 51131 and the features supported. This built-in function needs to be 51132 invoked along with the built-in functions to check CPU type and 51133 features, '__builtin_cpu_is' and '__builtin_cpu_supports', only 51134 when used in a function that is executed before any constructors 51135 are called. The CPU detection code is automatically executed in a 51136 very high priority constructor. 51137 51138 For example, this function has to be used in 'ifunc' resolvers that 51139 check for CPU type using the built-in functions '__builtin_cpu_is' 51140 and '__builtin_cpu_supports', or in constructors on targets that 51141 don't support constructor priority. 51142 51143 static void (*resolve_memcpy (void)) (void) 51144 { 51145 // ifunc resolvers fire before constructors, explicitly call the init 51146 // function. 51147 __builtin_cpu_init (); 51148 if (__builtin_cpu_supports ("ssse3")) 51149 return ssse3_memcpy; // super fast memcpy with ssse3 instructions. 51150 else 51151 return default_memcpy; 51152 } 51153 51154 void *memcpy (void *, const void *, size_t) 51155 __attribute__ ((ifunc ("resolve_memcpy"))); 51156 51157 -- Built-in Function: int __builtin_cpu_is (const char *CPUNAME) 51158 This function returns a positive integer if the run-time CPU is of 51159 type CPUNAME and returns '0' otherwise. The following CPU names 51160 can be detected: 51161 51162 'amd' 51163 AMD CPU. 51164 51165 'intel' 51166 Intel CPU. 51167 51168 'atom' 51169 Intel Atom CPU. 51170 51171 'slm' 51172 Intel Silvermont CPU. 51173 51174 'core2' 51175 Intel Core 2 CPU. 51176 51177 'corei7' 51178 Intel Core i7 CPU. 51179 51180 'nehalem' 51181 Intel Core i7 Nehalem CPU. 51182 51183 'westmere' 51184 Intel Core i7 Westmere CPU. 51185 51186 'sandybridge' 51187 Intel Core i7 Sandy Bridge CPU. 51188 51189 'ivybridge' 51190 Intel Core i7 Ivy Bridge CPU. 51191 51192 'haswell' 51193 Intel Core i7 Haswell CPU. 51194 51195 'broadwell' 51196 Intel Core i7 Broadwell CPU. 51197 51198 'skylake' 51199 Intel Core i7 Skylake CPU. 51200 51201 'skylake-avx512' 51202 Intel Core i7 Skylake AVX512 CPU. 51203 51204 'cannonlake' 51205 Intel Core i7 Cannon Lake CPU. 51206 51207 'icelake-client' 51208 Intel Core i7 Ice Lake Client CPU. 51209 51210 'icelake-server' 51211 Intel Core i7 Ice Lake Server CPU. 51212 51213 'cascadelake' 51214 Intel Core i7 Cascadelake CPU. 51215 51216 'tigerlake' 51217 Intel Core i7 Tigerlake CPU. 51218 51219 'cooperlake' 51220 Intel Core i7 Cooperlake CPU. 51221 51222 'bonnell' 51223 Intel Atom Bonnell CPU. 51224 51225 'silvermont' 51226 Intel Atom Silvermont CPU. 51227 51228 'goldmont' 51229 Intel Atom Goldmont CPU. 51230 51231 'goldmont-plus' 51232 Intel Atom Goldmont Plus CPU. 51233 51234 'tremont' 51235 Intel Atom Tremont CPU. 51236 51237 'knl' 51238 Intel Knights Landing CPU. 51239 51240 'knm' 51241 Intel Knights Mill CPU. 51242 51243 'amdfam10h' 51244 AMD Family 10h CPU. 51245 51246 'barcelona' 51247 AMD Family 10h Barcelona CPU. 51248 51249 'shanghai' 51250 AMD Family 10h Shanghai CPU. 51251 51252 'istanbul' 51253 AMD Family 10h Istanbul CPU. 51254 51255 'btver1' 51256 AMD Family 14h CPU. 51257 51258 'amdfam15h' 51259 AMD Family 15h CPU. 51260 51261 'bdver1' 51262 AMD Family 15h Bulldozer version 1. 51263 51264 'bdver2' 51265 AMD Family 15h Bulldozer version 2. 51266 51267 'bdver3' 51268 AMD Family 15h Bulldozer version 3. 51269 51270 'bdver4' 51271 AMD Family 15h Bulldozer version 4. 51272 51273 'btver2' 51274 AMD Family 16h CPU. 51275 51276 'amdfam17h' 51277 AMD Family 17h CPU. 51278 51279 'znver1' 51280 AMD Family 17h Zen version 1. 51281 51282 'znver2' 51283 AMD Family 17h Zen version 2. 51284 51285 'amdfam19h' 51286 AMD Family 19h CPU. 51287 51288 'znver3' 51289 AMD Family 19h Zen version 3. 51290 51291 Here is an example: 51292 if (__builtin_cpu_is ("corei7")) 51293 { 51294 do_corei7 (); // Core i7 specific implementation. 51295 } 51296 else 51297 { 51298 do_generic (); // Generic implementation. 51299 } 51300 51301 -- Built-in Function: int __builtin_cpu_supports (const char *FEATURE) 51302 This function returns a positive integer if the run-time CPU 51303 supports FEATURE and returns '0' otherwise. The following features 51304 can be detected: 51305 51306 'cmov' 51307 CMOV instruction. 51308 'mmx' 51309 MMX instructions. 51310 'popcnt' 51311 POPCNT instruction. 51312 'sse' 51313 SSE instructions. 51314 'sse2' 51315 SSE2 instructions. 51316 'sse3' 51317 SSE3 instructions. 51318 'ssse3' 51319 SSSE3 instructions. 51320 'sse4.1' 51321 SSE4.1 instructions. 51322 'sse4.2' 51323 SSE4.2 instructions. 51324 'avx' 51325 AVX instructions. 51326 'avx2' 51327 AVX2 instructions. 51328 'sse4a' 51329 SSE4A instructions. 51330 'fma4' 51331 FMA4 instructions. 51332 'xop' 51333 XOP instructions. 51334 'fma' 51335 FMA instructions. 51336 'avx512f' 51337 AVX512F instructions. 51338 'bmi' 51339 BMI instructions. 51340 'bmi2' 51341 BMI2 instructions. 51342 'aes' 51343 AES instructions. 51344 'pclmul' 51345 PCLMUL instructions. 51346 'avx512vl' 51347 AVX512VL instructions. 51348 'avx512bw' 51349 AVX512BW instructions. 51350 'avx512dq' 51351 AVX512DQ instructions. 51352 'avx512cd' 51353 AVX512CD instructions. 51354 'avx512er' 51355 AVX512ER instructions. 51356 'avx512pf' 51357 AVX512PF instructions. 51358 'avx512vbmi' 51359 AVX512VBMI instructions. 51360 'avx512ifma' 51361 AVX512IFMA instructions. 51362 'avx5124vnniw' 51363 AVX5124VNNIW instructions. 51364 'avx5124fmaps' 51365 AVX5124FMAPS instructions. 51366 'avx512vpopcntdq' 51367 AVX512VPOPCNTDQ instructions. 51368 'avx512vbmi2' 51369 AVX512VBMI2 instructions. 51370 'gfni' 51371 GFNI instructions. 51372 'vpclmulqdq' 51373 VPCLMULQDQ instructions. 51374 'avx512vnni' 51375 AVX512VNNI instructions. 51376 'avx512bitalg' 51377 AVX512BITALG instructions. 51378 51379 Here is an example: 51380 if (__builtin_cpu_supports ("popcnt")) 51381 { 51382 asm("popcnt %1,%0" : "=r"(count) : "rm"(n) : "cc"); 51383 } 51384 else 51385 { 51386 count = generic_countbits (n); //generic implementation. 51387 } 51388 51389 The following built-in functions are made available by '-mmmx'. All of 51390them generate the machine instruction that is part of the name. 51391 51392 v8qi __builtin_ia32_paddb (v8qi, v8qi) 51393 v4hi __builtin_ia32_paddw (v4hi, v4hi) 51394 v2si __builtin_ia32_paddd (v2si, v2si) 51395 v8qi __builtin_ia32_psubb (v8qi, v8qi) 51396 v4hi __builtin_ia32_psubw (v4hi, v4hi) 51397 v2si __builtin_ia32_psubd (v2si, v2si) 51398 v8qi __builtin_ia32_paddsb (v8qi, v8qi) 51399 v4hi __builtin_ia32_paddsw (v4hi, v4hi) 51400 v8qi __builtin_ia32_psubsb (v8qi, v8qi) 51401 v4hi __builtin_ia32_psubsw (v4hi, v4hi) 51402 v8qi __builtin_ia32_paddusb (v8qi, v8qi) 51403 v4hi __builtin_ia32_paddusw (v4hi, v4hi) 51404 v8qi __builtin_ia32_psubusb (v8qi, v8qi) 51405 v4hi __builtin_ia32_psubusw (v4hi, v4hi) 51406 v4hi __builtin_ia32_pmullw (v4hi, v4hi) 51407 v4hi __builtin_ia32_pmulhw (v4hi, v4hi) 51408 di __builtin_ia32_pand (di, di) 51409 di __builtin_ia32_pandn (di,di) 51410 di __builtin_ia32_por (di, di) 51411 di __builtin_ia32_pxor (di, di) 51412 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi) 51413 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi) 51414 v2si __builtin_ia32_pcmpeqd (v2si, v2si) 51415 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi) 51416 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi) 51417 v2si __builtin_ia32_pcmpgtd (v2si, v2si) 51418 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi) 51419 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi) 51420 v2si __builtin_ia32_punpckhdq (v2si, v2si) 51421 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi) 51422 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi) 51423 v2si __builtin_ia32_punpckldq (v2si, v2si) 51424 v8qi __builtin_ia32_packsswb (v4hi, v4hi) 51425 v4hi __builtin_ia32_packssdw (v2si, v2si) 51426 v8qi __builtin_ia32_packuswb (v4hi, v4hi) 51427 51428 v4hi __builtin_ia32_psllw (v4hi, v4hi) 51429 v2si __builtin_ia32_pslld (v2si, v2si) 51430 v1di __builtin_ia32_psllq (v1di, v1di) 51431 v4hi __builtin_ia32_psrlw (v4hi, v4hi) 51432 v2si __builtin_ia32_psrld (v2si, v2si) 51433 v1di __builtin_ia32_psrlq (v1di, v1di) 51434 v4hi __builtin_ia32_psraw (v4hi, v4hi) 51435 v2si __builtin_ia32_psrad (v2si, v2si) 51436 v4hi __builtin_ia32_psllwi (v4hi, int) 51437 v2si __builtin_ia32_pslldi (v2si, int) 51438 v1di __builtin_ia32_psllqi (v1di, int) 51439 v4hi __builtin_ia32_psrlwi (v4hi, int) 51440 v2si __builtin_ia32_psrldi (v2si, int) 51441 v1di __builtin_ia32_psrlqi (v1di, int) 51442 v4hi __builtin_ia32_psrawi (v4hi, int) 51443 v2si __builtin_ia32_psradi (v2si, int) 51444 51445 51446 The following built-in functions are made available either with 51447'-msse', or with '-m3dnowa'. All of them generate the machine 51448instruction that is part of the name. 51449 51450 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi) 51451 v8qi __builtin_ia32_pavgb (v8qi, v8qi) 51452 v4hi __builtin_ia32_pavgw (v4hi, v4hi) 51453 v1di __builtin_ia32_psadbw (v8qi, v8qi) 51454 v8qi __builtin_ia32_pmaxub (v8qi, v8qi) 51455 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi) 51456 v8qi __builtin_ia32_pminub (v8qi, v8qi) 51457 v4hi __builtin_ia32_pminsw (v4hi, v4hi) 51458 int __builtin_ia32_pmovmskb (v8qi) 51459 void __builtin_ia32_maskmovq (v8qi, v8qi, char *) 51460 void __builtin_ia32_movntq (di *, di) 51461 void __builtin_ia32_sfence (void) 51462 51463 The following built-in functions are available when '-msse' is used. 51464All of them generate the machine instruction that is part of the name. 51465 51466 int __builtin_ia32_comieq (v4sf, v4sf) 51467 int __builtin_ia32_comineq (v4sf, v4sf) 51468 int __builtin_ia32_comilt (v4sf, v4sf) 51469 int __builtin_ia32_comile (v4sf, v4sf) 51470 int __builtin_ia32_comigt (v4sf, v4sf) 51471 int __builtin_ia32_comige (v4sf, v4sf) 51472 int __builtin_ia32_ucomieq (v4sf, v4sf) 51473 int __builtin_ia32_ucomineq (v4sf, v4sf) 51474 int __builtin_ia32_ucomilt (v4sf, v4sf) 51475 int __builtin_ia32_ucomile (v4sf, v4sf) 51476 int __builtin_ia32_ucomigt (v4sf, v4sf) 51477 int __builtin_ia32_ucomige (v4sf, v4sf) 51478 v4sf __builtin_ia32_addps (v4sf, v4sf) 51479 v4sf __builtin_ia32_subps (v4sf, v4sf) 51480 v4sf __builtin_ia32_mulps (v4sf, v4sf) 51481 v4sf __builtin_ia32_divps (v4sf, v4sf) 51482 v4sf __builtin_ia32_addss (v4sf, v4sf) 51483 v4sf __builtin_ia32_subss (v4sf, v4sf) 51484 v4sf __builtin_ia32_mulss (v4sf, v4sf) 51485 v4sf __builtin_ia32_divss (v4sf, v4sf) 51486 v4sf __builtin_ia32_cmpeqps (v4sf, v4sf) 51487 v4sf __builtin_ia32_cmpltps (v4sf, v4sf) 51488 v4sf __builtin_ia32_cmpleps (v4sf, v4sf) 51489 v4sf __builtin_ia32_cmpgtps (v4sf, v4sf) 51490 v4sf __builtin_ia32_cmpgeps (v4sf, v4sf) 51491 v4sf __builtin_ia32_cmpunordps (v4sf, v4sf) 51492 v4sf __builtin_ia32_cmpneqps (v4sf, v4sf) 51493 v4sf __builtin_ia32_cmpnltps (v4sf, v4sf) 51494 v4sf __builtin_ia32_cmpnleps (v4sf, v4sf) 51495 v4sf __builtin_ia32_cmpngtps (v4sf, v4sf) 51496 v4sf __builtin_ia32_cmpngeps (v4sf, v4sf) 51497 v4sf __builtin_ia32_cmpordps (v4sf, v4sf) 51498 v4sf __builtin_ia32_cmpeqss (v4sf, v4sf) 51499 v4sf __builtin_ia32_cmpltss (v4sf, v4sf) 51500 v4sf __builtin_ia32_cmpless (v4sf, v4sf) 51501 v4sf __builtin_ia32_cmpunordss (v4sf, v4sf) 51502 v4sf __builtin_ia32_cmpneqss (v4sf, v4sf) 51503 v4sf __builtin_ia32_cmpnltss (v4sf, v4sf) 51504 v4sf __builtin_ia32_cmpnless (v4sf, v4sf) 51505 v4sf __builtin_ia32_cmpordss (v4sf, v4sf) 51506 v4sf __builtin_ia32_maxps (v4sf, v4sf) 51507 v4sf __builtin_ia32_maxss (v4sf, v4sf) 51508 v4sf __builtin_ia32_minps (v4sf, v4sf) 51509 v4sf __builtin_ia32_minss (v4sf, v4sf) 51510 v4sf __builtin_ia32_andps (v4sf, v4sf) 51511 v4sf __builtin_ia32_andnps (v4sf, v4sf) 51512 v4sf __builtin_ia32_orps (v4sf, v4sf) 51513 v4sf __builtin_ia32_xorps (v4sf, v4sf) 51514 v4sf __builtin_ia32_movss (v4sf, v4sf) 51515 v4sf __builtin_ia32_movhlps (v4sf, v4sf) 51516 v4sf __builtin_ia32_movlhps (v4sf, v4sf) 51517 v4sf __builtin_ia32_unpckhps (v4sf, v4sf) 51518 v4sf __builtin_ia32_unpcklps (v4sf, v4sf) 51519 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si) 51520 v4sf __builtin_ia32_cvtsi2ss (v4sf, int) 51521 v2si __builtin_ia32_cvtps2pi (v4sf) 51522 int __builtin_ia32_cvtss2si (v4sf) 51523 v2si __builtin_ia32_cvttps2pi (v4sf) 51524 int __builtin_ia32_cvttss2si (v4sf) 51525 v4sf __builtin_ia32_rcpps (v4sf) 51526 v4sf __builtin_ia32_rsqrtps (v4sf) 51527 v4sf __builtin_ia32_sqrtps (v4sf) 51528 v4sf __builtin_ia32_rcpss (v4sf) 51529 v4sf __builtin_ia32_rsqrtss (v4sf) 51530 v4sf __builtin_ia32_sqrtss (v4sf) 51531 v4sf __builtin_ia32_shufps (v4sf, v4sf, int) 51532 void __builtin_ia32_movntps (float *, v4sf) 51533 int __builtin_ia32_movmskps (v4sf) 51534 51535 The following built-in functions are available when '-msse' is used. 51536 51537'v4sf __builtin_ia32_loadups (float *)' 51538 Generates the 'movups' machine instruction as a load from memory. 51539'void __builtin_ia32_storeups (float *, v4sf)' 51540 Generates the 'movups' machine instruction as a store to memory. 51541'v4sf __builtin_ia32_loadss (float *)' 51542 Generates the 'movss' machine instruction as a load from memory. 51543'v4sf __builtin_ia32_loadhps (v4sf, const v2sf *)' 51544 Generates the 'movhps' machine instruction as a load from memory. 51545'v4sf __builtin_ia32_loadlps (v4sf, const v2sf *)' 51546 Generates the 'movlps' machine instruction as a load from memory 51547'void __builtin_ia32_storehps (v2sf *, v4sf)' 51548 Generates the 'movhps' machine instruction as a store to memory. 51549'void __builtin_ia32_storelps (v2sf *, v4sf)' 51550 Generates the 'movlps' machine instruction as a store to memory. 51551 51552 The following built-in functions are available when '-msse2' is used. 51553All of them generate the machine instruction that is part of the name. 51554 51555 int __builtin_ia32_comisdeq (v2df, v2df) 51556 int __builtin_ia32_comisdlt (v2df, v2df) 51557 int __builtin_ia32_comisdle (v2df, v2df) 51558 int __builtin_ia32_comisdgt (v2df, v2df) 51559 int __builtin_ia32_comisdge (v2df, v2df) 51560 int __builtin_ia32_comisdneq (v2df, v2df) 51561 int __builtin_ia32_ucomisdeq (v2df, v2df) 51562 int __builtin_ia32_ucomisdlt (v2df, v2df) 51563 int __builtin_ia32_ucomisdle (v2df, v2df) 51564 int __builtin_ia32_ucomisdgt (v2df, v2df) 51565 int __builtin_ia32_ucomisdge (v2df, v2df) 51566 int __builtin_ia32_ucomisdneq (v2df, v2df) 51567 v2df __builtin_ia32_cmpeqpd (v2df, v2df) 51568 v2df __builtin_ia32_cmpltpd (v2df, v2df) 51569 v2df __builtin_ia32_cmplepd (v2df, v2df) 51570 v2df __builtin_ia32_cmpgtpd (v2df, v2df) 51571 v2df __builtin_ia32_cmpgepd (v2df, v2df) 51572 v2df __builtin_ia32_cmpunordpd (v2df, v2df) 51573 v2df __builtin_ia32_cmpneqpd (v2df, v2df) 51574 v2df __builtin_ia32_cmpnltpd (v2df, v2df) 51575 v2df __builtin_ia32_cmpnlepd (v2df, v2df) 51576 v2df __builtin_ia32_cmpngtpd (v2df, v2df) 51577 v2df __builtin_ia32_cmpngepd (v2df, v2df) 51578 v2df __builtin_ia32_cmpordpd (v2df, v2df) 51579 v2df __builtin_ia32_cmpeqsd (v2df, v2df) 51580 v2df __builtin_ia32_cmpltsd (v2df, v2df) 51581 v2df __builtin_ia32_cmplesd (v2df, v2df) 51582 v2df __builtin_ia32_cmpunordsd (v2df, v2df) 51583 v2df __builtin_ia32_cmpneqsd (v2df, v2df) 51584 v2df __builtin_ia32_cmpnltsd (v2df, v2df) 51585 v2df __builtin_ia32_cmpnlesd (v2df, v2df) 51586 v2df __builtin_ia32_cmpordsd (v2df, v2df) 51587 v2di __builtin_ia32_paddq (v2di, v2di) 51588 v2di __builtin_ia32_psubq (v2di, v2di) 51589 v2df __builtin_ia32_addpd (v2df, v2df) 51590 v2df __builtin_ia32_subpd (v2df, v2df) 51591 v2df __builtin_ia32_mulpd (v2df, v2df) 51592 v2df __builtin_ia32_divpd (v2df, v2df) 51593 v2df __builtin_ia32_addsd (v2df, v2df) 51594 v2df __builtin_ia32_subsd (v2df, v2df) 51595 v2df __builtin_ia32_mulsd (v2df, v2df) 51596 v2df __builtin_ia32_divsd (v2df, v2df) 51597 v2df __builtin_ia32_minpd (v2df, v2df) 51598 v2df __builtin_ia32_maxpd (v2df, v2df) 51599 v2df __builtin_ia32_minsd (v2df, v2df) 51600 v2df __builtin_ia32_maxsd (v2df, v2df) 51601 v2df __builtin_ia32_andpd (v2df, v2df) 51602 v2df __builtin_ia32_andnpd (v2df, v2df) 51603 v2df __builtin_ia32_orpd (v2df, v2df) 51604 v2df __builtin_ia32_xorpd (v2df, v2df) 51605 v2df __builtin_ia32_movsd (v2df, v2df) 51606 v2df __builtin_ia32_unpckhpd (v2df, v2df) 51607 v2df __builtin_ia32_unpcklpd (v2df, v2df) 51608 v16qi __builtin_ia32_paddb128 (v16qi, v16qi) 51609 v8hi __builtin_ia32_paddw128 (v8hi, v8hi) 51610 v4si __builtin_ia32_paddd128 (v4si, v4si) 51611 v2di __builtin_ia32_paddq128 (v2di, v2di) 51612 v16qi __builtin_ia32_psubb128 (v16qi, v16qi) 51613 v8hi __builtin_ia32_psubw128 (v8hi, v8hi) 51614 v4si __builtin_ia32_psubd128 (v4si, v4si) 51615 v2di __builtin_ia32_psubq128 (v2di, v2di) 51616 v8hi __builtin_ia32_pmullw128 (v8hi, v8hi) 51617 v8hi __builtin_ia32_pmulhw128 (v8hi, v8hi) 51618 v2di __builtin_ia32_pand128 (v2di, v2di) 51619 v2di __builtin_ia32_pandn128 (v2di, v2di) 51620 v2di __builtin_ia32_por128 (v2di, v2di) 51621 v2di __builtin_ia32_pxor128 (v2di, v2di) 51622 v16qi __builtin_ia32_pavgb128 (v16qi, v16qi) 51623 v8hi __builtin_ia32_pavgw128 (v8hi, v8hi) 51624 v16qi __builtin_ia32_pcmpeqb128 (v16qi, v16qi) 51625 v8hi __builtin_ia32_pcmpeqw128 (v8hi, v8hi) 51626 v4si __builtin_ia32_pcmpeqd128 (v4si, v4si) 51627 v16qi __builtin_ia32_pcmpgtb128 (v16qi, v16qi) 51628 v8hi __builtin_ia32_pcmpgtw128 (v8hi, v8hi) 51629 v4si __builtin_ia32_pcmpgtd128 (v4si, v4si) 51630 v16qi __builtin_ia32_pmaxub128 (v16qi, v16qi) 51631 v8hi __builtin_ia32_pmaxsw128 (v8hi, v8hi) 51632 v16qi __builtin_ia32_pminub128 (v16qi, v16qi) 51633 v8hi __builtin_ia32_pminsw128 (v8hi, v8hi) 51634 v16qi __builtin_ia32_punpckhbw128 (v16qi, v16qi) 51635 v8hi __builtin_ia32_punpckhwd128 (v8hi, v8hi) 51636 v4si __builtin_ia32_punpckhdq128 (v4si, v4si) 51637 v2di __builtin_ia32_punpckhqdq128 (v2di, v2di) 51638 v16qi __builtin_ia32_punpcklbw128 (v16qi, v16qi) 51639 v8hi __builtin_ia32_punpcklwd128 (v8hi, v8hi) 51640 v4si __builtin_ia32_punpckldq128 (v4si, v4si) 51641 v2di __builtin_ia32_punpcklqdq128 (v2di, v2di) 51642 v16qi __builtin_ia32_packsswb128 (v8hi, v8hi) 51643 v8hi __builtin_ia32_packssdw128 (v4si, v4si) 51644 v16qi __builtin_ia32_packuswb128 (v8hi, v8hi) 51645 v8hi __builtin_ia32_pmulhuw128 (v8hi, v8hi) 51646 void __builtin_ia32_maskmovdqu (v16qi, v16qi) 51647 v2df __builtin_ia32_loadupd (double *) 51648 void __builtin_ia32_storeupd (double *, v2df) 51649 v2df __builtin_ia32_loadhpd (v2df, double const *) 51650 v2df __builtin_ia32_loadlpd (v2df, double const *) 51651 int __builtin_ia32_movmskpd (v2df) 51652 int __builtin_ia32_pmovmskb128 (v16qi) 51653 void __builtin_ia32_movnti (int *, int) 51654 void __builtin_ia32_movnti64 (long long int *, long long int) 51655 void __builtin_ia32_movntpd (double *, v2df) 51656 void __builtin_ia32_movntdq (v2df *, v2df) 51657 v4si __builtin_ia32_pshufd (v4si, int) 51658 v8hi __builtin_ia32_pshuflw (v8hi, int) 51659 v8hi __builtin_ia32_pshufhw (v8hi, int) 51660 v2di __builtin_ia32_psadbw128 (v16qi, v16qi) 51661 v2df __builtin_ia32_sqrtpd (v2df) 51662 v2df __builtin_ia32_sqrtsd (v2df) 51663 v2df __builtin_ia32_shufpd (v2df, v2df, int) 51664 v2df __builtin_ia32_cvtdq2pd (v4si) 51665 v4sf __builtin_ia32_cvtdq2ps (v4si) 51666 v4si __builtin_ia32_cvtpd2dq (v2df) 51667 v2si __builtin_ia32_cvtpd2pi (v2df) 51668 v4sf __builtin_ia32_cvtpd2ps (v2df) 51669 v4si __builtin_ia32_cvttpd2dq (v2df) 51670 v2si __builtin_ia32_cvttpd2pi (v2df) 51671 v2df __builtin_ia32_cvtpi2pd (v2si) 51672 int __builtin_ia32_cvtsd2si (v2df) 51673 int __builtin_ia32_cvttsd2si (v2df) 51674 long long __builtin_ia32_cvtsd2si64 (v2df) 51675 long long __builtin_ia32_cvttsd2si64 (v2df) 51676 v4si __builtin_ia32_cvtps2dq (v4sf) 51677 v2df __builtin_ia32_cvtps2pd (v4sf) 51678 v4si __builtin_ia32_cvttps2dq (v4sf) 51679 v2df __builtin_ia32_cvtsi2sd (v2df, int) 51680 v2df __builtin_ia32_cvtsi642sd (v2df, long long) 51681 v4sf __builtin_ia32_cvtsd2ss (v4sf, v2df) 51682 v2df __builtin_ia32_cvtss2sd (v2df, v4sf) 51683 void __builtin_ia32_clflush (const void *) 51684 void __builtin_ia32_lfence (void) 51685 void __builtin_ia32_mfence (void) 51686 v16qi __builtin_ia32_loaddqu (const char *) 51687 void __builtin_ia32_storedqu (char *, v16qi) 51688 v1di __builtin_ia32_pmuludq (v2si, v2si) 51689 v2di __builtin_ia32_pmuludq128 (v4si, v4si) 51690 v8hi __builtin_ia32_psllw128 (v8hi, v8hi) 51691 v4si __builtin_ia32_pslld128 (v4si, v4si) 51692 v2di __builtin_ia32_psllq128 (v2di, v2di) 51693 v8hi __builtin_ia32_psrlw128 (v8hi, v8hi) 51694 v4si __builtin_ia32_psrld128 (v4si, v4si) 51695 v2di __builtin_ia32_psrlq128 (v2di, v2di) 51696 v8hi __builtin_ia32_psraw128 (v8hi, v8hi) 51697 v4si __builtin_ia32_psrad128 (v4si, v4si) 51698 v2di __builtin_ia32_pslldqi128 (v2di, int) 51699 v8hi __builtin_ia32_psllwi128 (v8hi, int) 51700 v4si __builtin_ia32_pslldi128 (v4si, int) 51701 v2di __builtin_ia32_psllqi128 (v2di, int) 51702 v2di __builtin_ia32_psrldqi128 (v2di, int) 51703 v8hi __builtin_ia32_psrlwi128 (v8hi, int) 51704 v4si __builtin_ia32_psrldi128 (v4si, int) 51705 v2di __builtin_ia32_psrlqi128 (v2di, int) 51706 v8hi __builtin_ia32_psrawi128 (v8hi, int) 51707 v4si __builtin_ia32_psradi128 (v4si, int) 51708 v4si __builtin_ia32_pmaddwd128 (v8hi, v8hi) 51709 v2di __builtin_ia32_movq128 (v2di) 51710 51711 The following built-in functions are available when '-msse3' is used. 51712All of them generate the machine instruction that is part of the name. 51713 51714 v2df __builtin_ia32_addsubpd (v2df, v2df) 51715 v4sf __builtin_ia32_addsubps (v4sf, v4sf) 51716 v2df __builtin_ia32_haddpd (v2df, v2df) 51717 v4sf __builtin_ia32_haddps (v4sf, v4sf) 51718 v2df __builtin_ia32_hsubpd (v2df, v2df) 51719 v4sf __builtin_ia32_hsubps (v4sf, v4sf) 51720 v16qi __builtin_ia32_lddqu (char const *) 51721 void __builtin_ia32_monitor (void *, unsigned int, unsigned int) 51722 v4sf __builtin_ia32_movshdup (v4sf) 51723 v4sf __builtin_ia32_movsldup (v4sf) 51724 void __builtin_ia32_mwait (unsigned int, unsigned int) 51725 51726 The following built-in functions are available when '-mssse3' is used. 51727All of them generate the machine instruction that is part of the name. 51728 51729 v2si __builtin_ia32_phaddd (v2si, v2si) 51730 v4hi __builtin_ia32_phaddw (v4hi, v4hi) 51731 v4hi __builtin_ia32_phaddsw (v4hi, v4hi) 51732 v2si __builtin_ia32_phsubd (v2si, v2si) 51733 v4hi __builtin_ia32_phsubw (v4hi, v4hi) 51734 v4hi __builtin_ia32_phsubsw (v4hi, v4hi) 51735 v4hi __builtin_ia32_pmaddubsw (v8qi, v8qi) 51736 v4hi __builtin_ia32_pmulhrsw (v4hi, v4hi) 51737 v8qi __builtin_ia32_pshufb (v8qi, v8qi) 51738 v8qi __builtin_ia32_psignb (v8qi, v8qi) 51739 v2si __builtin_ia32_psignd (v2si, v2si) 51740 v4hi __builtin_ia32_psignw (v4hi, v4hi) 51741 v1di __builtin_ia32_palignr (v1di, v1di, int) 51742 v8qi __builtin_ia32_pabsb (v8qi) 51743 v2si __builtin_ia32_pabsd (v2si) 51744 v4hi __builtin_ia32_pabsw (v4hi) 51745 51746 The following built-in functions are available when '-mssse3' is used. 51747All of them generate the machine instruction that is part of the name. 51748 51749 v4si __builtin_ia32_phaddd128 (v4si, v4si) 51750 v8hi __builtin_ia32_phaddw128 (v8hi, v8hi) 51751 v8hi __builtin_ia32_phaddsw128 (v8hi, v8hi) 51752 v4si __builtin_ia32_phsubd128 (v4si, v4si) 51753 v8hi __builtin_ia32_phsubw128 (v8hi, v8hi) 51754 v8hi __builtin_ia32_phsubsw128 (v8hi, v8hi) 51755 v8hi __builtin_ia32_pmaddubsw128 (v16qi, v16qi) 51756 v8hi __builtin_ia32_pmulhrsw128 (v8hi, v8hi) 51757 v16qi __builtin_ia32_pshufb128 (v16qi, v16qi) 51758 v16qi __builtin_ia32_psignb128 (v16qi, v16qi) 51759 v4si __builtin_ia32_psignd128 (v4si, v4si) 51760 v8hi __builtin_ia32_psignw128 (v8hi, v8hi) 51761 v2di __builtin_ia32_palignr128 (v2di, v2di, int) 51762 v16qi __builtin_ia32_pabsb128 (v16qi) 51763 v4si __builtin_ia32_pabsd128 (v4si) 51764 v8hi __builtin_ia32_pabsw128 (v8hi) 51765 51766 The following built-in functions are available when '-msse4.1' is used. 51767All of them generate the machine instruction that is part of the name. 51768 51769 v2df __builtin_ia32_blendpd (v2df, v2df, const int) 51770 v4sf __builtin_ia32_blendps (v4sf, v4sf, const int) 51771 v2df __builtin_ia32_blendvpd (v2df, v2df, v2df) 51772 v4sf __builtin_ia32_blendvps (v4sf, v4sf, v4sf) 51773 v2df __builtin_ia32_dppd (v2df, v2df, const int) 51774 v4sf __builtin_ia32_dpps (v4sf, v4sf, const int) 51775 v4sf __builtin_ia32_insertps128 (v4sf, v4sf, const int) 51776 v2di __builtin_ia32_movntdqa (v2di *); 51777 v16qi __builtin_ia32_mpsadbw128 (v16qi, v16qi, const int) 51778 v8hi __builtin_ia32_packusdw128 (v4si, v4si) 51779 v16qi __builtin_ia32_pblendvb128 (v16qi, v16qi, v16qi) 51780 v8hi __builtin_ia32_pblendw128 (v8hi, v8hi, const int) 51781 v2di __builtin_ia32_pcmpeqq (v2di, v2di) 51782 v8hi __builtin_ia32_phminposuw128 (v8hi) 51783 v16qi __builtin_ia32_pmaxsb128 (v16qi, v16qi) 51784 v4si __builtin_ia32_pmaxsd128 (v4si, v4si) 51785 v4si __builtin_ia32_pmaxud128 (v4si, v4si) 51786 v8hi __builtin_ia32_pmaxuw128 (v8hi, v8hi) 51787 v16qi __builtin_ia32_pminsb128 (v16qi, v16qi) 51788 v4si __builtin_ia32_pminsd128 (v4si, v4si) 51789 v4si __builtin_ia32_pminud128 (v4si, v4si) 51790 v8hi __builtin_ia32_pminuw128 (v8hi, v8hi) 51791 v4si __builtin_ia32_pmovsxbd128 (v16qi) 51792 v2di __builtin_ia32_pmovsxbq128 (v16qi) 51793 v8hi __builtin_ia32_pmovsxbw128 (v16qi) 51794 v2di __builtin_ia32_pmovsxdq128 (v4si) 51795 v4si __builtin_ia32_pmovsxwd128 (v8hi) 51796 v2di __builtin_ia32_pmovsxwq128 (v8hi) 51797 v4si __builtin_ia32_pmovzxbd128 (v16qi) 51798 v2di __builtin_ia32_pmovzxbq128 (v16qi) 51799 v8hi __builtin_ia32_pmovzxbw128 (v16qi) 51800 v2di __builtin_ia32_pmovzxdq128 (v4si) 51801 v4si __builtin_ia32_pmovzxwd128 (v8hi) 51802 v2di __builtin_ia32_pmovzxwq128 (v8hi) 51803 v2di __builtin_ia32_pmuldq128 (v4si, v4si) 51804 v4si __builtin_ia32_pmulld128 (v4si, v4si) 51805 int __builtin_ia32_ptestc128 (v2di, v2di) 51806 int __builtin_ia32_ptestnzc128 (v2di, v2di) 51807 int __builtin_ia32_ptestz128 (v2di, v2di) 51808 v2df __builtin_ia32_roundpd (v2df, const int) 51809 v4sf __builtin_ia32_roundps (v4sf, const int) 51810 v2df __builtin_ia32_roundsd (v2df, v2df, const int) 51811 v4sf __builtin_ia32_roundss (v4sf, v4sf, const int) 51812 51813 The following built-in functions are available when '-msse4.1' is used. 51814 51815'v4sf __builtin_ia32_vec_set_v4sf (v4sf, float, const int)' 51816 Generates the 'insertps' machine instruction. 51817'int __builtin_ia32_vec_ext_v16qi (v16qi, const int)' 51818 Generates the 'pextrb' machine instruction. 51819'v16qi __builtin_ia32_vec_set_v16qi (v16qi, int, const int)' 51820 Generates the 'pinsrb' machine instruction. 51821'v4si __builtin_ia32_vec_set_v4si (v4si, int, const int)' 51822 Generates the 'pinsrd' machine instruction. 51823'v2di __builtin_ia32_vec_set_v2di (v2di, long long, const int)' 51824 Generates the 'pinsrq' machine instruction in 64bit mode. 51825 51826 The following built-in functions are changed to generate new SSE4.1 51827instructions when '-msse4.1' is used. 51828 51829'float __builtin_ia32_vec_ext_v4sf (v4sf, const int)' 51830 Generates the 'extractps' machine instruction. 51831'int __builtin_ia32_vec_ext_v4si (v4si, const int)' 51832 Generates the 'pextrd' machine instruction. 51833'long long __builtin_ia32_vec_ext_v2di (v2di, const int)' 51834 Generates the 'pextrq' machine instruction in 64bit mode. 51835 51836 The following built-in functions are available when '-msse4.2' is used. 51837All of them generate the machine instruction that is part of the name. 51838 51839 v16qi __builtin_ia32_pcmpestrm128 (v16qi, int, v16qi, int, const int) 51840 int __builtin_ia32_pcmpestri128 (v16qi, int, v16qi, int, const int) 51841 int __builtin_ia32_pcmpestria128 (v16qi, int, v16qi, int, const int) 51842 int __builtin_ia32_pcmpestric128 (v16qi, int, v16qi, int, const int) 51843 int __builtin_ia32_pcmpestrio128 (v16qi, int, v16qi, int, const int) 51844 int __builtin_ia32_pcmpestris128 (v16qi, int, v16qi, int, const int) 51845 int __builtin_ia32_pcmpestriz128 (v16qi, int, v16qi, int, const int) 51846 v16qi __builtin_ia32_pcmpistrm128 (v16qi, v16qi, const int) 51847 int __builtin_ia32_pcmpistri128 (v16qi, v16qi, const int) 51848 int __builtin_ia32_pcmpistria128 (v16qi, v16qi, const int) 51849 int __builtin_ia32_pcmpistric128 (v16qi, v16qi, const int) 51850 int __builtin_ia32_pcmpistrio128 (v16qi, v16qi, const int) 51851 int __builtin_ia32_pcmpistris128 (v16qi, v16qi, const int) 51852 int __builtin_ia32_pcmpistriz128 (v16qi, v16qi, const int) 51853 v2di __builtin_ia32_pcmpgtq (v2di, v2di) 51854 51855 The following built-in functions are available when '-msse4.2' is used. 51856 51857'unsigned int __builtin_ia32_crc32qi (unsigned int, unsigned char)' 51858 Generates the 'crc32b' machine instruction. 51859'unsigned int __builtin_ia32_crc32hi (unsigned int, unsigned short)' 51860 Generates the 'crc32w' machine instruction. 51861'unsigned int __builtin_ia32_crc32si (unsigned int, unsigned int)' 51862 Generates the 'crc32l' machine instruction. 51863'unsigned long long __builtin_ia32_crc32di (unsigned long long, unsigned long long)' 51864 Generates the 'crc32q' machine instruction. 51865 51866 The following built-in functions are changed to generate new SSE4.2 51867instructions when '-msse4.2' is used. 51868 51869'int __builtin_popcount (unsigned int)' 51870 Generates the 'popcntl' machine instruction. 51871'int __builtin_popcountl (unsigned long)' 51872 Generates the 'popcntl' or 'popcntq' machine instruction, depending 51873 on the size of 'unsigned long'. 51874'int __builtin_popcountll (unsigned long long)' 51875 Generates the 'popcntq' machine instruction. 51876 51877 The following built-in functions are available when '-mavx' is used. 51878All of them generate the machine instruction that is part of the name. 51879 51880 v4df __builtin_ia32_addpd256 (v4df,v4df) 51881 v8sf __builtin_ia32_addps256 (v8sf,v8sf) 51882 v4df __builtin_ia32_addsubpd256 (v4df,v4df) 51883 v8sf __builtin_ia32_addsubps256 (v8sf,v8sf) 51884 v4df __builtin_ia32_andnpd256 (v4df,v4df) 51885 v8sf __builtin_ia32_andnps256 (v8sf,v8sf) 51886 v4df __builtin_ia32_andpd256 (v4df,v4df) 51887 v8sf __builtin_ia32_andps256 (v8sf,v8sf) 51888 v4df __builtin_ia32_blendpd256 (v4df,v4df,int) 51889 v8sf __builtin_ia32_blendps256 (v8sf,v8sf,int) 51890 v4df __builtin_ia32_blendvpd256 (v4df,v4df,v4df) 51891 v8sf __builtin_ia32_blendvps256 (v8sf,v8sf,v8sf) 51892 v2df __builtin_ia32_cmppd (v2df,v2df,int) 51893 v4df __builtin_ia32_cmppd256 (v4df,v4df,int) 51894 v4sf __builtin_ia32_cmpps (v4sf,v4sf,int) 51895 v8sf __builtin_ia32_cmpps256 (v8sf,v8sf,int) 51896 v2df __builtin_ia32_cmpsd (v2df,v2df,int) 51897 v4sf __builtin_ia32_cmpss (v4sf,v4sf,int) 51898 v4df __builtin_ia32_cvtdq2pd256 (v4si) 51899 v8sf __builtin_ia32_cvtdq2ps256 (v8si) 51900 v4si __builtin_ia32_cvtpd2dq256 (v4df) 51901 v4sf __builtin_ia32_cvtpd2ps256 (v4df) 51902 v8si __builtin_ia32_cvtps2dq256 (v8sf) 51903 v4df __builtin_ia32_cvtps2pd256 (v4sf) 51904 v4si __builtin_ia32_cvttpd2dq256 (v4df) 51905 v8si __builtin_ia32_cvttps2dq256 (v8sf) 51906 v4df __builtin_ia32_divpd256 (v4df,v4df) 51907 v8sf __builtin_ia32_divps256 (v8sf,v8sf) 51908 v8sf __builtin_ia32_dpps256 (v8sf,v8sf,int) 51909 v4df __builtin_ia32_haddpd256 (v4df,v4df) 51910 v8sf __builtin_ia32_haddps256 (v8sf,v8sf) 51911 v4df __builtin_ia32_hsubpd256 (v4df,v4df) 51912 v8sf __builtin_ia32_hsubps256 (v8sf,v8sf) 51913 v32qi __builtin_ia32_lddqu256 (pcchar) 51914 v32qi __builtin_ia32_loaddqu256 (pcchar) 51915 v4df __builtin_ia32_loadupd256 (pcdouble) 51916 v8sf __builtin_ia32_loadups256 (pcfloat) 51917 v2df __builtin_ia32_maskloadpd (pcv2df,v2df) 51918 v4df __builtin_ia32_maskloadpd256 (pcv4df,v4df) 51919 v4sf __builtin_ia32_maskloadps (pcv4sf,v4sf) 51920 v8sf __builtin_ia32_maskloadps256 (pcv8sf,v8sf) 51921 void __builtin_ia32_maskstorepd (pv2df,v2df,v2df) 51922 void __builtin_ia32_maskstorepd256 (pv4df,v4df,v4df) 51923 void __builtin_ia32_maskstoreps (pv4sf,v4sf,v4sf) 51924 void __builtin_ia32_maskstoreps256 (pv8sf,v8sf,v8sf) 51925 v4df __builtin_ia32_maxpd256 (v4df,v4df) 51926 v8sf __builtin_ia32_maxps256 (v8sf,v8sf) 51927 v4df __builtin_ia32_minpd256 (v4df,v4df) 51928 v8sf __builtin_ia32_minps256 (v8sf,v8sf) 51929 v4df __builtin_ia32_movddup256 (v4df) 51930 int __builtin_ia32_movmskpd256 (v4df) 51931 int __builtin_ia32_movmskps256 (v8sf) 51932 v8sf __builtin_ia32_movshdup256 (v8sf) 51933 v8sf __builtin_ia32_movsldup256 (v8sf) 51934 v4df __builtin_ia32_mulpd256 (v4df,v4df) 51935 v8sf __builtin_ia32_mulps256 (v8sf,v8sf) 51936 v4df __builtin_ia32_orpd256 (v4df,v4df) 51937 v8sf __builtin_ia32_orps256 (v8sf,v8sf) 51938 v2df __builtin_ia32_pd_pd256 (v4df) 51939 v4df __builtin_ia32_pd256_pd (v2df) 51940 v4sf __builtin_ia32_ps_ps256 (v8sf) 51941 v8sf __builtin_ia32_ps256_ps (v4sf) 51942 int __builtin_ia32_ptestc256 (v4di,v4di,ptest) 51943 int __builtin_ia32_ptestnzc256 (v4di,v4di,ptest) 51944 int __builtin_ia32_ptestz256 (v4di,v4di,ptest) 51945 v8sf __builtin_ia32_rcpps256 (v8sf) 51946 v4df __builtin_ia32_roundpd256 (v4df,int) 51947 v8sf __builtin_ia32_roundps256 (v8sf,int) 51948 v8sf __builtin_ia32_rsqrtps_nr256 (v8sf) 51949 v8sf __builtin_ia32_rsqrtps256 (v8sf) 51950 v4df __builtin_ia32_shufpd256 (v4df,v4df,int) 51951 v8sf __builtin_ia32_shufps256 (v8sf,v8sf,int) 51952 v4si __builtin_ia32_si_si256 (v8si) 51953 v8si __builtin_ia32_si256_si (v4si) 51954 v4df __builtin_ia32_sqrtpd256 (v4df) 51955 v8sf __builtin_ia32_sqrtps_nr256 (v8sf) 51956 v8sf __builtin_ia32_sqrtps256 (v8sf) 51957 void __builtin_ia32_storedqu256 (pchar,v32qi) 51958 void __builtin_ia32_storeupd256 (pdouble,v4df) 51959 void __builtin_ia32_storeups256 (pfloat,v8sf) 51960 v4df __builtin_ia32_subpd256 (v4df,v4df) 51961 v8sf __builtin_ia32_subps256 (v8sf,v8sf) 51962 v4df __builtin_ia32_unpckhpd256 (v4df,v4df) 51963 v8sf __builtin_ia32_unpckhps256 (v8sf,v8sf) 51964 v4df __builtin_ia32_unpcklpd256 (v4df,v4df) 51965 v8sf __builtin_ia32_unpcklps256 (v8sf,v8sf) 51966 v4df __builtin_ia32_vbroadcastf128_pd256 (pcv2df) 51967 v8sf __builtin_ia32_vbroadcastf128_ps256 (pcv4sf) 51968 v4df __builtin_ia32_vbroadcastsd256 (pcdouble) 51969 v4sf __builtin_ia32_vbroadcastss (pcfloat) 51970 v8sf __builtin_ia32_vbroadcastss256 (pcfloat) 51971 v2df __builtin_ia32_vextractf128_pd256 (v4df,int) 51972 v4sf __builtin_ia32_vextractf128_ps256 (v8sf,int) 51973 v4si __builtin_ia32_vextractf128_si256 (v8si,int) 51974 v4df __builtin_ia32_vinsertf128_pd256 (v4df,v2df,int) 51975 v8sf __builtin_ia32_vinsertf128_ps256 (v8sf,v4sf,int) 51976 v8si __builtin_ia32_vinsertf128_si256 (v8si,v4si,int) 51977 v4df __builtin_ia32_vperm2f128_pd256 (v4df,v4df,int) 51978 v8sf __builtin_ia32_vperm2f128_ps256 (v8sf,v8sf,int) 51979 v8si __builtin_ia32_vperm2f128_si256 (v8si,v8si,int) 51980 v2df __builtin_ia32_vpermil2pd (v2df,v2df,v2di,int) 51981 v4df __builtin_ia32_vpermil2pd256 (v4df,v4df,v4di,int) 51982 v4sf __builtin_ia32_vpermil2ps (v4sf,v4sf,v4si,int) 51983 v8sf __builtin_ia32_vpermil2ps256 (v8sf,v8sf,v8si,int) 51984 v2df __builtin_ia32_vpermilpd (v2df,int) 51985 v4df __builtin_ia32_vpermilpd256 (v4df,int) 51986 v4sf __builtin_ia32_vpermilps (v4sf,int) 51987 v8sf __builtin_ia32_vpermilps256 (v8sf,int) 51988 v2df __builtin_ia32_vpermilvarpd (v2df,v2di) 51989 v4df __builtin_ia32_vpermilvarpd256 (v4df,v4di) 51990 v4sf __builtin_ia32_vpermilvarps (v4sf,v4si) 51991 v8sf __builtin_ia32_vpermilvarps256 (v8sf,v8si) 51992 int __builtin_ia32_vtestcpd (v2df,v2df,ptest) 51993 int __builtin_ia32_vtestcpd256 (v4df,v4df,ptest) 51994 int __builtin_ia32_vtestcps (v4sf,v4sf,ptest) 51995 int __builtin_ia32_vtestcps256 (v8sf,v8sf,ptest) 51996 int __builtin_ia32_vtestnzcpd (v2df,v2df,ptest) 51997 int __builtin_ia32_vtestnzcpd256 (v4df,v4df,ptest) 51998 int __builtin_ia32_vtestnzcps (v4sf,v4sf,ptest) 51999 int __builtin_ia32_vtestnzcps256 (v8sf,v8sf,ptest) 52000 int __builtin_ia32_vtestzpd (v2df,v2df,ptest) 52001 int __builtin_ia32_vtestzpd256 (v4df,v4df,ptest) 52002 int __builtin_ia32_vtestzps (v4sf,v4sf,ptest) 52003 int __builtin_ia32_vtestzps256 (v8sf,v8sf,ptest) 52004 void __builtin_ia32_vzeroall (void) 52005 void __builtin_ia32_vzeroupper (void) 52006 v4df __builtin_ia32_xorpd256 (v4df,v4df) 52007 v8sf __builtin_ia32_xorps256 (v8sf,v8sf) 52008 52009 The following built-in functions are available when '-mavx2' is used. 52010All of them generate the machine instruction that is part of the name. 52011 52012 v32qi __builtin_ia32_mpsadbw256 (v32qi,v32qi,int) 52013 v32qi __builtin_ia32_pabsb256 (v32qi) 52014 v16hi __builtin_ia32_pabsw256 (v16hi) 52015 v8si __builtin_ia32_pabsd256 (v8si) 52016 v16hi __builtin_ia32_packssdw256 (v8si,v8si) 52017 v32qi __builtin_ia32_packsswb256 (v16hi,v16hi) 52018 v16hi __builtin_ia32_packusdw256 (v8si,v8si) 52019 v32qi __builtin_ia32_packuswb256 (v16hi,v16hi) 52020 v32qi __builtin_ia32_paddb256 (v32qi,v32qi) 52021 v16hi __builtin_ia32_paddw256 (v16hi,v16hi) 52022 v8si __builtin_ia32_paddd256 (v8si,v8si) 52023 v4di __builtin_ia32_paddq256 (v4di,v4di) 52024 v32qi __builtin_ia32_paddsb256 (v32qi,v32qi) 52025 v16hi __builtin_ia32_paddsw256 (v16hi,v16hi) 52026 v32qi __builtin_ia32_paddusb256 (v32qi,v32qi) 52027 v16hi __builtin_ia32_paddusw256 (v16hi,v16hi) 52028 v4di __builtin_ia32_palignr256 (v4di,v4di,int) 52029 v4di __builtin_ia32_andsi256 (v4di,v4di) 52030 v4di __builtin_ia32_andnotsi256 (v4di,v4di) 52031 v32qi __builtin_ia32_pavgb256 (v32qi,v32qi) 52032 v16hi __builtin_ia32_pavgw256 (v16hi,v16hi) 52033 v32qi __builtin_ia32_pblendvb256 (v32qi,v32qi,v32qi) 52034 v16hi __builtin_ia32_pblendw256 (v16hi,v16hi,int) 52035 v32qi __builtin_ia32_pcmpeqb256 (v32qi,v32qi) 52036 v16hi __builtin_ia32_pcmpeqw256 (v16hi,v16hi) 52037 v8si __builtin_ia32_pcmpeqd256 (c8si,v8si) 52038 v4di __builtin_ia32_pcmpeqq256 (v4di,v4di) 52039 v32qi __builtin_ia32_pcmpgtb256 (v32qi,v32qi) 52040 v16hi __builtin_ia32_pcmpgtw256 (16hi,v16hi) 52041 v8si __builtin_ia32_pcmpgtd256 (v8si,v8si) 52042 v4di __builtin_ia32_pcmpgtq256 (v4di,v4di) 52043 v16hi __builtin_ia32_phaddw256 (v16hi,v16hi) 52044 v8si __builtin_ia32_phaddd256 (v8si,v8si) 52045 v16hi __builtin_ia32_phaddsw256 (v16hi,v16hi) 52046 v16hi __builtin_ia32_phsubw256 (v16hi,v16hi) 52047 v8si __builtin_ia32_phsubd256 (v8si,v8si) 52048 v16hi __builtin_ia32_phsubsw256 (v16hi,v16hi) 52049 v32qi __builtin_ia32_pmaddubsw256 (v32qi,v32qi) 52050 v16hi __builtin_ia32_pmaddwd256 (v16hi,v16hi) 52051 v32qi __builtin_ia32_pmaxsb256 (v32qi,v32qi) 52052 v16hi __builtin_ia32_pmaxsw256 (v16hi,v16hi) 52053 v8si __builtin_ia32_pmaxsd256 (v8si,v8si) 52054 v32qi __builtin_ia32_pmaxub256 (v32qi,v32qi) 52055 v16hi __builtin_ia32_pmaxuw256 (v16hi,v16hi) 52056 v8si __builtin_ia32_pmaxud256 (v8si,v8si) 52057 v32qi __builtin_ia32_pminsb256 (v32qi,v32qi) 52058 v16hi __builtin_ia32_pminsw256 (v16hi,v16hi) 52059 v8si __builtin_ia32_pminsd256 (v8si,v8si) 52060 v32qi __builtin_ia32_pminub256 (v32qi,v32qi) 52061 v16hi __builtin_ia32_pminuw256 (v16hi,v16hi) 52062 v8si __builtin_ia32_pminud256 (v8si,v8si) 52063 int __builtin_ia32_pmovmskb256 (v32qi) 52064 v16hi __builtin_ia32_pmovsxbw256 (v16qi) 52065 v8si __builtin_ia32_pmovsxbd256 (v16qi) 52066 v4di __builtin_ia32_pmovsxbq256 (v16qi) 52067 v8si __builtin_ia32_pmovsxwd256 (v8hi) 52068 v4di __builtin_ia32_pmovsxwq256 (v8hi) 52069 v4di __builtin_ia32_pmovsxdq256 (v4si) 52070 v16hi __builtin_ia32_pmovzxbw256 (v16qi) 52071 v8si __builtin_ia32_pmovzxbd256 (v16qi) 52072 v4di __builtin_ia32_pmovzxbq256 (v16qi) 52073 v8si __builtin_ia32_pmovzxwd256 (v8hi) 52074 v4di __builtin_ia32_pmovzxwq256 (v8hi) 52075 v4di __builtin_ia32_pmovzxdq256 (v4si) 52076 v4di __builtin_ia32_pmuldq256 (v8si,v8si) 52077 v16hi __builtin_ia32_pmulhrsw256 (v16hi, v16hi) 52078 v16hi __builtin_ia32_pmulhuw256 (v16hi,v16hi) 52079 v16hi __builtin_ia32_pmulhw256 (v16hi,v16hi) 52080 v16hi __builtin_ia32_pmullw256 (v16hi,v16hi) 52081 v8si __builtin_ia32_pmulld256 (v8si,v8si) 52082 v4di __builtin_ia32_pmuludq256 (v8si,v8si) 52083 v4di __builtin_ia32_por256 (v4di,v4di) 52084 v16hi __builtin_ia32_psadbw256 (v32qi,v32qi) 52085 v32qi __builtin_ia32_pshufb256 (v32qi,v32qi) 52086 v8si __builtin_ia32_pshufd256 (v8si,int) 52087 v16hi __builtin_ia32_pshufhw256 (v16hi,int) 52088 v16hi __builtin_ia32_pshuflw256 (v16hi,int) 52089 v32qi __builtin_ia32_psignb256 (v32qi,v32qi) 52090 v16hi __builtin_ia32_psignw256 (v16hi,v16hi) 52091 v8si __builtin_ia32_psignd256 (v8si,v8si) 52092 v4di __builtin_ia32_pslldqi256 (v4di,int) 52093 v16hi __builtin_ia32_psllwi256 (16hi,int) 52094 v16hi __builtin_ia32_psllw256(v16hi,v8hi) 52095 v8si __builtin_ia32_pslldi256 (v8si,int) 52096 v8si __builtin_ia32_pslld256(v8si,v4si) 52097 v4di __builtin_ia32_psllqi256 (v4di,int) 52098 v4di __builtin_ia32_psllq256(v4di,v2di) 52099 v16hi __builtin_ia32_psrawi256 (v16hi,int) 52100 v16hi __builtin_ia32_psraw256 (v16hi,v8hi) 52101 v8si __builtin_ia32_psradi256 (v8si,int) 52102 v8si __builtin_ia32_psrad256 (v8si,v4si) 52103 v4di __builtin_ia32_psrldqi256 (v4di, int) 52104 v16hi __builtin_ia32_psrlwi256 (v16hi,int) 52105 v16hi __builtin_ia32_psrlw256 (v16hi,v8hi) 52106 v8si __builtin_ia32_psrldi256 (v8si,int) 52107 v8si __builtin_ia32_psrld256 (v8si,v4si) 52108 v4di __builtin_ia32_psrlqi256 (v4di,int) 52109 v4di __builtin_ia32_psrlq256(v4di,v2di) 52110 v32qi __builtin_ia32_psubb256 (v32qi,v32qi) 52111 v32hi __builtin_ia32_psubw256 (v16hi,v16hi) 52112 v8si __builtin_ia32_psubd256 (v8si,v8si) 52113 v4di __builtin_ia32_psubq256 (v4di,v4di) 52114 v32qi __builtin_ia32_psubsb256 (v32qi,v32qi) 52115 v16hi __builtin_ia32_psubsw256 (v16hi,v16hi) 52116 v32qi __builtin_ia32_psubusb256 (v32qi,v32qi) 52117 v16hi __builtin_ia32_psubusw256 (v16hi,v16hi) 52118 v32qi __builtin_ia32_punpckhbw256 (v32qi,v32qi) 52119 v16hi __builtin_ia32_punpckhwd256 (v16hi,v16hi) 52120 v8si __builtin_ia32_punpckhdq256 (v8si,v8si) 52121 v4di __builtin_ia32_punpckhqdq256 (v4di,v4di) 52122 v32qi __builtin_ia32_punpcklbw256 (v32qi,v32qi) 52123 v16hi __builtin_ia32_punpcklwd256 (v16hi,v16hi) 52124 v8si __builtin_ia32_punpckldq256 (v8si,v8si) 52125 v4di __builtin_ia32_punpcklqdq256 (v4di,v4di) 52126 v4di __builtin_ia32_pxor256 (v4di,v4di) 52127 v4di __builtin_ia32_movntdqa256 (pv4di) 52128 v4sf __builtin_ia32_vbroadcastss_ps (v4sf) 52129 v8sf __builtin_ia32_vbroadcastss_ps256 (v4sf) 52130 v4df __builtin_ia32_vbroadcastsd_pd256 (v2df) 52131 v4di __builtin_ia32_vbroadcastsi256 (v2di) 52132 v4si __builtin_ia32_pblendd128 (v4si,v4si) 52133 v8si __builtin_ia32_pblendd256 (v8si,v8si) 52134 v32qi __builtin_ia32_pbroadcastb256 (v16qi) 52135 v16hi __builtin_ia32_pbroadcastw256 (v8hi) 52136 v8si __builtin_ia32_pbroadcastd256 (v4si) 52137 v4di __builtin_ia32_pbroadcastq256 (v2di) 52138 v16qi __builtin_ia32_pbroadcastb128 (v16qi) 52139 v8hi __builtin_ia32_pbroadcastw128 (v8hi) 52140 v4si __builtin_ia32_pbroadcastd128 (v4si) 52141 v2di __builtin_ia32_pbroadcastq128 (v2di) 52142 v8si __builtin_ia32_permvarsi256 (v8si,v8si) 52143 v4df __builtin_ia32_permdf256 (v4df,int) 52144 v8sf __builtin_ia32_permvarsf256 (v8sf,v8sf) 52145 v4di __builtin_ia32_permdi256 (v4di,int) 52146 v4di __builtin_ia32_permti256 (v4di,v4di,int) 52147 v4di __builtin_ia32_extract128i256 (v4di,int) 52148 v4di __builtin_ia32_insert128i256 (v4di,v2di,int) 52149 v8si __builtin_ia32_maskloadd256 (pcv8si,v8si) 52150 v4di __builtin_ia32_maskloadq256 (pcv4di,v4di) 52151 v4si __builtin_ia32_maskloadd (pcv4si,v4si) 52152 v2di __builtin_ia32_maskloadq (pcv2di,v2di) 52153 void __builtin_ia32_maskstored256 (pv8si,v8si,v8si) 52154 void __builtin_ia32_maskstoreq256 (pv4di,v4di,v4di) 52155 void __builtin_ia32_maskstored (pv4si,v4si,v4si) 52156 void __builtin_ia32_maskstoreq (pv2di,v2di,v2di) 52157 v8si __builtin_ia32_psllv8si (v8si,v8si) 52158 v4si __builtin_ia32_psllv4si (v4si,v4si) 52159 v4di __builtin_ia32_psllv4di (v4di,v4di) 52160 v2di __builtin_ia32_psllv2di (v2di,v2di) 52161 v8si __builtin_ia32_psrav8si (v8si,v8si) 52162 v4si __builtin_ia32_psrav4si (v4si,v4si) 52163 v8si __builtin_ia32_psrlv8si (v8si,v8si) 52164 v4si __builtin_ia32_psrlv4si (v4si,v4si) 52165 v4di __builtin_ia32_psrlv4di (v4di,v4di) 52166 v2di __builtin_ia32_psrlv2di (v2di,v2di) 52167 v2df __builtin_ia32_gathersiv2df (v2df, pcdouble,v4si,v2df,int) 52168 v4df __builtin_ia32_gathersiv4df (v4df, pcdouble,v4si,v4df,int) 52169 v2df __builtin_ia32_gatherdiv2df (v2df, pcdouble,v2di,v2df,int) 52170 v4df __builtin_ia32_gatherdiv4df (v4df, pcdouble,v4di,v4df,int) 52171 v4sf __builtin_ia32_gathersiv4sf (v4sf, pcfloat,v4si,v4sf,int) 52172 v8sf __builtin_ia32_gathersiv8sf (v8sf, pcfloat,v8si,v8sf,int) 52173 v4sf __builtin_ia32_gatherdiv4sf (v4sf, pcfloat,v2di,v4sf,int) 52174 v4sf __builtin_ia32_gatherdiv4sf256 (v4sf, pcfloat,v4di,v4sf,int) 52175 v2di __builtin_ia32_gathersiv2di (v2di, pcint64,v4si,v2di,int) 52176 v4di __builtin_ia32_gathersiv4di (v4di, pcint64,v4si,v4di,int) 52177 v2di __builtin_ia32_gatherdiv2di (v2di, pcint64,v2di,v2di,int) 52178 v4di __builtin_ia32_gatherdiv4di (v4di, pcint64,v4di,v4di,int) 52179 v4si __builtin_ia32_gathersiv4si (v4si, pcint,v4si,v4si,int) 52180 v8si __builtin_ia32_gathersiv8si (v8si, pcint,v8si,v8si,int) 52181 v4si __builtin_ia32_gatherdiv4si (v4si, pcint,v2di,v4si,int) 52182 v4si __builtin_ia32_gatherdiv4si256 (v4si, pcint,v4di,v4si,int) 52183 52184 The following built-in functions are available when '-maes' is used. 52185All of them generate the machine instruction that is part of the name. 52186 52187 v2di __builtin_ia32_aesenc128 (v2di, v2di) 52188 v2di __builtin_ia32_aesenclast128 (v2di, v2di) 52189 v2di __builtin_ia32_aesdec128 (v2di, v2di) 52190 v2di __builtin_ia32_aesdeclast128 (v2di, v2di) 52191 v2di __builtin_ia32_aeskeygenassist128 (v2di, const int) 52192 v2di __builtin_ia32_aesimc128 (v2di) 52193 52194 The following built-in function is available when '-mpclmul' is used. 52195 52196'v2di __builtin_ia32_pclmulqdq128 (v2di, v2di, const int)' 52197 Generates the 'pclmulqdq' machine instruction. 52198 52199 The following built-in function is available when '-mfsgsbase' is used. 52200All of them generate the machine instruction that is part of the name. 52201 52202 unsigned int __builtin_ia32_rdfsbase32 (void) 52203 unsigned long long __builtin_ia32_rdfsbase64 (void) 52204 unsigned int __builtin_ia32_rdgsbase32 (void) 52205 unsigned long long __builtin_ia32_rdgsbase64 (void) 52206 void _writefsbase_u32 (unsigned int) 52207 void _writefsbase_u64 (unsigned long long) 52208 void _writegsbase_u32 (unsigned int) 52209 void _writegsbase_u64 (unsigned long long) 52210 52211 The following built-in function is available when '-mrdrnd' is used. 52212All of them generate the machine instruction that is part of the name. 52213 52214 unsigned int __builtin_ia32_rdrand16_step (unsigned short *) 52215 unsigned int __builtin_ia32_rdrand32_step (unsigned int *) 52216 unsigned int __builtin_ia32_rdrand64_step (unsigned long long *) 52217 52218 The following built-in function is available when '-mptwrite' is used. 52219All of them generate the machine instruction that is part of the name. 52220 52221 void __builtin_ia32_ptwrite32 (unsigned) 52222 void __builtin_ia32_ptwrite64 (unsigned long long) 52223 52224 The following built-in functions are available when '-msse4a' is used. 52225All of them generate the machine instruction that is part of the name. 52226 52227 void __builtin_ia32_movntsd (double *, v2df) 52228 void __builtin_ia32_movntss (float *, v4sf) 52229 v2di __builtin_ia32_extrq (v2di, v16qi) 52230 v2di __builtin_ia32_extrqi (v2di, const unsigned int, const unsigned int) 52231 v2di __builtin_ia32_insertq (v2di, v2di) 52232 v2di __builtin_ia32_insertqi (v2di, v2di, const unsigned int, const unsigned int) 52233 52234 The following built-in functions are available when '-mxop' is used. 52235 v2df __builtin_ia32_vfrczpd (v2df) 52236 v4sf __builtin_ia32_vfrczps (v4sf) 52237 v2df __builtin_ia32_vfrczsd (v2df) 52238 v4sf __builtin_ia32_vfrczss (v4sf) 52239 v4df __builtin_ia32_vfrczpd256 (v4df) 52240 v8sf __builtin_ia32_vfrczps256 (v8sf) 52241 v2di __builtin_ia32_vpcmov (v2di, v2di, v2di) 52242 v2di __builtin_ia32_vpcmov_v2di (v2di, v2di, v2di) 52243 v4si __builtin_ia32_vpcmov_v4si (v4si, v4si, v4si) 52244 v8hi __builtin_ia32_vpcmov_v8hi (v8hi, v8hi, v8hi) 52245 v16qi __builtin_ia32_vpcmov_v16qi (v16qi, v16qi, v16qi) 52246 v2df __builtin_ia32_vpcmov_v2df (v2df, v2df, v2df) 52247 v4sf __builtin_ia32_vpcmov_v4sf (v4sf, v4sf, v4sf) 52248 v4di __builtin_ia32_vpcmov_v4di256 (v4di, v4di, v4di) 52249 v8si __builtin_ia32_vpcmov_v8si256 (v8si, v8si, v8si) 52250 v16hi __builtin_ia32_vpcmov_v16hi256 (v16hi, v16hi, v16hi) 52251 v32qi __builtin_ia32_vpcmov_v32qi256 (v32qi, v32qi, v32qi) 52252 v4df __builtin_ia32_vpcmov_v4df256 (v4df, v4df, v4df) 52253 v8sf __builtin_ia32_vpcmov_v8sf256 (v8sf, v8sf, v8sf) 52254 v16qi __builtin_ia32_vpcomeqb (v16qi, v16qi) 52255 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi) 52256 v4si __builtin_ia32_vpcomeqd (v4si, v4si) 52257 v2di __builtin_ia32_vpcomeqq (v2di, v2di) 52258 v16qi __builtin_ia32_vpcomequb (v16qi, v16qi) 52259 v4si __builtin_ia32_vpcomequd (v4si, v4si) 52260 v2di __builtin_ia32_vpcomequq (v2di, v2di) 52261 v8hi __builtin_ia32_vpcomequw (v8hi, v8hi) 52262 v8hi __builtin_ia32_vpcomeqw (v8hi, v8hi) 52263 v16qi __builtin_ia32_vpcomfalseb (v16qi, v16qi) 52264 v4si __builtin_ia32_vpcomfalsed (v4si, v4si) 52265 v2di __builtin_ia32_vpcomfalseq (v2di, v2di) 52266 v16qi __builtin_ia32_vpcomfalseub (v16qi, v16qi) 52267 v4si __builtin_ia32_vpcomfalseud (v4si, v4si) 52268 v2di __builtin_ia32_vpcomfalseuq (v2di, v2di) 52269 v8hi __builtin_ia32_vpcomfalseuw (v8hi, v8hi) 52270 v8hi __builtin_ia32_vpcomfalsew (v8hi, v8hi) 52271 v16qi __builtin_ia32_vpcomgeb (v16qi, v16qi) 52272 v4si __builtin_ia32_vpcomged (v4si, v4si) 52273 v2di __builtin_ia32_vpcomgeq (v2di, v2di) 52274 v16qi __builtin_ia32_vpcomgeub (v16qi, v16qi) 52275 v4si __builtin_ia32_vpcomgeud (v4si, v4si) 52276 v2di __builtin_ia32_vpcomgeuq (v2di, v2di) 52277 v8hi __builtin_ia32_vpcomgeuw (v8hi, v8hi) 52278 v8hi __builtin_ia32_vpcomgew (v8hi, v8hi) 52279 v16qi __builtin_ia32_vpcomgtb (v16qi, v16qi) 52280 v4si __builtin_ia32_vpcomgtd (v4si, v4si) 52281 v2di __builtin_ia32_vpcomgtq (v2di, v2di) 52282 v16qi __builtin_ia32_vpcomgtub (v16qi, v16qi) 52283 v4si __builtin_ia32_vpcomgtud (v4si, v4si) 52284 v2di __builtin_ia32_vpcomgtuq (v2di, v2di) 52285 v8hi __builtin_ia32_vpcomgtuw (v8hi, v8hi) 52286 v8hi __builtin_ia32_vpcomgtw (v8hi, v8hi) 52287 v16qi __builtin_ia32_vpcomleb (v16qi, v16qi) 52288 v4si __builtin_ia32_vpcomled (v4si, v4si) 52289 v2di __builtin_ia32_vpcomleq (v2di, v2di) 52290 v16qi __builtin_ia32_vpcomleub (v16qi, v16qi) 52291 v4si __builtin_ia32_vpcomleud (v4si, v4si) 52292 v2di __builtin_ia32_vpcomleuq (v2di, v2di) 52293 v8hi __builtin_ia32_vpcomleuw (v8hi, v8hi) 52294 v8hi __builtin_ia32_vpcomlew (v8hi, v8hi) 52295 v16qi __builtin_ia32_vpcomltb (v16qi, v16qi) 52296 v4si __builtin_ia32_vpcomltd (v4si, v4si) 52297 v2di __builtin_ia32_vpcomltq (v2di, v2di) 52298 v16qi __builtin_ia32_vpcomltub (v16qi, v16qi) 52299 v4si __builtin_ia32_vpcomltud (v4si, v4si) 52300 v2di __builtin_ia32_vpcomltuq (v2di, v2di) 52301 v8hi __builtin_ia32_vpcomltuw (v8hi, v8hi) 52302 v8hi __builtin_ia32_vpcomltw (v8hi, v8hi) 52303 v16qi __builtin_ia32_vpcomneb (v16qi, v16qi) 52304 v4si __builtin_ia32_vpcomned (v4si, v4si) 52305 v2di __builtin_ia32_vpcomneq (v2di, v2di) 52306 v16qi __builtin_ia32_vpcomneub (v16qi, v16qi) 52307 v4si __builtin_ia32_vpcomneud (v4si, v4si) 52308 v2di __builtin_ia32_vpcomneuq (v2di, v2di) 52309 v8hi __builtin_ia32_vpcomneuw (v8hi, v8hi) 52310 v8hi __builtin_ia32_vpcomnew (v8hi, v8hi) 52311 v16qi __builtin_ia32_vpcomtrueb (v16qi, v16qi) 52312 v4si __builtin_ia32_vpcomtrued (v4si, v4si) 52313 v2di __builtin_ia32_vpcomtrueq (v2di, v2di) 52314 v16qi __builtin_ia32_vpcomtrueub (v16qi, v16qi) 52315 v4si __builtin_ia32_vpcomtrueud (v4si, v4si) 52316 v2di __builtin_ia32_vpcomtrueuq (v2di, v2di) 52317 v8hi __builtin_ia32_vpcomtrueuw (v8hi, v8hi) 52318 v8hi __builtin_ia32_vpcomtruew (v8hi, v8hi) 52319 v4si __builtin_ia32_vphaddbd (v16qi) 52320 v2di __builtin_ia32_vphaddbq (v16qi) 52321 v8hi __builtin_ia32_vphaddbw (v16qi) 52322 v2di __builtin_ia32_vphadddq (v4si) 52323 v4si __builtin_ia32_vphaddubd (v16qi) 52324 v2di __builtin_ia32_vphaddubq (v16qi) 52325 v8hi __builtin_ia32_vphaddubw (v16qi) 52326 v2di __builtin_ia32_vphaddudq (v4si) 52327 v4si __builtin_ia32_vphadduwd (v8hi) 52328 v2di __builtin_ia32_vphadduwq (v8hi) 52329 v4si __builtin_ia32_vphaddwd (v8hi) 52330 v2di __builtin_ia32_vphaddwq (v8hi) 52331 v8hi __builtin_ia32_vphsubbw (v16qi) 52332 v2di __builtin_ia32_vphsubdq (v4si) 52333 v4si __builtin_ia32_vphsubwd (v8hi) 52334 v4si __builtin_ia32_vpmacsdd (v4si, v4si, v4si) 52335 v2di __builtin_ia32_vpmacsdqh (v4si, v4si, v2di) 52336 v2di __builtin_ia32_vpmacsdql (v4si, v4si, v2di) 52337 v4si __builtin_ia32_vpmacssdd (v4si, v4si, v4si) 52338 v2di __builtin_ia32_vpmacssdqh (v4si, v4si, v2di) 52339 v2di __builtin_ia32_vpmacssdql (v4si, v4si, v2di) 52340 v4si __builtin_ia32_vpmacsswd (v8hi, v8hi, v4si) 52341 v8hi __builtin_ia32_vpmacssww (v8hi, v8hi, v8hi) 52342 v4si __builtin_ia32_vpmacswd (v8hi, v8hi, v4si) 52343 v8hi __builtin_ia32_vpmacsww (v8hi, v8hi, v8hi) 52344 v4si __builtin_ia32_vpmadcsswd (v8hi, v8hi, v4si) 52345 v4si __builtin_ia32_vpmadcswd (v8hi, v8hi, v4si) 52346 v16qi __builtin_ia32_vpperm (v16qi, v16qi, v16qi) 52347 v16qi __builtin_ia32_vprotb (v16qi, v16qi) 52348 v4si __builtin_ia32_vprotd (v4si, v4si) 52349 v2di __builtin_ia32_vprotq (v2di, v2di) 52350 v8hi __builtin_ia32_vprotw (v8hi, v8hi) 52351 v16qi __builtin_ia32_vpshab (v16qi, v16qi) 52352 v4si __builtin_ia32_vpshad (v4si, v4si) 52353 v2di __builtin_ia32_vpshaq (v2di, v2di) 52354 v8hi __builtin_ia32_vpshaw (v8hi, v8hi) 52355 v16qi __builtin_ia32_vpshlb (v16qi, v16qi) 52356 v4si __builtin_ia32_vpshld (v4si, v4si) 52357 v2di __builtin_ia32_vpshlq (v2di, v2di) 52358 v8hi __builtin_ia32_vpshlw (v8hi, v8hi) 52359 52360 The following built-in functions are available when '-mfma4' is used. 52361All of them generate the machine instruction that is part of the name. 52362 52363 v2df __builtin_ia32_vfmaddpd (v2df, v2df, v2df) 52364 v4sf __builtin_ia32_vfmaddps (v4sf, v4sf, v4sf) 52365 v2df __builtin_ia32_vfmaddsd (v2df, v2df, v2df) 52366 v4sf __builtin_ia32_vfmaddss (v4sf, v4sf, v4sf) 52367 v2df __builtin_ia32_vfmsubpd (v2df, v2df, v2df) 52368 v4sf __builtin_ia32_vfmsubps (v4sf, v4sf, v4sf) 52369 v2df __builtin_ia32_vfmsubsd (v2df, v2df, v2df) 52370 v4sf __builtin_ia32_vfmsubss (v4sf, v4sf, v4sf) 52371 v2df __builtin_ia32_vfnmaddpd (v2df, v2df, v2df) 52372 v4sf __builtin_ia32_vfnmaddps (v4sf, v4sf, v4sf) 52373 v2df __builtin_ia32_vfnmaddsd (v2df, v2df, v2df) 52374 v4sf __builtin_ia32_vfnmaddss (v4sf, v4sf, v4sf) 52375 v2df __builtin_ia32_vfnmsubpd (v2df, v2df, v2df) 52376 v4sf __builtin_ia32_vfnmsubps (v4sf, v4sf, v4sf) 52377 v2df __builtin_ia32_vfnmsubsd (v2df, v2df, v2df) 52378 v4sf __builtin_ia32_vfnmsubss (v4sf, v4sf, v4sf) 52379 v2df __builtin_ia32_vfmaddsubpd (v2df, v2df, v2df) 52380 v4sf __builtin_ia32_vfmaddsubps (v4sf, v4sf, v4sf) 52381 v2df __builtin_ia32_vfmsubaddpd (v2df, v2df, v2df) 52382 v4sf __builtin_ia32_vfmsubaddps (v4sf, v4sf, v4sf) 52383 v4df __builtin_ia32_vfmaddpd256 (v4df, v4df, v4df) 52384 v8sf __builtin_ia32_vfmaddps256 (v8sf, v8sf, v8sf) 52385 v4df __builtin_ia32_vfmsubpd256 (v4df, v4df, v4df) 52386 v8sf __builtin_ia32_vfmsubps256 (v8sf, v8sf, v8sf) 52387 v4df __builtin_ia32_vfnmaddpd256 (v4df, v4df, v4df) 52388 v8sf __builtin_ia32_vfnmaddps256 (v8sf, v8sf, v8sf) 52389 v4df __builtin_ia32_vfnmsubpd256 (v4df, v4df, v4df) 52390 v8sf __builtin_ia32_vfnmsubps256 (v8sf, v8sf, v8sf) 52391 v4df __builtin_ia32_vfmaddsubpd256 (v4df, v4df, v4df) 52392 v8sf __builtin_ia32_vfmaddsubps256 (v8sf, v8sf, v8sf) 52393 v4df __builtin_ia32_vfmsubaddpd256 (v4df, v4df, v4df) 52394 v8sf __builtin_ia32_vfmsubaddps256 (v8sf, v8sf, v8sf) 52395 52396 52397 The following built-in functions are available when '-mlwp' is used. 52398 52399 void __builtin_ia32_llwpcb16 (void *); 52400 void __builtin_ia32_llwpcb32 (void *); 52401 void __builtin_ia32_llwpcb64 (void *); 52402 void * __builtin_ia32_llwpcb16 (void); 52403 void * __builtin_ia32_llwpcb32 (void); 52404 void * __builtin_ia32_llwpcb64 (void); 52405 void __builtin_ia32_lwpval16 (unsigned short, unsigned int, unsigned short) 52406 void __builtin_ia32_lwpval32 (unsigned int, unsigned int, unsigned int) 52407 void __builtin_ia32_lwpval64 (unsigned __int64, unsigned int, unsigned int) 52408 unsigned char __builtin_ia32_lwpins16 (unsigned short, unsigned int, unsigned short) 52409 unsigned char __builtin_ia32_lwpins32 (unsigned int, unsigned int, unsigned int) 52410 unsigned char __builtin_ia32_lwpins64 (unsigned __int64, unsigned int, unsigned int) 52411 52412 The following built-in functions are available when '-mbmi' is used. 52413All of them generate the machine instruction that is part of the name. 52414 unsigned int __builtin_ia32_bextr_u32(unsigned int, unsigned int); 52415 unsigned long long __builtin_ia32_bextr_u64 (unsigned long long, unsigned long long); 52416 52417 The following built-in functions are available when '-mbmi2' is used. 52418All of them generate the machine instruction that is part of the name. 52419 unsigned int _bzhi_u32 (unsigned int, unsigned int) 52420 unsigned int _pdep_u32 (unsigned int, unsigned int) 52421 unsigned int _pext_u32 (unsigned int, unsigned int) 52422 unsigned long long _bzhi_u64 (unsigned long long, unsigned long long) 52423 unsigned long long _pdep_u64 (unsigned long long, unsigned long long) 52424 unsigned long long _pext_u64 (unsigned long long, unsigned long long) 52425 52426 The following built-in functions are available when '-mlzcnt' is used. 52427All of them generate the machine instruction that is part of the name. 52428 unsigned short __builtin_ia32_lzcnt_u16(unsigned short); 52429 unsigned int __builtin_ia32_lzcnt_u32(unsigned int); 52430 unsigned long long __builtin_ia32_lzcnt_u64 (unsigned long long); 52431 52432 The following built-in functions are available when '-mfxsr' is used. 52433All of them generate the machine instruction that is part of the name. 52434 void __builtin_ia32_fxsave (void *) 52435 void __builtin_ia32_fxrstor (void *) 52436 void __builtin_ia32_fxsave64 (void *) 52437 void __builtin_ia32_fxrstor64 (void *) 52438 52439 The following built-in functions are available when '-mxsave' is used. 52440All of them generate the machine instruction that is part of the name. 52441 void __builtin_ia32_xsave (void *, long long) 52442 void __builtin_ia32_xrstor (void *, long long) 52443 void __builtin_ia32_xsave64 (void *, long long) 52444 void __builtin_ia32_xrstor64 (void *, long long) 52445 52446 The following built-in functions are available when '-mxsaveopt' is 52447used. All of them generate the machine instruction that is part of the 52448name. 52449 void __builtin_ia32_xsaveopt (void *, long long) 52450 void __builtin_ia32_xsaveopt64 (void *, long long) 52451 52452 The following built-in functions are available when '-mtbm' is used. 52453Both of them generate the immediate form of the bextr machine 52454instruction. 52455 unsigned int __builtin_ia32_bextri_u32 (unsigned int, 52456 const unsigned int); 52457 unsigned long long __builtin_ia32_bextri_u64 (unsigned long long, 52458 const unsigned long long); 52459 52460 The following built-in functions are available when '-m3dnow' is used. 52461All of them generate the machine instruction that is part of the name. 52462 52463 void __builtin_ia32_femms (void) 52464 v8qi __builtin_ia32_pavgusb (v8qi, v8qi) 52465 v2si __builtin_ia32_pf2id (v2sf) 52466 v2sf __builtin_ia32_pfacc (v2sf, v2sf) 52467 v2sf __builtin_ia32_pfadd (v2sf, v2sf) 52468 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf) 52469 v2si __builtin_ia32_pfcmpge (v2sf, v2sf) 52470 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf) 52471 v2sf __builtin_ia32_pfmax (v2sf, v2sf) 52472 v2sf __builtin_ia32_pfmin (v2sf, v2sf) 52473 v2sf __builtin_ia32_pfmul (v2sf, v2sf) 52474 v2sf __builtin_ia32_pfrcp (v2sf) 52475 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf) 52476 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf) 52477 v2sf __builtin_ia32_pfrsqrt (v2sf) 52478 v2sf __builtin_ia32_pfsub (v2sf, v2sf) 52479 v2sf __builtin_ia32_pfsubr (v2sf, v2sf) 52480 v2sf __builtin_ia32_pi2fd (v2si) 52481 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi) 52482 52483 The following built-in functions are available when '-m3dnowa' is used. 52484All of them generate the machine instruction that is part of the name. 52485 52486 v2si __builtin_ia32_pf2iw (v2sf) 52487 v2sf __builtin_ia32_pfnacc (v2sf, v2sf) 52488 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf) 52489 v2sf __builtin_ia32_pi2fw (v2si) 52490 v2sf __builtin_ia32_pswapdsf (v2sf) 52491 v2si __builtin_ia32_pswapdsi (v2si) 52492 52493 The following built-in functions are available when '-mrtm' is used 52494They are used for restricted transactional memory. These are the 52495internal low level functions. Normally the functions in *note x86 52496transactional memory intrinsics:: should be used instead. 52497 52498 int __builtin_ia32_xbegin () 52499 void __builtin_ia32_xend () 52500 void __builtin_ia32_xabort (status) 52501 int __builtin_ia32_xtest () 52502 52503 The following built-in functions are available when '-mmwaitx' is used. 52504All of them generate the machine instruction that is part of the name. 52505 void __builtin_ia32_monitorx (void *, unsigned int, unsigned int) 52506 void __builtin_ia32_mwaitx (unsigned int, unsigned int, unsigned int) 52507 52508 The following built-in functions are available when '-mclzero' is used. 52509All of them generate the machine instruction that is part of the name. 52510 void __builtin_i32_clzero (void *) 52511 52512 The following built-in functions are available when '-mpku' is used. 52513They generate reads and writes to PKRU. 52514 void __builtin_ia32_wrpkru (unsigned int) 52515 unsigned int __builtin_ia32_rdpkru () 52516 52517 The following built-in functions are available when '-mcet' or 52518'-mshstk' option is used. They support shadow stack machine 52519instructions from Intel Control-flow Enforcement Technology (CET). Each 52520built-in function generates the machine instruction that is part of the 52521function's name. These are the internal low-level functions. Normally 52522the functions in *note x86 control-flow protection intrinsics:: should 52523be used instead. 52524 52525 unsigned int __builtin_ia32_rdsspd (void) 52526 unsigned long long __builtin_ia32_rdsspq (void) 52527 void __builtin_ia32_incsspd (unsigned int) 52528 void __builtin_ia32_incsspq (unsigned long long) 52529 void __builtin_ia32_saveprevssp(void); 52530 void __builtin_ia32_rstorssp(void *); 52531 void __builtin_ia32_wrssd(unsigned int, void *); 52532 void __builtin_ia32_wrssq(unsigned long long, void *); 52533 void __builtin_ia32_wrussd(unsigned int, void *); 52534 void __builtin_ia32_wrussq(unsigned long long, void *); 52535 void __builtin_ia32_setssbsy(void); 52536 void __builtin_ia32_clrssbsy(void *); 52537 52538 52539File: gcc.info, Node: x86 transactional memory intrinsics, Next: x86 control-flow protection intrinsics, Prev: x86 Built-in Functions, Up: Target Builtins 52540 525416.60.36 x86 Transactional Memory Intrinsics 52542------------------------------------------- 52543 52544These hardware transactional memory intrinsics for x86 allow you to use 52545memory transactions with RTM (Restricted Transactional Memory). This 52546support is enabled with the '-mrtm' option. For using HLE (Hardware 52547Lock Elision) see *note x86 specific memory model extensions for 52548transactional memory:: instead. 52549 52550 A memory transaction commits all changes to memory in an atomic way, as 52551visible to other threads. If the transaction fails it is rolled back 52552and all side effects discarded. 52553 52554 Generally there is no guarantee that a memory transaction ever succeeds 52555and suitable fallback code always needs to be supplied. 52556 52557 -- RTM Function: unsigned _xbegin () 52558 Start a RTM (Restricted Transactional Memory) transaction. Returns 52559 '_XBEGIN_STARTED' when the transaction started successfully (note 52560 this is not 0, so the constant has to be explicitly tested). 52561 52562 If the transaction aborts, all side effects are undone and an abort 52563 code encoded as a bit mask is returned. The following macros are 52564 defined: 52565 52566 '_XABORT_EXPLICIT' 52567 Transaction was explicitly aborted with '_xabort'. The 52568 parameter passed to '_xabort' is available with 52569 '_XABORT_CODE(status)'. 52570 '_XABORT_RETRY' 52571 Transaction retry is possible. 52572 '_XABORT_CONFLICT' 52573 Transaction abort due to a memory conflict with another 52574 thread. 52575 '_XABORT_CAPACITY' 52576 Transaction abort due to the transaction using too much 52577 memory. 52578 '_XABORT_DEBUG' 52579 Transaction abort due to a debug trap. 52580 '_XABORT_NESTED' 52581 Transaction abort in an inner nested transaction. 52582 52583 There is no guarantee any transaction ever succeeds, so there 52584 always needs to be a valid fallback path. 52585 52586 -- RTM Function: void _xend () 52587 Commit the current transaction. When no transaction is active this 52588 faults. All memory side effects of the transaction become visible 52589 to other threads in an atomic manner. 52590 52591 -- RTM Function: int _xtest () 52592 Return a nonzero value if a transaction is currently active, 52593 otherwise 0. 52594 52595 -- RTM Function: void _xabort (status) 52596 Abort the current transaction. When no transaction is active this 52597 is a no-op. The STATUS is an 8-bit constant; its value is encoded 52598 in the return value from '_xbegin'. 52599 52600 Here is an example showing handling for '_XABORT_RETRY' and a fallback 52601path for other failures: 52602 52603 #include <immintrin.h> 52604 52605 int n_tries, max_tries; 52606 unsigned status = _XABORT_EXPLICIT; 52607 ... 52608 52609 for (n_tries = 0; n_tries < max_tries; n_tries++) 52610 { 52611 status = _xbegin (); 52612 if (status == _XBEGIN_STARTED || !(status & _XABORT_RETRY)) 52613 break; 52614 } 52615 if (status == _XBEGIN_STARTED) 52616 { 52617 ... transaction code... 52618 _xend (); 52619 } 52620 else 52621 { 52622 ... non-transactional fallback path... 52623 } 52624 52625Note that, in most cases, the transactional and non-transactional code 52626must synchronize together to ensure consistency. 52627 52628 52629File: gcc.info, Node: x86 control-flow protection intrinsics, Prev: x86 transactional memory intrinsics, Up: Target Builtins 52630 526316.60.37 x86 Control-Flow Protection Intrinsics 52632---------------------------------------------- 52633 52634 -- CET Function: ret_type _get_ssp (void) 52635 Get the current value of shadow stack pointer if shadow stack 52636 support from Intel CET is enabled in the hardware or '0' otherwise. 52637 The 'ret_type' is 'unsigned long long' for 64-bit targets and 52638 'unsigned int' for 32-bit targets. 52639 52640 -- CET Function: void _inc_ssp (unsigned int) 52641 Increment the current shadow stack pointer by the size specified by 52642 the function argument. The argument is masked to a byte value for 52643 security reasons, so to increment by more than 255 bytes you must 52644 call the function multiple times. 52645 52646 The shadow stack unwind code looks like: 52647 52648 #include <immintrin.h> 52649 52650 /* Unwind the shadow stack for EH. */ 52651 #define _Unwind_Frames_Extra(x) \ 52652 do \ 52653 { \ 52654 _Unwind_Word ssp = _get_ssp (); \ 52655 if (ssp != 0) \ 52656 { \ 52657 _Unwind_Word tmp = (x); \ 52658 while (tmp > 255) \ 52659 { \ 52660 _inc_ssp (tmp); \ 52661 tmp -= 255; \ 52662 } \ 52663 _inc_ssp (tmp); \ 52664 } \ 52665 } \ 52666 while (0) 52667 52668This code runs unconditionally on all 64-bit processors. For 32-bit 52669processors the code runs on those that support multi-byte NOP 52670instructions. 52671 52672 52673File: gcc.info, Node: Target Format Checks, Next: Pragmas, Prev: Target Builtins, Up: C Extensions 52674 526756.61 Format Checks Specific to Particular Target Machines 52676========================================================= 52677 52678For some target machines, GCC supports additional options to the format 52679attribute (*note Declaring Attributes of Functions: Function 52680Attributes.). 52681 52682* Menu: 52683 52684* Solaris Format Checks:: 52685* Darwin Format Checks:: 52686 52687 52688File: gcc.info, Node: Solaris Format Checks, Next: Darwin Format Checks, Up: Target Format Checks 52689 526906.61.1 Solaris Format Checks 52691---------------------------- 52692 52693Solaris targets support the 'cmn_err' (or '__cmn_err__') format check. 52694'cmn_err' accepts a subset of the standard 'printf' conversions, and the 52695two-argument '%b' conversion for displaying bit-fields. See the Solaris 52696man page for 'cmn_err' for more information. 52697 52698 52699File: gcc.info, Node: Darwin Format Checks, Prev: Solaris Format Checks, Up: Target Format Checks 52700 527016.61.2 Darwin Format Checks 52702--------------------------- 52703 52704In addition to the full set of format archetypes (attribute format style 52705arguments such as 'printf', 'scanf', 'strftime', and 'strfmon'), Darwin 52706targets also support the 'CFString' (or '__CFString__') archetype in the 52707'format' attribute. Declarations with this archetype are parsed for 52708correct syntax and argument types. However, parsing of the format 52709string itself and validating arguments against it in calls to such 52710functions is currently not performed. 52711 52712 Additionally, 'CFStringRefs' (defined by the 'CoreFoundation' headers) 52713may also be used as format arguments. Note that the relevant headers 52714are only likely to be available on Darwin (OSX) installations. On such 52715installations, the XCode and system documentation provide descriptions 52716of 'CFString', 'CFStringRefs' and associated functions. 52717 52718 52719File: gcc.info, Node: Pragmas, Next: Unnamed Fields, Prev: Target Format Checks, Up: C Extensions 52720 527216.62 Pragmas Accepted by GCC 52722============================ 52723 52724GCC supports several types of pragmas, primarily in order to compile 52725code originally written for other compilers. Note that in general we do 52726not recommend the use of pragmas; *Note Function Attributes::, for 52727further explanation. 52728 52729 The GNU C preprocessor recognizes several pragmas in addition to the 52730compiler pragmas documented here. Refer to the CPP manual for more 52731information. 52732 52733* Menu: 52734 52735* AArch64 Pragmas:: 52736* ARM Pragmas:: 52737* M32C Pragmas:: 52738* MeP Pragmas:: 52739* PRU Pragmas:: 52740* RS/6000 and PowerPC Pragmas:: 52741* S/390 Pragmas:: 52742* Darwin Pragmas:: 52743* Solaris Pragmas:: 52744* Symbol-Renaming Pragmas:: 52745* Structure-Layout Pragmas:: 52746* Weak Pragmas:: 52747* Diagnostic Pragmas:: 52748* Visibility Pragmas:: 52749* Push/Pop Macro Pragmas:: 52750* Function Specific Option Pragmas:: 52751* Loop-Specific Pragmas:: 52752 52753 52754File: gcc.info, Node: AArch64 Pragmas, Next: ARM Pragmas, Up: Pragmas 52755 527566.62.1 AArch64 Pragmas 52757---------------------- 52758 52759The pragmas defined by the AArch64 target correspond to the AArch64 52760target function attributes. They can be specified as below: 52761 #pragma GCC target("string") 52762 52763 where 'STRING' can be any string accepted as an AArch64 target 52764attribute. *Note AArch64 Function Attributes::, for more details on the 52765permissible values of 'string'. 52766 52767 52768File: gcc.info, Node: ARM Pragmas, Next: M32C Pragmas, Prev: AArch64 Pragmas, Up: Pragmas 52769 527706.62.2 ARM Pragmas 52771------------------ 52772 52773The ARM target defines pragmas for controlling the default addition of 52774'long_call' and 'short_call' attributes to functions. *Note Function 52775Attributes::, for information about the effects of these attributes. 52776 52777'long_calls' 52778 Set all subsequent functions to have the 'long_call' attribute. 52779 52780'no_long_calls' 52781 Set all subsequent functions to have the 'short_call' attribute. 52782 52783'long_calls_off' 52784 Do not affect the 'long_call' or 'short_call' attributes of 52785 subsequent functions. 52786 52787 52788File: gcc.info, Node: M32C Pragmas, Next: MeP Pragmas, Prev: ARM Pragmas, Up: Pragmas 52789 527906.62.3 M32C Pragmas 52791------------------- 52792 52793'GCC memregs NUMBER' 52794 Overrides the command-line option '-memregs=' for the current file. 52795 Use with care! This pragma must be before any function in the 52796 file, and mixing different memregs values in different objects may 52797 make them incompatible. This pragma is useful when a 52798 performance-critical function uses a memreg for temporary values, 52799 as it may allow you to reduce the number of memregs used. 52800 52801'ADDRESS NAME ADDRESS' 52802 For any declared symbols matching NAME, this does three things to 52803 that symbol: it forces the symbol to be located at the given 52804 address (a number), it forces the symbol to be volatile, and it 52805 changes the symbol's scope to be static. This pragma exists for 52806 compatibility with other compilers, but note that the common 52807 '1234H' numeric syntax is not supported (use '0x1234' instead). 52808 Example: 52809 52810 #pragma ADDRESS port3 0x103 52811 char port3; 52812 52813 52814File: gcc.info, Node: MeP Pragmas, Next: PRU Pragmas, Prev: M32C Pragmas, Up: Pragmas 52815 528166.62.4 MeP Pragmas 52817------------------ 52818 52819'custom io_volatile (on|off)' 52820 Overrides the command-line option '-mio-volatile' for the current 52821 file. Note that for compatibility with future GCC releases, this 52822 option should only be used once before any 'io' variables in each 52823 file. 52824 52825'GCC coprocessor available REGISTERS' 52826 Specifies which coprocessor registers are available to the register 52827 allocator. REGISTERS may be a single register, register range 52828 separated by ellipses, or comma-separated list of those. Example: 52829 52830 #pragma GCC coprocessor available $c0...$c10, $c28 52831 52832'GCC coprocessor call_saved REGISTERS' 52833 Specifies which coprocessor registers are to be saved and restored 52834 by any function using them. REGISTERS may be a single register, 52835 register range separated by ellipses, or comma-separated list of 52836 those. Example: 52837 52838 #pragma GCC coprocessor call_saved $c4...$c6, $c31 52839 52840'GCC coprocessor subclass '(A|B|C|D)' = REGISTERS' 52841 Creates and defines a register class. These register classes can 52842 be used by inline 'asm' constructs. REGISTERS may be a single 52843 register, register range separated by ellipses, or comma-separated 52844 list of those. Example: 52845 52846 #pragma GCC coprocessor subclass 'B' = $c2, $c4, $c6 52847 52848 asm ("cpfoo %0" : "=B" (x)); 52849 52850'GCC disinterrupt NAME , NAME ...' 52851 For the named functions, the compiler adds code to disable 52852 interrupts for the duration of those functions. If any functions 52853 so named are not encountered in the source, a warning is emitted 52854 that the pragma is not used. Examples: 52855 52856 #pragma disinterrupt foo 52857 #pragma disinterrupt bar, grill 52858 int foo () { ... } 52859 52860'GCC call NAME , NAME ...' 52861 For the named functions, the compiler always uses a 52862 register-indirect call model when calling the named functions. 52863 Examples: 52864 52865 extern int foo (); 52866 #pragma call foo 52867 52868 52869File: gcc.info, Node: PRU Pragmas, Next: RS/6000 and PowerPC Pragmas, Prev: MeP Pragmas, Up: Pragmas 52870 528716.62.5 PRU Pragmas 52872------------------ 52873 52874'ctable_entry INDEX CONSTANT_ADDRESS' 52875 Specifies that the PRU CTABLE entry given by INDEX has the value 52876 CONSTANT_ADDRESS. This enables GCC to emit LBCO/SBCO instructions 52877 when the load/store address is known and can be addressed with some 52878 CTABLE entry. For example: 52879 52880 /* will compile to "sbco Rx, 2, 0x10, 4" */ 52881 #pragma ctable_entry 2 0x4802a000 52882 *(unsigned int *)0x4802a010 = val; 52883 52884 52885File: gcc.info, Node: RS/6000 and PowerPC Pragmas, Next: S/390 Pragmas, Prev: PRU Pragmas, Up: Pragmas 52886 528876.62.6 RS/6000 and PowerPC Pragmas 52888---------------------------------- 52889 52890The RS/6000 and PowerPC targets define one pragma for controlling 52891whether or not the 'longcall' attribute is added to function 52892declarations by default. This pragma overrides the '-mlongcall' option, 52893but not the 'longcall' and 'shortcall' attributes. *Note RS/6000 and 52894PowerPC Options::, for more information about when long calls are and 52895are not necessary. 52896 52897'longcall (1)' 52898 Apply the 'longcall' attribute to all subsequent function 52899 declarations. 52900 52901'longcall (0)' 52902 Do not apply the 'longcall' attribute to subsequent function 52903 declarations. 52904 52905 52906File: gcc.info, Node: S/390 Pragmas, Next: Darwin Pragmas, Prev: RS/6000 and PowerPC Pragmas, Up: Pragmas 52907 529086.62.7 S/390 Pragmas 52909-------------------- 52910 52911The pragmas defined by the S/390 target correspond to the S/390 target 52912function attributes and some the additional options: 52913 52914'zvector' 52915'no-zvector' 52916 52917 Note that options of the pragma, unlike options of the target 52918attribute, do change the value of preprocessor macros like '__VEC__'. 52919They can be specified as below: 52920 52921 #pragma GCC target("string[,string]...") 52922 #pragma GCC target("string"[,"string"]...) 52923 52924 52925File: gcc.info, Node: Darwin Pragmas, Next: Solaris Pragmas, Prev: S/390 Pragmas, Up: Pragmas 52926 529276.62.8 Darwin Pragmas 52928--------------------- 52929 52930The following pragmas are available for all architectures running the 52931Darwin operating system. These are useful for compatibility with other 52932Mac OS compilers. 52933 52934'mark TOKENS...' 52935 This pragma is accepted, but has no effect. 52936 52937'options align=ALIGNMENT' 52938 This pragma sets the alignment of fields in structures. The values 52939 of ALIGNMENT may be 'mac68k', to emulate m68k alignment, or 52940 'power', to emulate PowerPC alignment. Uses of this pragma nest 52941 properly; to restore the previous setting, use 'reset' for the 52942 ALIGNMENT. 52943 52944'segment TOKENS...' 52945 This pragma is accepted, but has no effect. 52946 52947'unused (VAR [, VAR]...)' 52948 This pragma declares variables to be possibly unused. GCC does not 52949 produce warnings for the listed variables. The effect is similar 52950 to that of the 'unused' attribute, except that this pragma may 52951 appear anywhere within the variables' scopes. 52952 52953 52954File: gcc.info, Node: Solaris Pragmas, Next: Symbol-Renaming Pragmas, Prev: Darwin Pragmas, Up: Pragmas 52955 529566.62.9 Solaris Pragmas 52957---------------------- 52958 52959The Solaris target supports '#pragma redefine_extname' (*note 52960Symbol-Renaming Pragmas::). It also supports additional '#pragma' 52961directives for compatibility with the system compiler. 52962 52963'align ALIGNMENT (VARIABLE [, VARIABLE]...)' 52964 52965 Increase the minimum alignment of each VARIABLE to ALIGNMENT. This 52966 is the same as GCC's 'aligned' attribute *note Variable 52967 Attributes::). Macro expansion occurs on the arguments to this 52968 pragma when compiling C and Objective-C. It does not currently 52969 occur when compiling C++, but this is a bug which may be fixed in a 52970 future release. 52971 52972'fini (FUNCTION [, FUNCTION]...)' 52973 52974 This pragma causes each listed FUNCTION to be called after main, or 52975 during shared module unloading, by adding a call to the '.fini' 52976 section. 52977 52978'init (FUNCTION [, FUNCTION]...)' 52979 52980 This pragma causes each listed FUNCTION to be called during 52981 initialization (before 'main') or during shared module loading, by 52982 adding a call to the '.init' section. 52983 52984 52985File: gcc.info, Node: Symbol-Renaming Pragmas, Next: Structure-Layout Pragmas, Prev: Solaris Pragmas, Up: Pragmas 52986 529876.62.10 Symbol-Renaming Pragmas 52988------------------------------- 52989 52990GCC supports a '#pragma' directive that changes the name used in 52991assembly for a given declaration. While this pragma is supported on all 52992platforms, it is intended primarily to provide compatibility with the 52993Solaris system headers. This effect can also be achieved using the asm 52994labels extension (*note Asm Labels::). 52995 52996'redefine_extname OLDNAME NEWNAME' 52997 52998 This pragma gives the C function OLDNAME the assembly symbol 52999 NEWNAME. The preprocessor macro '__PRAGMA_REDEFINE_EXTNAME' is 53000 defined if this pragma is available (currently on all platforms). 53001 53002 This pragma and the 'asm' labels extension interact in a complicated 53003manner. Here are some corner cases you may want to be aware of: 53004 53005 1. This pragma silently applies only to declarations with external 53006 linkage. The 'asm' label feature does not have this restriction. 53007 53008 2. In C++, this pragma silently applies only to declarations with "C" 53009 linkage. Again, 'asm' labels do not have this restriction. 53010 53011 3. If either of the ways of changing the assembly name of a 53012 declaration are applied to a declaration whose assembly name has 53013 already been determined (either by a previous use of one of these 53014 features, or because the compiler needed the assembly name in order 53015 to generate code), and the new name is different, a warning issues 53016 and the name does not change. 53017 53018 4. The OLDNAME used by '#pragma redefine_extname' is always the 53019 C-language name. 53020 53021 53022File: gcc.info, Node: Structure-Layout Pragmas, Next: Weak Pragmas, Prev: Symbol-Renaming Pragmas, Up: Pragmas 53023 530246.62.11 Structure-Layout Pragmas 53025-------------------------------- 53026 53027For compatibility with Microsoft Windows compilers, GCC supports a set 53028of '#pragma' directives that change the maximum alignment of members of 53029structures (other than zero-width bit-fields), unions, and classes 53030subsequently defined. The N value below always is required to be a 53031small power of two and specifies the new alignment in bytes. 53032 53033 1. '#pragma pack(N)' simply sets the new alignment. 53034 2. '#pragma pack()' sets the alignment to the one that was in effect 53035 when compilation started (see also command-line option 53036 '-fpack-struct[=N]' *note Code Gen Options::). 53037 3. '#pragma pack(push[,N])' pushes the current alignment setting on an 53038 internal stack and then optionally sets the new alignment. 53039 4. '#pragma pack(pop)' restores the alignment setting to the one saved 53040 at the top of the internal stack (and removes that stack entry). 53041 Note that '#pragma pack([N])' does not influence this internal 53042 stack; thus it is possible to have '#pragma pack(push)' followed by 53043 multiple '#pragma pack(N)' instances and finalized by a single 53044 '#pragma pack(pop)'. 53045 53046 Some targets, e.g. x86 and PowerPC, support the '#pragma ms_struct' 53047directive which lays out structures and unions subsequently defined as 53048the documented '__attribute__ ((ms_struct))'. 53049 53050 1. '#pragma ms_struct on' turns on the Microsoft layout. 53051 2. '#pragma ms_struct off' turns off the Microsoft layout. 53052 3. '#pragma ms_struct reset' goes back to the default layout. 53053 53054 Most targets also support the '#pragma scalar_storage_order' directive 53055which lays out structures and unions subsequently defined as the 53056documented '__attribute__ ((scalar_storage_order))'. 53057 53058 1. '#pragma scalar_storage_order big-endian' sets the storage order of 53059 the scalar fields to big-endian. 53060 2. '#pragma scalar_storage_order little-endian' sets the storage order 53061 of the scalar fields to little-endian. 53062 3. '#pragma scalar_storage_order default' goes back to the endianness 53063 that was in effect when compilation started (see also command-line 53064 option '-fsso-struct=ENDIANNESS' *note C Dialect Options::). 53065 53066 53067File: gcc.info, Node: Weak Pragmas, Next: Diagnostic Pragmas, Prev: Structure-Layout Pragmas, Up: Pragmas 53068 530696.62.12 Weak Pragmas 53070-------------------- 53071 53072For compatibility with SVR4, GCC supports a set of '#pragma' directives 53073for declaring symbols to be weak, and defining weak aliases. 53074 53075'#pragma weak SYMBOL' 53076 This pragma declares SYMBOL to be weak, as if the declaration had 53077 the attribute of the same name. The pragma may appear before or 53078 after the declaration of SYMBOL. It is not an error for SYMBOL to 53079 never be defined at all. 53080 53081'#pragma weak SYMBOL1 = SYMBOL2' 53082 This pragma declares SYMBOL1 to be a weak alias of SYMBOL2. It is 53083 an error if SYMBOL2 is not defined in the current translation unit. 53084 53085 53086File: gcc.info, Node: Diagnostic Pragmas, Next: Visibility Pragmas, Prev: Weak Pragmas, Up: Pragmas 53087 530886.62.13 Diagnostic Pragmas 53089-------------------------- 53090 53091GCC allows the user to selectively enable or disable certain types of 53092diagnostics, and change the kind of the diagnostic. For example, a 53093project's policy might require that all sources compile with '-Werror' 53094but certain files might have exceptions allowing specific types of 53095warnings. Or, a project might selectively enable diagnostics and treat 53096them as errors depending on which preprocessor macros are defined. 53097 53098'#pragma GCC diagnostic KIND OPTION' 53099 53100 Modifies the disposition of a diagnostic. Note that not all 53101 diagnostics are modifiable; at the moment only warnings (normally 53102 controlled by '-W...') can be controlled, and not all of them. Use 53103 '-fdiagnostics-show-option' to determine which diagnostics are 53104 controllable and which option controls them. 53105 53106 KIND is 'error' to treat this diagnostic as an error, 'warning' to 53107 treat it like a warning (even if '-Werror' is in effect), or 53108 'ignored' if the diagnostic is to be ignored. OPTION is a double 53109 quoted string that matches the command-line option. 53110 53111 #pragma GCC diagnostic warning "-Wformat" 53112 #pragma GCC diagnostic error "-Wformat" 53113 #pragma GCC diagnostic ignored "-Wformat" 53114 53115 Note that these pragmas override any command-line options. GCC 53116 keeps track of the location of each pragma, and issues diagnostics 53117 according to the state as of that point in the source file. Thus, 53118 pragmas occurring after a line do not affect diagnostics caused by 53119 that line. 53120 53121'#pragma GCC diagnostic push' 53122'#pragma GCC diagnostic pop' 53123 53124 Causes GCC to remember the state of the diagnostics as of each 53125 'push', and restore to that point at each 'pop'. If a 'pop' has no 53126 matching 'push', the command-line options are restored. 53127 53128 #pragma GCC diagnostic error "-Wuninitialized" 53129 foo(a); /* error is given for this one */ 53130 #pragma GCC diagnostic push 53131 #pragma GCC diagnostic ignored "-Wuninitialized" 53132 foo(b); /* no diagnostic for this one */ 53133 #pragma GCC diagnostic pop 53134 foo(c); /* error is given for this one */ 53135 #pragma GCC diagnostic pop 53136 foo(d); /* depends on command-line options */ 53137 53138 GCC also offers a simple mechanism for printing messages during 53139compilation. 53140 53141'#pragma message STRING' 53142 53143 Prints STRING as a compiler message on compilation. The message is 53144 informational only, and is neither a compilation warning nor an 53145 error. Newlines can be included in the string by using the '\n' 53146 escape sequence. 53147 53148 #pragma message "Compiling " __FILE__ "..." 53149 53150 STRING may be parenthesized, and is printed with location 53151 information. For example, 53152 53153 #define DO_PRAGMA(x) _Pragma (#x) 53154 #define TODO(x) DO_PRAGMA(message ("TODO - " #x)) 53155 53156 TODO(Remember to fix this) 53157 53158 prints '/tmp/file.c:4: note: #pragma message: TODO - Remember to 53159 fix this'. 53160 53161'#pragma GCC error MESSAGE' 53162 Generates an error message. This pragma _is_ considered to 53163 indicate an error in the compilation, and it will be treated as 53164 such. 53165 53166 Newlines can be included in the string by using the '\n' escape 53167 sequence. They will be displayed as newlines even if the 53168 '-fmessage-length' option is set to zero. 53169 53170 The error is only generated if the pragma is present in the code 53171 after pre-processing has been completed. It does not matter 53172 however if the code containing the pragma is unreachable: 53173 53174 #if 0 53175 #pragma GCC error "this error is not seen" 53176 #endif 53177 void foo (void) 53178 { 53179 return; 53180 #pragma GCC error "this error is seen" 53181 } 53182 53183'#pragma GCC warning MESSAGE' 53184 This is just like 'pragma GCC error' except that a warning message 53185 is issued instead of an error message. Unless '-Werror' is in 53186 effect, in which case this pragma will generate an error as well. 53187 53188 53189File: gcc.info, Node: Visibility Pragmas, Next: Push/Pop Macro Pragmas, Prev: Diagnostic Pragmas, Up: Pragmas 53190 531916.62.14 Visibility Pragmas 53192-------------------------- 53193 53194'#pragma GCC visibility push(VISIBILITY)' 53195'#pragma GCC visibility pop' 53196 53197 This pragma allows the user to set the visibility for multiple 53198 declarations without having to give each a visibility attribute 53199 (*note Function Attributes::). 53200 53201 In C++, '#pragma GCC visibility' affects only namespace-scope 53202 declarations. Class members and template specializations are not 53203 affected; if you want to override the visibility for a particular 53204 member or instantiation, you must use an attribute. 53205 53206 53207File: gcc.info, Node: Push/Pop Macro Pragmas, Next: Function Specific Option Pragmas, Prev: Visibility Pragmas, Up: Pragmas 53208 532096.62.15 Push/Pop Macro Pragmas 53210------------------------------ 53211 53212For compatibility with Microsoft Windows compilers, GCC supports 53213'#pragma push_macro("MACRO_NAME")' and '#pragma 53214pop_macro("MACRO_NAME")'. 53215 53216'#pragma push_macro("MACRO_NAME")' 53217 This pragma saves the value of the macro named as MACRO_NAME to the 53218 top of the stack for this macro. 53219 53220'#pragma pop_macro("MACRO_NAME")' 53221 This pragma sets the value of the macro named as MACRO_NAME to the 53222 value on top of the stack for this macro. If the stack for 53223 MACRO_NAME is empty, the value of the macro remains unchanged. 53224 53225 For example: 53226 53227 #define X 1 53228 #pragma push_macro("X") 53229 #undef X 53230 #define X -1 53231 #pragma pop_macro("X") 53232 int x [X]; 53233 53234In this example, the definition of X as 1 is saved by '#pragma 53235push_macro' and restored by '#pragma pop_macro'. 53236 53237 53238File: gcc.info, Node: Function Specific Option Pragmas, Next: Loop-Specific Pragmas, Prev: Push/Pop Macro Pragmas, Up: Pragmas 53239 532406.62.16 Function Specific Option Pragmas 53241---------------------------------------- 53242 53243'#pragma GCC target (STRING, ...)' 53244 53245 This pragma allows you to set target-specific options for functions 53246 defined later in the source file. One or more strings can be 53247 specified. Each function that is defined after this point is 53248 treated as if it had been declared with one 'target('STRING')' 53249 attribute for each STRING argument. The parentheses around the 53250 strings in the pragma are optional. *Note Function Attributes::, 53251 for more information about the 'target' attribute and the attribute 53252 syntax. 53253 53254 The '#pragma GCC target' pragma is presently implemented for x86, 53255 ARM, AArch64, PowerPC, S/390, and Nios II targets only. 53256 53257'#pragma GCC optimize (STRING, ...)' 53258 53259 This pragma allows you to set global optimization options for 53260 functions defined later in the source file. One or more strings 53261 can be specified. Each function that is defined after this point 53262 is treated as if it had been declared with one 'optimize('STRING')' 53263 attribute for each STRING argument. The parentheses around the 53264 strings in the pragma are optional. *Note Function Attributes::, 53265 for more information about the 'optimize' attribute and the 53266 attribute syntax. 53267 53268'#pragma GCC push_options' 53269'#pragma GCC pop_options' 53270 53271 These pragmas maintain a stack of the current target and 53272 optimization options. It is intended for include files where you 53273 temporarily want to switch to using a different '#pragma GCC 53274 target' or '#pragma GCC optimize' and then to pop back to the 53275 previous options. 53276 53277'#pragma GCC reset_options' 53278 53279 This pragma clears the current '#pragma GCC target' and '#pragma 53280 GCC optimize' to use the default switches as specified on the 53281 command line. 53282 53283 53284File: gcc.info, Node: Loop-Specific Pragmas, Prev: Function Specific Option Pragmas, Up: Pragmas 53285 532866.62.17 Loop-Specific Pragmas 53287----------------------------- 53288 53289'#pragma GCC ivdep' 53290 53291 With this pragma, the programmer asserts that there are no 53292 loop-carried dependencies which would prevent consecutive 53293 iterations of the following loop from executing concurrently with 53294 SIMD (single instruction multiple data) instructions. 53295 53296 For example, the compiler can only unconditionally vectorize the 53297 following loop with the pragma: 53298 53299 void foo (int n, int *a, int *b, int *c) 53300 { 53301 int i, j; 53302 #pragma GCC ivdep 53303 for (i = 0; i < n; ++i) 53304 a[i] = b[i] + c[i]; 53305 } 53306 53307 In this example, using the 'restrict' qualifier had the same 53308 effect. In the following example, that would not be possible. 53309 Assume k < -m or k >= m. Only with the pragma, the compiler knows 53310 that it can unconditionally vectorize the following loop: 53311 53312 void ignore_vec_dep (int *a, int k, int c, int m) 53313 { 53314 #pragma GCC ivdep 53315 for (int i = 0; i < m; i++) 53316 a[i] = a[i + k] * c; 53317 } 53318 53319'#pragma GCC unroll N' 53320 53321 You can use this pragma to control how many times a loop should be 53322 unrolled. It must be placed immediately before a 'for', 'while' or 53323 'do' loop or a '#pragma GCC ivdep', and applies only to the loop 53324 that follows. N is an integer constant expression specifying the 53325 unrolling factor. The values of 0 and 1 block any unrolling of the 53326 loop. 53327 53328 53329File: gcc.info, Node: Unnamed Fields, Next: Thread-Local, Prev: Pragmas, Up: C Extensions 53330 533316.63 Unnamed Structure and Union Fields 53332======================================= 53333 53334As permitted by ISO C11 and for compatibility with other compilers, GCC 53335allows you to define a structure or union that contains, as fields, 53336structures and unions without names. For example: 53337 53338 struct { 53339 int a; 53340 union { 53341 int b; 53342 float c; 53343 }; 53344 int d; 53345 } foo; 53346 53347In this example, you are able to access members of the unnamed union 53348with code like 'foo.b'. Note that only unnamed structs and unions are 53349allowed, you may not have, for example, an unnamed 'int'. 53350 53351 You must never create such structures that cause ambiguous field 53352definitions. For example, in this structure: 53353 53354 struct { 53355 int a; 53356 struct { 53357 int a; 53358 }; 53359 } foo; 53360 53361it is ambiguous which 'a' is being referred to with 'foo.a'. The 53362compiler gives errors for such constructs. 53363 53364 Unless '-fms-extensions' is used, the unnamed field must be a structure 53365or union definition without a tag (for example, 'struct { int a; };'). 53366If '-fms-extensions' is used, the field may also be a definition with a 53367tag such as 'struct foo { int a; };', a reference to a previously 53368defined structure or union such as 'struct foo;', or a reference to a 53369'typedef' name for a previously defined structure or union type. 53370 53371 The option '-fplan9-extensions' enables '-fms-extensions' as well as 53372two other extensions. First, a pointer to a structure is automatically 53373converted to a pointer to an anonymous field for assignments and 53374function calls. For example: 53375 53376 struct s1 { int a; }; 53377 struct s2 { struct s1; }; 53378 extern void f1 (struct s1 *); 53379 void f2 (struct s2 *p) { f1 (p); } 53380 53381In the call to 'f1' inside 'f2', the pointer 'p' is converted into a 53382pointer to the anonymous field. 53383 53384 Second, when the type of an anonymous field is a 'typedef' for a 53385'struct' or 'union', code may refer to the field using the name of the 53386'typedef'. 53387 53388 typedef struct { int a; } s1; 53389 struct s2 { s1; }; 53390 s1 f1 (struct s2 *p) { return p->s1; } 53391 53392 These usages are only permitted when they are not ambiguous. 53393 53394 53395File: gcc.info, Node: Thread-Local, Next: Binary constants, Prev: Unnamed Fields, Up: C Extensions 53396 533976.64 Thread-Local Storage 53398========================= 53399 53400Thread-local storage (TLS) is a mechanism by which variables are 53401allocated such that there is one instance of the variable per extant 53402thread. The runtime model GCC uses to implement this originates in the 53403IA-64 processor-specific ABI, but has since been migrated to other 53404processors as well. It requires significant support from the linker 53405('ld'), dynamic linker ('ld.so'), and system libraries ('libc.so' and 53406'libpthread.so'), so it is not available everywhere. 53407 53408 At the user level, the extension is visible with a new storage class 53409keyword: '__thread'. For example: 53410 53411 __thread int i; 53412 extern __thread struct state s; 53413 static __thread char *p; 53414 53415 The '__thread' specifier may be used alone, with the 'extern' or 53416'static' specifiers, but with no other storage class specifier. When 53417used with 'extern' or 'static', '__thread' must appear immediately after 53418the other storage class specifier. 53419 53420 The '__thread' specifier may be applied to any global, file-scoped 53421static, function-scoped static, or static data member of a class. It 53422may not be applied to block-scoped automatic or non-static data member. 53423 53424 When the address-of operator is applied to a thread-local variable, it 53425is evaluated at run time and returns the address of the current thread's 53426instance of that variable. An address so obtained may be used by any 53427thread. When a thread terminates, any pointers to thread-local 53428variables in that thread become invalid. 53429 53430 No static initialization may refer to the address of a thread-local 53431variable. 53432 53433 In C++, if an initializer is present for a thread-local variable, it 53434must be a CONSTANT-EXPRESSION, as defined in 5.19.2 of the ANSI/ISO C++ 53435standard. 53436 53437 See ELF Handling For Thread-Local Storage 53438(https://www.akkadia.org/drepper/tls.pdf) for a detailed explanation of 53439the four thread-local storage addressing models, and how the runtime is 53440expected to function. 53441 53442* Menu: 53443 53444* C99 Thread-Local Edits:: 53445* C++98 Thread-Local Edits:: 53446 53447 53448File: gcc.info, Node: C99 Thread-Local Edits, Next: C++98 Thread-Local Edits, Up: Thread-Local 53449 534506.64.1 ISO/IEC 9899:1999 Edits for Thread-Local Storage 53451------------------------------------------------------- 53452 53453The following are a set of changes to ISO/IEC 9899:1999 (aka C99) that 53454document the exact semantics of the language extension. 53455 53456 * '5.1.2 Execution environments' 53457 53458 Add new text after paragraph 1 53459 53460 Within either execution environment, a "thread" is a flow of 53461 control within a program. It is implementation defined 53462 whether or not there may be more than one thread associated 53463 with a program. It is implementation defined how threads 53464 beyond the first are created, the name and type of the 53465 function called at thread startup, and how threads may be 53466 terminated. However, objects with thread storage duration 53467 shall be initialized before thread startup. 53468 53469 * '6.2.4 Storage durations of objects' 53470 53471 Add new text before paragraph 3 53472 53473 An object whose identifier is declared with the storage-class 53474 specifier '__thread' has "thread storage duration". Its 53475 lifetime is the entire execution of the thread, and its stored 53476 value is initialized only once, prior to thread startup. 53477 53478 * '6.4.1 Keywords' 53479 53480 Add '__thread'. 53481 53482 * '6.7.1 Storage-class specifiers' 53483 53484 Add '__thread' to the list of storage class specifiers in paragraph 53485 1. 53486 53487 Change paragraph 2 to 53488 53489 With the exception of '__thread', at most one storage-class 53490 specifier may be given [...]. The '__thread' specifier may be 53491 used alone, or immediately following 'extern' or 'static'. 53492 53493 Add new text after paragraph 6 53494 53495 The declaration of an identifier for a variable that has block 53496 scope that specifies '__thread' shall also specify either 53497 'extern' or 'static'. 53498 53499 The '__thread' specifier shall be used only with variables. 53500 53501 53502File: gcc.info, Node: C++98 Thread-Local Edits, Prev: C99 Thread-Local Edits, Up: Thread-Local 53503 535046.64.2 ISO/IEC 14882:1998 Edits for Thread-Local Storage 53505-------------------------------------------------------- 53506 53507The following are a set of changes to ISO/IEC 14882:1998 (aka C++98) 53508that document the exact semantics of the language extension. 53509 53510 * [intro.execution] 53511 53512 New text after paragraph 4 53513 53514 A "thread" is a flow of control within the abstract machine. 53515 It is implementation defined whether or not there may be more 53516 than one thread. 53517 53518 New text after paragraph 7 53519 53520 It is unspecified whether additional action must be taken to 53521 ensure when and whether side effects are visible to other 53522 threads. 53523 53524 * [lex.key] 53525 53526 Add '__thread'. 53527 53528 * [basic.start.main] 53529 53530 Add after paragraph 5 53531 53532 The thread that begins execution at the 'main' function is 53533 called the "main thread". It is implementation defined how 53534 functions beginning threads other than the main thread are 53535 designated or typed. A function so designated, as well as the 53536 'main' function, is called a "thread startup function". It is 53537 implementation defined what happens if a thread startup 53538 function returns. It is implementation defined what happens 53539 to other threads when any thread calls 'exit'. 53540 53541 * [basic.start.init] 53542 53543 Add after paragraph 4 53544 53545 The storage for an object of thread storage duration shall be 53546 statically initialized before the first statement of the 53547 thread startup function. An object of thread storage duration 53548 shall not require dynamic initialization. 53549 53550 * [basic.start.term] 53551 53552 Add after paragraph 3 53553 53554 The type of an object with thread storage duration shall not 53555 have a non-trivial destructor, nor shall it be an array type 53556 whose elements (directly or indirectly) have non-trivial 53557 destructors. 53558 53559 * [basic.stc] 53560 53561 Add "thread storage duration" to the list in paragraph 1. 53562 53563 Change paragraph 2 53564 53565 Thread, static, and automatic storage durations are associated 53566 with objects introduced by declarations [...]. 53567 53568 Add '__thread' to the list of specifiers in paragraph 3. 53569 53570 * [basic.stc.thread] 53571 53572 New section before [basic.stc.static] 53573 53574 The keyword '__thread' applied to a non-local object gives the 53575 object thread storage duration. 53576 53577 A local variable or class data member declared both 'static' 53578 and '__thread' gives the variable or member thread storage 53579 duration. 53580 53581 * [basic.stc.static] 53582 53583 Change paragraph 1 53584 53585 All objects that have neither thread storage duration, dynamic 53586 storage duration nor are local [...]. 53587 53588 * [dcl.stc] 53589 53590 Add '__thread' to the list in paragraph 1. 53591 53592 Change paragraph 1 53593 53594 With the exception of '__thread', at most one 53595 STORAGE-CLASS-SPECIFIER shall appear in a given 53596 DECL-SPECIFIER-SEQ. The '__thread' specifier may be used 53597 alone, or immediately following the 'extern' or 'static' 53598 specifiers. [...] 53599 53600 Add after paragraph 5 53601 53602 The '__thread' specifier can be applied only to the names of 53603 objects and to anonymous unions. 53604 53605 * [class.mem] 53606 53607 Add after paragraph 6 53608 53609 Non-'static' members shall not be '__thread'. 53610 53611 53612File: gcc.info, Node: Binary constants, Prev: Thread-Local, Up: C Extensions 53613 536146.65 Binary Constants using the '0b' Prefix 53615=========================================== 53616 53617Integer constants can be written as binary constants, consisting of a 53618sequence of '0' and '1' digits, prefixed by '0b' or '0B'. This is 53619particularly useful in environments that operate a lot on the bit level 53620(like microcontrollers). 53621 53622 The following statements are identical: 53623 53624 i = 42; 53625 i = 0x2a; 53626 i = 052; 53627 i = 0b101010; 53628 53629 The type of these constants follows the same rules as for octal or 53630hexadecimal integer constants, so suffixes like 'L' or 'UL' can be 53631applied. 53632 53633 53634File: gcc.info, Node: C++ Extensions, Next: Objective-C, Prev: C Extensions, Up: Top 53635 536367 Extensions to the C++ Language 53637******************************** 53638 53639The GNU compiler provides these extensions to the C++ language (and you 53640can also use most of the C language extensions in your C++ programs). 53641If you want to write code that checks whether these features are 53642available, you can test for the GNU compiler the same way as for C 53643programs: check for a predefined macro '__GNUC__'. You can also use 53644'__GNUG__' to test specifically for GNU C++ (*note Predefined Macros: 53645(cpp)Common Predefined Macros.). 53646 53647* Menu: 53648 53649* C++ Volatiles:: What constitutes an access to a volatile object. 53650* Restricted Pointers:: C99 restricted pointers and references. 53651* Vague Linkage:: Where G++ puts inlines, vtables and such. 53652* C++ Interface:: You can use a single C++ header file for both 53653 declarations and definitions. 53654* Template Instantiation:: Methods for ensuring that exactly one copy of 53655 each needed template instantiation is emitted. 53656* Bound member functions:: You can extract a function pointer to the 53657 method denoted by a '->*' or '.*' expression. 53658* C++ Attributes:: Variable, function, and type attributes for C++ only. 53659* Function Multiversioning:: Declaring multiple function versions. 53660* Type Traits:: Compiler support for type traits. 53661* C++ Concepts:: Improved support for generic programming. 53662* Deprecated Features:: Things will disappear from G++. 53663* Backwards Compatibility:: Compatibilities with earlier definitions of C++. 53664 53665 53666File: gcc.info, Node: C++ Volatiles, Next: Restricted Pointers, Up: C++ Extensions 53667 536687.1 When is a Volatile C++ Object Accessed? 53669=========================================== 53670 53671The C++ standard differs from the C standard in its treatment of 53672volatile objects. It fails to specify what constitutes a volatile 53673access, except to say that C++ should behave in a similar manner to C 53674with respect to volatiles, where possible. However, the different 53675lvalueness of expressions between C and C++ complicate the behavior. 53676G++ behaves the same as GCC for volatile access, *Note Volatiles: C 53677Extensions, for a description of GCC's behavior. 53678 53679 The C and C++ language specifications differ when an object is accessed 53680in a void context: 53681 53682 volatile int *src = SOMEVALUE; 53683 *src; 53684 53685 The C++ standard specifies that such expressions do not undergo lvalue 53686to rvalue conversion, and that the type of the dereferenced object may 53687be incomplete. The C++ standard does not specify explicitly that it is 53688lvalue to rvalue conversion that is responsible for causing an access. 53689There is reason to believe that it is, because otherwise certain simple 53690expressions become undefined. However, because it would surprise most 53691programmers, G++ treats dereferencing a pointer to volatile object of 53692complete type as GCC would do for an equivalent type in C. When the 53693object has incomplete type, G++ issues a warning; if you wish to force 53694an error, you must force a conversion to rvalue with, for instance, a 53695static cast. 53696 53697 When using a reference to volatile, G++ does not treat equivalent 53698expressions as accesses to volatiles, but instead issues a warning that 53699no volatile is accessed. The rationale for this is that otherwise it 53700becomes difficult to determine where volatile access occur, and not 53701possible to ignore the return value from functions returning volatile 53702references. Again, if you wish to force a read, cast the reference to 53703an rvalue. 53704 53705 G++ implements the same behavior as GCC does when assigning to a 53706volatile object--there is no reread of the assigned-to object, the 53707assigned rvalue is reused. Note that in C++ assignment expressions are 53708lvalues, and if used as an lvalue, the volatile object is referred to. 53709For instance, VREF refers to VOBJ, as expected, in the following 53710example: 53711 53712 volatile int vobj; 53713 volatile int &vref = vobj = SOMETHING; 53714 53715 53716File: gcc.info, Node: Restricted Pointers, Next: Vague Linkage, Prev: C++ Volatiles, Up: C++ Extensions 53717 537187.2 Restricting Pointer Aliasing 53719================================ 53720 53721As with the C front end, G++ understands the C99 feature of restricted 53722pointers, specified with the '__restrict__', or '__restrict' type 53723qualifier. Because you cannot compile C++ by specifying the '-std=c99' 53724language flag, 'restrict' is not a keyword in C++. 53725 53726 In addition to allowing restricted pointers, you can specify restricted 53727references, which indicate that the reference is not aliased in the 53728local context. 53729 53730 void fn (int *__restrict__ rptr, int &__restrict__ rref) 53731 { 53732 /* ... */ 53733 } 53734 53735In the body of 'fn', RPTR points to an unaliased integer and RREF refers 53736to a (different) unaliased integer. 53737 53738 You may also specify whether a member function's THIS pointer is 53739unaliased by using '__restrict__' as a member function qualifier. 53740 53741 void T::fn () __restrict__ 53742 { 53743 /* ... */ 53744 } 53745 53746Within the body of 'T::fn', THIS has the effective definition 'T 53747*__restrict__ const this'. Notice that the interpretation of a 53748'__restrict__' member function qualifier is different to that of 'const' 53749or 'volatile' qualifier, in that it is applied to the pointer rather 53750than the object. This is consistent with other compilers that implement 53751restricted pointers. 53752 53753 As with all outermost parameter qualifiers, '__restrict__' is ignored 53754in function definition matching. This means you only need to specify 53755'__restrict__' in a function definition, rather than in a function 53756prototype as well. 53757 53758 53759File: gcc.info, Node: Vague Linkage, Next: C++ Interface, Prev: Restricted Pointers, Up: C++ Extensions 53760 537617.3 Vague Linkage 53762================= 53763 53764There are several constructs in C++ that require space in the object 53765file but are not clearly tied to a single translation unit. We say that 53766these constructs have "vague linkage". Typically such constructs are 53767emitted wherever they are needed, though sometimes we can be more 53768clever. 53769 53770Inline Functions 53771 Inline functions are typically defined in a header file which can 53772 be included in many different compilations. Hopefully they can 53773 usually be inlined, but sometimes an out-of-line copy is necessary, 53774 if the address of the function is taken or if inlining fails. In 53775 general, we emit an out-of-line copy in all translation units where 53776 one is needed. As an exception, we only emit inline virtual 53777 functions with the vtable, since it always requires a copy. 53778 53779 Local static variables and string constants used in an inline 53780 function are also considered to have vague linkage, since they must 53781 be shared between all inlined and out-of-line instances of the 53782 function. 53783 53784VTables 53785 C++ virtual functions are implemented in most compilers using a 53786 lookup table, known as a vtable. The vtable contains pointers to 53787 the virtual functions provided by a class, and each object of the 53788 class contains a pointer to its vtable (or vtables, in some 53789 multiple-inheritance situations). If the class declares any 53790 non-inline, non-pure virtual functions, the first one is chosen as 53791 the "key method" for the class, and the vtable is only emitted in 53792 the translation unit where the key method is defined. 53793 53794 _Note:_ If the chosen key method is later defined as inline, the 53795 vtable is still emitted in every translation unit that defines it. 53796 Make sure that any inline virtuals are declared inline in the class 53797 body, even if they are not defined there. 53798 53799'type_info' objects 53800 C++ requires information about types to be written out in order to 53801 implement 'dynamic_cast', 'typeid' and exception handling. For 53802 polymorphic classes (classes with virtual functions), the 53803 'type_info' object is written out along with the vtable so that 53804 'dynamic_cast' can determine the dynamic type of a class object at 53805 run time. For all other types, we write out the 'type_info' object 53806 when it is used: when applying 'typeid' to an expression, throwing 53807 an object, or referring to a type in a catch clause or exception 53808 specification. 53809 53810Template Instantiations 53811 Most everything in this section also applies to template 53812 instantiations, but there are other options as well. *Note Where's 53813 the Template?: Template Instantiation. 53814 53815 When used with GNU ld version 2.8 or later on an ELF system such as 53816GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of 53817these constructs will be discarded at link time. This is known as 53818COMDAT support. 53819 53820 On targets that don't support COMDAT, but do support weak symbols, GCC 53821uses them. This way one copy overrides all the others, but the unused 53822copies still take up space in the executable. 53823 53824 For targets that do not support either COMDAT or weak symbols, most 53825entities with vague linkage are emitted as local symbols to avoid 53826duplicate definition errors from the linker. This does not happen for 53827local statics in inlines, however, as having multiple copies almost 53828certainly breaks things. 53829 53830 *Note Declarations and Definitions in One Header: C++ Interface, for 53831another way to control placement of these constructs. 53832 53833 53834File: gcc.info, Node: C++ Interface, Next: Template Instantiation, Prev: Vague Linkage, Up: C++ Extensions 53835 538367.4 C++ Interface and Implementation Pragmas 53837============================================ 53838 53839'#pragma interface' and '#pragma implementation' provide the user with a 53840way of explicitly directing the compiler to emit entities with vague 53841linkage (and debugging information) in a particular translation unit. 53842 53843 _Note:_ These '#pragma's have been superceded as of GCC 2.7.2 by COMDAT 53844support and the "key method" heuristic mentioned in *note Vague 53845Linkage::. Using them can actually cause your program to grow due to 53846unnecessary out-of-line copies of inline functions. 53847 53848'#pragma interface' 53849'#pragma interface "SUBDIR/OBJECTS.h"' 53850 Use this directive in _header files_ that define object classes, to 53851 save space in most of the object files that use those classes. 53852 Normally, local copies of certain information (backup copies of 53853 inline member functions, debugging information, and the internal 53854 tables that implement virtual functions) must be kept in each 53855 object file that includes class definitions. You can use this 53856 pragma to avoid such duplication. When a header file containing 53857 '#pragma interface' is included in a compilation, this auxiliary 53858 information is not generated (unless the main input source file 53859 itself uses '#pragma implementation'). Instead, the object files 53860 contain references to be resolved at link time. 53861 53862 The second form of this directive is useful for the case where you 53863 have multiple headers with the same name in different directories. 53864 If you use this form, you must specify the same string to '#pragma 53865 implementation'. 53866 53867'#pragma implementation' 53868'#pragma implementation "OBJECTS.h"' 53869 Use this pragma in a _main input file_, when you want full output 53870 from included header files to be generated (and made globally 53871 visible). The included header file, in turn, should use '#pragma 53872 interface'. Backup copies of inline member functions, debugging 53873 information, and the internal tables used to implement virtual 53874 functions are all generated in implementation files. 53875 53876 If you use '#pragma implementation' with no argument, it applies to 53877 an include file with the same basename(1) as your source file. For 53878 example, in 'allclass.cc', giving just '#pragma implementation' by 53879 itself is equivalent to '#pragma implementation "allclass.h"'. 53880 53881 Use the string argument if you want a single implementation file to 53882 include code from multiple header files. (You must also use 53883 '#include' to include the header file; '#pragma implementation' 53884 only specifies how to use the file--it doesn't actually include 53885 it.) 53886 53887 There is no way to split up the contents of a single header file 53888 into multiple implementation files. 53889 53890 '#pragma implementation' and '#pragma interface' also have an effect on 53891function inlining. 53892 53893 If you define a class in a header file marked with '#pragma interface', 53894the effect on an inline function defined in that class is similar to an 53895explicit 'extern' declaration--the compiler emits no code at all to 53896define an independent version of the function. Its definition is used 53897only for inlining with its callers. 53898 53899 Conversely, when you include the same header file in a main source file 53900that declares it as '#pragma implementation', the compiler emits code 53901for the function itself; this defines a version of the function that can 53902be found via pointers (or by callers compiled without inlining). If all 53903calls to the function can be inlined, you can avoid emitting the 53904function by compiling with '-fno-implement-inlines'. If any calls are 53905not inlined, you will get linker errors. 53906 53907 ---------- Footnotes ---------- 53908 53909 (1) A file's "basename" is the name stripped of all leading path 53910information and of trailing suffixes, such as '.h' or '.C' or '.cc'. 53911 53912 53913File: gcc.info, Node: Template Instantiation, Next: Bound member functions, Prev: C++ Interface, Up: C++ Extensions 53914 539157.5 Where's the Template? 53916========================= 53917 53918C++ templates were the first language feature to require more 53919intelligence from the environment than was traditionally found on a UNIX 53920system. Somehow the compiler and linker have to make sure that each 53921template instance occurs exactly once in the executable if it is needed, 53922and not at all otherwise. There are two basic approaches to this 53923problem, which are referred to as the Borland model and the Cfront 53924model. 53925 53926Borland model 53927 Borland C++ solved the template instantiation problem by adding the 53928 code equivalent of common blocks to their linker; the compiler 53929 emits template instances in each translation unit that uses them, 53930 and the linker collapses them together. The advantage of this 53931 model is that the linker only has to consider the object files 53932 themselves; there is no external complexity to worry about. The 53933 disadvantage is that compilation time is increased because the 53934 template code is being compiled repeatedly. Code written for this 53935 model tends to include definitions of all templates in the header 53936 file, since they must be seen to be instantiated. 53937 53938Cfront model 53939 The AT&T C++ translator, Cfront, solved the template instantiation 53940 problem by creating the notion of a template repository, an 53941 automatically maintained place where template instances are stored. 53942 A more modern version of the repository works as follows: As 53943 individual object files are built, the compiler places any template 53944 definitions and instantiations encountered in the repository. At 53945 link time, the link wrapper adds in the objects in the repository 53946 and compiles any needed instances that were not previously emitted. 53947 The advantages of this model are more optimal compilation speed and 53948 the ability to use the system linker; to implement the Borland 53949 model a compiler vendor also needs to replace the linker. The 53950 disadvantages are vastly increased complexity, and thus potential 53951 for error; for some code this can be just as transparent, but in 53952 practice it can been very difficult to build multiple programs in 53953 one directory and one program in multiple directories. Code 53954 written for this model tends to separate definitions of non-inline 53955 member templates into a separate file, which should be compiled 53956 separately. 53957 53958 G++ implements the Borland model on targets where the linker supports 53959it, including ELF targets (such as GNU/Linux), Mac OS X and Microsoft 53960Windows. Otherwise G++ implements neither automatic model. 53961 53962 You have the following options for dealing with template 53963instantiations: 53964 53965 1. Do nothing. Code written for the Borland model works fine, but 53966 each translation unit contains instances of each of the templates 53967 it uses. The duplicate instances will be discarded by the linker, 53968 but in a large program, this can lead to an unacceptable amount of 53969 code duplication in object files or shared libraries. 53970 53971 Duplicate instances of a template can be avoided by defining an 53972 explicit instantiation in one object file, and preventing the 53973 compiler from doing implicit instantiations in any other object 53974 files by using an explicit instantiation declaration, using the 53975 'extern template' syntax: 53976 53977 extern template int max (int, int); 53978 53979 This syntax is defined in the C++ 2011 standard, but has been 53980 supported by G++ and other compilers since well before 2011. 53981 53982 Explicit instantiations can be used for the largest or most 53983 frequently duplicated instances, without having to know exactly 53984 which other instances are used in the rest of the program. You can 53985 scatter the explicit instantiations throughout your program, 53986 perhaps putting them in the translation units where the instances 53987 are used or the translation units that define the templates 53988 themselves; you can put all of the explicit instantiations you need 53989 into one big file; or you can create small files like 53990 53991 #include "Foo.h" 53992 #include "Foo.cc" 53993 53994 template class Foo<int>; 53995 template ostream& operator << 53996 (ostream&, const Foo<int>&); 53997 53998 for each of the instances you need, and create a template 53999 instantiation library from those. 54000 54001 This is the simplest option, but also offers flexibility and 54002 fine-grained control when necessary. It is also the most portable 54003 alternative and programs using this approach will work with most 54004 modern compilers. 54005 54006 2. Compile your code with '-fno-implicit-templates' to disable the 54007 implicit generation of template instances, and explicitly 54008 instantiate all the ones you use. This approach requires more 54009 knowledge of exactly which instances you need than do the others, 54010 but it's less mysterious and allows greater control if you want to 54011 ensure that only the intended instances are used. 54012 54013 If you are using Cfront-model code, you can probably get away with 54014 not using '-fno-implicit-templates' when compiling files that don't 54015 '#include' the member template definitions. 54016 54017 If you use one big file to do the instantiations, you may want to 54018 compile it without '-fno-implicit-templates' so you get all of the 54019 instances required by your explicit instantiations (but not by any 54020 other files) without having to specify them as well. 54021 54022 In addition to forward declaration of explicit instantiations (with 54023 'extern'), G++ has extended the template instantiation syntax to 54024 support instantiation of the compiler support data for a template 54025 class (i.e. the vtable) without instantiating any of its members 54026 (with 'inline'), and instantiation of only the static data members 54027 of a template class, without the support data or member functions 54028 (with 'static'): 54029 54030 inline template class Foo<int>; 54031 static template class Foo<int>; 54032 54033 54034File: gcc.info, Node: Bound member functions, Next: C++ Attributes, Prev: Template Instantiation, Up: C++ Extensions 54035 540367.6 Extracting the Function Pointer from a Bound Pointer to Member Function 54037=========================================================================== 54038 54039In C++, pointer to member functions (PMFs) are implemented using a wide 54040pointer of sorts to handle all the possible call mechanisms; the PMF 54041needs to store information about how to adjust the 'this' pointer, and 54042if the function pointed to is virtual, where to find the vtable, and 54043where in the vtable to look for the member function. If you are using 54044PMFs in an inner loop, you should really reconsider that decision. If 54045that is not an option, you can extract the pointer to the function that 54046would be called for a given object/PMF pair and call it directly inside 54047the inner loop, to save a bit of time. 54048 54049 Note that you still pay the penalty for the call through a function 54050pointer; on most modern architectures, such a call defeats the branch 54051prediction features of the CPU. This is also true of normal virtual 54052function calls. 54053 54054 The syntax for this extension is 54055 54056 extern A a; 54057 extern int (A::*fp)(); 54058 typedef int (*fptr)(A *); 54059 54060 fptr p = (fptr)(a.*fp); 54061 54062 For PMF constants (i.e. expressions of the form '&Klasse::Member'), no 54063object is needed to obtain the address of the function. They can be 54064converted to function pointers directly: 54065 54066 fptr p1 = (fptr)(&A::foo); 54067 54068 You must specify '-Wno-pmf-conversions' to use this extension. 54069 54070 54071File: gcc.info, Node: C++ Attributes, Next: Function Multiversioning, Prev: Bound member functions, Up: C++ Extensions 54072 540737.7 C++-Specific Variable, Function, and Type Attributes 54074======================================================== 54075 54076Some attributes only make sense for C++ programs. 54077 54078'abi_tag ("TAG", ...)' 54079 The 'abi_tag' attribute can be applied to a function, variable, or 54080 class declaration. It modifies the mangled name of the entity to 54081 incorporate the tag name, in order to distinguish the function or 54082 class from an earlier version with a different ABI; perhaps the 54083 class has changed size, or the function has a different return type 54084 that is not encoded in the mangled name. 54085 54086 The attribute can also be applied to an inline namespace, but does 54087 not affect the mangled name of the namespace; in this case it is 54088 only used for '-Wabi-tag' warnings and automatic tagging of 54089 functions and variables. Tagging inline namespaces is generally 54090 preferable to tagging individual declarations, but the latter is 54091 sometimes necessary, such as when only certain members of a class 54092 need to be tagged. 54093 54094 The argument can be a list of strings of arbitrary length. The 54095 strings are sorted on output, so the order of the list is 54096 unimportant. 54097 54098 A redeclaration of an entity must not add new ABI tags, since doing 54099 so would change the mangled name. 54100 54101 The ABI tags apply to a name, so all instantiations and 54102 specializations of a template have the same tags. The attribute 54103 will be ignored if applied to an explicit specialization or 54104 instantiation. 54105 54106 The '-Wabi-tag' flag enables a warning about a class which does not 54107 have all the ABI tags used by its subobjects and virtual functions; 54108 for users with code that needs to coexist with an earlier ABI, 54109 using this option can help to find all affected types that need to 54110 be tagged. 54111 54112 When a type involving an ABI tag is used as the type of a variable 54113 or return type of a function where that tag is not already present 54114 in the signature of the function, the tag is automatically applied 54115 to the variable or function. '-Wabi-tag' also warns about this 54116 situation; this warning can be avoided by explicitly tagging the 54117 variable or function or moving it into a tagged inline namespace. 54118 54119'init_priority (PRIORITY)' 54120 54121 In Standard C++, objects defined at namespace scope are guaranteed 54122 to be initialized in an order in strict accordance with that of 54123 their definitions _in a given translation unit_. No guarantee is 54124 made for initializations across translation units. However, GNU 54125 C++ allows users to control the order of initialization of objects 54126 defined at namespace scope with the 'init_priority' attribute by 54127 specifying a relative PRIORITY, a constant integral expression 54128 currently bounded between 101 and 65535 inclusive. Lower numbers 54129 indicate a higher priority. 54130 54131 In the following example, 'A' would normally be created before 'B', 54132 but the 'init_priority' attribute reverses that order: 54133 54134 Some_Class A __attribute__ ((init_priority (2000))); 54135 Some_Class B __attribute__ ((init_priority (543))); 54136 54137 Note that the particular values of PRIORITY do not matter; only 54138 their relative ordering. 54139 54140'warn_unused' 54141 54142 For C++ types with non-trivial constructors and/or destructors it 54143 is impossible for the compiler to determine whether a variable of 54144 this type is truly unused if it is not referenced. This type 54145 attribute informs the compiler that variables of this type should 54146 be warned about if they appear to be unused, just like variables of 54147 fundamental types. 54148 54149 This attribute is appropriate for types which just represent a 54150 value, such as 'std::string'; it is not appropriate for types which 54151 control a resource, such as 'std::lock_guard'. 54152 54153 This attribute is also accepted in C, but it is unnecessary because 54154 C does not have constructors or destructors. 54155 54156 54157File: gcc.info, Node: Function Multiversioning, Next: Type Traits, Prev: C++ Attributes, Up: C++ Extensions 54158 541597.8 Function Multiversioning 54160============================ 54161 54162With the GNU C++ front end, for x86 targets, you may specify multiple 54163versions of a function, where each function is specialized for a 54164specific target feature. At runtime, the appropriate version of the 54165function is automatically executed depending on the characteristics of 54166the execution platform. Here is an example. 54167 54168 __attribute__ ((target ("default"))) 54169 int foo () 54170 { 54171 // The default version of foo. 54172 return 0; 54173 } 54174 54175 __attribute__ ((target ("sse4.2"))) 54176 int foo () 54177 { 54178 // foo version for SSE4.2 54179 return 1; 54180 } 54181 54182 __attribute__ ((target ("arch=atom"))) 54183 int foo () 54184 { 54185 // foo version for the Intel ATOM processor 54186 return 2; 54187 } 54188 54189 __attribute__ ((target ("arch=amdfam10"))) 54190 int foo () 54191 { 54192 // foo version for the AMD Family 0x10 processors. 54193 return 3; 54194 } 54195 54196 int main () 54197 { 54198 int (*p)() = &foo; 54199 assert ((*p) () == foo ()); 54200 return 0; 54201 } 54202 54203 In the above example, four versions of function foo are created. The 54204first version of foo with the target attribute "default" is the default 54205version. This version gets executed when no other target specific 54206version qualifies for execution on a particular platform. A new version 54207of foo is created by using the same function signature but with a 54208different target string. Function foo is called or a pointer to it is 54209taken just like a regular function. GCC takes care of doing the 54210dispatching to call the right version at runtime. Refer to the GCC wiki 54211on Function Multiversioning 54212(http://gcc.gnu.org/wiki/FunctionMultiVersioning) for more details. 54213 54214 54215File: gcc.info, Node: Type Traits, Next: C++ Concepts, Prev: Function Multiversioning, Up: C++ Extensions 54216 542177.9 Type Traits 54218=============== 54219 54220The C++ front end implements syntactic extensions that allow 54221compile-time determination of various characteristics of a type (or of a 54222pair of types). 54223 54224'__has_nothrow_assign (type)' 54225 If 'type' is 'const'-qualified or is a reference type then the 54226 trait is 'false'. Otherwise if '__has_trivial_assign (type)' is 54227 'true' then the trait is 'true', else if 'type' is a cv-qualified 54228 class or union type with copy assignment operators that are known 54229 not to throw an exception then the trait is 'true', else it is 54230 'false'. Requires: 'type' shall be a complete type, (possibly 54231 cv-qualified) 'void', or an array of unknown bound. 54232 54233'__has_nothrow_copy (type)' 54234 If '__has_trivial_copy (type)' is 'true' then the trait is 'true', 54235 else if 'type' is a cv-qualified class or union type with copy 54236 constructors that are known not to throw an exception then the 54237 trait is 'true', else it is 'false'. Requires: 'type' shall be a 54238 complete type, (possibly cv-qualified) 'void', or an array of 54239 unknown bound. 54240 54241'__has_nothrow_constructor (type)' 54242 If '__has_trivial_constructor (type)' is 'true' then the trait is 54243 'true', else if 'type' is a cv class or union type (or array 54244 thereof) with a default constructor that is known not to throw an 54245 exception then the trait is 'true', else it is 'false'. Requires: 54246 'type' shall be a complete type, (possibly cv-qualified) 'void', or 54247 an array of unknown bound. 54248 54249'__has_trivial_assign (type)' 54250 If 'type' is 'const'- qualified or is a reference type then the 54251 trait is 'false'. Otherwise if '__is_pod (type)' is 'true' then 54252 the trait is 'true', else if 'type' is a cv-qualified class or 54253 union type with a trivial copy assignment ([class.copy]) then the 54254 trait is 'true', else it is 'false'. Requires: 'type' shall be a 54255 complete type, (possibly cv-qualified) 'void', or an array of 54256 unknown bound. 54257 54258'__has_trivial_copy (type)' 54259 If '__is_pod (type)' is 'true' or 'type' is a reference type then 54260 the trait is 'true', else if 'type' is a cv class or union type 54261 with a trivial copy constructor ([class.copy]) then the trait is 54262 'true', else it is 'false'. Requires: 'type' shall be a complete 54263 type, (possibly cv-qualified) 'void', or an array of unknown bound. 54264 54265'__has_trivial_constructor (type)' 54266 If '__is_pod (type)' is 'true' then the trait is 'true', else if 54267 'type' is a cv-qualified class or union type (or array thereof) 54268 with a trivial default constructor ([class.ctor]) then the trait is 54269 'true', else it is 'false'. Requires: 'type' shall be a complete 54270 type, (possibly cv-qualified) 'void', or an array of unknown bound. 54271 54272'__has_trivial_destructor (type)' 54273 If '__is_pod (type)' is 'true' or 'type' is a reference type then 54274 the trait is 'true', else if 'type' is a cv class or union type (or 54275 array thereof) with a trivial destructor ([class.dtor]) then the 54276 trait is 'true', else it is 'false'. Requires: 'type' shall be a 54277 complete type, (possibly cv-qualified) 'void', or an array of 54278 unknown bound. 54279 54280'__has_virtual_destructor (type)' 54281 If 'type' is a class type with a virtual destructor ([class.dtor]) 54282 then the trait is 'true', else it is 'false'. Requires: 'type' 54283 shall be a complete type, (possibly cv-qualified) 'void', or an 54284 array of unknown bound. 54285 54286'__is_abstract (type)' 54287 If 'type' is an abstract class ([class.abstract]) then the trait is 54288 'true', else it is 'false'. Requires: 'type' shall be a complete 54289 type, (possibly cv-qualified) 'void', or an array of unknown bound. 54290 54291'__is_base_of (base_type, derived_type)' 54292 If 'base_type' is a base class of 'derived_type' ([class.derived]) 54293 then the trait is 'true', otherwise it is 'false'. Top-level 54294 cv-qualifications of 'base_type' and 'derived_type' are ignored. 54295 For the purposes of this trait, a class type is considered is own 54296 base. Requires: if '__is_class (base_type)' and '__is_class 54297 (derived_type)' are 'true' and 'base_type' and 'derived_type' are 54298 not the same type (disregarding cv-qualifiers), 'derived_type' 54299 shall be a complete type. A diagnostic is produced if this 54300 requirement is not met. 54301 54302'__is_class (type)' 54303 If 'type' is a cv-qualified class type, and not a union type 54304 ([basic.compound]) the trait is 'true', else it is 'false'. 54305 54306'__is_empty (type)' 54307 If '__is_class (type)' is 'false' then the trait is 'false'. 54308 Otherwise 'type' is considered empty if and only if: 'type' has no 54309 non-static data members, or all non-static data members, if any, 54310 are bit-fields of length 0, and 'type' has no virtual members, and 54311 'type' has no virtual base classes, and 'type' has no base classes 54312 'base_type' for which '__is_empty (base_type)' is 'false'. 54313 Requires: 'type' shall be a complete type, (possibly cv-qualified) 54314 'void', or an array of unknown bound. 54315 54316'__is_enum (type)' 54317 If 'type' is a cv enumeration type ([basic.compound]) the trait is 54318 'true', else it is 'false'. 54319 54320'__is_literal_type (type)' 54321 If 'type' is a literal type ([basic.types]) the trait is 'true', 54322 else it is 'false'. Requires: 'type' shall be a complete type, 54323 (possibly cv-qualified) 'void', or an array of unknown bound. 54324 54325'__is_pod (type)' 54326 If 'type' is a cv POD type ([basic.types]) then the trait is 54327 'true', else it is 'false'. Requires: 'type' shall be a complete 54328 type, (possibly cv-qualified) 'void', or an array of unknown bound. 54329 54330'__is_polymorphic (type)' 54331 If 'type' is a polymorphic class ([class.virtual]) then the trait 54332 is 'true', else it is 'false'. Requires: 'type' shall be a 54333 complete type, (possibly cv-qualified) 'void', or an array of 54334 unknown bound. 54335 54336'__is_standard_layout (type)' 54337 If 'type' is a standard-layout type ([basic.types]) the trait is 54338 'true', else it is 'false'. Requires: 'type' shall be a complete 54339 type, (possibly cv-qualified) 'void', or an array of unknown bound. 54340 54341'__is_trivial (type)' 54342 If 'type' is a trivial type ([basic.types]) the trait is 'true', 54343 else it is 'false'. Requires: 'type' shall be a complete type, 54344 (possibly cv-qualified) 'void', or an array of unknown bound. 54345 54346'__is_union (type)' 54347 If 'type' is a cv union type ([basic.compound]) the trait is 54348 'true', else it is 'false'. 54349 54350'__underlying_type (type)' 54351 The underlying type of 'type'. Requires: 'type' shall be an 54352 enumeration type ([dcl.enum]). 54353 54354'__integer_pack (length)' 54355 When used as the pattern of a pack expansion within a template 54356 definition, expands to a template argument pack containing integers 54357 from '0' to 'length-1'. This is provided for efficient 54358 implementation of 'std::make_integer_sequence'. 54359 54360 54361File: gcc.info, Node: C++ Concepts, Next: Deprecated Features, Prev: Type Traits, Up: C++ Extensions 54362 543637.10 C++ Concepts 54364================= 54365 54366C++ concepts provide much-improved support for generic programming. In 54367particular, they allow the specification of constraints on template 54368arguments. The constraints are used to extend the usual overloading and 54369partial specialization capabilities of the language, allowing generic 54370data structures and algorithms to be "refined" based on their properties 54371rather than their type names. 54372 54373 The following keywords are reserved for concepts. 54374 54375'assumes' 54376 States an expression as an assumption, and if possible, verifies 54377 that the assumption is valid. For example, 'assume(n > 0)'. 54378 54379'axiom' 54380 Introduces an axiom definition. Axioms introduce requirements on 54381 values. 54382 54383'forall' 54384 Introduces a universally quantified object in an axiom. For 54385 example, 'forall (int n) n + 0 == n'). 54386 54387'concept' 54388 Introduces a concept definition. Concepts are sets of syntactic 54389 and semantic requirements on types and their values. 54390 54391'requires' 54392 Introduces constraints on template arguments or requirements for a 54393 member function of a class template. 54394 54395 The front end also exposes a number of internal mechanism that can be 54396used to simplify the writing of type traits. Note that some of these 54397traits are likely to be removed in the future. 54398 54399'__is_same (type1, type2)' 54400 A binary type trait: 'true' whenever the type arguments are the 54401 same. 54402 54403 54404File: gcc.info, Node: Deprecated Features, Next: Backwards Compatibility, Prev: C++ Concepts, Up: C++ Extensions 54405 544067.11 Deprecated Features 54407======================== 54408 54409In the past, the GNU C++ compiler was extended to experiment with new 54410features, at a time when the C++ language was still evolving. Now that 54411the C++ standard is complete, some of those features are superseded by 54412superior alternatives. Using the old features might cause a warning in 54413some cases that the feature will be dropped in the future. In other 54414cases, the feature might be gone already. 54415 54416 G++ allows a virtual function returning 'void *' to be overridden by 54417one returning a different pointer type. This extension to the covariant 54418return type rules is now deprecated and will be removed from a future 54419version. 54420 54421 The use of default arguments in function pointers, function typedefs 54422and other places where they are not permitted by the standard is 54423deprecated and will be removed from a future version of G++. 54424 54425 G++ allows floating-point literals to appear in integral constant 54426expressions, e.g. ' enum E { e = int(2.2 * 3.7) } ' This extension is 54427deprecated and will be removed from a future version. 54428 54429 G++ allows static data members of const floating-point type to be 54430declared with an initializer in a class definition. The standard only 54431allows initializers for static members of const integral types and const 54432enumeration types so this extension has been deprecated and will be 54433removed from a future version. 54434 54435 G++ allows attributes to follow a parenthesized direct initializer, 54436e.g. ' int f (0) __attribute__ ((something)); ' This extension has been 54437ignored since G++ 3.3 and is deprecated. 54438 54439 G++ allows anonymous structs and unions to have members that are not 54440public non-static data members (i.e. fields). These extensions are 54441deprecated. 54442 54443 54444File: gcc.info, Node: Backwards Compatibility, Prev: Deprecated Features, Up: C++ Extensions 54445 544467.12 Backwards Compatibility 54447============================ 54448 54449Now that there is a definitive ISO standard C++, G++ has a specification 54450to adhere to. The C++ language evolved over time, and features that 54451used to be acceptable in previous drafts of the standard, such as the 54452ARM [Annotated C++ Reference Manual], are no longer accepted. In order 54453to allow compilation of C++ written to such drafts, G++ contains some 54454backwards compatibilities. _All such backwards compatibility features 54455are liable to disappear in future versions of G++._ They should be 54456considered deprecated. *Note Deprecated Features::. 54457 54458'Implicit C language' 54459 Old C system header files did not contain an 'extern "C" {...}' 54460 scope to set the language. On such systems, all system header 54461 files are implicitly scoped inside a C language scope. Such 54462 headers must correctly prototype function argument types, there is 54463 no leeway for '()' to indicate an unspecified set of arguments. 54464 54465 54466File: gcc.info, Node: Objective-C, Next: Compatibility, Prev: C++ Extensions, Up: Top 54467 544688 GNU Objective-C Features 54469************************** 54470 54471This document is meant to describe some of the GNU Objective-C features. 54472It is not intended to teach you Objective-C. There are several resources 54473on the Internet that present the language. 54474 54475* Menu: 54476 54477* GNU Objective-C runtime API:: 54478* Executing code before main:: 54479* Type encoding:: 54480* Garbage Collection:: 54481* Constant string objects:: 54482* compatibility_alias:: 54483* Exceptions:: 54484* Synchronization:: 54485* Fast enumeration:: 54486* Messaging with the GNU Objective-C runtime:: 54487 54488 54489File: gcc.info, Node: GNU Objective-C runtime API, Next: Executing code before main, Up: Objective-C 54490 544918.1 GNU Objective-C Runtime API 54492=============================== 54493 54494This section is specific for the GNU Objective-C runtime. If you are 54495using a different runtime, you can skip it. 54496 54497 The GNU Objective-C runtime provides an API that allows you to interact 54498with the Objective-C runtime system, querying the live runtime 54499structures and even manipulating them. This allows you for example to 54500inspect and navigate classes, methods and protocols; to define new 54501classes or new methods, and even to modify existing classes or 54502protocols. 54503 54504 If you are using a "Foundation" library such as GNUstep-Base, this 54505library will provide you with a rich set of functionality to do most of 54506the inspection tasks, and you probably will only need direct access to 54507the GNU Objective-C runtime API to define new classes or methods. 54508 54509* Menu: 54510 54511* Modern GNU Objective-C runtime API:: 54512* Traditional GNU Objective-C runtime API:: 54513 54514 54515File: gcc.info, Node: Modern GNU Objective-C runtime API, Next: Traditional GNU Objective-C runtime API, Up: GNU Objective-C runtime API 54516 545178.1.1 Modern GNU Objective-C Runtime API 54518---------------------------------------- 54519 54520The GNU Objective-C runtime provides an API which is similar to the one 54521provided by the "Objective-C 2.0" Apple/NeXT Objective-C runtime. The 54522API is documented in the public header files of the GNU Objective-C 54523runtime: 54524 54525 * 'objc/objc.h': this is the basic Objective-C header file, defining 54526 the basic Objective-C types such as 'id', 'Class' and 'BOOL'. You 54527 have to include this header to do almost anything with Objective-C. 54528 54529 * 'objc/runtime.h': this header declares most of the public runtime 54530 API functions allowing you to inspect and manipulate the 54531 Objective-C runtime data structures. These functions are fairly 54532 standardized across Objective-C runtimes and are almost identical 54533 to the Apple/NeXT Objective-C runtime ones. It does not declare 54534 functions in some specialized areas (constructing and forwarding 54535 message invocations, threading) which are in the other headers 54536 below. You have to include 'objc/objc.h' and 'objc/runtime.h' to 54537 use any of the functions, such as 'class_getName()', declared in 54538 'objc/runtime.h'. 54539 54540 * 'objc/message.h': this header declares public functions used to 54541 construct, deconstruct and forward message invocations. Because 54542 messaging is done in quite a different way on different runtimes, 54543 functions in this header are specific to the GNU Objective-C 54544 runtime implementation. 54545 54546 * 'objc/objc-exception.h': this header declares some public functions 54547 related to Objective-C exceptions. For example functions in this 54548 header allow you to throw an Objective-C exception from plain C/C++ 54549 code. 54550 54551 * 'objc/objc-sync.h': this header declares some public functions 54552 related to the Objective-C '@synchronized()' syntax, allowing you 54553 to emulate an Objective-C '@synchronized()' block in plain C/C++ 54554 code. 54555 54556 * 'objc/thr.h': this header declares a public runtime API threading 54557 layer that is only provided by the GNU Objective-C runtime. It 54558 declares functions such as 'objc_mutex_lock()', which provide a 54559 platform-independent set of threading functions. 54560 54561 The header files contain detailed documentation for each function in 54562the GNU Objective-C runtime API. 54563 54564 54565File: gcc.info, Node: Traditional GNU Objective-C runtime API, Prev: Modern GNU Objective-C runtime API, Up: GNU Objective-C runtime API 54566 545678.1.2 Traditional GNU Objective-C Runtime API 54568--------------------------------------------- 54569 54570The GNU Objective-C runtime used to provide a different API, which we 54571call the "traditional" GNU Objective-C runtime API. Functions belonging 54572to this API are easy to recognize because they use a different naming 54573convention, such as 'class_get_super_class()' (traditional API) instead 54574of 'class_getSuperclass()' (modern API). Software using this API 54575includes the file 'objc/objc-api.h' where it is declared. 54576 54577 Starting with GCC 4.7.0, the traditional GNU runtime API is no longer 54578available. 54579 54580 54581File: gcc.info, Node: Executing code before main, Next: Type encoding, Prev: GNU Objective-C runtime API, Up: Objective-C 54582 545838.2 '+load': Executing Code before 'main' 54584========================================= 54585 54586This section is specific for the GNU Objective-C runtime. If you are 54587using a different runtime, you can skip it. 54588 54589 The GNU Objective-C runtime provides a way that allows you to execute 54590code before the execution of the program enters the 'main' function. 54591The code is executed on a per-class and a per-category basis, through a 54592special class method '+load'. 54593 54594 This facility is very useful if you want to initialize global variables 54595which can be accessed by the program directly, without sending a message 54596to the class first. The usual way to initialize global variables, in 54597the '+initialize' method, might not be useful because '+initialize' is 54598only called when the first message is sent to a class object, which in 54599some cases could be too late. 54600 54601 Suppose for example you have a 'FileStream' class that declares 54602'Stdin', 'Stdout' and 'Stderr' as global variables, like below: 54603 54604 54605 FileStream *Stdin = nil; 54606 FileStream *Stdout = nil; 54607 FileStream *Stderr = nil; 54608 54609 @implementation FileStream 54610 54611 + (void)initialize 54612 { 54613 Stdin = [[FileStream new] initWithFd:0]; 54614 Stdout = [[FileStream new] initWithFd:1]; 54615 Stderr = [[FileStream new] initWithFd:2]; 54616 } 54617 54618 /* Other methods here */ 54619 @end 54620 54621 54622 In this example, the initialization of 'Stdin', 'Stdout' and 'Stderr' 54623in '+initialize' occurs too late. The programmer can send a message to 54624one of these objects before the variables are actually initialized, thus 54625sending messages to the 'nil' object. The '+initialize' method which 54626actually initializes the global variables is not invoked until the first 54627message is sent to the class object. The solution would require these 54628variables to be initialized just before entering 'main'. 54629 54630 The correct solution of the above problem is to use the '+load' method 54631instead of '+initialize': 54632 54633 54634 @implementation FileStream 54635 54636 + (void)load 54637 { 54638 Stdin = [[FileStream new] initWithFd:0]; 54639 Stdout = [[FileStream new] initWithFd:1]; 54640 Stderr = [[FileStream new] initWithFd:2]; 54641 } 54642 54643 /* Other methods here */ 54644 @end 54645 54646 54647 The '+load' is a method that is not overridden by categories. If a 54648class and a category of it both implement '+load', both methods are 54649invoked. This allows some additional initializations to be performed in 54650a category. 54651 54652 This mechanism is not intended to be a replacement for '+initialize'. 54653You should be aware of its limitations when you decide to use it instead 54654of '+initialize'. 54655 54656* Menu: 54657 54658* What you can and what you cannot do in +load:: 54659 54660 54661File: gcc.info, Node: What you can and what you cannot do in +load, Up: Executing code before main 54662 546638.2.1 What You Can and Cannot Do in '+load' 54664------------------------------------------- 54665 54666'+load' is to be used only as a last resort. Because it is executed 54667very early, most of the Objective-C runtime machinery will not be ready 54668when '+load' is executed; hence '+load' works best for executing C code 54669that is independent on the Objective-C runtime. 54670 54671 The '+load' implementation in the GNU runtime guarantees you the 54672following things: 54673 54674 * you can write whatever C code you like; 54675 54676 * you can allocate and send messages to objects whose class is 54677 implemented in the same file; 54678 54679 * the '+load' implementation of all super classes of a class are 54680 executed before the '+load' of that class is executed; 54681 54682 * the '+load' implementation of a class is executed before the 54683 '+load' implementation of any category. 54684 54685 In particular, the following things, even if they can work in a 54686particular case, are not guaranteed: 54687 54688 * allocation of or sending messages to arbitrary objects; 54689 54690 * allocation of or sending messages to objects whose classes have a 54691 category implemented in the same file; 54692 54693 * sending messages to Objective-C constant strings ('@"this is a 54694 constant string"'); 54695 54696 You should make no assumptions about receiving '+load' in sibling 54697classes when you write '+load' of a class. The order in which sibling 54698classes receive '+load' is not guaranteed. 54699 54700 The order in which '+load' and '+initialize' are called could be 54701problematic if this matters. If you don't allocate objects inside 54702'+load', it is guaranteed that '+load' is called before '+initialize'. 54703If you create an object inside '+load' the '+initialize' method of 54704object's class is invoked even if '+load' was not invoked. Note if you 54705explicitly call '+load' on a class, '+initialize' will be called first. 54706To avoid possible problems try to implement only one of these methods. 54707 54708 The '+load' method is also invoked when a bundle is dynamically loaded 54709into your running program. This happens automatically without any 54710intervening operation from you. When you write bundles and you need to 54711write '+load' you can safely create and send messages to objects whose 54712classes already exist in the running program. The same restrictions as 54713above apply to classes defined in bundle. 54714 54715 54716File: gcc.info, Node: Type encoding, Next: Garbage Collection, Prev: Executing code before main, Up: Objective-C 54717 547188.3 Type Encoding 54719================= 54720 54721This is an advanced section. Type encodings are used extensively by the 54722compiler and by the runtime, but you generally do not need to know about 54723them to use Objective-C. 54724 54725 The Objective-C compiler generates type encodings for all the types. 54726These type encodings are used at runtime to find out information about 54727selectors and methods and about objects and classes. 54728 54729 The types are encoded in the following way: 54730 54731'_Bool' 'B' 54732'char' 'c' 54733'unsigned char' 'C' 54734'short' 's' 54735'unsigned short' 'S' 54736'int' 'i' 54737'unsigned int' 'I' 54738'long' 'l' 54739'unsigned long' 'L' 54740'long long' 'q' 54741'unsigned long 'Q' 54742long' 54743'float' 'f' 54744'double' 'd' 54745'long double' 'D' 54746'void' 'v' 54747'id' '@' 54748'Class' '#' 54749'SEL' ':' 54750'char*' '*' 54751'enum' an 'enum' is encoded exactly as the integer type 54752 that the compiler uses for it, which depends on the 54753 enumeration values. Often the compiler users 54754 'unsigned int', which is then encoded as 'I'. 54755unknown type '?' 54756Complex types 'j' followed by the inner type. For example 54757 '_Complex double' is encoded as "jd". 54758bit-fields 'b' followed by the starting position of the 54759 bit-field, the type of the bit-field and the size of 54760 the bit-field (the bit-fields encoding was changed 54761 from the NeXT's compiler encoding, see below) 54762 54763 The encoding of bit-fields has changed to allow bit-fields to be 54764properly handled by the runtime functions that compute sizes and 54765alignments of types that contain bit-fields. The previous encoding 54766contained only the size of the bit-field. Using only this information 54767it is not possible to reliably compute the size occupied by the 54768bit-field. This is very important in the presence of the Boehm's 54769garbage collector because the objects are allocated using the typed 54770memory facility available in this collector. The typed memory 54771allocation requires information about where the pointers are located 54772inside the object. 54773 54774 The position in the bit-field is the position, counting in bits, of the 54775bit closest to the beginning of the structure. 54776 54777 The non-atomic types are encoded as follows: 54778 54779pointers '^' followed by the pointed type. 54780arrays '[' followed by the number of elements in the array 54781 followed by the type of the elements followed by ']' 54782structures '{' followed by the name of the structure (or '?' if the 54783 structure is unnamed), the '=' sign, the type of the 54784 members and by '}' 54785unions '(' followed by the name of the structure (or '?' if the 54786 union is unnamed), the '=' sign, the type of the members 54787 followed by ')' 54788vectors '![' followed by the vector_size (the number of bytes 54789 composing the vector) followed by a comma, followed by 54790 the alignment (in bytes) of the vector, followed by the 54791 type of the elements followed by ']' 54792 54793 Here are some types and their encodings, as they are generated by the 54794compiler on an i386 machine: 54795 54796 54797Objective-C type Compiler encoding 54798 int a[10]; '[10i]' 54799 struct { '{?=i[3f]b128i3b131i2c}' 54800 int i; 54801 float f[3]; 54802 int a:3; 54803 int b:2; 54804 char c; 54805 } 54806 int a __attribute__ ((vector_size (16)));'![16,16i]' (alignment 54807 depends on the machine) 54808 54809 54810 In addition to the types the compiler also encodes the type specifiers. 54811The table below describes the encoding of the current Objective-C type 54812specifiers: 54813 54814 54815Specifier Encoding 54816'const' 'r' 54817'in' 'n' 54818'inout' 'N' 54819'out' 'o' 54820'bycopy' 'O' 54821'byref' 'R' 54822'oneway' 'V' 54823 54824 54825 The type specifiers are encoded just before the type. Unlike types 54826however, the type specifiers are only encoded when they appear in method 54827argument types. 54828 54829 Note how 'const' interacts with pointers: 54830 54831 54832Objective-C type Compiler encoding 54833 const int 'ri' 54834 const int* '^ri' 54835 int *const 'r^i' 54836 54837 54838 'const int*' is a pointer to a 'const int', and so is encoded as '^ri'. 54839'int* const', instead, is a 'const' pointer to an 'int', and so is 54840encoded as 'r^i'. 54841 54842 Finally, there is a complication when encoding 'const char *' versus 54843'char * const'. Because 'char *' is encoded as '*' and not as '^c', 54844there is no way to express the fact that 'r' applies to the pointer or 54845to the pointee. 54846 54847 Hence, it is assumed as a convention that 'r*' means 'const char *' 54848(since it is what is most often meant), and there is no way to encode 54849'char *const'. 'char *const' would simply be encoded as '*', and the 54850'const' is lost. 54851 54852* Menu: 54853 54854* Legacy type encoding:: 54855* @encode:: 54856* Method signatures:: 54857 54858 54859File: gcc.info, Node: Legacy type encoding, Next: @encode, Up: Type encoding 54860 548618.3.1 Legacy Type Encoding 54862-------------------------- 54863 54864Unfortunately, historically GCC used to have a number of bugs in its 54865encoding code. The NeXT runtime expects GCC to emit type encodings in 54866this historical format (compatible with GCC-3.3), so when using the NeXT 54867runtime, GCC will introduce on purpose a number of incorrect encodings: 54868 54869 * the read-only qualifier of the pointee gets emitted before the '^'. 54870 The read-only qualifier of the pointer itself gets ignored, unless 54871 it is a typedef. Also, the 'r' is only emitted for the outermost 54872 type. 54873 54874 * 32-bit longs are encoded as 'l' or 'L', but not always. For 54875 typedefs, the compiler uses 'i' or 'I' instead if encoding a struct 54876 field or a pointer. 54877 54878 * 'enum's are always encoded as 'i' (int) even if they are actually 54879 unsigned or long. 54880 54881 In addition to that, the NeXT runtime uses a different encoding for 54882bitfields. It encodes them as 'b' followed by the size, without a bit 54883offset or the underlying field type. 54884 54885 54886File: gcc.info, Node: @encode, Next: Method signatures, Prev: Legacy type encoding, Up: Type encoding 54887 548888.3.2 '@encode' 54889--------------- 54890 54891GNU Objective-C supports the '@encode' syntax that allows you to create 54892a type encoding from a C/Objective-C type. For example, '@encode(int)' 54893is compiled by the compiler into '"i"'. 54894 54895 '@encode' does not support type qualifiers other than 'const'. For 54896example, '@encode(const char*)' is valid and is compiled into '"r*"', 54897while '@encode(bycopy char *)' is invalid and will cause a compilation 54898error. 54899 54900 54901File: gcc.info, Node: Method signatures, Prev: @encode, Up: Type encoding 54902 549038.3.3 Method Signatures 54904----------------------- 54905 54906This section documents the encoding of method types, which is rarely 54907needed to use Objective-C. You should skip it at a first reading; the 54908runtime provides functions that will work on methods and can walk 54909through the list of parameters and interpret them for you. These 54910functions are part of the public "API" and are the preferred way to 54911interact with method signatures from user code. 54912 54913 But if you need to debug a problem with method signatures and need to 54914know how they are implemented (i.e., the "ABI"), read on. 54915 54916 Methods have their "signature" encoded and made available to the 54917runtime. The "signature" encodes all the information required to 54918dynamically build invocations of the method at runtime: return type and 54919arguments. 54920 54921 The "signature" is a null-terminated string, composed of the following: 54922 54923 * The return type, including type qualifiers. For example, a method 54924 returning 'int' would have 'i' here. 54925 54926 * The total size (in bytes) required to pass all the parameters. 54927 This includes the two hidden parameters (the object 'self' and the 54928 method selector '_cmd'). 54929 54930 * Each argument, with the type encoding, followed by the offset (in 54931 bytes) of the argument in the list of parameters. 54932 54933 For example, a method with no arguments and returning 'int' would have 54934the signature 'i8@0:4' if the size of a pointer is 4. The signature is 54935interpreted as follows: the 'i' is the return type (an 'int'), the '8' 54936is the total size of the parameters in bytes (two pointers each of size 549374), the '@0' is the first parameter (an object at byte offset '0') and 54938':4' is the second parameter (a 'SEL' at byte offset '4'). 54939 54940 You can easily find more examples by running the "strings" program on 54941an Objective-C object file compiled by GCC. You'll see a lot of strings 54942that look very much like 'i8@0:4'. They are signatures of Objective-C 54943methods. 54944 54945 54946File: gcc.info, Node: Garbage Collection, Next: Constant string objects, Prev: Type encoding, Up: Objective-C 54947 549488.4 Garbage Collection 54949====================== 54950 54951This section is specific for the GNU Objective-C runtime. If you are 54952using a different runtime, you can skip it. 54953 54954 Support for garbage collection with the GNU runtime has been added by 54955using a powerful conservative garbage collector, known as the 54956Boehm-Demers-Weiser conservative garbage collector. 54957 54958 To enable the support for it you have to configure the compiler using 54959an additional argument, '--enable-objc-gc'. This will build the 54960boehm-gc library, and build an additional runtime library which has 54961several enhancements to support the garbage collector. The new library 54962has a new name, 'libobjc_gc.a' to not conflict with the 54963non-garbage-collected library. 54964 54965 When the garbage collector is used, the objects are allocated using the 54966so-called typed memory allocation mechanism available in the 54967Boehm-Demers-Weiser collector. This mode requires precise information 54968on where pointers are located inside objects. This information is 54969computed once per class, immediately after the class has been 54970initialized. 54971 54972 There is a new runtime function 'class_ivar_set_gcinvisible()' which 54973can be used to declare a so-called "weak pointer" reference. Such a 54974pointer is basically hidden for the garbage collector; this can be 54975useful in certain situations, especially when you want to keep track of 54976the allocated objects, yet allow them to be collected. This kind of 54977pointers can only be members of objects, you cannot declare a global 54978pointer as a weak reference. Every type which is a pointer type can be 54979declared a weak pointer, including 'id', 'Class' and 'SEL'. 54980 54981 Here is an example of how to use this feature. Suppose you want to 54982implement a class whose instances hold a weak pointer reference; the 54983following class does this: 54984 54985 54986 @interface WeakPointer : Object 54987 { 54988 const void* weakPointer; 54989 } 54990 54991 - initWithPointer:(const void*)p; 54992 - (const void*)weakPointer; 54993 @end 54994 54995 54996 @implementation WeakPointer 54997 54998 + (void)initialize 54999 { 55000 if (self == objc_lookUpClass ("WeakPointer")) 55001 class_ivar_set_gcinvisible (self, "weakPointer", YES); 55002 } 55003 55004 - initWithPointer:(const void*)p 55005 { 55006 weakPointer = p; 55007 return self; 55008 } 55009 55010 - (const void*)weakPointer 55011 { 55012 return weakPointer; 55013 } 55014 55015 @end 55016 55017 55018 Weak pointers are supported through a new type character specifier 55019represented by the '!' character. The 'class_ivar_set_gcinvisible()' 55020function adds or removes this specifier to the string type description 55021of the instance variable named as argument. 55022 55023 55024File: gcc.info, Node: Constant string objects, Next: compatibility_alias, Prev: Garbage Collection, Up: Objective-C 55025 550268.5 Constant String Objects 55027=========================== 55028 55029GNU Objective-C provides constant string objects that are generated 55030directly by the compiler. You declare a constant string object by 55031prefixing a C constant string with the character '@': 55032 55033 id myString = @"this is a constant string object"; 55034 55035 The constant string objects are by default instances of the 55036'NXConstantString' class which is provided by the GNU Objective-C 55037runtime. To get the definition of this class you must include the 55038'objc/NXConstStr.h' header file. 55039 55040 User defined libraries may want to implement their own constant string 55041class. To be able to support them, the GNU Objective-C compiler 55042provides a new command line options 55043'-fconstant-string-class=CLASS-NAME'. The provided class should adhere 55044to a strict structure, the same as 'NXConstantString''s structure: 55045 55046 55047 @interface MyConstantStringClass 55048 { 55049 Class isa; 55050 char *c_string; 55051 unsigned int len; 55052 } 55053 @end 55054 55055 55056 'NXConstantString' inherits from 'Object'; user class libraries may 55057choose to inherit the customized constant string class from a different 55058class than 'Object'. There is no requirement in the methods the 55059constant string class has to implement, but the final ivar layout of the 55060class must be the compatible with the given structure. 55061 55062 When the compiler creates the statically allocated constant string 55063object, the 'c_string' field will be filled by the compiler with the 55064string; the 'length' field will be filled by the compiler with the 55065string length; the 'isa' pointer will be filled with 'NULL' by the 55066compiler, and it will later be fixed up automatically at runtime by the 55067GNU Objective-C runtime library to point to the class which was set by 55068the '-fconstant-string-class' option when the object file is loaded (if 55069you wonder how it works behind the scenes, the name of the class to use, 55070and the list of static objects to fixup, are stored by the compiler in 55071the object file in a place where the GNU runtime library will find them 55072at runtime). 55073 55074 As a result, when a file is compiled with the '-fconstant-string-class' 55075option, all the constant string objects will be instances of the class 55076specified as argument to this option. It is possible to have multiple 55077compilation units referring to different constant string classes, 55078neither the compiler nor the linker impose any restrictions in doing 55079this. 55080 55081 55082File: gcc.info, Node: compatibility_alias, Next: Exceptions, Prev: Constant string objects, Up: Objective-C 55083 550848.6 'compatibility_alias' 55085========================= 55086 55087The keyword '@compatibility_alias' allows you to define a class name as 55088equivalent to another class name. For example: 55089 55090 @compatibility_alias WOApplication GSWApplication; 55091 55092 tells the compiler that each time it encounters 'WOApplication' as a 55093class name, it should replace it with 'GSWApplication' (that is, 55094'WOApplication' is just an alias for 'GSWApplication'). 55095 55096 There are some constraints on how this can be used-- 55097 55098 * 'WOApplication' (the alias) must not be an existing class; 55099 55100 * 'GSWApplication' (the real class) must be an existing class. 55101 55102 55103File: gcc.info, Node: Exceptions, Next: Synchronization, Prev: compatibility_alias, Up: Objective-C 55104 551058.7 Exceptions 55106============== 55107 55108GNU Objective-C provides exception support built into the language, as 55109in the following example: 55110 55111 @try { 55112 ... 55113 @throw expr; 55114 ... 55115 } 55116 @catch (AnObjCClass *exc) { 55117 ... 55118 @throw expr; 55119 ... 55120 @throw; 55121 ... 55122 } 55123 @catch (AnotherClass *exc) { 55124 ... 55125 } 55126 @catch (id allOthers) { 55127 ... 55128 } 55129 @finally { 55130 ... 55131 @throw expr; 55132 ... 55133 } 55134 55135 The '@throw' statement may appear anywhere in an Objective-C or 55136Objective-C++ program; when used inside of a '@catch' block, the 55137'@throw' may appear without an argument (as shown above), in which case 55138the object caught by the '@catch' will be rethrown. 55139 55140 Note that only (pointers to) Objective-C objects may be thrown and 55141caught using this scheme. When an object is thrown, it will be caught 55142by the nearest '@catch' clause capable of handling objects of that type, 55143analogously to how 'catch' blocks work in C++ and Java. A '@catch(id 55144...)' clause (as shown above) may also be provided to catch any and all 55145Objective-C exceptions not caught by previous '@catch' clauses (if any). 55146 55147 The '@finally' clause, if present, will be executed upon exit from the 55148immediately preceding '@try ... @catch' section. This will happen 55149regardless of whether any exceptions are thrown, caught or rethrown 55150inside the '@try ... @catch' section, analogously to the behavior of the 55151'finally' clause in Java. 55152 55153 There are several caveats to using the new exception mechanism: 55154 55155 * The '-fobjc-exceptions' command line option must be used when 55156 compiling Objective-C files that use exceptions. 55157 55158 * With the GNU runtime, exceptions are always implemented as "native" 55159 exceptions and it is recommended that the '-fexceptions' and 55160 '-shared-libgcc' options are used when linking. 55161 55162 * With the NeXT runtime, although currently designed to be binary 55163 compatible with 'NS_HANDLER'-style idioms provided by the 55164 'NSException' class, the new exceptions can only be used on Mac OS 55165 X 10.3 (Panther) and later systems, due to additional functionality 55166 needed in the NeXT Objective-C runtime. 55167 55168 * As mentioned above, the new exceptions do not support handling 55169 types other than Objective-C objects. Furthermore, when used from 55170 Objective-C++, the Objective-C exception model does not 55171 interoperate with C++ exceptions at this time. This means you 55172 cannot '@throw' an exception from Objective-C and 'catch' it in 55173 C++, or vice versa (i.e., 'throw ... @catch'). 55174 55175 55176File: gcc.info, Node: Synchronization, Next: Fast enumeration, Prev: Exceptions, Up: Objective-C 55177 551788.8 Synchronization 55179=================== 55180 55181GNU Objective-C provides support for synchronized blocks: 55182 55183 @synchronized (ObjCClass *guard) { 55184 ... 55185 } 55186 55187 Upon entering the '@synchronized' block, a thread of execution shall 55188first check whether a lock has been placed on the corresponding 'guard' 55189object by another thread. If it has, the current thread shall wait 55190until the other thread relinquishes its lock. Once 'guard' becomes 55191available, the current thread will place its own lock on it, execute the 55192code contained in the '@synchronized' block, and finally relinquish the 55193lock (thereby making 'guard' available to other threads). 55194 55195 Unlike Java, Objective-C does not allow for entire methods to be marked 55196'@synchronized'. Note that throwing exceptions out of '@synchronized' 55197blocks is allowed, and will cause the guarding object to be unlocked 55198properly. 55199 55200 Because of the interactions between synchronization and exception 55201handling, you can only use '@synchronized' when compiling with 55202exceptions enabled, that is with the command line option 55203'-fobjc-exceptions'. 55204 55205 55206File: gcc.info, Node: Fast enumeration, Next: Messaging with the GNU Objective-C runtime, Prev: Synchronization, Up: Objective-C 55207 552088.9 Fast Enumeration 55209==================== 55210 55211* Menu: 55212 55213* Using fast enumeration:: 55214* c99-like fast enumeration syntax:: 55215* Fast enumeration details:: 55216* Fast enumeration protocol:: 55217 55218 55219File: gcc.info, Node: Using fast enumeration, Next: c99-like fast enumeration syntax, Up: Fast enumeration 55220 552218.9.1 Using Fast Enumeration 55222---------------------------- 55223 55224GNU Objective-C provides support for the fast enumeration syntax: 55225 55226 id array = ...; 55227 id object; 55228 55229 for (object in array) 55230 { 55231 /* Do something with 'object' */ 55232 } 55233 55234 'array' needs to be an Objective-C object (usually a collection object, 55235for example an array, a dictionary or a set) which implements the "Fast 55236Enumeration Protocol" (see below). If you are using a Foundation 55237library such as GNUstep Base or Apple Cocoa Foundation, all collection 55238objects in the library implement this protocol and can be used in this 55239way. 55240 55241 The code above would iterate over all objects in 'array'. For each of 55242them, it assigns it to 'object', then executes the 'Do something with 55243'object'' statements. 55244 55245 Here is a fully worked-out example using a Foundation library (which 55246provides the implementation of 'NSArray', 'NSString' and 'NSLog'): 55247 55248 NSArray *array = [NSArray arrayWithObjects: @"1", @"2", @"3", nil]; 55249 NSString *object; 55250 55251 for (object in array) 55252 NSLog (@"Iterating over %@", object); 55253 55254 55255File: gcc.info, Node: c99-like fast enumeration syntax, Next: Fast enumeration details, Prev: Using fast enumeration, Up: Fast enumeration 55256 552578.9.2 C99-Like Fast Enumeration Syntax 55258-------------------------------------- 55259 55260A c99-like declaration syntax is also allowed: 55261 55262 id array = ...; 55263 55264 for (id object in array) 55265 { 55266 /* Do something with 'object' */ 55267 } 55268 55269 this is completely equivalent to: 55270 55271 id array = ...; 55272 55273 { 55274 id object; 55275 for (object in array) 55276 { 55277 /* Do something with 'object' */ 55278 } 55279 } 55280 55281 but can save some typing. 55282 55283 Note that the option '-std=c99' is not required to allow this syntax in 55284Objective-C. 55285 55286 55287File: gcc.info, Node: Fast enumeration details, Next: Fast enumeration protocol, Prev: c99-like fast enumeration syntax, Up: Fast enumeration 55288 552898.9.3 Fast Enumeration Details 55290------------------------------ 55291 55292Here is a more technical description with the gory details. Consider 55293the code 55294 55295 for (OBJECT EXPRESSION in COLLECTION EXPRESSION) 55296 { 55297 STATEMENTS 55298 } 55299 55300 here is what happens when you run it: 55301 55302 * 'COLLECTION EXPRESSION' is evaluated exactly once and the result is 55303 used as the collection object to iterate over. This means it is 55304 safe to write code such as 'for (object in [NSDictionary 55305 keyEnumerator]) ...'. 55306 55307 * the iteration is implemented by the compiler by repeatedly getting 55308 batches of objects from the collection object using the fast 55309 enumeration protocol (see below), then iterating over all objects 55310 in the batch. This is faster than a normal enumeration where 55311 objects are retrieved one by one (hence the name "fast 55312 enumeration"). 55313 55314 * if there are no objects in the collection, then 'OBJECT EXPRESSION' 55315 is set to 'nil' and the loop immediately terminates. 55316 55317 * if there are objects in the collection, then for each object in the 55318 collection (in the order they are returned) 'OBJECT EXPRESSION' is 55319 set to the object, then 'STATEMENTS' are executed. 55320 55321 * 'STATEMENTS' can contain 'break' and 'continue' commands, which 55322 will abort the iteration or skip to the next loop iteration as 55323 expected. 55324 55325 * when the iteration ends because there are no more objects to 55326 iterate over, 'OBJECT EXPRESSION' is set to 'nil'. This allows you 55327 to determine whether the iteration finished because a 'break' 55328 command was used (in which case 'OBJECT EXPRESSION' will remain set 55329 to the last object that was iterated over) or because it iterated 55330 over all the objects (in which case 'OBJECT EXPRESSION' will be set 55331 to 'nil'). 55332 55333 * 'STATEMENTS' must not make any changes to the collection object; if 55334 they do, it is a hard error and the fast enumeration terminates by 55335 invoking 'objc_enumerationMutation', a runtime function that 55336 normally aborts the program but which can be customized by 55337 Foundation libraries via 'objc_set_mutation_handler' to do 55338 something different, such as raising an exception. 55339 55340 55341File: gcc.info, Node: Fast enumeration protocol, Prev: Fast enumeration details, Up: Fast enumeration 55342 553438.9.4 Fast Enumeration Protocol 55344------------------------------- 55345 55346If you want your own collection object to be usable with fast 55347enumeration, you need to have it implement the method 55348 55349 - (unsigned long) countByEnumeratingWithState: (NSFastEnumerationState *)state 55350 objects: (id *)objects 55351 count: (unsigned long)len; 55352 55353 where 'NSFastEnumerationState' must be defined in your code as follows: 55354 55355 typedef struct 55356 { 55357 unsigned long state; 55358 id *itemsPtr; 55359 unsigned long *mutationsPtr; 55360 unsigned long extra[5]; 55361 } NSFastEnumerationState; 55362 55363 If no 'NSFastEnumerationState' is defined in your code, the compiler 55364will automatically replace 'NSFastEnumerationState *' with 'struct 55365__objcFastEnumerationState *', where that type is silently defined by 55366the compiler in an identical way. This can be confusing and we 55367recommend that you define 'NSFastEnumerationState' (as shown above) 55368instead. 55369 55370 The method is called repeatedly during a fast enumeration to retrieve 55371batches of objects. Each invocation of the method should retrieve the 55372next batch of objects. 55373 55374 The return value of the method is the number of objects in the current 55375batch; this should not exceed 'len', which is the maximum size of a 55376batch as requested by the caller. The batch itself is returned in the 55377'itemsPtr' field of the 'NSFastEnumerationState' struct. 55378 55379 To help with returning the objects, the 'objects' array is a C array 55380preallocated by the caller (on the stack) of size 'len'. In many cases 55381you can put the objects you want to return in that 'objects' array, then 55382do 'itemsPtr = objects'. But you don't have to; if your collection 55383already has the objects to return in some form of C array, it could 55384return them from there instead. 55385 55386 The 'state' and 'extra' fields of the 'NSFastEnumerationState' 55387structure allows your collection object to keep track of the state of 55388the enumeration. In a simple array implementation, 'state' may keep 55389track of the index of the last object that was returned, and 'extra' may 55390be unused. 55391 55392 The 'mutationsPtr' field of the 'NSFastEnumerationState' is used to 55393keep track of mutations. It should point to a number; before working on 55394each object, the fast enumeration loop will check that this number has 55395not changed. If it has, a mutation has happened and the fast 55396enumeration will abort. So, 'mutationsPtr' could be set to point to 55397some sort of version number of your collection, which is increased by 55398one every time there is a change (for example when an object is added or 55399removed). Or, if you are content with less strict mutation checks, it 55400could point to the number of objects in your collection or some other 55401value that can be checked to perform an approximate check that the 55402collection has not been mutated. 55403 55404 Finally, note how we declared the 'len' argument and the return value 55405to be of type 'unsigned long'. They could also be declared to be of 55406type 'unsigned int' and everything would still work. 55407 55408 55409File: gcc.info, Node: Messaging with the GNU Objective-C runtime, Prev: Fast enumeration, Up: Objective-C 55410 554118.10 Messaging with the GNU Objective-C Runtime 55412=============================================== 55413 55414This section is specific for the GNU Objective-C runtime. If you are 55415using a different runtime, you can skip it. 55416 55417 The implementation of messaging in the GNU Objective-C runtime is 55418designed to be portable, and so is based on standard C. 55419 55420 Sending a message in the GNU Objective-C runtime is composed of two 55421separate steps. First, there is a call to the lookup function, 55422'objc_msg_lookup ()' (or, in the case of messages to super, 55423'objc_msg_lookup_super ()'). This runtime function takes as argument 55424the receiver and the selector of the method to be called; it returns the 55425'IMP', that is a pointer to the function implementing the method. The 55426second step of method invocation consists of casting this pointer 55427function to the appropriate function pointer type, and calling the 55428function pointed to it with the right arguments. 55429 55430 For example, when the compiler encounters a method invocation such as 55431'[object init]', it compiles it into a call to 'objc_msg_lookup (object, 55432@selector(init))' followed by a cast of the returned value to the 55433appropriate function pointer type, and then it calls it. 55434 55435* Menu: 55436 55437* Dynamically registering methods:: 55438* Forwarding hook:: 55439 55440 55441File: gcc.info, Node: Dynamically registering methods, Next: Forwarding hook, Up: Messaging with the GNU Objective-C runtime 55442 554438.10.1 Dynamically Registering Methods 55444-------------------------------------- 55445 55446If 'objc_msg_lookup()' does not find a suitable method implementation, 55447because the receiver does not implement the required method, it tries to 55448see if the class can dynamically register the method. 55449 55450 To do so, the runtime checks if the class of the receiver implements 55451the method 55452 55453 + (BOOL) resolveInstanceMethod: (SEL)selector; 55454 55455 in the case of an instance method, or 55456 55457 + (BOOL) resolveClassMethod: (SEL)selector; 55458 55459 in the case of a class method. If the class implements it, the runtime 55460invokes it, passing as argument the selector of the original method, and 55461if it returns 'YES', the runtime tries the lookup again, which could now 55462succeed if a matching method was added dynamically by 55463'+resolveInstanceMethod:' or '+resolveClassMethod:'. 55464 55465 This allows classes to dynamically register methods (by adding them to 55466the class using 'class_addMethod') when they are first called. To do 55467so, a class should implement '+resolveInstanceMethod:' (or, depending on 55468the case, '+resolveClassMethod:') and have it recognize the selectors of 55469methods that can be registered dynamically at runtime, register them, 55470and return 'YES'. It should return 'NO' for methods that it does not 55471dynamically registered at runtime. 55472 55473 If '+resolveInstanceMethod:' (or '+resolveClassMethod:') is not 55474implemented or returns 'NO', the runtime then tries the forwarding hook. 55475 55476 Support for '+resolveInstanceMethod:' and 'resolveClassMethod:' was 55477added to the GNU Objective-C runtime in GCC version 4.6. 55478 55479 55480File: gcc.info, Node: Forwarding hook, Prev: Dynamically registering methods, Up: Messaging with the GNU Objective-C runtime 55481 554828.10.2 Forwarding Hook 55483---------------------- 55484 55485The GNU Objective-C runtime provides a hook, called 55486'__objc_msg_forward2', which is called by 'objc_msg_lookup()' when it 55487cannot find a method implementation in the runtime tables and after 55488calling '+resolveInstanceMethod:' and '+resolveClassMethod:' has been 55489attempted and did not succeed in dynamically registering the method. 55490 55491 To configure the hook, you set the global variable 55492'__objc_msg_forward2' to a function with the same argument and return 55493types of 'objc_msg_lookup()'. When 'objc_msg_lookup()' cannot find a 55494method implementation, it invokes the hook function you provided to get 55495a method implementation to return. So, in practice 55496'__objc_msg_forward2' allows you to extend 'objc_msg_lookup()' by adding 55497some custom code that is called to do a further lookup when no standard 55498method implementation can be found using the normal lookup. 55499 55500 This hook is generally reserved for "Foundation" libraries such as 55501GNUstep Base, which use it to implement their high-level method 55502forwarding API, typically based around the 'forwardInvocation:' method. 55503So, unless you are implementing your own "Foundation" library, you 55504should not set this hook. 55505 55506 In a typical forwarding implementation, the '__objc_msg_forward2' hook 55507function determines the argument and return type of the method that is 55508being looked up, and then creates a function that takes these arguments 55509and has that return type, and returns it to the caller. Creating this 55510function is non-trivial and is typically performed using a dedicated 55511library such as 'libffi'. 55512 55513 The forwarding method implementation thus created is returned by 55514'objc_msg_lookup()' and is executed as if it was a normal method 55515implementation. When the forwarding method implementation is called, it 55516is usually expected to pack all arguments into some sort of object 55517(typically, an 'NSInvocation' in a "Foundation" library), and hand it 55518over to the programmer ('forwardInvocation:') who is then allowed to 55519manipulate the method invocation using a high-level API provided by the 55520"Foundation" library. For example, the programmer may want to examine 55521the method invocation arguments and name and potentially change them 55522before forwarding the method invocation to one or more local objects 55523('performInvocation:') or even to remote objects (by using Distributed 55524Objects or some other mechanism). When all this completes, the return 55525value is passed back and must be returned correctly to the original 55526caller. 55527 55528 Note that the GNU Objective-C runtime currently provides no support for 55529method forwarding or method invocations other than the 55530'__objc_msg_forward2' hook. 55531 55532 If the forwarding hook does not exist or returns 'NULL', the runtime 55533currently attempts forwarding using an older, deprecated API, and if 55534that fails, it aborts the program. In future versions of the GNU 55535Objective-C runtime, the runtime will immediately abort. 55536 55537 55538File: gcc.info, Node: Compatibility, Next: Gcov, Prev: Objective-C, Up: Top 55539 555409 Binary Compatibility 55541********************** 55542 55543Binary compatibility encompasses several related concepts: 55544 55545"application binary interface (ABI)" 55546 The set of runtime conventions followed by all of the tools that 55547 deal with binary representations of a program, including compilers, 55548 assemblers, linkers, and language runtime support. Some ABIs are 55549 formal with a written specification, possibly designed by multiple 55550 interested parties. Others are simply the way things are actually 55551 done by a particular set of tools. 55552 55553"ABI conformance" 55554 A compiler conforms to an ABI if it generates code that follows all 55555 of the specifications enumerated by that ABI. A library conforms 55556 to an ABI if it is implemented according to that ABI. An 55557 application conforms to an ABI if it is built using tools that 55558 conform to that ABI and does not contain source code that 55559 specifically changes behavior specified by the ABI. 55560 55561"calling conventions" 55562 Calling conventions are a subset of an ABI that specify of how 55563 arguments are passed and function results are returned. 55564 55565"interoperability" 55566 Different sets of tools are interoperable if they generate files 55567 that can be used in the same program. The set of tools includes 55568 compilers, assemblers, linkers, libraries, header files, startup 55569 files, and debuggers. Binaries produced by different sets of tools 55570 are not interoperable unless they implement the same ABI. This 55571 applies to different versions of the same tools as well as tools 55572 from different vendors. 55573 55574"intercallability" 55575 Whether a function in a binary built by one set of tools can call a 55576 function in a binary built by a different set of tools is a subset 55577 of interoperability. 55578 55579"implementation-defined features" 55580 Language standards include lists of implementation-defined features 55581 whose behavior can vary from one implementation to another. Some 55582 of these features are normally covered by a platform's ABI and 55583 others are not. The features that are not covered by an ABI 55584 generally affect how a program behaves, but not intercallability. 55585 55586"compatibility" 55587 Conformance to the same ABI and the same behavior of 55588 implementation-defined features are both relevant for 55589 compatibility. 55590 55591 The application binary interface implemented by a C or C++ compiler 55592affects code generation and runtime support for: 55593 55594 * size and alignment of data types 55595 * layout of structured types 55596 * calling conventions 55597 * register usage conventions 55598 * interfaces for runtime arithmetic support 55599 * object file formats 55600 55601 In addition, the application binary interface implemented by a C++ 55602compiler affects code generation and runtime support for: 55603 * name mangling 55604 * exception handling 55605 * invoking constructors and destructors 55606 * layout, alignment, and padding of classes 55607 * layout and alignment of virtual tables 55608 55609 Some GCC compilation options cause the compiler to generate code that 55610does not conform to the platform's default ABI. Other options cause 55611different program behavior for implementation-defined features that are 55612not covered by an ABI. These options are provided for consistency with 55613other compilers that do not follow the platform's default ABI or the 55614usual behavior of implementation-defined features for the platform. Be 55615very careful about using such options. 55616 55617 Most platforms have a well-defined ABI that covers C code, but ABIs 55618that cover C++ functionality are not yet common. 55619 55620 Starting with GCC 3.2, GCC binary conventions for C++ are based on a 55621written, vendor-neutral C++ ABI that was designed to be specific to 5562264-bit Itanium but also includes generic specifications that apply to 55623any platform. This C++ ABI is also implemented by other compiler 55624vendors on some platforms, notably GNU/Linux and BSD systems. We have 55625tried hard to provide a stable ABI that will be compatible with future 55626GCC releases, but it is possible that we will encounter problems that 55627make this difficult. Such problems could include different 55628interpretations of the C++ ABI by different vendors, bugs in the ABI, or 55629bugs in the implementation of the ABI in different compilers. GCC's 55630'-Wabi' switch warns when G++ generates code that is probably not 55631compatible with the C++ ABI. 55632 55633 The C++ library used with a C++ compiler includes the Standard C++ 55634Library, with functionality defined in the C++ Standard, plus language 55635runtime support. The runtime support is included in a C++ ABI, but 55636there is no formal ABI for the Standard C++ Library. Two 55637implementations of that library are interoperable if one follows the 55638de-facto ABI of the other and if they are both built with the same 55639compiler, or with compilers that conform to the same ABI for C++ 55640compiler and runtime support. 55641 55642 When G++ and another C++ compiler conform to the same C++ ABI, but the 55643implementations of the Standard C++ Library that they normally use do 55644not follow the same ABI for the Standard C++ Library, object files built 55645with those compilers can be used in the same program only if they use 55646the same C++ library. This requires specifying the location of the C++ 55647library header files when invoking the compiler whose usual library is 55648not being used. The location of GCC's C++ header files depends on how 55649the GCC build was configured, but can be seen by using the G++ '-v' 55650option. With default configuration options for G++ 3.3 the compile line 55651for a different C++ compiler needs to include 55652 55653 -IGCC_INSTALL_DIRECTORY/include/c++/3.3 55654 55655 Similarly, compiling code with G++ that must use a C++ library other 55656than the GNU C++ library requires specifying the location of the header 55657files for that other library. 55658 55659 The most straightforward way to link a program to use a particular C++ 55660library is to use a C++ driver that specifies that C++ library by 55661default. The 'g++' driver, for example, tells the linker where to find 55662GCC's C++ library ('libstdc++') plus the other libraries and startup 55663files it needs, in the proper order. 55664 55665 If a program must use a different C++ library and it's not possible to 55666do the final link using a C++ driver that uses that library by default, 55667it is necessary to tell 'g++' the location and name of that library. It 55668might also be necessary to specify different startup files and other 55669runtime support libraries, and to suppress the use of GCC's support 55670libraries with one or more of the options '-nostdlib', '-nostartfiles', 55671and '-nodefaultlibs'. 55672 55673 55674File: gcc.info, Node: Gcov, Next: Gcov-tool, Prev: Compatibility, Up: Top 55675 5567610 'gcov'--a Test Coverage Program 55677********************************** 55678 55679'gcov' is a tool you can use in conjunction with GCC to test code 55680coverage in your programs. 55681 55682* Menu: 55683 55684* Gcov Intro:: Introduction to gcov. 55685* Invoking Gcov:: How to use gcov. 55686* Gcov and Optimization:: Using gcov with GCC optimization. 55687* Gcov Data Files:: The files used by gcov. 55688* Cross-profiling:: Data file relocation. 55689 55690 55691File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov 55692 5569310.1 Introduction to 'gcov' 55694=========================== 55695 55696'gcov' is a test coverage program. Use it in concert with GCC to 55697analyze your programs to help create more efficient, faster running code 55698and to discover untested parts of your program. You can use 'gcov' as a 55699profiling tool to help discover where your optimization efforts will 55700best affect your code. You can also use 'gcov' along with the other 55701profiling tool, 'gprof', to assess which parts of your code use the 55702greatest amount of computing time. 55703 55704 Profiling tools help you analyze your code's performance. Using a 55705profiler such as 'gcov' or 'gprof', you can find out some basic 55706performance statistics, such as: 55707 55708 * how often each line of code executes 55709 55710 * what lines of code are actually executed 55711 55712 * how much computing time each section of code uses 55713 55714 Once you know these things about how your code works when compiled, you 55715can look at each module to see which modules should be optimized. 55716'gcov' helps you determine where to work on optimization. 55717 55718 Software developers also use coverage testing in concert with 55719testsuites, to make sure software is actually good enough for a release. 55720Testsuites can verify that a program works as expected; a coverage 55721program tests to see how much of the program is exercised by the 55722testsuite. Developers can then determine what kinds of test cases need 55723to be added to the testsuites to create both better testing and a better 55724final product. 55725 55726 You should compile your code without optimization if you plan to use 55727'gcov' because the optimization, by combining some lines of code into 55728one function, may not give you as much information as you need to look 55729for 'hot spots' where the code is using a great deal of computer time. 55730Likewise, because 'gcov' accumulates statistics by line (at the lowest 55731resolution), it works best with a programming style that places only one 55732statement on each line. If you use complicated macros that expand to 55733loops or to other control structures, the statistics are less 55734helpful--they only report on the line where the macro call appears. If 55735your complex macros behave like functions, you can replace them with 55736inline functions to solve this problem. 55737 55738 'gcov' creates a logfile called 'SOURCEFILE.gcov' which indicates how 55739many times each line of a source file 'SOURCEFILE.c' has executed. You 55740can use these logfiles along with 'gprof' to aid in fine-tuning the 55741performance of your programs. 'gprof' gives timing information you can 55742use along with the information you get from 'gcov'. 55743 55744 'gcov' works only on code compiled with GCC. It is not compatible with 55745any other profiling or test coverage mechanism. 55746 55747 55748File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov 55749 5575010.2 Invoking 'gcov' 55751==================== 55752 55753 gcov [OPTIONS] FILES 55754 55755 'gcov' accepts the following options: 55756 55757'-a' 55758'--all-blocks' 55759 Write individual execution counts for every basic block. Normally 55760 gcov outputs execution counts only for the main blocks of a line. 55761 With this option you can determine if blocks within a single line 55762 are not being executed. 55763 55764'-b' 55765'--branch-probabilities' 55766 Write branch frequencies to the output file, and write branch 55767 summary info to the standard output. This option allows you to see 55768 how often each branch in your program was taken. Unconditional 55769 branches will not be shown, unless the '-u' option is given. 55770 55771'-c' 55772'--branch-counts' 55773 Write branch frequencies as the number of branches taken, rather 55774 than the percentage of branches taken. 55775 55776'-d' 55777'--display-progress' 55778 Display the progress on the standard output. 55779 55780'-f' 55781'--function-summaries' 55782 Output summaries for each function in addition to the file level 55783 summary. 55784 55785'-h' 55786'--help' 55787 Display help about using 'gcov' (on the standard output), and exit 55788 without doing any further processing. 55789 55790'-i' 55791'--json-format' 55792 Output gcov file in an easy-to-parse JSON intermediate format which 55793 does not require source code for generation. The JSON file is 55794 compressed with gzip compression algorithm and the files have 55795 '.gcov.json.gz' extension. 55796 55797 Structure of the JSON is following: 55798 55799 { 55800 "current_working_directory": CURRENT_WORKING_DIRECTORY, 55801 "data_file": DATA_FILE, 55802 "format_version": FORMAT_VERSION, 55803 "gcc_version": GCC_VERSION 55804 "files": [FILE] 55805 } 55806 55807 Fields of the root element have following semantics: 55808 55809 * CURRENT_WORKING_DIRECTORY: working directory where a 55810 compilation unit was compiled 55811 55812 * DATA_FILE: name of the data file (GCDA) 55813 55814 * FORMAT_VERSION: semantic version of the format 55815 55816 * GCC_VERSION: version of the GCC compiler 55817 55818 Each FILE has the following form: 55819 55820 { 55821 "file": FILE_NAME, 55822 "functions": [FUNCTION], 55823 "lines": [LINE] 55824 } 55825 55826 Fields of the FILE element have following semantics: 55827 55828 * FILE_NAME: name of the source file 55829 55830 Each FUNCTION has the following form: 55831 55832 { 55833 "blocks": BLOCKS, 55834 "blocks_executed": BLOCKS_EXECUTED, 55835 "demangled_name": "DEMANGLED_NAME, 55836 "end_column": END_COLUMN, 55837 "end_line": END_LINE, 55838 "execution_count": EXECUTION_COUNT, 55839 "name": NAME, 55840 "start_column": START_COLUMN 55841 "start_line": START_LINE 55842 } 55843 55844 Fields of the FUNCTION element have following semantics: 55845 55846 * BLOCKS: number of blocks that are in the function 55847 55848 * BLOCKS_EXECUTED: number of executed blocks of the function 55849 55850 * DEMANGLED_NAME: demangled name of the function 55851 55852 * END_COLUMN: column in the source file where the function ends 55853 55854 * END_LINE: line in the source file where the function ends 55855 55856 * EXECUTION_COUNT: number of executions of the function 55857 55858 * NAME: name of the function 55859 55860 * START_COLUMN: column in the source file where the function 55861 begins 55862 55863 * START_LINE: line in the source file where the function begins 55864 55865 Note that line numbers and column numbers number from 1. In the 55866 current implementation, START_LINE and START_COLUMN do not include 55867 any template parameters and the leading return type but that this 55868 is likely to be fixed in the future. 55869 55870 Each LINE has the following form: 55871 55872 { 55873 "branches": [BRANCH], 55874 "count": COUNT, 55875 "line_number": LINE_NUMBER, 55876 "unexecuted_block": UNEXECUTED_BLOCK 55877 "function_name": FUNCTION_NAME, 55878 } 55879 55880 Branches are present only with -B option. Fields of the LINE 55881 element have following semantics: 55882 55883 * COUNT: number of executions of the line 55884 55885 * LINE_NUMBER: line number 55886 55887 * UNEXECUTED_BLOCK: flag whether the line contains an unexecuted 55888 block (not all statements on the line are executed) 55889 55890 * FUNCTION_NAME: a name of a function this LINE belongs to (for 55891 a line with an inlined statements can be not set) 55892 55893 Each BRANCH has the following form: 55894 55895 { 55896 "count": COUNT, 55897 "fallthrough": FALLTHROUGH, 55898 "throw": THROW 55899 } 55900 55901 Fields of the BRANCH element have following semantics: 55902 55903 * COUNT: number of executions of the branch 55904 55905 * FALLTHROUGH: true when the branch is a fall through branch 55906 55907 * THROW: true when the branch is an exceptional branch 55908 55909'-j' 55910'--human-readable' 55911 Write counts in human readable format (like 24.6k). 55912 55913'-k' 55914'--use-colors' 55915 55916 Use colors for lines of code that have zero coverage. We use red 55917 color for non-exceptional lines and cyan for exceptional. Same 55918 colors are used for basic blocks with '-a' option. 55919 55920'-l' 55921'--long-file-names' 55922 Create long file names for included source files. For example, if 55923 the header file 'x.h' contains code, and was included in the file 55924 'a.c', then running 'gcov' on the file 'a.c' will produce an output 55925 file called 'a.c##x.h.gcov' instead of 'x.h.gcov'. This can be 55926 useful if 'x.h' is included in multiple source files and you want 55927 to see the individual contributions. If you use the '-p' option, 55928 both the including and included file names will be complete path 55929 names. 55930 55931'-m' 55932'--demangled-names' 55933 Display demangled function names in output. The default is to show 55934 mangled function names. 55935 55936'-n' 55937'--no-output' 55938 Do not create the 'gcov' output file. 55939 55940'-o DIRECTORY|FILE' 55941'--object-directory DIRECTORY' 55942'--object-file FILE' 55943 Specify either the directory containing the gcov data files, or the 55944 object path name. The '.gcno', and '.gcda' data files are searched 55945 for using this option. If a directory is specified, the data files 55946 are in that directory and named after the input file name, without 55947 its extension. If a file is specified here, the data files are 55948 named after that file, without its extension. 55949 55950'-p' 55951'--preserve-paths' 55952 Preserve complete path information in the names of generated 55953 '.gcov' files. Without this option, just the filename component is 55954 used. With this option, all directories are used, with '/' 55955 characters translated to '#' characters, '.' directory components 55956 removed and unremoveable '..' components renamed to '^'. This is 55957 useful if sourcefiles are in several different directories. 55958 55959'-q' 55960'--use-hotness-colors' 55961 55962 Emit perf-like colored output for hot lines. Legend of the color 55963 scale is printed at the very beginning of the output file. 55964 55965'-r' 55966'--relative-only' 55967 Only output information about source files with a relative pathname 55968 (after source prefix elision). Absolute paths are usually system 55969 header files and coverage of any inline functions therein is 55970 normally uninteresting. 55971 55972'-s DIRECTORY' 55973'--source-prefix DIRECTORY' 55974 A prefix for source file names to remove when generating the output 55975 coverage files. This option is useful when building in a separate 55976 directory, and the pathname to the source directory is not wanted 55977 when determining the output file names. Note that this prefix 55978 detection is applied before determining whether the source file is 55979 absolute. 55980 55981'-t' 55982'--stdout' 55983 Output to standard output instead of output files. 55984 55985'-u' 55986'--unconditional-branches' 55987 When branch probabilities are given, include those of unconditional 55988 branches. Unconditional branches are normally not interesting. 55989 55990'-v' 55991'--version' 55992 Display the 'gcov' version number (on the standard output), and 55993 exit without doing any further processing. 55994 55995'-w' 55996'--verbose' 55997 Print verbose informations related to basic blocks and arcs. 55998 55999'-x' 56000'--hash-filenames' 56001 When using -PRESERVE-PATHS, gcov uses the full pathname of the 56002 source files to create an output filename. This can lead to long 56003 filenames that can overflow filesystem limits. This option creates 56004 names of the form 'SOURCE-FILE##MD5.gcov', where the SOURCE-FILE 56005 component is the final filename part and the MD5 component is 56006 calculated from the full mangled name that would have been used 56007 otherwise. The option is an alternative to the -PRESERVE-PATHS on 56008 systems which have a filesystem limit. 56009 56010 'gcov' should be run with the current directory the same as that when 56011you invoked the compiler. Otherwise it will not be able to locate the 56012source files. 'gcov' produces files called 'MANGLEDNAME.gcov' in the 56013current directory. These contain the coverage information of the source 56014file they correspond to. One '.gcov' file is produced for each source 56015(or header) file containing code, which was compiled to produce the data 56016files. The MANGLEDNAME part of the output file name is usually simply 56017the source file name, but can be something more complicated if the '-l' 56018or '-p' options are given. Refer to those options for details. 56019 56020 If you invoke 'gcov' with multiple input files, the contributions from 56021each input file are summed. Typically you would invoke it with the same 56022list of files as the final link of your executable. 56023 56024 The '.gcov' files contain the ':' separated fields along with program 56025source code. The format is 56026 56027 EXECUTION_COUNT:LINE_NUMBER:SOURCE LINE TEXT 56028 56029 Additional block information may succeed each line, when requested by 56030command line option. The EXECUTION_COUNT is '-' for lines containing no 56031code. Unexecuted lines are marked '#####' or '=====', depending on 56032whether they are reachable by non-exceptional paths or only exceptional 56033paths such as C++ exception handlers, respectively. Given the '-a' 56034option, unexecuted blocks are marked '$$$$$' or '%%%%%', depending on 56035whether a basic block is reachable via non-exceptional or exceptional 56036paths. Executed basic blocks having a statement with zero 56037EXECUTION_COUNT end with '*' character and are colored with magenta 56038color with the '-k' option. This functionality is not supported in Ada. 56039 56040 Note that GCC can completely remove the bodies of functions that are 56041not needed - for instance if they are inlined everywhere. Such 56042functions are marked with '-', which can be confusing. Use the 56043'-fkeep-inline-functions' and '-fkeep-static-functions' options to 56044retain these functions and allow gcov to properly show their 56045EXECUTION_COUNT. 56046 56047 Some lines of information at the start have LINE_NUMBER of zero. These 56048preamble lines are of the form 56049 56050 -:0:TAG:VALUE 56051 56052 The ordering and number of these preamble lines will be augmented as 56053'gcov' development progresses -- do not rely on them remaining 56054unchanged. Use TAG to locate a particular preamble line. 56055 56056 The additional block information is of the form 56057 56058 TAG INFORMATION 56059 56060 The INFORMATION is human readable, but designed to be simple enough for 56061machine parsing too. 56062 56063 When printing percentages, 0% and 100% are only printed when the values 56064are _exactly_ 0% and 100% respectively. Other values which would 56065conventionally be rounded to 0% or 100% are instead printed as the 56066nearest non-boundary value. 56067 56068 When using 'gcov', you must first compile your program with a special 56069GCC option '--coverage'. This tells the compiler to generate additional 56070information needed by gcov (basically a flow graph of the program) and 56071also includes additional code in the object files for generating the 56072extra profiling information needed by gcov. These additional files are 56073placed in the directory where the object file is located. 56074 56075 Running the program will cause profile output to be generated. For 56076each source file compiled with '-fprofile-arcs', an accompanying '.gcda' 56077file will be placed in the object file directory. 56078 56079 Running 'gcov' with your program's source file names as arguments will 56080now produce a listing of the code along with frequency of execution for 56081each line. For example, if your program is called 'tmp.cpp', this is 56082what you see when you use the basic 'gcov' facility: 56083 56084 $ g++ --coverage tmp.cpp 56085 $ a.out 56086 $ gcov tmp.cpp -m 56087 File 'tmp.cpp' 56088 Lines executed:92.86% of 14 56089 Creating 'tmp.cpp.gcov' 56090 56091 The file 'tmp.cpp.gcov' contains output from 'gcov'. Here is a sample: 56092 56093 -: 0:Source:tmp.cpp 56094 -: 0:Working directory:/home/gcc/testcase 56095 -: 0:Graph:tmp.gcno 56096 -: 0:Data:tmp.gcda 56097 -: 0:Runs:1 56098 -: 0:Programs:1 56099 -: 1:#include <stdio.h> 56100 -: 2: 56101 -: 3:template<class T> 56102 -: 4:class Foo 56103 -: 5:{ 56104 -: 6: public: 56105 1*: 7: Foo(): b (1000) {} 56106 ------------------ 56107 Foo<char>::Foo(): 56108 #####: 7: Foo(): b (1000) {} 56109 ------------------ 56110 Foo<int>::Foo(): 56111 1: 7: Foo(): b (1000) {} 56112 ------------------ 56113 2*: 8: void inc () { b++; } 56114 ------------------ 56115 Foo<char>::inc(): 56116 #####: 8: void inc () { b++; } 56117 ------------------ 56118 Foo<int>::inc(): 56119 2: 8: void inc () { b++; } 56120 ------------------ 56121 -: 9: 56122 -: 10: private: 56123 -: 11: int b; 56124 -: 12:}; 56125 -: 13: 56126 -: 14:template class Foo<int>; 56127 -: 15:template class Foo<char>; 56128 -: 16: 56129 -: 17:int 56130 1: 18:main (void) 56131 -: 19:{ 56132 -: 20: int i, total; 56133 1: 21: Foo<int> counter; 56134 -: 22: 56135 1: 23: counter.inc(); 56136 1: 24: counter.inc(); 56137 1: 25: total = 0; 56138 -: 26: 56139 11: 27: for (i = 0; i < 10; i++) 56140 10: 28: total += i; 56141 -: 29: 56142 1*: 30: int v = total > 100 ? 1 : 2; 56143 -: 31: 56144 1: 32: if (total != 45) 56145 #####: 33: printf ("Failure\n"); 56146 -: 34: else 56147 1: 35: printf ("Success\n"); 56148 1: 36: return 0; 56149 -: 37:} 56150 56151 Note that line 7 is shown in the report multiple times. First 56152occurrence presents total number of execution of the line and the next 56153two belong to instances of class Foo constructors. As you can also see, 56154line 30 contains some unexecuted basic blocks and thus execution count 56155has asterisk symbol. 56156 56157 When you use the '-a' option, you will get individual block counts, and 56158the output looks like this: 56159 56160 -: 0:Source:tmp.cpp 56161 -: 0:Working directory:/home/gcc/testcase 56162 -: 0:Graph:tmp.gcno 56163 -: 0:Data:tmp.gcda 56164 -: 0:Runs:1 56165 -: 0:Programs:1 56166 -: 1:#include <stdio.h> 56167 -: 2: 56168 -: 3:template<class T> 56169 -: 4:class Foo 56170 -: 5:{ 56171 -: 6: public: 56172 1*: 7: Foo(): b (1000) {} 56173 ------------------ 56174 Foo<char>::Foo(): 56175 #####: 7: Foo(): b (1000) {} 56176 ------------------ 56177 Foo<int>::Foo(): 56178 1: 7: Foo(): b (1000) {} 56179 ------------------ 56180 2*: 8: void inc () { b++; } 56181 ------------------ 56182 Foo<char>::inc(): 56183 #####: 8: void inc () { b++; } 56184 ------------------ 56185 Foo<int>::inc(): 56186 2: 8: void inc () { b++; } 56187 ------------------ 56188 -: 9: 56189 -: 10: private: 56190 -: 11: int b; 56191 -: 12:}; 56192 -: 13: 56193 -: 14:template class Foo<int>; 56194 -: 15:template class Foo<char>; 56195 -: 16: 56196 -: 17:int 56197 1: 18:main (void) 56198 -: 19:{ 56199 -: 20: int i, total; 56200 1: 21: Foo<int> counter; 56201 1: 21-block 0 56202 -: 22: 56203 1: 23: counter.inc(); 56204 1: 23-block 0 56205 1: 24: counter.inc(); 56206 1: 24-block 0 56207 1: 25: total = 0; 56208 -: 26: 56209 11: 27: for (i = 0; i < 10; i++) 56210 1: 27-block 0 56211 11: 27-block 1 56212 10: 28: total += i; 56213 10: 28-block 0 56214 -: 29: 56215 1*: 30: int v = total > 100 ? 1 : 2; 56216 1: 30-block 0 56217 %%%%%: 30-block 1 56218 1: 30-block 2 56219 -: 31: 56220 1: 32: if (total != 45) 56221 1: 32-block 0 56222 #####: 33: printf ("Failure\n"); 56223 %%%%%: 33-block 0 56224 -: 34: else 56225 1: 35: printf ("Success\n"); 56226 1: 35-block 0 56227 1: 36: return 0; 56228 1: 36-block 0 56229 -: 37:} 56230 56231 In this mode, each basic block is only shown on one line - the last 56232line of the block. A multi-line block will only contribute to the 56233execution count of that last line, and other lines will not be shown to 56234contain code, unless previous blocks end on those lines. The total 56235execution count of a line is shown and subsequent lines show the 56236execution counts for individual blocks that end on that line. After 56237each block, the branch and call counts of the block will be shown, if 56238the '-b' option is given. 56239 56240 Because of the way GCC instruments calls, a call count can be shown 56241after a line with no individual blocks. As you can see, line 33 56242contains a basic block that was not executed. 56243 56244 When you use the '-b' option, your output looks like this: 56245 56246 -: 0:Source:tmp.cpp 56247 -: 0:Working directory:/home/gcc/testcase 56248 -: 0:Graph:tmp.gcno 56249 -: 0:Data:tmp.gcda 56250 -: 0:Runs:1 56251 -: 0:Programs:1 56252 -: 1:#include <stdio.h> 56253 -: 2: 56254 -: 3:template<class T> 56255 -: 4:class Foo 56256 -: 5:{ 56257 -: 6: public: 56258 1*: 7: Foo(): b (1000) {} 56259 ------------------ 56260 Foo<char>::Foo(): 56261 function Foo<char>::Foo() called 0 returned 0% blocks executed 0% 56262 #####: 7: Foo(): b (1000) {} 56263 ------------------ 56264 Foo<int>::Foo(): 56265 function Foo<int>::Foo() called 1 returned 100% blocks executed 100% 56266 1: 7: Foo(): b (1000) {} 56267 ------------------ 56268 2*: 8: void inc () { b++; } 56269 ------------------ 56270 Foo<char>::inc(): 56271 function Foo<char>::inc() called 0 returned 0% blocks executed 0% 56272 #####: 8: void inc () { b++; } 56273 ------------------ 56274 Foo<int>::inc(): 56275 function Foo<int>::inc() called 2 returned 100% blocks executed 100% 56276 2: 8: void inc () { b++; } 56277 ------------------ 56278 -: 9: 56279 -: 10: private: 56280 -: 11: int b; 56281 -: 12:}; 56282 -: 13: 56283 -: 14:template class Foo<int>; 56284 -: 15:template class Foo<char>; 56285 -: 16: 56286 -: 17:int 56287 function main called 1 returned 100% blocks executed 81% 56288 1: 18:main (void) 56289 -: 19:{ 56290 -: 20: int i, total; 56291 1: 21: Foo<int> counter; 56292 call 0 returned 100% 56293 branch 1 taken 100% (fallthrough) 56294 branch 2 taken 0% (throw) 56295 -: 22: 56296 1: 23: counter.inc(); 56297 call 0 returned 100% 56298 branch 1 taken 100% (fallthrough) 56299 branch 2 taken 0% (throw) 56300 1: 24: counter.inc(); 56301 call 0 returned 100% 56302 branch 1 taken 100% (fallthrough) 56303 branch 2 taken 0% (throw) 56304 1: 25: total = 0; 56305 -: 26: 56306 11: 27: for (i = 0; i < 10; i++) 56307 branch 0 taken 91% (fallthrough) 56308 branch 1 taken 9% 56309 10: 28: total += i; 56310 -: 29: 56311 1*: 30: int v = total > 100 ? 1 : 2; 56312 branch 0 taken 0% (fallthrough) 56313 branch 1 taken 100% 56314 -: 31: 56315 1: 32: if (total != 45) 56316 branch 0 taken 0% (fallthrough) 56317 branch 1 taken 100% 56318 #####: 33: printf ("Failure\n"); 56319 call 0 never executed 56320 branch 1 never executed 56321 branch 2 never executed 56322 -: 34: else 56323 1: 35: printf ("Success\n"); 56324 call 0 returned 100% 56325 branch 1 taken 100% (fallthrough) 56326 branch 2 taken 0% (throw) 56327 1: 36: return 0; 56328 -: 37:} 56329 56330 For each function, a line is printed showing how many times the 56331function is called, how many times it returns and what percentage of the 56332function's blocks were executed. 56333 56334 For each basic block, a line is printed after the last line of the 56335basic block describing the branch or call that ends the basic block. 56336There can be multiple branches and calls listed for a single source line 56337if there are multiple basic blocks that end on that line. In this case, 56338the branches and calls are each given a number. There is no simple way 56339to map these branches and calls back to source constructs. In general, 56340though, the lowest numbered branch or call will correspond to the 56341leftmost construct on the source line. 56342 56343 For a branch, if it was executed at least once, then a percentage 56344indicating the number of times the branch was taken divided by the 56345number of times the branch was executed will be printed. Otherwise, the 56346message "never executed" is printed. 56347 56348 For a call, if it was executed at least once, then a percentage 56349indicating the number of times the call returned divided by the number 56350of times the call was executed will be printed. This will usually be 56351100%, but may be less for functions that call 'exit' or 'longjmp', and 56352thus may not return every time they are called. 56353 56354 The execution counts are cumulative. If the example program were 56355executed again without removing the '.gcda' file, the count for the 56356number of times each line in the source was executed would be added to 56357the results of the previous run(s). This is potentially useful in 56358several ways. For example, it could be used to accumulate data over a 56359number of program runs as part of a test verification suite, or to 56360provide more accurate long-term information over a large number of 56361program runs. 56362 56363 The data in the '.gcda' files is saved immediately before the program 56364exits. For each source file compiled with '-fprofile-arcs', the 56365profiling code first attempts to read in an existing '.gcda' file; if 56366the file doesn't match the executable (differing number of basic block 56367counts) it will ignore the contents of the file. It then adds in the 56368new execution counts and finally writes the data to the file. 56369 56370 56371File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov 56372 5637310.3 Using 'gcov' with GCC Optimization 56374======================================= 56375 56376If you plan to use 'gcov' to help optimize your code, you must first 56377compile your program with a special GCC option '--coverage'. Aside from 56378that, you can use any other GCC options; but if you want to prove that 56379every single line in your program was executed, you should not compile 56380with optimization at the same time. On some machines the optimizer can 56381eliminate some simple code lines by combining them with other lines. 56382For example, code like this: 56383 56384 if (a != b) 56385 c = 1; 56386 else 56387 c = 0; 56388 56389can be compiled into one instruction on some machines. In this case, 56390there is no way for 'gcov' to calculate separate execution counts for 56391each line because there isn't separate code for each line. Hence the 56392'gcov' output looks like this if you compiled the program with 56393optimization: 56394 56395 100: 12:if (a != b) 56396 100: 13: c = 1; 56397 100: 14:else 56398 100: 15: c = 0; 56399 56400 The output shows that this block of code, combined by optimization, 56401executed 100 times. In one sense this result is correct, because there 56402was only one instruction representing all four of these lines. However, 56403the output does not indicate how many times the result was 0 and how 56404many times the result was 1. 56405 56406 Inlineable functions can create unexpected line counts. Line counts 56407are shown for the source code of the inlineable function, but what is 56408shown depends on where the function is inlined, or if it is not inlined 56409at all. 56410 56411 If the function is not inlined, the compiler must emit an out of line 56412copy of the function, in any object file that needs it. If 'fileA.o' 56413and 'fileB.o' both contain out of line bodies of a particular inlineable 56414function, they will also both contain coverage counts for that function. 56415When 'fileA.o' and 'fileB.o' are linked together, the linker will, on 56416many systems, select one of those out of line bodies for all calls to 56417that function, and remove or ignore the other. Unfortunately, it will 56418not remove the coverage counters for the unused function body. Hence 56419when instrumented, all but one use of that function will show zero 56420counts. 56421 56422 If the function is inlined in several places, the block structure in 56423each location might not be the same. For instance, a condition might 56424now be calculable at compile time in some instances. Because the 56425coverage of all the uses of the inline function will be shown for the 56426same source lines, the line counts themselves might seem inconsistent. 56427 56428 Long-running applications can use the '__gcov_reset' and '__gcov_dump' 56429facilities to restrict profile collection to the program region of 56430interest. Calling '__gcov_reset(void)' will clear all profile counters 56431to zero, and calling '__gcov_dump(void)' will cause the profile 56432information collected at that point to be dumped to '.gcda' output 56433files. Instrumented applications use a static destructor with priority 5643499 to invoke the '__gcov_dump' function. Thus '__gcov_dump' is executed 56435after all user defined static destructors, as well as handlers 56436registered with 'atexit'. If an executable loads a dynamic shared 56437object via dlopen functionality, '-Wl,--dynamic-list-data' is needed to 56438dump all profile data. 56439 56440 Profiling run-time library reports various errors related to profile 56441manipulation and profile saving. Errors are printed into standard error 56442output or 'GCOV_ERROR_FILE' file, if environment variable is used. In 56443order to terminate immediately after an errors occurs set 56444'GCOV_EXIT_AT_ERROR' environment variable. That can help users to find 56445profile clashing which leads to a misleading profile. 56446 56447 56448File: gcc.info, Node: Gcov Data Files, Next: Cross-profiling, Prev: Gcov and Optimization, Up: Gcov 56449 5645010.4 Brief Description of 'gcov' Data Files 56451=========================================== 56452 56453'gcov' uses two files for profiling. The names of these files are 56454derived from the original _object_ file by substituting the file suffix 56455with either '.gcno', or '.gcda'. The files contain coverage and profile 56456data stored in a platform-independent format. The '.gcno' files are 56457placed in the same directory as the object file. By default, the 56458'.gcda' files are also stored in the same directory as the object file, 56459but the GCC '-fprofile-dir' option may be used to store the '.gcda' 56460files in a separate directory. 56461 56462 The '.gcno' notes file is generated when the source file is compiled 56463with the GCC '-ftest-coverage' option. It contains information to 56464reconstruct the basic block graphs and assign source line numbers to 56465blocks. 56466 56467 The '.gcda' count data file is generated when a program containing 56468object files built with the GCC '-fprofile-arcs' option is executed. A 56469separate '.gcda' file is created for each object file compiled with this 56470option. It contains arc transition counts, value profile counts, and 56471some summary information. 56472 56473 It is not recommended to access the coverage files directly. Consumers 56474should use the intermediate format that is provided by 'gcov' tool via 56475'--json-format' option. 56476 56477 56478File: gcc.info, Node: Cross-profiling, Prev: Gcov Data Files, Up: Gcov 56479 5648010.5 Data File Relocation to Support Cross-Profiling 56481==================================================== 56482 56483Running the program will cause profile output to be generated. For each 56484source file compiled with '-fprofile-arcs', an accompanying '.gcda' file 56485will be placed in the object file directory. That implicitly requires 56486running the program on the same system as it was built or having the 56487same absolute directory structure on the target system. The program 56488will try to create the needed directory structure, if it is not already 56489present. 56490 56491 To support cross-profiling, a program compiled with '-fprofile-arcs' 56492can relocate the data files based on two environment variables: 56493 56494 * GCOV_PREFIX contains the prefix to add to the absolute paths in the 56495 object file. Prefix can be absolute, or relative. The default is 56496 no prefix. 56497 56498 * GCOV_PREFIX_STRIP indicates the how many initial directory names to 56499 strip off the hardwired absolute paths. Default value is 0. 56500 56501 _Note:_ If GCOV_PREFIX_STRIP is set without GCOV_PREFIX is 56502 undefined, then a relative path is made out of the hardwired 56503 absolute paths. 56504 56505 For example, if the object file '/user/build/foo.o' was built with 56506'-fprofile-arcs', the final executable will try to create the data file 56507'/user/build/foo.gcda' when running on the target system. This will 56508fail if the corresponding directory does not exist and it is unable to 56509create it. This can be overcome by, for example, setting the 56510environment as 'GCOV_PREFIX=/target/run' and 'GCOV_PREFIX_STRIP=1'. 56511Such a setting will name the data file '/target/run/build/foo.gcda'. 56512 56513 You must move the data files to the expected directory tree in order to 56514use them for profile directed optimizations ('-fprofile-use'), or to use 56515the 'gcov' tool. 56516 56517 56518File: gcc.info, Node: Gcov-tool, Next: Gcov-dump, Prev: Gcov, Up: Top 56519 5652011 'gcov-tool'--an Offline Gcda Profile Processing Tool 56521******************************************************* 56522 56523'gcov-tool' is a tool you can use in conjunction with GCC to manipulate 56524or process gcda profile files offline. 56525 56526* Menu: 56527 56528* Gcov-tool Intro:: Introduction to gcov-tool. 56529* Invoking Gcov-tool:: How to use gcov-tool. 56530 56531 56532File: gcc.info, Node: Gcov-tool Intro, Next: Invoking Gcov-tool, Up: Gcov-tool 56533 5653411.1 Introduction to 'gcov-tool' 56535================================ 56536 56537'gcov-tool' is an offline tool to process gcc's gcda profile files. 56538 56539 Current gcov-tool supports the following functionalities: 56540 56541 * merge two sets of profiles with weights. 56542 56543 * read one set of profile and rewrite profile contents. One can 56544 scale or normalize the count values. 56545 56546 Examples of the use cases for this tool are: 56547 * Collect the profiles for different set of inputs, and use this tool 56548 to merge them. One can specify the weight to factor in the 56549 relative importance of each input. 56550 56551 * Rewrite the profile after removing a subset of the gcda files, 56552 while maintaining the consistency of the summary and the histogram. 56553 56554 * It can also be used to debug or libgcov code as the tools shares 56555 the majority code as the runtime library. 56556 56557 Note that for the merging operation, this profile generated offline may 56558contain slight different values from the online merged profile. Here 56559are a list of typical differences: 56560 56561 * histogram difference: This offline tool recomputes the histogram 56562 after merging the counters. The resulting histogram, therefore, is 56563 precise. The online merging does not have this capability - the 56564 histogram is merged from two histograms and the result is an 56565 approximation. 56566 56567 * summary checksum difference: Summary checksum uses a CRC32 56568 operation. The value depends on the link list order of gcov-info 56569 objects. This order is different in gcov-tool from that in the 56570 online merge. It's expected to have different summary checksums. 56571 It does not really matter as the compiler does not use this 56572 checksum anywhere. 56573 56574 * value profile counter values difference: Some counter values for 56575 value profile are runtime dependent, like heap addresses. It's 56576 normal to see some difference in these kind of counters. 56577 56578 56579File: gcc.info, Node: Invoking Gcov-tool, Prev: Gcov-tool Intro, Up: Gcov-tool 56580 5658111.2 Invoking 'gcov-tool' 56582========================= 56583 56584 gcov-tool [GLOBAL-OPTIONS] SUB_COMMAND [SUB_COMMAND-OPTIONS] PROFILE_DIR 56585 56586 'gcov-tool' accepts the following options: 56587 56588'-h' 56589'--help' 56590 Display help about using 'gcov-tool' (on the standard output), and 56591 exit without doing any further processing. 56592 56593'-v' 56594'--version' 56595 Display the 'gcov-tool' version number (on the standard output), 56596 and exit without doing any further processing. 56597 56598'merge' 56599 Merge two profile directories. 56600 56601 '-o DIRECTORY' 56602 '--output DIRECTORY' 56603 Set the output profile directory. Default output directory 56604 name is MERGED_PROFILE. 56605 56606 '-v' 56607 '--verbose' 56608 Set the verbose mode. 56609 56610 '-w W1,W2' 56611 '--weight W1,W2' 56612 Set the merge weights of the DIRECTORY1 and DIRECTORY2, 56613 respectively. The default weights are 1 for both. 56614 56615'rewrite' 56616 Read the specified profile directory and rewrite to a new 56617 directory. 56618 56619 '-n LONG_LONG_VALUE' 56620 '--normalize <long_long_value>' 56621 Normalize the profile. The specified value is the max counter 56622 value in the new profile. 56623 56624 '-o DIRECTORY' 56625 '--output DIRECTORY' 56626 Set the output profile directory. Default output name is 56627 REWRITE_PROFILE. 56628 56629 '-s FLOAT_OR_SIMPLE-FRAC_VALUE' 56630 '--scale FLOAT_OR_SIMPLE-FRAC_VALUE' 56631 Scale the profile counters. The specified value can be in 56632 floating point value, or simple fraction value form, such 1, 56633 2, 2/3, and 5/3. 56634 56635 '-v' 56636 '--verbose' 56637 Set the verbose mode. 56638 56639'overlap' 56640 Compute the overlap score between the two specified profile 56641 directories. The overlap score is computed based on the arc 56642 profiles. It is defined as the sum of min (p1_counter[i] / 56643 p1_sum_all, p2_counter[i] / p2_sum_all), for all arc counter i, 56644 where p1_counter[i] and p2_counter[i] are two matched counters and 56645 p1_sum_all and p2_sum_all are the sum of counter values in profile 56646 1 and profile 2, respectively. 56647 56648 '-f' 56649 '--function' 56650 Print function level overlap score. 56651 56652 '-F' 56653 '--fullname' 56654 Print full gcda filename. 56655 56656 '-h' 56657 '--hotonly' 56658 Only print info for hot objects/functions. 56659 56660 '-o' 56661 '--object' 56662 Print object level overlap score. 56663 56664 '-t FLOAT' 56665 '--hot_threshold <float>' 56666 Set the threshold for hot counter value. 56667 56668 '-v' 56669 '--verbose' 56670 Set the verbose mode. 56671 56672 56673File: gcc.info, Node: Gcov-dump, Next: lto-dump, Prev: Gcov-tool, Up: Top 56674 5667512 'gcov-dump'--an Offline Gcda and Gcno Profile Dump Tool 56676********************************************************** 56677 56678* Menu: 56679 56680* Gcov-dump Intro:: Introduction to gcov-dump. 56681* Invoking Gcov-dump:: How to use gcov-dump. 56682 56683 56684File: gcc.info, Node: Gcov-dump Intro, Next: Invoking Gcov-dump, Up: Gcov-dump 56685 5668612.1 Introduction to 'gcov-dump' 56687================================ 56688 56689'gcov-dump' is a tool you can use in conjunction with GCC to dump 56690content of gcda and gcno profile files offline. 56691 56692 56693File: gcc.info, Node: Invoking Gcov-dump, Prev: Gcov-dump Intro, Up: Gcov-dump 56694 5669512.2 Invoking 'gcov-dump' 56696========================= 56697 56698 Usage: gcov-dump [OPTION] ... GCOVFILES 56699 56700 'gcov-dump' accepts the following options: 56701 56702'-h' 56703'--help' 56704 Display help about using 'gcov-dump' (on the standard output), and 56705 exit without doing any further processing. 56706 56707'-l' 56708'--long' 56709 Dump content of records. 56710 56711'-p' 56712'--positions' 56713 Dump positions of records. 56714 56715'-v' 56716'--version' 56717 Display the 'gcov-dump' version number (on the standard output), 56718 and exit without doing any further processing. 56719 56720 56721File: gcc.info, Node: lto-dump, Next: Trouble, Prev: Gcov-dump, Up: Top 56722 5672313 'lto-dump'--Tool for dumping LTO object files. 56724************************************************* 56725 56726* Menu: 56727 56728* lto-dump Intro:: Introduction to lto-dump. 56729* Invoking lto-dump:: How to use lto-dump. 56730 56731 56732File: gcc.info, Node: lto-dump Intro, Next: Invoking lto-dump, Up: lto-dump 56733 5673413.1 Introduction to 'lto-dump' 56735=============================== 56736 56737'lto-dump' is a tool you can use in conjunction with GCC to dump link 56738time optimization object files. 56739 56740 56741File: gcc.info, Node: Invoking lto-dump, Prev: lto-dump Intro, Up: lto-dump 56742 5674313.2 Invoking 'lto-dump' 56744======================== 56745 56746 Usage: lto-dump [OPTION] ... OBJFILES 56747 56748 'lto-dump' accepts the following options: 56749 56750'-list' 56751 Dumps list of details of functions and variables. 56752 56753'-demangle' 56754 Dump the demangled output. 56755 56756'-defined-only' 56757 Dump only the defined symbols. 56758 56759'-print-value' 56760 Dump initial values of the variables. 56761 56762'-name-sort' 56763 Sort the symbols alphabetically. 56764 56765'-size-sort' 56766 Sort the symbols according to size. 56767 56768'-reverse-sort' 56769 Dump the symbols in reverse order. 56770 56771'-no-sort' 56772 Dump the symbols in order of occurrence. 56773 56774'-symbol=' 56775 Dump the details of specific symbol. 56776 56777'-objects' 56778 Dump the details of LTO objects. 56779 56780'-type-stats' 56781 Dump the statistics of tree types. 56782 56783'-tree-stats' 56784 Dump the statistics of trees. 56785 56786'-gimple-stats' 56787 Dump the statistics of gimple statements. 56788 56789'-dump-level=' 56790 For deciding the optimization level of body. 56791 56792'-dump-body=' 56793 Dump the specific gimple body. 56794 56795'-help' 56796 Display the dump tool help. 56797 56798 56799File: gcc.info, Node: Trouble, Next: Bugs, Prev: lto-dump, Up: Top 56800 5680114 Known Causes of Trouble with GCC 56802*********************************** 56803 56804This section describes known problems that affect users of GCC. Most of 56805these are not GCC bugs per se--if they were, we would fix them. But the 56806result for a user may be like the result of a bug. 56807 56808 Some of these problems are due to bugs in other software, some are 56809missing features that are too much work to add, and some are places 56810where people's opinions differ as to what is best. 56811 56812* Menu: 56813 56814* Actual Bugs:: Bugs we will fix later. 56815* Interoperation:: Problems using GCC with other compilers, 56816 and with certain linkers, assemblers and debuggers. 56817* Incompatibilities:: GCC is incompatible with traditional C. 56818* Fixed Headers:: GCC uses corrected versions of system header files. 56819 This is necessary, but doesn't always work smoothly. 56820* Standard Libraries:: GCC uses the system C library, which might not be 56821 compliant with the ISO C standard. 56822* Disappointments:: Regrettable things we cannot change, but not quite bugs. 56823* C++ Misunderstandings:: Common misunderstandings with GNU C++. 56824* Non-bugs:: Things we think are right, but some others disagree. 56825* Warnings and Errors:: Which problems in your code get warnings, 56826 and which get errors. 56827 56828 56829File: gcc.info, Node: Actual Bugs, Next: Interoperation, Up: Trouble 56830 5683114.1 Actual Bugs We Haven't Fixed Yet 56832===================================== 56833 56834 * The 'fixincludes' script interacts badly with automounters; if the 56835 directory of system header files is automounted, it tends to be 56836 unmounted while 'fixincludes' is running. This would seem to be a 56837 bug in the automounter. We don't know any good way to work around 56838 it. 56839 56840 56841File: gcc.info, Node: Interoperation, Next: Incompatibilities, Prev: Actual Bugs, Up: Trouble 56842 5684314.2 Interoperation 56844=================== 56845 56846This section lists various difficulties encountered in using GCC 56847together with other compilers or with the assemblers, linkers, libraries 56848and debuggers on certain systems. 56849 56850 * On many platforms, GCC supports a different ABI for C++ than do 56851 other compilers, so the object files compiled by GCC cannot be used 56852 with object files generated by another C++ compiler. 56853 56854 An area where the difference is most apparent is name mangling. 56855 The use of different name mangling is intentional, to protect you 56856 from more subtle problems. Compilers differ as to many internal 56857 details of C++ implementation, including: how class instances are 56858 laid out, how multiple inheritance is implemented, and how virtual 56859 function calls are handled. If the name encoding were made the 56860 same, your programs would link against libraries provided from 56861 other compilers--but the programs would then crash when run. 56862 Incompatible libraries are then detected at link time, rather than 56863 at run time. 56864 56865 * On some BSD systems, including some versions of Ultrix, use of 56866 profiling causes static variable destructors (currently used only 56867 in C++) not to be run. 56868 56869 * On a SPARC, GCC aligns all values of type 'double' on an 8-byte 56870 boundary, and it expects every 'double' to be so aligned. The Sun 56871 compiler usually gives 'double' values 8-byte alignment, with one 56872 exception: function arguments of type 'double' may not be aligned. 56873 56874 As a result, if a function compiled with Sun CC takes the address 56875 of an argument of type 'double' and passes this pointer of type 56876 'double *' to a function compiled with GCC, dereferencing the 56877 pointer may cause a fatal signal. 56878 56879 One way to solve this problem is to compile your entire program 56880 with GCC. Another solution is to modify the function that is 56881 compiled with Sun CC to copy the argument into a local variable; 56882 local variables are always properly aligned. A third solution is 56883 to modify the function that uses the pointer to dereference it via 56884 the following function 'access_double' instead of directly with 56885 '*': 56886 56887 inline double 56888 access_double (double *unaligned_ptr) 56889 { 56890 union d2i { double d; int i[2]; }; 56891 56892 union d2i *p = (union d2i *) unaligned_ptr; 56893 union d2i u; 56894 56895 u.i[0] = p->i[0]; 56896 u.i[1] = p->i[1]; 56897 56898 return u.d; 56899 } 56900 56901 Storing into the pointer can be done likewise with the same union. 56902 56903 * On Solaris, the 'malloc' function in the 'libmalloc.a' library may 56904 allocate memory that is only 4 byte aligned. Since GCC on the 56905 SPARC assumes that doubles are 8 byte aligned, this may result in a 56906 fatal signal if doubles are stored in memory allocated by the 56907 'libmalloc.a' library. 56908 56909 The solution is to not use the 'libmalloc.a' library. Use instead 56910 'malloc' and related functions from 'libc.a'; they do not have this 56911 problem. 56912 56913 * On the HP PA machine, ADB sometimes fails to work on functions 56914 compiled with GCC. Specifically, it fails to work on functions 56915 that use 'alloca' or variable-size arrays. This is because GCC 56916 doesn't generate HP-UX unwind descriptors for such functions. It 56917 may even be impossible to generate them. 56918 56919 * Debugging ('-g') is not supported on the HP PA machine, unless you 56920 use the preliminary GNU tools. 56921 56922 * Taking the address of a label may generate errors from the HP-UX PA 56923 assembler. GAS for the PA does not have this problem. 56924 56925 * Using floating point parameters for indirect calls to static 56926 functions will not work when using the HP assembler. There simply 56927 is no way for GCC to specify what registers hold arguments for 56928 static functions when using the HP assembler. GAS for the PA does 56929 not have this problem. 56930 56931 * In extremely rare cases involving some very large functions you may 56932 receive errors from the HP linker complaining about an out of 56933 bounds unconditional branch offset. This used to occur more often 56934 in previous versions of GCC, but is now exceptionally rare. If you 56935 should run into it, you can work around by making your function 56936 smaller. 56937 56938 * GCC compiled code sometimes emits warnings from the HP-UX assembler 56939 of the form: 56940 56941 (warning) Use of GR3 when 56942 frame >= 8192 may cause conflict. 56943 56944 These warnings are harmless and can be safely ignored. 56945 56946 * In extremely rare cases involving some very large functions you may 56947 receive errors from the AIX Assembler complaining about a 56948 displacement that is too large. If you should run into it, you can 56949 work around by making your function smaller. 56950 56951 * The 'libstdc++.a' library in GCC relies on the SVR4 dynamic linker 56952 semantics which merges global symbols between libraries and 56953 applications, especially necessary for C++ streams functionality. 56954 This is not the default behavior of AIX shared libraries and 56955 dynamic linking. 'libstdc++.a' is built on AIX with 56956 "runtime-linking" enabled so that symbol merging can occur. To 56957 utilize this feature, the application linked with 'libstdc++.a' 56958 must include the '-Wl,-brtl' flag on the link line. G++ cannot 56959 impose this because this option may interfere with the semantics of 56960 the user program and users may not always use 'g++' to link his or 56961 her application. Applications are not required to use the 56962 '-Wl,-brtl' flag on the link line--the rest of the 'libstdc++.a' 56963 library which is not dependent on the symbol merging semantics will 56964 continue to function correctly. 56965 56966 * An application can interpose its own definition of functions for 56967 functions invoked by 'libstdc++.a' with "runtime-linking" enabled 56968 on AIX. To accomplish this the application must be linked with 56969 "runtime-linking" option and the functions explicitly must be 56970 exported by the application ('-Wl,-brtl,-bE:exportfile'). 56971 56972 * AIX on the RS/6000 provides support (NLS) for environments outside 56973 of the United States. Compilers and assemblers use NLS to support 56974 locale-specific representations of various objects including 56975 floating-point numbers ('.' vs ',' for separating decimal 56976 fractions). There have been problems reported where the library 56977 linked with GCC does not produce the same floating-point formats 56978 that the assembler accepts. If you have this problem, set the 56979 'LANG' environment variable to 'C' or 'En_US'. 56980 56981 * Even if you specify '-fdollars-in-identifiers', you cannot 56982 successfully use '$' in identifiers on the RS/6000 due to a 56983 restriction in the IBM assembler. GAS supports these identifiers. 56984 56985 56986File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: Interoperation, Up: Trouble 56987 5698814.3 Incompatibilities of GCC 56989============================= 56990 56991There are several noteworthy incompatibilities between GNU C and K&R 56992(non-ISO) versions of C. 56993 56994 * GCC normally makes string constants read-only. If several 56995 identical-looking string constants are used, GCC stores only one 56996 copy of the string. 56997 56998 One consequence is that you cannot call 'mktemp' with a string 56999 constant argument. The function 'mktemp' always alters the string 57000 its argument points to. 57001 57002 Another consequence is that 'sscanf' does not work on some very old 57003 systems when passed a string constant as its format control string 57004 or input. This is because 'sscanf' incorrectly tries to write into 57005 the string constant. Likewise 'fscanf' and 'scanf'. 57006 57007 The solution to these problems is to change the program to use 57008 'char'-array variables with initialization strings for these 57009 purposes instead of string constants. 57010 57011 * '-2147483648' is positive. 57012 57013 This is because 2147483648 cannot fit in the type 'int', so 57014 (following the ISO C rules) its data type is 'unsigned long int'. 57015 Negating this value yields 2147483648 again. 57016 57017 * GCC does not substitute macro arguments when they appear inside of 57018 string constants. For example, the following macro in GCC 57019 57020 #define foo(a) "a" 57021 57022 will produce output '"a"' regardless of what the argument A is. 57023 57024 * When you use 'setjmp' and 'longjmp', the only automatic variables 57025 guaranteed to remain valid are those declared 'volatile'. This is 57026 a consequence of automatic register allocation. Consider this 57027 function: 57028 57029 jmp_buf j; 57030 57031 foo () 57032 { 57033 int a, b; 57034 57035 a = fun1 (); 57036 if (setjmp (j)) 57037 return a; 57038 57039 a = fun2 (); 57040 /* 'longjmp (j)' may occur in 'fun3'. */ 57041 return a + fun3 (); 57042 } 57043 57044 Here 'a' may or may not be restored to its first value when the 57045 'longjmp' occurs. If 'a' is allocated in a register, then its 57046 first value is restored; otherwise, it keeps the last value stored 57047 in it. 57048 57049 If you use the '-W' option with the '-O' option, you will get a 57050 warning when GCC thinks such a problem might be possible. 57051 57052 * Programs that use preprocessing directives in the middle of macro 57053 arguments do not work with GCC. For example, a program like this 57054 will not work: 57055 57056 foobar ( 57057 #define luser 57058 hack) 57059 57060 ISO C does not permit such a construct. 57061 57062 * K&R compilers allow comments to cross over an inclusion boundary 57063 (i.e. started in an include file and ended in the including file). 57064 57065 * Declarations of external variables and functions within a block 57066 apply only to the block containing the declaration. In other 57067 words, they have the same scope as any other declaration in the 57068 same place. 57069 57070 In some other C compilers, an 'extern' declaration affects all the 57071 rest of the file even if it happens within a block. 57072 57073 * In traditional C, you can combine 'long', etc., with a typedef 57074 name, as shown here: 57075 57076 typedef int foo; 57077 typedef long foo bar; 57078 57079 In ISO C, this is not allowed: 'long' and other type modifiers 57080 require an explicit 'int'. 57081 57082 * PCC allows typedef names to be used as function parameters. 57083 57084 * Traditional C allows the following erroneous pair of declarations 57085 to appear together in a given scope: 57086 57087 typedef int foo; 57088 typedef foo foo; 57089 57090 * GCC treats all characters of identifiers as significant. According 57091 to K&R-1 (2.2), "No more than the first eight characters are 57092 significant, although more may be used.". Also according to K&R-1 57093 (2.2), "An identifier is a sequence of letters and digits; the 57094 first character must be a letter. The underscore _ counts as a 57095 letter.", but GCC also allows dollar signs in identifiers. 57096 57097 * PCC allows whitespace in the middle of compound assignment 57098 operators such as '+='. GCC, following the ISO standard, does not 57099 allow this. 57100 57101 * GCC complains about unterminated character constants inside of 57102 preprocessing conditionals that fail. Some programs have English 57103 comments enclosed in conditionals that are guaranteed to fail; if 57104 these comments contain apostrophes, GCC will probably report an 57105 error. For example, this code would produce an error: 57106 57107 #if 0 57108 You can't expect this to work. 57109 #endif 57110 57111 The best solution to such a problem is to put the text into an 57112 actual C comment delimited by '/*...*/'. 57113 57114 * Many user programs contain the declaration 'long time ();'. In the 57115 past, the system header files on many systems did not actually 57116 declare 'time', so it did not matter what type your program 57117 declared it to return. But in systems with ISO C headers, 'time' 57118 is declared to return 'time_t', and if that is not the same as 57119 'long', then 'long time ();' is erroneous. 57120 57121 The solution is to change your program to use appropriate system 57122 headers ('<time.h>' on systems with ISO C headers) and not to 57123 declare 'time' if the system header files declare it, or failing 57124 that to use 'time_t' as the return type of 'time'. 57125 57126 * When compiling functions that return 'float', PCC converts it to a 57127 double. GCC actually returns a 'float'. If you are concerned with 57128 PCC compatibility, you should declare your functions to return 57129 'double'; you might as well say what you mean. 57130 57131 * When compiling functions that return structures or unions, GCC 57132 output code normally uses a method different from that used on most 57133 versions of Unix. As a result, code compiled with GCC cannot call 57134 a structure-returning function compiled with PCC, and vice versa. 57135 57136 The method used by GCC is as follows: a structure or union which is 57137 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or 57138 union with any other size is stored into an address supplied by the 57139 caller (usually in a special, fixed register, but on some machines 57140 it is passed on the stack). The target hook 57141 'TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address. 57142 57143 By contrast, PCC on most target machines returns structures and 57144 unions of any size by copying the data into an area of static 57145 storage, and then returning the address of that storage as if it 57146 were a pointer value. The caller must copy the data from that 57147 memory area to the place where the value is wanted. GCC does not 57148 use this method because it is slower and nonreentrant. 57149 57150 On some newer machines, PCC uses a reentrant convention for all 57151 structure and union returning. GCC on most of these machines uses 57152 a compatible convention when returning structures and unions in 57153 memory, but still returns small structures and unions in registers. 57154 57155 You can tell GCC to use a compatible convention for all structure 57156 and union returning with the option '-fpcc-struct-return'. 57157 57158 * GCC complains about program fragments such as '0x74ae-0x4000' which 57159 appear to be two hexadecimal constants separated by the minus 57160 operator. Actually, this string is a single "preprocessing token". 57161 Each such token must correspond to one token in C. Since this does 57162 not, GCC prints an error message. Although it may appear obvious 57163 that what is meant is an operator and two values, the ISO C 57164 standard specifically requires that this be treated as erroneous. 57165 57166 A "preprocessing token" is a "preprocessing number" if it begins 57167 with a digit and is followed by letters, underscores, digits, 57168 periods and 'e+', 'e-', 'E+', 'E-', 'p+', 'p-', 'P+', or 'P-' 57169 character sequences. (In strict C90 mode, the sequences 'p+', 57170 'p-', 'P+' and 'P-' cannot appear in preprocessing numbers.) 57171 57172 To make the above program fragment valid, place whitespace in front 57173 of the minus sign. This whitespace will end the preprocessing 57174 number. 57175 57176 57177File: gcc.info, Node: Fixed Headers, Next: Standard Libraries, Prev: Incompatibilities, Up: Trouble 57178 5717914.4 Fixed Header Files 57180======================= 57181 57182GCC needs to install corrected versions of some system header files. 57183This is because most target systems have some header files that won't 57184work with GCC unless they are changed. Some have bugs, some are 57185incompatible with ISO C, and some depend on special features of other 57186compilers. 57187 57188 Installing GCC automatically creates and installs the fixed header 57189files, by running a program called 'fixincludes'. Normally, you don't 57190need to pay attention to this. But there are cases where it doesn't do 57191the right thing automatically. 57192 57193 * If you update the system's header files, such as by installing a 57194 new system version, the fixed header files of GCC are not 57195 automatically updated. They can be updated using the 'mkheaders' 57196 script installed in 'LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'. 57197 57198 * On some systems, header file directories contain machine-specific 57199 symbolic links in certain places. This makes it possible to share 57200 most of the header files among hosts running the same version of 57201 the system on different machine models. 57202 57203 The programs that fix the header files do not understand this 57204 special way of using symbolic links; therefore, the directory of 57205 fixed header files is good only for the machine model used to build 57206 it. 57207 57208 It is possible to make separate sets of fixed header files for the 57209 different machine models, and arrange a structure of symbolic links 57210 so as to use the proper set, but you'll have to do this by hand. 57211 57212 57213File: gcc.info, Node: Standard Libraries, Next: Disappointments, Prev: Fixed Headers, Up: Trouble 57214 5721514.5 Standard Libraries 57216======================= 57217 57218GCC by itself attempts to be a conforming freestanding implementation. 57219*Note Language Standards Supported by GCC: Standards, for details of 57220what this means. Beyond the library facilities required of such an 57221implementation, the rest of the C library is supplied by the vendor of 57222the operating system. If that C library doesn't conform to the C 57223standards, then your programs might get warnings (especially when using 57224'-Wall') that you don't expect. 57225 57226 For example, the 'sprintf' function on SunOS 4.1.3 returns 'char *' 57227while the C standard says that 'sprintf' returns an 'int'. The 57228'fixincludes' program could make the prototype for this function match 57229the Standard, but that would be wrong, since the function will still 57230return 'char *'. 57231 57232 If you need a Standard compliant library, then you need to find one, as 57233GCC does not provide one. The GNU C library (called 'glibc') provides 57234ISO C, POSIX, BSD, SystemV and X/Open compatibility for GNU/Linux and 57235HURD-based GNU systems; no recent version of it supports other systems, 57236though some very old versions did. Version 2.2 of the GNU C library 57237includes nearly complete C99 support. You could also ask your operating 57238system vendor if newer libraries are available. 57239 57240 57241File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Standard Libraries, Up: Trouble 57242 5724314.6 Disappointments and Misunderstandings 57244========================================== 57245 57246These problems are perhaps regrettable, but we don't know any practical 57247way around them. 57248 57249 * Certain local variables aren't recognized by debuggers when you 57250 compile with optimization. 57251 57252 This occurs because sometimes GCC optimizes the variable out of 57253 existence. There is no way to tell the debugger how to compute the 57254 value such a variable "would have had", and it is not clear that 57255 would be desirable anyway. So GCC simply does not mention the 57256 eliminated variable when it writes debugging information. 57257 57258 You have to expect a certain amount of disagreement between the 57259 executable and your source code, when you use optimization. 57260 57261 * Users often think it is a bug when GCC reports an error for code 57262 like this: 57263 57264 int foo (struct mumble *); 57265 57266 struct mumble { ... }; 57267 57268 int foo (struct mumble *x) 57269 { ... } 57270 57271 This code really is erroneous, because the scope of 'struct mumble' 57272 in the prototype is limited to the argument list containing it. It 57273 does not refer to the 'struct mumble' defined with file scope 57274 immediately below--they are two unrelated types with similar names 57275 in different scopes. 57276 57277 But in the definition of 'foo', the file-scope type is used because 57278 that is available to be inherited. Thus, the definition and the 57279 prototype do not match, and you get an error. 57280 57281 This behavior may seem silly, but it's what the ISO standard 57282 specifies. It is easy enough for you to make your code work by 57283 moving the definition of 'struct mumble' above the prototype. It's 57284 not worth being incompatible with ISO C just to avoid an error for 57285 the example shown above. 57286 57287 * Accesses to bit-fields even in volatile objects works by accessing 57288 larger objects, such as a byte or a word. You cannot rely on what 57289 size of object is accessed in order to read or write the bit-field; 57290 it may even vary for a given bit-field according to the precise 57291 usage. 57292 57293 If you care about controlling the amount of memory that is 57294 accessed, use volatile but do not use bit-fields. 57295 57296 * GCC comes with shell scripts to fix certain known problems in 57297 system header files. They install corrected copies of various 57298 header files in a special directory where only GCC will normally 57299 look for them. The scripts adapt to various systems by searching 57300 all the system header files for the problem cases that we know 57301 about. 57302 57303 If new system header files are installed, nothing automatically 57304 arranges to update the corrected header files. They can be updated 57305 using the 'mkheaders' script installed in 57306 'LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'. 57307 57308 * On 68000 and x86 systems, for instance, you can get paradoxical 57309 results if you test the precise values of floating point numbers. 57310 For example, you can find that a floating point value which is not 57311 a NaN is not equal to itself. This results from the fact that the 57312 floating point registers hold a few more bits of precision than fit 57313 in a 'double' in memory. Compiled code moves values between memory 57314 and floating point registers at its convenience, and moving them 57315 into memory truncates them. 57316 57317 You can partially avoid this problem by using the '-ffloat-store' 57318 option (*note Optimize Options::). 57319 57320 * On AIX and other platforms without weak symbol support, templates 57321 need to be instantiated explicitly and symbols for static members 57322 of templates will not be generated. 57323 57324 * On AIX, GCC scans object files and library archives for static 57325 constructors and destructors when linking an application before the 57326 linker prunes unreferenced symbols. This is necessary to prevent 57327 the AIX linker from mistakenly assuming that static constructor or 57328 destructor are unused and removing them before the scanning can 57329 occur. All static constructors and destructors found will be 57330 referenced even though the modules in which they occur may not be 57331 used by the program. This may lead to both increased executable 57332 size and unexpected symbol references. 57333 57334 57335File: gcc.info, Node: C++ Misunderstandings, Next: Non-bugs, Prev: Disappointments, Up: Trouble 57336 5733714.7 Common Misunderstandings with GNU C++ 57338========================================== 57339 57340C++ is a complex language and an evolving one, and its standard 57341definition (the ISO C++ standard) was only recently completed. As a 57342result, your C++ compiler may occasionally surprise you, even when its 57343behavior is correct. This section discusses some areas that frequently 57344give rise to questions of this sort. 57345 57346* Menu: 57347 57348* Static Definitions:: Static member declarations are not definitions 57349* Name lookup:: Name lookup, templates, and accessing members of base classes 57350* Temporaries:: Temporaries may vanish before you expect 57351* Copy Assignment:: Copy Assignment operators copy virtual bases twice 57352 57353 57354File: gcc.info, Node: Static Definitions, Next: Name lookup, Up: C++ Misunderstandings 57355 5735614.7.1 Declare _and_ Define Static Members 57357------------------------------------------ 57358 57359When a class has static data members, it is not enough to _declare_ the 57360static member; you must also _define_ it. For example: 57361 57362 class Foo 57363 { 57364 ... 57365 void method(); 57366 static int bar; 57367 }; 57368 57369 This declaration only establishes that the class 'Foo' has an 'int' 57370named 'Foo::bar', and a member function named 'Foo::method'. But you 57371still need to define _both_ 'method' and 'bar' elsewhere. According to 57372the ISO standard, you must supply an initializer in one (and only one) 57373source file, such as: 57374 57375 int Foo::bar = 0; 57376 57377 Other C++ compilers may not correctly implement the standard behavior. 57378As a result, when you switch to 'g++' from one of these compilers, you 57379may discover that a program that appeared to work correctly in fact does 57380not conform to the standard: 'g++' reports as undefined symbols any 57381static data members that lack definitions. 57382 57383 57384File: gcc.info, Node: Name lookup, Next: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings 57385 5738614.7.2 Name Lookup, Templates, and Accessing Members of Base Classes 57387-------------------------------------------------------------------- 57388 57389The C++ standard prescribes that all names that are not dependent on 57390template parameters are bound to their present definitions when parsing 57391a template function or class.(1) Only names that are dependent are 57392looked up at the point of instantiation. For example, consider 57393 57394 void foo(double); 57395 57396 struct A { 57397 template <typename T> 57398 void f () { 57399 foo (1); // 1 57400 int i = N; // 2 57401 T t; 57402 t.bar(); // 3 57403 foo (t); // 4 57404 } 57405 57406 static const int N; 57407 }; 57408 57409 Here, the names 'foo' and 'N' appear in a context that does not depend 57410on the type of 'T'. The compiler will thus require that they are 57411defined in the context of use in the template, not only before the point 57412of instantiation, and will here use '::foo(double)' and 'A::N', 57413respectively. In particular, it will convert the integer value to a 57414'double' when passing it to '::foo(double)'. 57415 57416 Conversely, 'bar' and the call to 'foo' in the fourth marked line are 57417used in contexts that do depend on the type of 'T', so they are only 57418looked up at the point of instantiation, and you can provide 57419declarations for them after declaring the template, but before 57420instantiating it. In particular, if you instantiate 'A::f<int>', the 57421last line will call an overloaded '::foo(int)' if one was provided, even 57422if after the declaration of 'struct A'. 57423 57424 This distinction between lookup of dependent and non-dependent names is 57425called two-stage (or dependent) name lookup. G++ implements it since 57426version 3.4. 57427 57428 Two-stage name lookup sometimes leads to situations with behavior 57429different from non-template codes. The most common is probably this: 57430 57431 template <typename T> struct Base { 57432 int i; 57433 }; 57434 57435 template <typename T> struct Derived : public Base<T> { 57436 int get_i() { return i; } 57437 }; 57438 57439 In 'get_i()', 'i' is not used in a dependent context, so the compiler 57440will look for a name declared at the enclosing namespace scope (which is 57441the global scope here). It will not look into the base class, since 57442that is dependent and you may declare specializations of 'Base' even 57443after declaring 'Derived', so the compiler cannot really know what 'i' 57444would refer to. If there is no global variable 'i', then you will get 57445an error message. 57446 57447 In order to make it clear that you want the member of the base class, 57448you need to defer lookup until instantiation time, at which the base 57449class is known. For this, you need to access 'i' in a dependent 57450context, by either using 'this->i' (remember that 'this' is of type 57451'Derived<T>*', so is obviously dependent), or using 'Base<T>::i'. 57452Alternatively, 'Base<T>::i' might be brought into scope by a 57453'using'-declaration. 57454 57455 Another, similar example involves calling member functions of a base 57456class: 57457 57458 template <typename T> struct Base { 57459 int f(); 57460 }; 57461 57462 template <typename T> struct Derived : Base<T> { 57463 int g() { return f(); }; 57464 }; 57465 57466 Again, the call to 'f()' is not dependent on template arguments (there 57467are no arguments that depend on the type 'T', and it is also not 57468otherwise specified that the call should be in a dependent context). 57469Thus a global declaration of such a function must be available, since 57470the one in the base class is not visible until instantiation time. The 57471compiler will consequently produce the following error message: 57472 57473 x.cc: In member function `int Derived<T>::g()': 57474 x.cc:6: error: there are no arguments to `f' that depend on a template 57475 parameter, so a declaration of `f' must be available 57476 x.cc:6: error: (if you use `-fpermissive', G++ will accept your code, but 57477 allowing the use of an undeclared name is deprecated) 57478 57479 To make the code valid either use 'this->f()', or 'Base<T>::f()'. 57480Using the '-fpermissive' flag will also let the compiler accept the 57481code, by marking all function calls for which no declaration is visible 57482at the time of definition of the template for later lookup at 57483instantiation time, as if it were a dependent call. We do not recommend 57484using '-fpermissive' to work around invalid code, and it will also only 57485catch cases where functions in base classes are called, not where 57486variables in base classes are used (as in the example above). 57487 57488 Note that some compilers (including G++ versions prior to 3.4) get 57489these examples wrong and accept above code without an error. Those 57490compilers do not implement two-stage name lookup correctly. 57491 57492 ---------- Footnotes ---------- 57493 57494 (1) The C++ standard just uses the term "dependent" for names that 57495depend on the type or value of template parameters. This shorter term 57496will also be used in the rest of this section. 57497 57498 57499File: gcc.info, Node: Temporaries, Next: Copy Assignment, Prev: Name lookup, Up: C++ Misunderstandings 57500 5750114.7.3 Temporaries May Vanish Before You Expect 57502----------------------------------------------- 57503 57504It is dangerous to use pointers or references to _portions_ of a 57505temporary object. The compiler may very well delete the object before 57506you expect it to, leaving a pointer to garbage. The most common place 57507where this problem crops up is in classes like string classes, 57508especially ones that define a conversion function to type 'char *' or 57509'const char *'--which is one reason why the standard 'string' class 57510requires you to call the 'c_str' member function. However, any class 57511that returns a pointer to some internal structure is potentially subject 57512to this problem. 57513 57514 For example, a program may use a function 'strfunc' that returns 57515'string' objects, and another function 'charfunc' that operates on 57516pointers to 'char': 57517 57518 string strfunc (); 57519 void charfunc (const char *); 57520 57521 void 57522 f () 57523 { 57524 const char *p = strfunc().c_str(); 57525 ... 57526 charfunc (p); 57527 ... 57528 charfunc (p); 57529 } 57530 57531In this situation, it may seem reasonable to save a pointer to the C 57532string returned by the 'c_str' member function and use that rather than 57533call 'c_str' repeatedly. However, the temporary string created by the 57534call to 'strfunc' is destroyed after 'p' is initialized, at which point 57535'p' is left pointing to freed memory. 57536 57537 Code like this may run successfully under some other compilers, 57538particularly obsolete cfront-based compilers that delete temporaries 57539along with normal local variables. However, the GNU C++ behavior is 57540standard-conforming, so if your program depends on late destruction of 57541temporaries it is not portable. 57542 57543 The safe way to write such code is to give the temporary a name, which 57544forces it to remain until the end of the scope of the name. For 57545example: 57546 57547 const string& tmp = strfunc (); 57548 charfunc (tmp.c_str ()); 57549 57550 57551File: gcc.info, Node: Copy Assignment, Prev: Temporaries, Up: C++ Misunderstandings 57552 5755314.7.4 Implicit Copy-Assignment for Virtual Bases 57554------------------------------------------------- 57555 57556When a base class is virtual, only one subobject of the base class 57557belongs to each full object. Also, the constructors and destructors are 57558invoked only once, and called from the most-derived class. However, 57559such objects behave unspecified when being assigned. For example: 57560 57561 struct Base{ 57562 char *name; 57563 Base(const char *n) : name(strdup(n)){} 57564 Base& operator= (const Base& other){ 57565 free (name); 57566 name = strdup (other.name); 57567 return *this; 57568 } 57569 }; 57570 57571 struct A:virtual Base{ 57572 int val; 57573 A():Base("A"){} 57574 }; 57575 57576 struct B:virtual Base{ 57577 int bval; 57578 B():Base("B"){} 57579 }; 57580 57581 struct Derived:public A, public B{ 57582 Derived():Base("Derived"){} 57583 }; 57584 57585 void func(Derived &d1, Derived &d2) 57586 { 57587 d1 = d2; 57588 } 57589 57590 The C++ standard specifies that 'Base::Base' is only called once when 57591constructing or copy-constructing a Derived object. It is unspecified 57592whether 'Base::operator=' is called more than once when the implicit 57593copy-assignment for Derived objects is invoked (as it is inside 'func' 57594in the example). 57595 57596 G++ implements the "intuitive" algorithm for copy-assignment: assign 57597all direct bases, then assign all members. In that algorithm, the 57598virtual base subobject can be encountered more than once. In the 57599example, copying proceeds in the following order: 'name' (via 'strdup'), 57600'val', 'name' again, and 'bval'. 57601 57602 If application code relies on copy-assignment, a user-defined 57603copy-assignment operator removes any uncertainties. With such an 57604operator, the application can define whether and how the virtual base 57605subobject is assigned. 57606 57607 57608File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: C++ Misunderstandings, Up: Trouble 57609 5761014.8 Certain Changes We Don't Want to Make 57611========================================== 57612 57613This section lists changes that people frequently request, but which we 57614do not make because we think GCC is better without them. 57615 57616 * Checking the number and type of arguments to a function which has 57617 an old-fashioned definition and no prototype. 57618 57619 Such a feature would work only occasionally--only for calls that 57620 appear in the same file as the called function, following the 57621 definition. The only way to check all calls reliably is to add a 57622 prototype for the function. But adding a prototype eliminates the 57623 motivation for this feature. So the feature is not worthwhile. 57624 57625 * Warning about using an expression whose type is signed as a shift 57626 count. 57627 57628 Shift count operands are probably signed more often than unsigned. 57629 Warning about this would cause far more annoyance than good. 57630 57631 * Warning about assigning a signed value to an unsigned variable. 57632 57633 Such assignments must be very common; warning about them would 57634 cause more annoyance than good. 57635 57636 * Warning when a non-void function value is ignored. 57637 57638 C contains many standard functions that return a value that most 57639 programs choose to ignore. One obvious example is 'printf'. 57640 Warning about this practice only leads the defensive programmer to 57641 clutter programs with dozens of casts to 'void'. Such casts are 57642 required so frequently that they become visual noise. Writing 57643 those casts becomes so automatic that they no longer convey useful 57644 information about the intentions of the programmer. For functions 57645 where the return value should never be ignored, use the 57646 'warn_unused_result' function attribute (*note Function 57647 Attributes::). 57648 57649 * Making '-fshort-enums' the default. 57650 57651 This would cause storage layout to be incompatible with most other 57652 C compilers. And it doesn't seem very important, given that you 57653 can get the same result in other ways. The case where it matters 57654 most is when the enumeration-valued object is inside a structure, 57655 and in that case you can specify a field width explicitly. 57656 57657 * Making bit-fields unsigned by default on particular machines where 57658 "the ABI standard" says to do so. 57659 57660 The ISO C standard leaves it up to the implementation whether a 57661 bit-field declared plain 'int' is signed or not. This in effect 57662 creates two alternative dialects of C. 57663 57664 The GNU C compiler supports both dialects; you can specify the 57665 signed dialect with '-fsigned-bitfields' and the unsigned dialect 57666 with '-funsigned-bitfields'. However, this leaves open the 57667 question of which dialect to use by default. 57668 57669 Currently, the preferred dialect makes plain bit-fields signed, 57670 because this is simplest. Since 'int' is the same as 'signed int' 57671 in every other context, it is cleanest for them to be the same in 57672 bit-fields as well. 57673 57674 Some computer manufacturers have published Application Binary 57675 Interface standards which specify that plain bit-fields should be 57676 unsigned. It is a mistake, however, to say anything about this 57677 issue in an ABI. This is because the handling of plain bit-fields 57678 distinguishes two dialects of C. Both dialects are meaningful on 57679 every type of machine. Whether a particular object file was 57680 compiled using signed bit-fields or unsigned is of no concern to 57681 other object files, even if they access the same bit-fields in the 57682 same data structures. 57683 57684 A given program is written in one or the other of these two 57685 dialects. The program stands a chance to work on most any machine 57686 if it is compiled with the proper dialect. It is unlikely to work 57687 at all if compiled with the wrong dialect. 57688 57689 Many users appreciate the GNU C compiler because it provides an 57690 environment that is uniform across machines. These users would be 57691 inconvenienced if the compiler treated plain bit-fields differently 57692 on certain machines. 57693 57694 Occasionally users write programs intended only for a particular 57695 machine type. On these occasions, the users would benefit if the 57696 GNU C compiler were to support by default the same dialect as the 57697 other compilers on that machine. But such applications are rare. 57698 And users writing a program to run on more than one type of machine 57699 cannot possibly benefit from this kind of compatibility. 57700 57701 This is why GCC does and will treat plain bit-fields in the same 57702 fashion on all types of machines (by default). 57703 57704 There are some arguments for making bit-fields unsigned by default 57705 on all machines. If, for example, this becomes a universal de 57706 facto standard, it would make sense for GCC to go along with it. 57707 This is something to be considered in the future. 57708 57709 (Of course, users strongly concerned about portability should 57710 indicate explicitly in each bit-field whether it is signed or not. 57711 In this way, they write programs which have the same meaning in 57712 both C dialects.) 57713 57714 * Undefining '__STDC__' when '-ansi' is not used. 57715 57716 Currently, GCC defines '__STDC__' unconditionally. This provides 57717 good results in practice. 57718 57719 Programmers normally use conditionals on '__STDC__' to ask whether 57720 it is safe to use certain features of ISO C, such as function 57721 prototypes or ISO token concatenation. Since plain 'gcc' supports 57722 all the features of ISO C, the correct answer to these questions is 57723 "yes". 57724 57725 Some users try to use '__STDC__' to check for the availability of 57726 certain library facilities. This is actually incorrect usage in an 57727 ISO C program, because the ISO C standard says that a conforming 57728 freestanding implementation should define '__STDC__' even though it 57729 does not have the library facilities. 'gcc -ansi -pedantic' is a 57730 conforming freestanding implementation, and it is therefore 57731 required to define '__STDC__', even though it does not come with an 57732 ISO C library. 57733 57734 Sometimes people say that defining '__STDC__' in a compiler that 57735 does not completely conform to the ISO C standard somehow violates 57736 the standard. This is illogical. The standard is a standard for 57737 compilers that claim to support ISO C, such as 'gcc -ansi'--not for 57738 other compilers such as plain 'gcc'. Whatever the ISO C standard 57739 says is relevant to the design of plain 'gcc' without '-ansi' only 57740 for pragmatic reasons, not as a requirement. 57741 57742 GCC normally defines '__STDC__' to be 1, and in addition defines 57743 '__STRICT_ANSI__' if you specify the '-ansi' option, or a '-std' 57744 option for strict conformance to some version of ISO C. On some 57745 hosts, system include files use a different convention, where 57746 '__STDC__' is normally 0, but is 1 if the user specifies strict 57747 conformance to the C Standard. GCC follows the host convention 57748 when processing system include files, but when processing user 57749 files it follows the usual GNU C convention. 57750 57751 * Undefining '__STDC__' in C++. 57752 57753 Programs written to compile with C++-to-C translators get the value 57754 of '__STDC__' that goes with the C compiler that is subsequently 57755 used. These programs must test '__STDC__' to determine what kind 57756 of C preprocessor that compiler uses: whether they should 57757 concatenate tokens in the ISO C fashion or in the traditional 57758 fashion. 57759 57760 These programs work properly with GNU C++ if '__STDC__' is defined. 57761 They would not work otherwise. 57762 57763 In addition, many header files are written to provide prototypes in 57764 ISO C but not in traditional C. Many of these header files can 57765 work without change in C++ provided '__STDC__' is defined. If 57766 '__STDC__' is not defined, they will all fail, and will all need to 57767 be changed to test explicitly for C++ as well. 57768 57769 * Deleting "empty" loops. 57770 57771 Historically, GCC has not deleted "empty" loops under the 57772 assumption that the most likely reason you would put one in a 57773 program is to have a delay, so deleting them will not make real 57774 programs run any faster. 57775 57776 However, the rationale here is that optimization of a nonempty loop 57777 cannot produce an empty one. This held for carefully written C 57778 compiled with less powerful optimizers but is not always the case 57779 for carefully written C++ or with more powerful optimizers. Thus 57780 GCC will remove operations from loops whenever it can determine 57781 those operations are not externally visible (apart from the time 57782 taken to execute them, of course). In case the loop can be proved 57783 to be finite, GCC will also remove the loop itself. 57784 57785 Be aware of this when performing timing tests, for instance the 57786 following loop can be completely removed, provided 57787 'some_expression' can provably not change any global state. 57788 57789 { 57790 int sum = 0; 57791 int ix; 57792 57793 for (ix = 0; ix != 10000; ix++) 57794 sum += some_expression; 57795 } 57796 57797 Even though 'sum' is accumulated in the loop, no use is made of 57798 that summation, so the accumulation can be removed. 57799 57800 * Making side effects happen in the same order as in some other 57801 compiler. 57802 57803 It is never safe to depend on the order of evaluation of side 57804 effects. For example, a function call like this may very well 57805 behave differently from one compiler to another: 57806 57807 void func (int, int); 57808 57809 int i = 2; 57810 func (i++, i++); 57811 57812 There is no guarantee (in either the C or the C++ standard language 57813 definitions) that the increments will be evaluated in any 57814 particular order. Either increment might happen first. 'func' 57815 might get the arguments '2, 3', or it might get '3, 2', or even '2, 57816 2'. 57817 57818 * Making certain warnings into errors by default. 57819 57820 Some ISO C testsuites report failure when the compiler does not 57821 produce an error message for a certain program. 57822 57823 ISO C requires a "diagnostic" message for certain kinds of invalid 57824 programs, but a warning is defined by GCC to count as a diagnostic. 57825 If GCC produces a warning but not an error, that is correct ISO C 57826 support. If testsuites call this "failure", they should be run 57827 with the GCC option '-pedantic-errors', which will turn these 57828 warnings into errors. 57829 57830 57831File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble 57832 5783314.9 Warning Messages and Error Messages 57834======================================== 57835 57836The GNU compiler can produce two kinds of diagnostics: errors and 57837warnings. Each kind has a different purpose: 57838 57839 "Errors" report problems that make it impossible to compile your 57840 program. GCC reports errors with the source file name and line 57841 number where the problem is apparent. 57842 57843 "Warnings" report other unusual conditions in your code that _may_ 57844 indicate a problem, although compilation can (and does) proceed. 57845 Warning messages also report the source file name and line number, 57846 but include the text 'warning:' to distinguish them from error 57847 messages. 57848 57849 Warnings may indicate danger points where you should check to make sure 57850that your program really does what you intend; or the use of obsolete 57851features; or the use of nonstandard features of GNU C or C++. Many 57852warnings are issued only if you ask for them, with one of the '-W' 57853options (for instance, '-Wall' requests a variety of useful warnings). 57854 57855 GCC always tries to compile your program if possible; it never 57856gratuitously rejects a program whose meaning is clear merely because 57857(for instance) it fails to conform to a standard. In some cases, 57858however, the C and C++ standards specify that certain extensions are 57859forbidden, and a diagnostic _must_ be issued by a conforming compiler. 57860The '-pedantic' option tells GCC to issue warnings in such cases; 57861'-pedantic-errors' says to make them errors instead. This does not mean 57862that _all_ non-ISO constructs get warnings or errors. 57863 57864 *Note Options to Request or Suppress Warnings: Warning Options, for 57865more detail on these and related command-line options. 57866 57867 57868File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top 57869 5787015 Reporting Bugs 57871***************** 57872 57873Your bug reports play an essential role in making GCC reliable. 57874 57875 When you encounter a problem, the first thing to do is to see if it is 57876already known. *Note Trouble::. If it isn't known, then you should 57877report the problem. 57878 57879* Menu: 57880 57881* Criteria: Bug Criteria. Have you really found a bug? 57882* Reporting: Bug Reporting. How to report a bug effectively. 57883 57884 57885File: gcc.info, Node: Bug Criteria, Next: Bug Reporting, Up: Bugs 57886 5788715.1 Have You Found a Bug? 57888========================== 57889 57890If you are not sure whether you have found a bug, here are some 57891guidelines: 57892 57893 * If the compiler gets a fatal signal, for any input whatever, that 57894 is a compiler bug. Reliable compilers never crash. 57895 57896 * If the compiler produces invalid assembly code, for any input 57897 whatever (except an 'asm' statement), that is a compiler bug, 57898 unless the compiler reports errors (not just warnings) which would 57899 ordinarily prevent the assembler from being run. 57900 57901 * If the compiler produces valid assembly code that does not 57902 correctly execute the input source code, that is a compiler bug. 57903 57904 However, you must double-check to make sure, because you may have a 57905 program whose behavior is undefined, which happened by chance to 57906 give the desired results with another C or C++ compiler. 57907 57908 For example, in many nonoptimizing compilers, you can write 'x;' at 57909 the end of a function instead of 'return x;', with the same 57910 results. But the value of the function is undefined if 'return' is 57911 omitted; it is not a bug when GCC produces different results. 57912 57913 Problems often result from expressions with two increment 57914 operators, as in 'f (*p++, *p++)'. Your previous compiler might 57915 have interpreted that expression the way you intended; GCC might 57916 interpret it another way. Neither compiler is wrong. The bug is 57917 in your code. 57918 57919 After you have localized the error to a single source line, it 57920 should be easy to check for these things. If your program is 57921 correct and well defined, you have found a compiler bug. 57922 57923 * If the compiler produces an error message for valid input, that is 57924 a compiler bug. 57925 57926 * If the compiler does not produce an error message for invalid 57927 input, that is a compiler bug. However, you should note that your 57928 idea of "invalid input" might be someone else's idea of "an 57929 extension" or "support for traditional practice". 57930 57931 * If you are an experienced user of one of the languages GCC 57932 supports, your suggestions for improvement of GCC are welcome in 57933 any case. 57934 57935 57936File: gcc.info, Node: Bug Reporting, Prev: Bug Criteria, Up: Bugs 57937 5793815.2 How and Where to Report Bugs 57939================================= 57940 57941Bugs should be reported to the bug database at 57942<https://gcc.gnu.org/bugs/>. 57943 57944 57945File: gcc.info, Node: Service, Next: Contributing, Prev: Bugs, Up: Top 57946 5794716 How To Get Help with GCC 57948*************************** 57949 57950If you need help installing, using or changing GCC, there are two ways 57951to find it: 57952 57953 * Send a message to a suitable network mailing list. First try 57954 <gcc-help@gcc.gnu.org> (for help installing or using GCC), and if 57955 that brings no response, try <gcc@gcc.gnu.org>. For help changing 57956 GCC, ask <gcc@gcc.gnu.org>. If you think you have found a bug in 57957 GCC, please report it following the instructions at *note Bug 57958 Reporting::. 57959 57960 * Look in the service directory for someone who might help you for a 57961 fee. The service directory is found at 57962 <https://www.fsf.org/resources/service>. 57963 57964 For further information, see <http://gcc.gnu.org/faq.html#support>. 57965 57966 57967File: gcc.info, Node: Contributing, Next: Funding, Prev: Service, Up: Top 57968 5796917 Contributing to GCC Development 57970********************************** 57971 57972If you would like to help pretest GCC releases to assure they work well, 57973current development sources are available via Git (see 57974<http://gcc.gnu.org/git.html>). Source and binary snapshots are also 57975available for FTP; see <http://gcc.gnu.org/snapshots.html>. 57976 57977 If you would like to work on improvements to GCC, please read the 57978advice at these URLs: 57979 57980 <http://gcc.gnu.org/contribute.html> 57981 <http://gcc.gnu.org/contributewhy.html> 57982 57983for information on how to make useful contributions and avoid 57984duplication of effort. Suggested projects are listed at 57985<http://gcc.gnu.org/projects/>. 57986 57987 57988File: gcc.info, Node: Funding, Next: GNU Project, Prev: Contributing, Up: Top 57989 57990Funding Free Software 57991********************* 57992 57993If you want to have more free software a few years from now, it makes 57994sense for you to help encourage people to contribute funds for its 57995development. The most effective approach known is to encourage 57996commercial redistributors to donate. 57997 57998 Users of free software systems can boost the pace of development by 57999encouraging for-a-fee distributors to donate part of their selling price 58000to free software developers--the Free Software Foundation, and others. 58001 58002 The way to convince distributors to do this is to demand it and expect 58003it from them. So when you compare distributors, judge them partly by 58004how much they give to free software development. Show distributors they 58005must compete to be the one who gives the most. 58006 58007 To make this approach work, you must insist on numbers that you can 58008compare, such as, "We will donate ten dollars to the Frobnitz project 58009for each disk sold." Don't be satisfied with a vague promise, such as 58010"A portion of the profits are donated," since it doesn't give a basis 58011for comparison. 58012 58013 Even a precise fraction "of the profits from this disk" is not very 58014meaningful, since creative accounting and unrelated business decisions 58015can greatly alter what fraction of the sales price counts as profit. If 58016the price you pay is $50, ten percent of the profit is probably less 58017than a dollar; it might be a few cents, or nothing at all. 58018 58019 Some redistributors do development work themselves. This is useful 58020too; but to keep everyone honest, you need to inquire how much they do, 58021and what kind. Some kinds of development make much more long-term 58022difference than others. For example, maintaining a separate version of 58023a program contributes very little; maintaining the standard version of a 58024program for the whole community contributes much. Easy new ports 58025contribute little, since someone else would surely do them; difficult 58026ports such as adding a new CPU to the GNU Compiler Collection contribute 58027more; major new features or packages contribute the most. 58028 58029 By establishing the idea that supporting further development is "the 58030proper thing to do" when distributing free software for a fee, we can 58031assure a steady flow of resources into making more free software. 58032 58033 Copyright (C) 1994 Free Software Foundation, Inc. 58034 Verbatim copying and redistribution of this section is permitted 58035 without royalty; alteration is not permitted. 58036 58037 58038File: gcc.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top 58039 58040The GNU Project and GNU/Linux 58041***************************** 58042 58043The GNU Project was launched in 1984 to develop a complete Unix-like 58044operating system which is free software: the GNU system. (GNU is a 58045recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".) 58046Variants of the GNU operating system, which use the kernel Linux, are 58047now widely used; though these systems are often referred to as "Linux", 58048they are more accurately called GNU/Linux systems. 58049 58050 For more information, see: 58051 <http://www.gnu.org/> 58052 <http://www.gnu.org/gnu/linux-and-gnu.html> 58053 58054 58055File: gcc.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top 58056 58057GNU General Public License 58058************************** 58059 58060 Version 3, 29 June 2007 58061 58062 Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/> 58063 58064 Everyone is permitted to copy and distribute verbatim copies of this 58065 license document, but changing it is not allowed. 58066 58067Preamble 58068======== 58069 58070The GNU General Public License is a free, copyleft license for software 58071and other kinds of works. 58072 58073 The licenses for most software and other practical works are designed 58074to take away your freedom to share and change the works. By contrast, 58075the GNU General Public License is intended to guarantee your freedom to 58076share and change all versions of a program-to make sure it remains free 58077software for all its users. We, the Free Software Foundation, use the 58078GNU General Public License for most of our software; it applies also to 58079any other work released this way by its authors. You can apply it to 58080your programs, too. 58081 58082 When we speak of free software, we are referring to freedom, not price. 58083Our General Public Licenses are designed to make sure that you have the 58084freedom to distribute copies of free software (and charge for them if 58085you wish), that you receive source code or can get it if you want it, 58086that you can change the software or use pieces of it in new free 58087programs, and that you know you can do these things. 58088 58089 To protect your rights, we need to prevent others from denying you 58090these rights or asking you to surrender the rights. Therefore, you have 58091certain responsibilities if you distribute copies of the software, or if 58092you modify it: responsibilities to respect the freedom of others. 58093 58094 For example, if you distribute copies of such a program, whether gratis 58095or for a fee, you must pass on to the recipients the same freedoms that 58096you received. You must make sure that they, too, receive or can get the 58097source code. And you must show them these terms so they know their 58098rights. 58099 58100 Developers that use the GNU GPL protect your rights with two steps: (1) 58101assert copyright on the software, and (2) offer you this License giving 58102you legal permission to copy, distribute and/or modify it. 58103 58104 For the developers' and authors' protection, the GPL clearly explains 58105that there is no warranty for this free software. For both users' and 58106authors' sake, the GPL requires that modified versions be marked as 58107changed, so that their problems will not be attributed erroneously to 58108authors of previous versions. 58109 58110 Some devices are designed to deny users access to install or run 58111modified versions of the software inside them, although the manufacturer 58112can do so. This is fundamentally incompatible with the aim of 58113protecting users' freedom to change the software. The systematic 58114pattern of such abuse occurs in the area of products for individuals to 58115use, which is precisely where it is most unacceptable. Therefore, we 58116have designed this version of the GPL to prohibit the practice for those 58117products. If such problems arise substantially in other domains, we 58118stand ready to extend this provision to those domains in future versions 58119of the GPL, as needed to protect the freedom of users. 58120 58121 Finally, every program is threatened constantly by software patents. 58122States should not allow patents to restrict development and use of 58123software on general-purpose computers, but in those that do, we wish to 58124avoid the special danger that patents applied to a free program could 58125make it effectively proprietary. To prevent this, the GPL assures that 58126patents cannot be used to render the program non-free. 58127 58128 The precise terms and conditions for copying, distribution and 58129modification follow. 58130 58131TERMS AND CONDITIONS 58132==================== 58133 58134 0. Definitions. 58135 58136 "This License" refers to version 3 of the GNU General Public 58137 License. 58138 58139 "Copyright" also means copyright-like laws that apply to other 58140 kinds of works, such as semiconductor masks. 58141 58142 "The Program" refers to any copyrightable work licensed under this 58143 License. Each licensee is addressed as "you". "Licensees" and 58144 "recipients" may be individuals or organizations. 58145 58146 To "modify" a work means to copy from or adapt all or part of the 58147 work in a fashion requiring copyright permission, other than the 58148 making of an exact copy. The resulting work is called a "modified 58149 version" of the earlier work or a work "based on" the earlier work. 58150 58151 A "covered work" means either the unmodified Program or a work 58152 based on the Program. 58153 58154 To "propagate" a work means to do anything with it that, without 58155 permission, would make you directly or secondarily liable for 58156 infringement under applicable copyright law, except executing it on 58157 a computer or modifying a private copy. Propagation includes 58158 copying, distribution (with or without modification), making 58159 available to the public, and in some countries other activities as 58160 well. 58161 58162 To "convey" a work means any kind of propagation that enables other 58163 parties to make or receive copies. Mere interaction with a user 58164 through a computer network, with no transfer of a copy, is not 58165 conveying. 58166 58167 An interactive user interface displays "Appropriate Legal Notices" 58168 to the extent that it includes a convenient and prominently visible 58169 feature that (1) displays an appropriate copyright notice, and (2) 58170 tells the user that there is no warranty for the work (except to 58171 the extent that warranties are provided), that licensees may convey 58172 the work under this License, and how to view a copy of this 58173 License. If the interface presents a list of user commands or 58174 options, such as a menu, a prominent item in the list meets this 58175 criterion. 58176 58177 1. Source Code. 58178 58179 The "source code" for a work means the preferred form of the work 58180 for making modifications to it. "Object code" means any non-source 58181 form of a work. 58182 58183 A "Standard Interface" means an interface that either is an 58184 official standard defined by a recognized standards body, or, in 58185 the case of interfaces specified for a particular programming 58186 language, one that is widely used among developers working in that 58187 language. 58188 58189 The "System Libraries" of an executable work include anything, 58190 other than the work as a whole, that (a) is included in the normal 58191 form of packaging a Major Component, but which is not part of that 58192 Major Component, and (b) serves only to enable use of the work with 58193 that Major Component, or to implement a Standard Interface for 58194 which an implementation is available to the public in source code 58195 form. A "Major Component", in this context, means a major 58196 essential component (kernel, window system, and so on) of the 58197 specific operating system (if any) on which the executable work 58198 runs, or a compiler used to produce the work, or an object code 58199 interpreter used to run it. 58200 58201 The "Corresponding Source" for a work in object code form means all 58202 the source code needed to generate, install, and (for an executable 58203 work) run the object code and to modify the work, including scripts 58204 to control those activities. However, it does not include the 58205 work's System Libraries, or general-purpose tools or generally 58206 available free programs which are used unmodified in performing 58207 those activities but which are not part of the work. For example, 58208 Corresponding Source includes interface definition files associated 58209 with source files for the work, and the source code for shared 58210 libraries and dynamically linked subprograms that the work is 58211 specifically designed to require, such as by intimate data 58212 communication or control flow between those subprograms and other 58213 parts of the work. 58214 58215 The Corresponding Source need not include anything that users can 58216 regenerate automatically from other parts of the Corresponding 58217 Source. 58218 58219 The Corresponding Source for a work in source code form is that 58220 same work. 58221 58222 2. Basic Permissions. 58223 58224 All rights granted under this License are granted for the term of 58225 copyright on the Program, and are irrevocable provided the stated 58226 conditions are met. This License explicitly affirms your unlimited 58227 permission to run the unmodified Program. The output from running 58228 a covered work is covered by this License only if the output, given 58229 its content, constitutes a covered work. This License acknowledges 58230 your rights of fair use or other equivalent, as provided by 58231 copyright law. 58232 58233 You may make, run and propagate covered works that you do not 58234 convey, without conditions so long as your license otherwise 58235 remains in force. You may convey covered works to others for the 58236 sole purpose of having them make modifications exclusively for you, 58237 or provide you with facilities for running those works, provided 58238 that you comply with the terms of this License in conveying all 58239 material for which you do not control copyright. Those thus making 58240 or running the covered works for you must do so exclusively on your 58241 behalf, under your direction and control, on terms that prohibit 58242 them from making any copies of your copyrighted material outside 58243 their relationship with you. 58244 58245 Conveying under any other circumstances is permitted solely under 58246 the conditions stated below. Sublicensing is not allowed; section 58247 10 makes it unnecessary. 58248 58249 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 58250 58251 No covered work shall be deemed part of an effective technological 58252 measure under any applicable law fulfilling obligations under 58253 article 11 of the WIPO copyright treaty adopted on 20 December 58254 1996, or similar laws prohibiting or restricting circumvention of 58255 such measures. 58256 58257 When you convey a covered work, you waive any legal power to forbid 58258 circumvention of technological measures to the extent such 58259 circumvention is effected by exercising rights under this License 58260 with respect to the covered work, and you disclaim any intention to 58261 limit operation or modification of the work as a means of 58262 enforcing, against the work's users, your or third parties' legal 58263 rights to forbid circumvention of technological measures. 58264 58265 4. Conveying Verbatim Copies. 58266 58267 You may convey verbatim copies of the Program's source code as you 58268 receive it, in any medium, provided that you conspicuously and 58269 appropriately publish on each copy an appropriate copyright notice; 58270 keep intact all notices stating that this License and any 58271 non-permissive terms added in accord with section 7 apply to the 58272 code; keep intact all notices of the absence of any warranty; and 58273 give all recipients a copy of this License along with the Program. 58274 58275 You may charge any price or no price for each copy that you convey, 58276 and you may offer support or warranty protection for a fee. 58277 58278 5. Conveying Modified Source Versions. 58279 58280 You may convey a work based on the Program, or the modifications to 58281 produce it from the Program, in the form of source code under the 58282 terms of section 4, provided that you also meet all of these 58283 conditions: 58284 58285 a. The work must carry prominent notices stating that you 58286 modified it, and giving a relevant date. 58287 58288 b. The work must carry prominent notices stating that it is 58289 released under this License and any conditions added under 58290 section 7. This requirement modifies the requirement in 58291 section 4 to "keep intact all notices". 58292 58293 c. You must license the entire work, as a whole, under this 58294 License to anyone who comes into possession of a copy. This 58295 License will therefore apply, along with any applicable 58296 section 7 additional terms, to the whole of the work, and all 58297 its parts, regardless of how they are packaged. This License 58298 gives no permission to license the work in any other way, but 58299 it does not invalidate such permission if you have separately 58300 received it. 58301 58302 d. If the work has interactive user interfaces, each must display 58303 Appropriate Legal Notices; however, if the Program has 58304 interactive interfaces that do not display Appropriate Legal 58305 Notices, your work need not make them do so. 58306 58307 A compilation of a covered work with other separate and independent 58308 works, which are not by their nature extensions of the covered 58309 work, and which are not combined with it such as to form a larger 58310 program, in or on a volume of a storage or distribution medium, is 58311 called an "aggregate" if the compilation and its resulting 58312 copyright are not used to limit the access or legal rights of the 58313 compilation's users beyond what the individual works permit. 58314 Inclusion of a covered work in an aggregate does not cause this 58315 License to apply to the other parts of the aggregate. 58316 58317 6. Conveying Non-Source Forms. 58318 58319 You may convey a covered work in object code form under the terms 58320 of sections 4 and 5, provided that you also convey the 58321 machine-readable Corresponding Source under the terms of this 58322 License, in one of these ways: 58323 58324 a. Convey the object code in, or embodied in, a physical product 58325 (including a physical distribution medium), accompanied by the 58326 Corresponding Source fixed on a durable physical medium 58327 customarily used for software interchange. 58328 58329 b. Convey the object code in, or embodied in, a physical product 58330 (including a physical distribution medium), accompanied by a 58331 written offer, valid for at least three years and valid for as 58332 long as you offer spare parts or customer support for that 58333 product model, to give anyone who possesses the object code 58334 either (1) a copy of the Corresponding Source for all the 58335 software in the product that is covered by this License, on a 58336 durable physical medium customarily used for software 58337 interchange, for a price no more than your reasonable cost of 58338 physically performing this conveying of source, or (2) access 58339 to copy the Corresponding Source from a network server at no 58340 charge. 58341 58342 c. Convey individual copies of the object code with a copy of the 58343 written offer to provide the Corresponding Source. This 58344 alternative is allowed only occasionally and noncommercially, 58345 and only if you received the object code with such an offer, 58346 in accord with subsection 6b. 58347 58348 d. Convey the object code by offering access from a designated 58349 place (gratis or for a charge), and offer equivalent access to 58350 the Corresponding Source in the same way through the same 58351 place at no further charge. You need not require recipients 58352 to copy the Corresponding Source along with the object code. 58353 If the place to copy the object code is a network server, the 58354 Corresponding Source may be on a different server (operated by 58355 you or a third party) that supports equivalent copying 58356 facilities, provided you maintain clear directions next to the 58357 object code saying where to find the Corresponding Source. 58358 Regardless of what server hosts the Corresponding Source, you 58359 remain obligated to ensure that it is available for as long as 58360 needed to satisfy these requirements. 58361 58362 e. Convey the object code using peer-to-peer transmission, 58363 provided you inform other peers where the object code and 58364 Corresponding Source of the work are being offered to the 58365 general public at no charge under subsection 6d. 58366 58367 A separable portion of the object code, whose source code is 58368 excluded from the Corresponding Source as a System Library, need 58369 not be included in conveying the object code work. 58370 58371 A "User Product" is either (1) a "consumer product", which means 58372 any tangible personal property which is normally used for personal, 58373 family, or household purposes, or (2) anything designed or sold for 58374 incorporation into a dwelling. In determining whether a product is 58375 a consumer product, doubtful cases shall be resolved in favor of 58376 coverage. For a particular product received by a particular user, 58377 "normally used" refers to a typical or common use of that class of 58378 product, regardless of the status of the particular user or of the 58379 way in which the particular user actually uses, or expects or is 58380 expected to use, the product. A product is a consumer product 58381 regardless of whether the product has substantial commercial, 58382 industrial or non-consumer uses, unless such uses represent the 58383 only significant mode of use of the product. 58384 58385 "Installation Information" for a User Product means any methods, 58386 procedures, authorization keys, or other information required to 58387 install and execute modified versions of a covered work in that 58388 User Product from a modified version of its Corresponding Source. 58389 The information must suffice to ensure that the continued 58390 functioning of the modified object code is in no case prevented or 58391 interfered with solely because modification has been made. 58392 58393 If you convey an object code work under this section in, or with, 58394 or specifically for use in, a User Product, and the conveying 58395 occurs as part of a transaction in which the right of possession 58396 and use of the User Product is transferred to the recipient in 58397 perpetuity or for a fixed term (regardless of how the transaction 58398 is characterized), the Corresponding Source conveyed under this 58399 section must be accompanied by the Installation Information. But 58400 this requirement does not apply if neither you nor any third party 58401 retains the ability to install modified object code on the User 58402 Product (for example, the work has been installed in ROM). 58403 58404 The requirement to provide Installation Information does not 58405 include a requirement to continue to provide support service, 58406 warranty, or updates for a work that has been modified or installed 58407 by the recipient, or for the User Product in which it has been 58408 modified or installed. Access to a network may be denied when the 58409 modification itself materially and adversely affects the operation 58410 of the network or violates the rules and protocols for 58411 communication across the network. 58412 58413 Corresponding Source conveyed, and Installation Information 58414 provided, in accord with this section must be in a format that is 58415 publicly documented (and with an implementation available to the 58416 public in source code form), and must require no special password 58417 or key for unpacking, reading or copying. 58418 58419 7. Additional Terms. 58420 58421 "Additional permissions" are terms that supplement the terms of 58422 this License by making exceptions from one or more of its 58423 conditions. Additional permissions that are applicable to the 58424 entire Program shall be treated as though they were included in 58425 this License, to the extent that they are valid under applicable 58426 law. If additional permissions apply only to part of the Program, 58427 that part may be used separately under those permissions, but the 58428 entire Program remains governed by this License without regard to 58429 the additional permissions. 58430 58431 When you convey a copy of a covered work, you may at your option 58432 remove any additional permissions from that copy, or from any part 58433 of it. (Additional permissions may be written to require their own 58434 removal in certain cases when you modify the work.) You may place 58435 additional permissions on material, added by you to a covered work, 58436 for which you have or can give appropriate copyright permission. 58437 58438 Notwithstanding any other provision of this License, for material 58439 you add to a covered work, you may (if authorized by the copyright 58440 holders of that material) supplement the terms of this License with 58441 terms: 58442 58443 a. Disclaiming warranty or limiting liability differently from 58444 the terms of sections 15 and 16 of this License; or 58445 58446 b. Requiring preservation of specified reasonable legal notices 58447 or author attributions in that material or in the Appropriate 58448 Legal Notices displayed by works containing it; or 58449 58450 c. Prohibiting misrepresentation of the origin of that material, 58451 or requiring that modified versions of such material be marked 58452 in reasonable ways as different from the original version; or 58453 58454 d. Limiting the use for publicity purposes of names of licensors 58455 or authors of the material; or 58456 58457 e. Declining to grant rights under trademark law for use of some 58458 trade names, trademarks, or service marks; or 58459 58460 f. Requiring indemnification of licensors and authors of that 58461 material by anyone who conveys the material (or modified 58462 versions of it) with contractual assumptions of liability to 58463 the recipient, for any liability that these contractual 58464 assumptions directly impose on those licensors and authors. 58465 58466 All other non-permissive additional terms are considered "further 58467 restrictions" within the meaning of section 10. If the Program as 58468 you received it, or any part of it, contains a notice stating that 58469 it is governed by this License along with a term that is a further 58470 restriction, you may remove that term. If a license document 58471 contains a further restriction but permits relicensing or conveying 58472 under this License, you may add to a covered work material governed 58473 by the terms of that license document, provided that the further 58474 restriction does not survive such relicensing or conveying. 58475 58476 If you add terms to a covered work in accord with this section, you 58477 must place, in the relevant source files, a statement of the 58478 additional terms that apply to those files, or a notice indicating 58479 where to find the applicable terms. 58480 58481 Additional terms, permissive or non-permissive, may be stated in 58482 the form of a separately written license, or stated as exceptions; 58483 the above requirements apply either way. 58484 58485 8. Termination. 58486 58487 You may not propagate or modify a covered work except as expressly 58488 provided under this License. Any attempt otherwise to propagate or 58489 modify it is void, and will automatically terminate your rights 58490 under this License (including any patent licenses granted under the 58491 third paragraph of section 11). 58492 58493 However, if you cease all violation of this License, then your 58494 license from a particular copyright holder is reinstated (a) 58495 provisionally, unless and until the copyright holder explicitly and 58496 finally terminates your license, and (b) permanently, if the 58497 copyright holder fails to notify you of the violation by some 58498 reasonable means prior to 60 days after the cessation. 58499 58500 Moreover, your license from a particular copyright holder is 58501 reinstated permanently if the copyright holder notifies you of the 58502 violation by some reasonable means, this is the first time you have 58503 received notice of violation of this License (for any work) from 58504 that copyright holder, and you cure the violation prior to 30 days 58505 after your receipt of the notice. 58506 58507 Termination of your rights under this section does not terminate 58508 the licenses of parties who have received copies or rights from you 58509 under this License. If your rights have been terminated and not 58510 permanently reinstated, you do not qualify to receive new licenses 58511 for the same material under section 10. 58512 58513 9. Acceptance Not Required for Having Copies. 58514 58515 You are not required to accept this License in order to receive or 58516 run a copy of the Program. Ancillary propagation of a covered work 58517 occurring solely as a consequence of using peer-to-peer 58518 transmission to receive a copy likewise does not require 58519 acceptance. However, nothing other than this License grants you 58520 permission to propagate or modify any covered work. These actions 58521 infringe copyright if you do not accept this License. Therefore, 58522 by modifying or propagating a covered work, you indicate your 58523 acceptance of this License to do so. 58524 58525 10. Automatic Licensing of Downstream Recipients. 58526 58527 Each time you convey a covered work, the recipient automatically 58528 receives a license from the original licensors, to run, modify and 58529 propagate that work, subject to this License. You are not 58530 responsible for enforcing compliance by third parties with this 58531 License. 58532 58533 An "entity transaction" is a transaction transferring control of an 58534 organization, or substantially all assets of one, or subdividing an 58535 organization, or merging organizations. If propagation of a 58536 covered work results from an entity transaction, each party to that 58537 transaction who receives a copy of the work also receives whatever 58538 licenses to the work the party's predecessor in interest had or 58539 could give under the previous paragraph, plus a right to possession 58540 of the Corresponding Source of the work from the predecessor in 58541 interest, if the predecessor has it or can get it with reasonable 58542 efforts. 58543 58544 You may not impose any further restrictions on the exercise of the 58545 rights granted or affirmed under this License. For example, you 58546 may not impose a license fee, royalty, or other charge for exercise 58547 of rights granted under this License, and you may not initiate 58548 litigation (including a cross-claim or counterclaim in a lawsuit) 58549 alleging that any patent claim is infringed by making, using, 58550 selling, offering for sale, or importing the Program or any portion 58551 of it. 58552 58553 11. Patents. 58554 58555 A "contributor" is a copyright holder who authorizes use under this 58556 License of the Program or a work on which the Program is based. 58557 The work thus licensed is called the contributor's "contributor 58558 version". 58559 58560 A contributor's "essential patent claims" are all patent claims 58561 owned or controlled by the contributor, whether already acquired or 58562 hereafter acquired, that would be infringed by some manner, 58563 permitted by this License, of making, using, or selling its 58564 contributor version, but do not include claims that would be 58565 infringed only as a consequence of further modification of the 58566 contributor version. For purposes of this definition, "control" 58567 includes the right to grant patent sublicenses in a manner 58568 consistent with the requirements of this License. 58569 58570 Each contributor grants you a non-exclusive, worldwide, 58571 royalty-free patent license under the contributor's essential 58572 patent claims, to make, use, sell, offer for sale, import and 58573 otherwise run, modify and propagate the contents of its contributor 58574 version. 58575 58576 In the following three paragraphs, a "patent license" is any 58577 express agreement or commitment, however denominated, not to 58578 enforce a patent (such as an express permission to practice a 58579 patent or covenant not to sue for patent infringement). To "grant" 58580 such a patent license to a party means to make such an agreement or 58581 commitment not to enforce a patent against the party. 58582 58583 If you convey a covered work, knowingly relying on a patent 58584 license, and the Corresponding Source of the work is not available 58585 for anyone to copy, free of charge and under the terms of this 58586 License, through a publicly available network server or other 58587 readily accessible means, then you must either (1) cause the 58588 Corresponding Source to be so available, or (2) arrange to deprive 58589 yourself of the benefit of the patent license for this particular 58590 work, or (3) arrange, in a manner consistent with the requirements 58591 of this License, to extend the patent license to downstream 58592 recipients. "Knowingly relying" means you have actual knowledge 58593 that, but for the patent license, your conveying the covered work 58594 in a country, or your recipient's use of the covered work in a 58595 country, would infringe one or more identifiable patents in that 58596 country that you have reason to believe are valid. 58597 58598 If, pursuant to or in connection with a single transaction or 58599 arrangement, you convey, or propagate by procuring conveyance of, a 58600 covered work, and grant a patent license to some of the parties 58601 receiving the covered work authorizing them to use, propagate, 58602 modify or convey a specific copy of the covered work, then the 58603 patent license you grant is automatically extended to all 58604 recipients of the covered work and works based on it. 58605 58606 A patent license is "discriminatory" if it does not include within 58607 the scope of its coverage, prohibits the exercise of, or is 58608 conditioned on the non-exercise of one or more of the rights that 58609 are specifically granted under this License. You may not convey a 58610 covered work if you are a party to an arrangement with a third 58611 party that is in the business of distributing software, under which 58612 you make payment to the third party based on the extent of your 58613 activity of conveying the work, and under which the third party 58614 grants, to any of the parties who would receive the covered work 58615 from you, a discriminatory patent license (a) in connection with 58616 copies of the covered work conveyed by you (or copies made from 58617 those copies), or (b) primarily for and in connection with specific 58618 products or compilations that contain the covered work, unless you 58619 entered into that arrangement, or that patent license was granted, 58620 prior to 28 March 2007. 58621 58622 Nothing in this License shall be construed as excluding or limiting 58623 any implied license or other defenses to infringement that may 58624 otherwise be available to you under applicable patent law. 58625 58626 12. No Surrender of Others' Freedom. 58627 58628 If conditions are imposed on you (whether by court order, agreement 58629 or otherwise) that contradict the conditions of this License, they 58630 do not excuse you from the conditions of this License. If you 58631 cannot convey a covered work so as to satisfy simultaneously your 58632 obligations under this License and any other pertinent obligations, 58633 then as a consequence you may not convey it at all. For example, 58634 if you agree to terms that obligate you to collect a royalty for 58635 further conveying from those to whom you convey the Program, the 58636 only way you could satisfy both those terms and this License would 58637 be to refrain entirely from conveying the Program. 58638 58639 13. Use with the GNU Affero General Public License. 58640 58641 Notwithstanding any other provision of this License, you have 58642 permission to link or combine any covered work with a work licensed 58643 under version 3 of the GNU Affero General Public License into a 58644 single combined work, and to convey the resulting work. The terms 58645 of this License will continue to apply to the part which is the 58646 covered work, but the special requirements of the GNU Affero 58647 General Public License, section 13, concerning interaction through 58648 a network will apply to the combination as such. 58649 58650 14. Revised Versions of this License. 58651 58652 The Free Software Foundation may publish revised and/or new 58653 versions of the GNU General Public License from time to time. Such 58654 new versions will be similar in spirit to the present version, but 58655 may differ in detail to address new problems or concerns. 58656 58657 Each version is given a distinguishing version number. If the 58658 Program specifies that a certain numbered version of the GNU 58659 General Public License "or any later version" applies to it, you 58660 have the option of following the terms and conditions either of 58661 that numbered version or of any later version published by the Free 58662 Software Foundation. If the Program does not specify a version 58663 number of the GNU General Public License, you may choose any 58664 version ever published by the Free Software Foundation. 58665 58666 If the Program specifies that a proxy can decide which future 58667 versions of the GNU General Public License can be used, that 58668 proxy's public statement of acceptance of a version permanently 58669 authorizes you to choose that version for the Program. 58670 58671 Later license versions may give you additional or different 58672 permissions. However, no additional obligations are imposed on any 58673 author or copyright holder as a result of your choosing to follow a 58674 later version. 58675 58676 15. Disclaimer of Warranty. 58677 58678 THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 58679 APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE 58680 COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" 58681 WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, 58682 INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 58683 MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE 58684 RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. 58685 SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL 58686 NECESSARY SERVICING, REPAIR OR CORRECTION. 58687 58688 16. Limitation of Liability. 58689 58690 IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN 58691 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES 58692 AND/OR CONVEYS THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR 58693 DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR 58694 CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE 58695 THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA 58696 BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 58697 PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER 58698 PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF 58699 THE POSSIBILITY OF SUCH DAMAGES. 58700 58701 17. Interpretation of Sections 15 and 16. 58702 58703 If the disclaimer of warranty and limitation of liability provided 58704 above cannot be given local legal effect according to their terms, 58705 reviewing courts shall apply local law that most closely 58706 approximates an absolute waiver of all civil liability in 58707 connection with the Program, unless a warranty or assumption of 58708 liability accompanies a copy of the Program in return for a fee. 58709 58710END OF TERMS AND CONDITIONS 58711=========================== 58712 58713How to Apply These Terms to Your New Programs 58714============================================= 58715 58716If you develop a new program, and you want it to be of the greatest 58717possible use to the public, the best way to achieve this is to make it 58718free software which everyone can redistribute and change under these 58719terms. 58720 58721 To do so, attach the following notices to the program. It is safest to 58722attach them to the start of each source file to most effectively state 58723the exclusion of warranty; and each file should have at least the 58724"copyright" line and a pointer to where the full notice is found. 58725 58726 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES. 58727 Copyright (C) YEAR NAME OF AUTHOR 58728 58729 This program is free software: you can redistribute it and/or modify 58730 it under the terms of the GNU General Public License as published by 58731 the Free Software Foundation, either version 3 of the License, or (at 58732 your option) any later version. 58733 58734 This program is distributed in the hope that it will be useful, but 58735 WITHOUT ANY WARRANTY; without even the implied warranty of 58736 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 58737 General Public License for more details. 58738 58739 You should have received a copy of the GNU General Public License 58740 along with this program. If not, see <http://www.gnu.org/licenses/>. 58741 58742 Also add information on how to contact you by electronic and paper 58743mail. 58744 58745 If the program does terminal interaction, make it output a short notice 58746like this when it starts in an interactive mode: 58747 58748 PROGRAM Copyright (C) YEAR NAME OF AUTHOR 58749 This program comes with ABSOLUTELY NO WARRANTY; for details type 'show w'. 58750 This is free software, and you are welcome to redistribute it 58751 under certain conditions; type 'show c' for details. 58752 58753 The hypothetical commands 'show w' and 'show c' should show the 58754appropriate parts of the General Public License. Of course, your 58755program's commands might be different; for a GUI interface, you would 58756use an "about box". 58757 58758 You should also get your employer (if you work as a programmer) or 58759school, if any, to sign a "copyright disclaimer" for the program, if 58760necessary. For more information on this, and how to apply and follow 58761the GNU GPL, see <http://www.gnu.org/licenses/>. 58762 58763 The GNU General Public License does not permit incorporating your 58764program into proprietary programs. If your program is a subroutine 58765library, you may consider it more useful to permit linking proprietary 58766applications with the library. If this is what you want to do, use the 58767GNU Lesser General Public License instead of this License. But first, 58768please read <https://www.gnu.org/licenses/why-not-lgpl.html>. 58769 58770 58771File: gcc.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top 58772 58773GNU Free Documentation License 58774****************************** 58775 58776 Version 1.3, 3 November 2008 58777 58778 Copyright (C) 2000, 2001, 2002, 2007, 2008 Free Software Foundation, Inc. 58779 <http://fsf.org/> 58780 58781 Everyone is permitted to copy and distribute verbatim copies 58782 of this license document, but changing it is not allowed. 58783 58784 0. PREAMBLE 58785 58786 The purpose of this License is to make a manual, textbook, or other 58787 functional and useful document "free" in the sense of freedom: to 58788 assure everyone the effective freedom to copy and redistribute it, 58789 with or without modifying it, either commercially or 58790 noncommercially. Secondarily, this License preserves for the 58791 author and publisher a way to get credit for their work, while not 58792 being considered responsible for modifications made by others. 58793 58794 This License is a kind of "copyleft", which means that derivative 58795 works of the document must themselves be free in the same sense. 58796 It complements the GNU General Public License, which is a copyleft 58797 license designed for free software. 58798 58799 We have designed this License in order to use it for manuals for 58800 free software, because free software needs free documentation: a 58801 free program should come with manuals providing the same freedoms 58802 that the software does. But this License is not limited to 58803 software manuals; it can be used for any textual work, regardless 58804 of subject matter or whether it is published as a printed book. We 58805 recommend this License principally for works whose purpose is 58806 instruction or reference. 58807 58808 1. APPLICABILITY AND DEFINITIONS 58809 58810 This License applies to any manual or other work, in any medium, 58811 that contains a notice placed by the copyright holder saying it can 58812 be distributed under the terms of this License. Such a notice 58813 grants a world-wide, royalty-free license, unlimited in duration, 58814 to use that work under the conditions stated herein. The 58815 "Document", below, refers to any such manual or work. Any member 58816 of the public is a licensee, and is addressed as "you". You accept 58817 the license if you copy, modify or distribute the work in a way 58818 requiring permission under copyright law. 58819 58820 A "Modified Version" of the Document means any work containing the 58821 Document or a portion of it, either copied verbatim, or with 58822 modifications and/or translated into another language. 58823 58824 A "Secondary Section" is a named appendix or a front-matter section 58825 of the Document that deals exclusively with the relationship of the 58826 publishers or authors of the Document to the Document's overall 58827 subject (or to related matters) and contains nothing that could 58828 fall directly within that overall subject. (Thus, if the Document 58829 is in part a textbook of mathematics, a Secondary Section may not 58830 explain any mathematics.) The relationship could be a matter of 58831 historical connection with the subject or with related matters, or 58832 of legal, commercial, philosophical, ethical or political position 58833 regarding them. 58834 58835 The "Invariant Sections" are certain Secondary Sections whose 58836 titles are designated, as being those of Invariant Sections, in the 58837 notice that says that the Document is released under this License. 58838 If a section does not fit the above definition of Secondary then it 58839 is not allowed to be designated as Invariant. The Document may 58840 contain zero Invariant Sections. If the Document does not identify 58841 any Invariant Sections then there are none. 58842 58843 The "Cover Texts" are certain short passages of text that are 58844 listed, as Front-Cover Texts or Back-Cover Texts, in the notice 58845 that says that the Document is released under this License. A 58846 Front-Cover Text may be at most 5 words, and a Back-Cover Text may 58847 be at most 25 words. 58848 58849 A "Transparent" copy of the Document means a machine-readable copy, 58850 represented in a format whose specification is available to the 58851 general public, that is suitable for revising the document 58852 straightforwardly with generic text editors or (for images composed 58853 of pixels) generic paint programs or (for drawings) some widely 58854 available drawing editor, and that is suitable for input to text 58855 formatters or for automatic translation to a variety of formats 58856 suitable for input to text formatters. A copy made in an otherwise 58857 Transparent file format whose markup, or absence of markup, has 58858 been arranged to thwart or discourage subsequent modification by 58859 readers is not Transparent. An image format is not Transparent if 58860 used for any substantial amount of text. A copy that is not 58861 "Transparent" is called "Opaque". 58862 58863 Examples of suitable formats for Transparent copies include plain 58864 ASCII without markup, Texinfo input format, LaTeX input format, 58865 SGML or XML using a publicly available DTD, and standard-conforming 58866 simple HTML, PostScript or PDF designed for human modification. 58867 Examples of transparent image formats include PNG, XCF and JPG. 58868 Opaque formats include proprietary formats that can be read and 58869 edited only by proprietary word processors, SGML or XML for which 58870 the DTD and/or processing tools are not generally available, and 58871 the machine-generated HTML, PostScript or PDF produced by some word 58872 processors for output purposes only. 58873 58874 The "Title Page" means, for a printed book, the title page itself, 58875 plus such following pages as are needed to hold, legibly, the 58876 material this License requires to appear in the title page. For 58877 works in formats which do not have any title page as such, "Title 58878 Page" means the text near the most prominent appearance of the 58879 work's title, preceding the beginning of the body of the text. 58880 58881 The "publisher" means any person or entity that distributes copies 58882 of the Document to the public. 58883 58884 A section "Entitled XYZ" means a named subunit of the Document 58885 whose title either is precisely XYZ or contains XYZ in parentheses 58886 following text that translates XYZ in another language. (Here XYZ 58887 stands for a specific section name mentioned below, such as 58888 "Acknowledgements", "Dedications", "Endorsements", or "History".) 58889 To "Preserve the Title" of such a section when you modify the 58890 Document means that it remains a section "Entitled XYZ" according 58891 to this definition. 58892 58893 The Document may include Warranty Disclaimers next to the notice 58894 which states that this License applies to the Document. These 58895 Warranty Disclaimers are considered to be included by reference in 58896 this License, but only as regards disclaiming warranties: any other 58897 implication that these Warranty Disclaimers may have is void and 58898 has no effect on the meaning of this License. 58899 58900 2. VERBATIM COPYING 58901 58902 You may copy and distribute the Document in any medium, either 58903 commercially or noncommercially, provided that this License, the 58904 copyright notices, and the license notice saying this License 58905 applies to the Document are reproduced in all copies, and that you 58906 add no other conditions whatsoever to those of this License. You 58907 may not use technical measures to obstruct or control the reading 58908 or further copying of the copies you make or distribute. However, 58909 you may accept compensation in exchange for copies. If you 58910 distribute a large enough number of copies you must also follow the 58911 conditions in section 3. 58912 58913 You may also lend copies, under the same conditions stated above, 58914 and you may publicly display copies. 58915 58916 3. COPYING IN QUANTITY 58917 58918 If you publish printed copies (or copies in media that commonly 58919 have printed covers) of the Document, numbering more than 100, and 58920 the Document's license notice requires Cover Texts, you must 58921 enclose the copies in covers that carry, clearly and legibly, all 58922 these Cover Texts: Front-Cover Texts on the front cover, and 58923 Back-Cover Texts on the back cover. Both covers must also clearly 58924 and legibly identify you as the publisher of these copies. The 58925 front cover must present the full title with all words of the title 58926 equally prominent and visible. You may add other material on the 58927 covers in addition. Copying with changes limited to the covers, as 58928 long as they preserve the title of the Document and satisfy these 58929 conditions, can be treated as verbatim copying in other respects. 58930 58931 If the required texts for either cover are too voluminous to fit 58932 legibly, you should put the first ones listed (as many as fit 58933 reasonably) on the actual cover, and continue the rest onto 58934 adjacent pages. 58935 58936 If you publish or distribute Opaque copies of the Document 58937 numbering more than 100, you must either include a machine-readable 58938 Transparent copy along with each Opaque copy, or state in or with 58939 each Opaque copy a computer-network location from which the general 58940 network-using public has access to download using public-standard 58941 network protocols a complete Transparent copy of the Document, free 58942 of added material. If you use the latter option, you must take 58943 reasonably prudent steps, when you begin distribution of Opaque 58944 copies in quantity, to ensure that this Transparent copy will 58945 remain thus accessible at the stated location until at least one 58946 year after the last time you distribute an Opaque copy (directly or 58947 through your agents or retailers) of that edition to the public. 58948 58949 It is requested, but not required, that you contact the authors of 58950 the Document well before redistributing any large number of copies, 58951 to give them a chance to provide you with an updated version of the 58952 Document. 58953 58954 4. MODIFICATIONS 58955 58956 You may copy and distribute a Modified Version of the Document 58957 under the conditions of sections 2 and 3 above, provided that you 58958 release the Modified Version under precisely this License, with the 58959 Modified Version filling the role of the Document, thus licensing 58960 distribution and modification of the Modified Version to whoever 58961 possesses a copy of it. In addition, you must do these things in 58962 the Modified Version: 58963 58964 A. Use in the Title Page (and on the covers, if any) a title 58965 distinct from that of the Document, and from those of previous 58966 versions (which should, if there were any, be listed in the 58967 History section of the Document). You may use the same title 58968 as a previous version if the original publisher of that 58969 version gives permission. 58970 58971 B. List on the Title Page, as authors, one or more persons or 58972 entities responsible for authorship of the modifications in 58973 the Modified Version, together with at least five of the 58974 principal authors of the Document (all of its principal 58975 authors, if it has fewer than five), unless they release you 58976 from this requirement. 58977 58978 C. State on the Title page the name of the publisher of the 58979 Modified Version, as the publisher. 58980 58981 D. Preserve all the copyright notices of the Document. 58982 58983 E. Add an appropriate copyright notice for your modifications 58984 adjacent to the other copyright notices. 58985 58986 F. Include, immediately after the copyright notices, a license 58987 notice giving the public permission to use the Modified 58988 Version under the terms of this License, in the form shown in 58989 the Addendum below. 58990 58991 G. Preserve in that license notice the full lists of Invariant 58992 Sections and required Cover Texts given in the Document's 58993 license notice. 58994 58995 H. Include an unaltered copy of this License. 58996 58997 I. Preserve the section Entitled "History", Preserve its Title, 58998 and add to it an item stating at least the title, year, new 58999 authors, and publisher of the Modified Version as given on the 59000 Title Page. If there is no section Entitled "History" in the 59001 Document, create one stating the title, year, authors, and 59002 publisher of the Document as given on its Title Page, then add 59003 an item describing the Modified Version as stated in the 59004 previous sentence. 59005 59006 J. Preserve the network location, if any, given in the Document 59007 for public access to a Transparent copy of the Document, and 59008 likewise the network locations given in the Document for 59009 previous versions it was based on. These may be placed in the 59010 "History" section. You may omit a network location for a work 59011 that was published at least four years before the Document 59012 itself, or if the original publisher of the version it refers 59013 to gives permission. 59014 59015 K. For any section Entitled "Acknowledgements" or "Dedications", 59016 Preserve the Title of the section, and preserve in the section 59017 all the substance and tone of each of the contributor 59018 acknowledgements and/or dedications given therein. 59019 59020 L. Preserve all the Invariant Sections of the Document, unaltered 59021 in their text and in their titles. Section numbers or the 59022 equivalent are not considered part of the section titles. 59023 59024 M. Delete any section Entitled "Endorsements". Such a section 59025 may not be included in the Modified Version. 59026 59027 N. Do not retitle any existing section to be Entitled 59028 "Endorsements" or to conflict in title with any Invariant 59029 Section. 59030 59031 O. Preserve any Warranty Disclaimers. 59032 59033 If the Modified Version includes new front-matter sections or 59034 appendices that qualify as Secondary Sections and contain no 59035 material copied from the Document, you may at your option designate 59036 some or all of these sections as invariant. To do this, add their 59037 titles to the list of Invariant Sections in the Modified Version's 59038 license notice. These titles must be distinct from any other 59039 section titles. 59040 59041 You may add a section Entitled "Endorsements", provided it contains 59042 nothing but endorsements of your Modified Version by various 59043 parties--for example, statements of peer review or that the text 59044 has been approved by an organization as the authoritative 59045 definition of a standard. 59046 59047 You may add a passage of up to five words as a Front-Cover Text, 59048 and a passage of up to 25 words as a Back-Cover Text, to the end of 59049 the list of Cover Texts in the Modified Version. Only one passage 59050 of Front-Cover Text and one of Back-Cover Text may be added by (or 59051 through arrangements made by) any one entity. If the Document 59052 already includes a cover text for the same cover, previously added 59053 by you or by arrangement made by the same entity you are acting on 59054 behalf of, you may not add another; but you may replace the old 59055 one, on explicit permission from the previous publisher that added 59056 the old one. 59057 59058 The author(s) and publisher(s) of the Document do not by this 59059 License give permission to use their names for publicity for or to 59060 assert or imply endorsement of any Modified Version. 59061 59062 5. COMBINING DOCUMENTS 59063 59064 You may combine the Document with other documents released under 59065 this License, under the terms defined in section 4 above for 59066 modified versions, provided that you include in the combination all 59067 of the Invariant Sections of all of the original documents, 59068 unmodified, and list them all as Invariant Sections of your 59069 combined work in its license notice, and that you preserve all 59070 their Warranty Disclaimers. 59071 59072 The combined work need only contain one copy of this License, and 59073 multiple identical Invariant Sections may be replaced with a single 59074 copy. If there are multiple Invariant Sections with the same name 59075 but different contents, make the title of each such section unique 59076 by adding at the end of it, in parentheses, the name of the 59077 original author or publisher of that section if known, or else a 59078 unique number. Make the same adjustment to the section titles in 59079 the list of Invariant Sections in the license notice of the 59080 combined work. 59081 59082 In the combination, you must combine any sections Entitled 59083 "History" in the various original documents, forming one section 59084 Entitled "History"; likewise combine any sections Entitled 59085 "Acknowledgements", and any sections Entitled "Dedications". You 59086 must delete all sections Entitled "Endorsements." 59087 59088 6. COLLECTIONS OF DOCUMENTS 59089 59090 You may make a collection consisting of the Document and other 59091 documents released under this License, and replace the individual 59092 copies of this License in the various documents with a single copy 59093 that is included in the collection, provided that you follow the 59094 rules of this License for verbatim copying of each of the documents 59095 in all other respects. 59096 59097 You may extract a single document from such a collection, and 59098 distribute it individually under this License, provided you insert 59099 a copy of this License into the extracted document, and follow this 59100 License in all other respects regarding verbatim copying of that 59101 document. 59102 59103 7. AGGREGATION WITH INDEPENDENT WORKS 59104 59105 A compilation of the Document or its derivatives with other 59106 separate and independent documents or works, in or on a volume of a 59107 storage or distribution medium, is called an "aggregate" if the 59108 copyright resulting from the compilation is not used to limit the 59109 legal rights of the compilation's users beyond what the individual 59110 works permit. When the Document is included in an aggregate, this 59111 License does not apply to the other works in the aggregate which 59112 are not themselves derivative works of the Document. 59113 59114 If the Cover Text requirement of section 3 is applicable to these 59115 copies of the Document, then if the Document is less than one half 59116 of the entire aggregate, the Document's Cover Texts may be placed 59117 on covers that bracket the Document within the aggregate, or the 59118 electronic equivalent of covers if the Document is in electronic 59119 form. Otherwise they must appear on printed covers that bracket 59120 the whole aggregate. 59121 59122 8. TRANSLATION 59123 59124 Translation is considered a kind of modification, so you may 59125 distribute translations of the Document under the terms of section 59126 4. Replacing Invariant Sections with translations requires special 59127 permission from their copyright holders, but you may include 59128 translations of some or all Invariant Sections in addition to the 59129 original versions of these Invariant Sections. You may include a 59130 translation of this License, and all the license notices in the 59131 Document, and any Warranty Disclaimers, provided that you also 59132 include the original English version of this License and the 59133 original versions of those notices and disclaimers. In case of a 59134 disagreement between the translation and the original version of 59135 this License or a notice or disclaimer, the original version will 59136 prevail. 59137 59138 If a section in the Document is Entitled "Acknowledgements", 59139 "Dedications", or "History", the requirement (section 4) to 59140 Preserve its Title (section 1) will typically require changing the 59141 actual title. 59142 59143 9. TERMINATION 59144 59145 You may not copy, modify, sublicense, or distribute the Document 59146 except as expressly provided under this License. Any attempt 59147 otherwise to copy, modify, sublicense, or distribute it is void, 59148 and will automatically terminate your rights under this License. 59149 59150 However, if you cease all violation of this License, then your 59151 license from a particular copyright holder is reinstated (a) 59152 provisionally, unless and until the copyright holder explicitly and 59153 finally terminates your license, and (b) permanently, if the 59154 copyright holder fails to notify you of the violation by some 59155 reasonable means prior to 60 days after the cessation. 59156 59157 Moreover, your license from a particular copyright holder is 59158 reinstated permanently if the copyright holder notifies you of the 59159 violation by some reasonable means, this is the first time you have 59160 received notice of violation of this License (for any work) from 59161 that copyright holder, and you cure the violation prior to 30 days 59162 after your receipt of the notice. 59163 59164 Termination of your rights under this section does not terminate 59165 the licenses of parties who have received copies or rights from you 59166 under this License. If your rights have been terminated and not 59167 permanently reinstated, receipt of a copy of some or all of the 59168 same material does not give you any rights to use it. 59169 59170 10. FUTURE REVISIONS OF THIS LICENSE 59171 59172 The Free Software Foundation may publish new, revised versions of 59173 the GNU Free Documentation License from time to time. Such new 59174 versions will be similar in spirit to the present version, but may 59175 differ in detail to address new problems or concerns. See 59176 <http://www.gnu.org/copyleft/>. 59177 59178 Each version of the License is given a distinguishing version 59179 number. If the Document specifies that a particular numbered 59180 version of this License "or any later version" applies to it, you 59181 have the option of following the terms and conditions either of 59182 that specified version or of any later version that has been 59183 published (not as a draft) by the Free Software Foundation. If the 59184 Document does not specify a version number of this License, you may 59185 choose any version ever published (not as a draft) by the Free 59186 Software Foundation. If the Document specifies that a proxy can 59187 decide which future versions of this License can be used, that 59188 proxy's public statement of acceptance of a version permanently 59189 authorizes you to choose that version for the Document. 59190 59191 11. RELICENSING 59192 59193 "Massive Multiauthor Collaboration Site" (or "MMC Site") means any 59194 World Wide Web server that publishes copyrightable works and also 59195 provides prominent facilities for anybody to edit those works. A 59196 public wiki that anybody can edit is an example of such a server. 59197 A "Massive Multiauthor Collaboration" (or "MMC") contained in the 59198 site means any set of copyrightable works thus published on the MMC 59199 site. 59200 59201 "CC-BY-SA" means the Creative Commons Attribution-Share Alike 3.0 59202 license published by Creative Commons Corporation, a not-for-profit 59203 corporation with a principal place of business in San Francisco, 59204 California, as well as future copyleft versions of that license 59205 published by that same organization. 59206 59207 "Incorporate" means to publish or republish a Document, in whole or 59208 in part, as part of another Document. 59209 59210 An MMC is "eligible for relicensing" if it is licensed under this 59211 License, and if all works that were first published under this 59212 License somewhere other than this MMC, and subsequently 59213 incorporated in whole or in part into the MMC, (1) had no cover 59214 texts or invariant sections, and (2) were thus incorporated prior 59215 to November 1, 2008. 59216 59217 The operator of an MMC Site may republish an MMC contained in the 59218 site under CC-BY-SA on the same site at any time before August 1, 59219 2009, provided the MMC is eligible for relicensing. 59220 59221ADDENDUM: How to use this License for your documents 59222==================================================== 59223 59224To use this License in a document you have written, include a copy of 59225the License in the document and put the following copyright and license 59226notices just after the title page: 59227 59228 Copyright (C) YEAR YOUR NAME. 59229 Permission is granted to copy, distribute and/or modify this document 59230 under the terms of the GNU Free Documentation License, Version 1.3 59231 or any later version published by the Free Software Foundation; 59232 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover 59233 Texts. A copy of the license is included in the section entitled ``GNU 59234 Free Documentation License''. 59235 59236 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, 59237replace the "with...Texts." line with this: 59238 59239 with the Invariant Sections being LIST THEIR TITLES, with 59240 the Front-Cover Texts being LIST, and with the Back-Cover Texts 59241 being LIST. 59242 59243 If you have Invariant Sections without Cover Texts, or some other 59244combination of the three, merge those two alternatives to suit the 59245situation. 59246 59247 If your document contains nontrivial examples of program code, we 59248recommend releasing these examples in parallel under your choice of free 59249software license, such as the GNU General Public License, to permit 59250their use in free software. 59251 59252 59253File: gcc.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top 59254 59255Contributors to GCC 59256******************* 59257 59258The GCC project would like to thank its many contributors. Without them 59259the project would not have been nearly as successful as it has been. 59260Any omissions in this list are accidental. Feel free to contact 59261<law@redhat.com> or <gerald@pfeifer.com> if you have been left out or 59262some of your contributions are not listed. Please keep this list in 59263alphabetical order. 59264 59265 * Analog Devices helped implement the support for complex data types 59266 and iterators. 59267 59268 * John David Anglin for threading-related fixes and improvements to 59269 libstdc++-v3, and the HP-UX port. 59270 59271 * James van Artsdalen wrote the code that makes efficient use of the 59272 Intel 80387 register stack. 59273 59274 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta 59275 Series port. 59276 59277 * Alasdair Baird for various bug fixes. 59278 59279 * Giovanni Bajo for analyzing lots of complicated C++ problem 59280 reports. 59281 59282 * Peter Barada for his work to improve code generation for new 59283 ColdFire cores. 59284 59285 * Gerald Baumgartner added the signature extension to the C++ front 59286 end. 59287 59288 * Godmar Back for his Java improvements and encouragement. 59289 59290 * Scott Bambrough for help porting the Java compiler. 59291 59292 * Wolfgang Bangerth for processing tons of bug reports. 59293 59294 * Jon Beniston for his Microsoft Windows port of Java and port to 59295 Lattice Mico32. 59296 59297 * Daniel Berlin for better DWARF 2 support, faster/better 59298 optimizations, improved alias analysis, plus migrating GCC to 59299 Bugzilla. 59300 59301 * Geoff Berry for his Java object serialization work and various 59302 patches. 59303 59304 * David Binderman tests weekly snapshots of GCC trunk against Fedora 59305 Rawhide for several architectures. 59306 59307 * Laurynas Biveinis for memory management work and DJGPP port fixes. 59308 59309 * Uros Bizjak for the implementation of x87 math built-in functions 59310 and for various middle end and i386 back end improvements and bug 59311 fixes. 59312 59313 * Eric Blake for helping to make GCJ and libgcj conform to the 59314 specifications. 59315 59316 * Janne Blomqvist for contributions to GNU Fortran. 59317 59318 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and 59319 other Java work. 59320 59321 * Segher Boessenkool for helping maintain the PowerPC port and the 59322 instruction combiner plus various contributions to the middle end. 59323 59324 * Neil Booth for work on cpplib, lang hooks, debug hooks and other 59325 miscellaneous clean-ups. 59326 59327 * Steven Bosscher for integrating the GNU Fortran front end into GCC 59328 and for contributing to the tree-ssa branch. 59329 59330 * Eric Botcazou for fixing middle- and backend bugs left and right. 59331 59332 * Per Bothner for his direction via the steering committee and 59333 various improvements to the infrastructure for supporting new 59334 languages. Chill front end implementation. Initial 59335 implementations of cpplib, fix-header, config.guess, libio, and 59336 past C++ library (libg++) maintainer. Dreaming up, designing and 59337 implementing much of GCJ. 59338 59339 * Devon Bowen helped port GCC to the Tahoe. 59340 59341 * Don Bowman for mips-vxworks contributions. 59342 59343 * James Bowman for the FT32 port. 59344 59345 * Dave Brolley for work on cpplib and Chill. 59346 59347 * Paul Brook for work on the ARM architecture and maintaining GNU 59348 Fortran. 59349 59350 * Robert Brown implemented the support for Encore 32000 systems. 59351 59352 * Christian Bruel for improvements to local store elimination. 59353 59354 * Herman A.J. ten Brugge for various fixes. 59355 59356 * Joerg Brunsmann for Java compiler hacking and help with the GCJ 59357 FAQ. 59358 59359 * Joe Buck for his direction via the steering committee from its 59360 creation to 2013. 59361 59362 * Iain Buclaw for the D frontend. 59363 59364 * Craig Burley for leadership of the G77 Fortran effort. 59365 59366 * Tobias Burnus for contributions to GNU Fortran. 59367 59368 * Stephan Buys for contributing Doxygen notes for libstdc++. 59369 59370 * Paolo Carlini for libstdc++ work: lots of efficiency improvements 59371 to the C++ strings, streambufs and formatted I/O, hard detective 59372 work on the frustrating localization issues, and keeping up with 59373 the problem reports. 59374 59375 * John Carr for his alias work, SPARC hacking, infrastructure 59376 improvements, previous contributions to the steering committee, 59377 loop optimizations, etc. 59378 59379 * Stephane Carrez for 68HC11 and 68HC12 ports. 59380 59381 * Steve Chamberlain for support for the Renesas SH and H8 processors 59382 and the PicoJava processor, and for GCJ config fixes. 59383 59384 * Glenn Chambers for help with the GCJ FAQ. 59385 59386 * John-Marc Chandonia for various libgcj patches. 59387 59388 * Denis Chertykov for contributing and maintaining the AVR port, the 59389 first GCC port for an 8-bit architecture. 59390 59391 * Kito Cheng for his work on the RISC-V port, including bringing up 59392 the test suite and maintenance. 59393 59394 * Scott Christley for his Objective-C contributions. 59395 59396 * Eric Christopher for his Java porting help and clean-ups. 59397 59398 * Branko Cibej for more warning contributions. 59399 59400 * The GNU Classpath project for all of their merged runtime code. 59401 59402 * Nick Clifton for arm, mcore, fr30, v850, m32r, msp430 rx work, 59403 '--help', and other random hacking. 59404 59405 * Michael Cook for libstdc++ cleanup patches to reduce warnings. 59406 59407 * R. Kelley Cook for making GCC buildable from a read-only directory 59408 as well as other miscellaneous build process and documentation 59409 clean-ups. 59410 59411 * Ralf Corsepius for SH testing and minor bug fixing. 59412 59413 * Franc,ois-Xavier Coudert for contributions to GNU Fortran. 59414 59415 * Stan Cox for care and feeding of the x86 port and lots of behind 59416 the scenes hacking. 59417 59418 * Alex Crain provided changes for the 3b1. 59419 59420 * Ian Dall for major improvements to the NS32k port. 59421 59422 * Paul Dale for his work to add uClinux platform support to the m68k 59423 backend. 59424 59425 * Palmer Dabbelt for his work maintaining the RISC-V port. 59426 59427 * Dario Dariol contributed the four varieties of sample programs that 59428 print a copy of their source. 59429 59430 * Russell Davidson for fstream and stringstream fixes in libstdc++. 59431 59432 * Bud Davis for work on the G77 and GNU Fortran compilers. 59433 59434 * Mo DeJong for GCJ and libgcj bug fixes. 59435 59436 * Jerry DeLisle for contributions to GNU Fortran. 59437 59438 * DJ Delorie for the DJGPP port, build and libiberty maintenance, 59439 various bug fixes, and the M32C, MeP, MSP430, and RL78 ports. 59440 59441 * Arnaud Desitter for helping to debug GNU Fortran. 59442 59443 * Gabriel Dos Reis for contributions to G++, contributions and 59444 maintenance of GCC diagnostics infrastructure, libstdc++-v3, 59445 including 'valarray<>', 'complex<>', maintaining the numerics 59446 library (including that pesky '<limits>' :-) and keeping up-to-date 59447 anything to do with numbers. 59448 59449 * Ulrich Drepper for his work on glibc, testing of GCC using glibc, 59450 ISO C99 support, CFG dumping support, etc., plus support of the C++ 59451 runtime libraries including for all kinds of C interface issues, 59452 contributing and maintaining 'complex<>', sanity checking and 59453 disbursement, configuration architecture, libio maintenance, and 59454 early math work. 59455 59456 * Franc,ois Dumont for his work on libstdc++-v3, especially 59457 maintaining and improving 'debug-mode' and associative and 59458 unordered containers. 59459 59460 * Zdenek Dvorak for a new loop unroller and various fixes. 59461 59462 * Michael Eager for his work on the Xilinx MicroBlaze port. 59463 59464 * Richard Earnshaw for his ongoing work with the ARM. 59465 59466 * David Edelsohn for his direction via the steering committee, 59467 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa 59468 loop changes, doing the entire AIX port of libstdc++ with his bare 59469 hands, and for ensuring GCC properly keeps working on AIX. 59470 59471 * Kevin Ediger for the floating point formatting of num_put::do_put 59472 in libstdc++. 59473 59474 * Phil Edwards for libstdc++ work including configuration hackery, 59475 documentation maintainer, chief breaker of the web pages, the 59476 occasional iostream bug fix, and work on shared library symbol 59477 versioning. 59478 59479 * Paul Eggert for random hacking all over GCC. 59480 59481 * Mark Elbrecht for various DJGPP improvements, and for libstdc++ 59482 configuration support for locales and fstream-related fixes. 59483 59484 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and 59485 iostreams. 59486 59487 * Christian Ehrhardt for dealing with bug reports. 59488 59489 * Ben Elliston for his work to move the Objective-C runtime into its 59490 own subdirectory and for his work on autoconf. 59491 59492 * Revital Eres for work on the PowerPC 750CL port. 59493 59494 * Marc Espie for OpenBSD support. 59495 59496 * Doug Evans for much of the global optimization framework, arc, 59497 m32r, and SPARC work. 59498 59499 * Christopher Faylor for his work on the Cygwin port and for caring 59500 and feeding the gcc.gnu.org box and saving its users tons of spam. 59501 59502 * Fred Fish for BeOS support and Ada fixes. 59503 59504 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ. 59505 59506 * Peter Gerwinski for various bug fixes and the Pascal front end. 59507 59508 * Kaveh R. Ghazi for his direction via the steering committee, 59509 amazing work to make '-W -Wall -W* -Werror' useful, and testing GCC 59510 on a plethora of platforms. Kaveh extends his gratitude to the 59511 CAIP Center at Rutgers University for providing him with computing 59512 resources to work on Free Software from the late 1980s to 2010. 59513 59514 * John Gilmore for a donation to the FSF earmarked improving GNU 59515 Java. 59516 59517 * Judy Goldberg for c++ contributions. 59518 59519 * Torbjorn Granlund for various fixes and the c-torture testsuite, 59520 multiply- and divide-by-constant optimization, improved long long 59521 support, improved leaf function register allocation, and his 59522 direction via the steering committee. 59523 59524 * Jonny Grant for improvements to 'collect2's' '--help' 59525 documentation. 59526 59527 * Anthony Green for his '-Os' contributions, the moxie port, and Java 59528 front end work. 59529 59530 * Stu Grossman for gdb hacking, allowing GCJ developers to debug Java 59531 code. 59532 59533 * Michael K. Gschwind contributed the port to the PDP-11. 59534 59535 * Richard Biener for his ongoing middle-end contributions and bug 59536 fixes and for release management. 59537 59538 * Ron Guilmette implemented the 'protoize' and 'unprotoize' tools, 59539 the support for DWARF 1 symbolic debugging information, and much of 59540 the support for System V Release 4. He has also worked heavily on 59541 the Intel 386 and 860 support. 59542 59543 * Sumanth Gundapaneni for contributing the CR16 port. 59544 59545 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload 59546 GCSE. 59547 59548 * Bruno Haible for improvements in the runtime overhead for EH, new 59549 warnings and assorted bug fixes. 59550 59551 * Andrew Haley for his amazing Java compiler and library efforts. 59552 59553 * Chris Hanson assisted in making GCC work on HP-UX for the 9000 59554 series 300. 59555 59556 * Michael Hayes for various thankless work he's done trying to get 59557 the c30/c40 ports functional. Lots of loop and unroll improvements 59558 and fixes. 59559 59560 * Dara Hazeghi for wading through myriads of target-specific bug 59561 reports. 59562 59563 * Kate Hedstrom for staking the G77 folks with an initial testsuite. 59564 59565 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64 59566 work, loop opts, and generally fixing lots of old problems we've 59567 ignored for years, flow rewrite and lots of further stuff, 59568 including reviewing tons of patches. 59569 59570 * Aldy Hernandez for working on the PowerPC port, SIMD support, and 59571 various fixes. 59572 59573 * Nobuyuki Hikichi of Software Research Associates, Tokyo, 59574 contributed the support for the Sony NEWS machine. 59575 59576 * Kazu Hirata for caring and feeding the Renesas H8/300 port and 59577 various fixes. 59578 59579 * Katherine Holcomb for work on GNU Fortran. 59580 59581 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots 59582 of testing and bug fixing, particularly of GCC configury code. 59583 59584 * Steve Holmgren for MachTen patches. 59585 59586 * Mat Hostetter for work on the TILE-Gx and TILEPro ports. 59587 59588 * Jan Hubicka for his x86 port improvements. 59589 59590 * Falk Hueffner for working on C and optimization bug reports. 59591 59592 * Bernardo Innocenti for his m68k work, including merging of ColdFire 59593 improvements and uClinux support. 59594 59595 * Christian Iseli for various bug fixes. 59596 59597 * Kamil Iskra for general m68k hacking. 59598 59599 * Lee Iverson for random fixes and MIPS testing. 59600 59601 * Balaji V. Iyer for Cilk+ development and merging. 59602 59603 * Andreas Jaeger for testing and benchmarking of GCC and various bug 59604 fixes. 59605 59606 * Martin Jambor for his work on inter-procedural optimizations, the 59607 switch conversion pass, and scalar replacement of aggregates. 59608 59609 * Jakub Jelinek for his SPARC work and sibling call optimizations as 59610 well as lots of bug fixes and test cases, and for improving the 59611 Java build system. 59612 59613 * Janis Johnson for ia64 testing and fixes, her quality improvement 59614 sidetracks, and web page maintenance. 59615 59616 * Kean Johnston for SCO OpenServer support and various fixes. 59617 59618 * Tim Josling for the sample language treelang based originally on 59619 Richard Kenner's "toy" language. 59620 59621 * Nicolai Josuttis for additional libstdc++ documentation. 59622 59623 * Klaus Kaempf for his ongoing work to make alpha-vms a viable 59624 target. 59625 59626 * Steven G. Kargl for work on GNU Fortran. 59627 59628 * David Kashtan of SRI adapted GCC to VMS. 59629 59630 * Ryszard Kabatek for many, many libstdc++ bug fixes and 59631 optimizations of strings, especially member functions, and for 59632 auto_ptr fixes. 59633 59634 * Geoffrey Keating for his ongoing work to make the PPC work for 59635 GNU/Linux and his automatic regression tester. 59636 59637 * Brendan Kehoe for his ongoing work with G++ and for a lot of early 59638 work in just about every part of libstdc++. 59639 59640 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the 59641 MIL-STD-1750A. 59642 59643 * Richard Kenner of the New York University Ultracomputer Research 59644 Laboratory wrote the machine descriptions for the AMD 29000, the 59645 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the 59646 support for instruction attributes. He also made changes to better 59647 support RISC processors including changes to common subexpression 59648 elimination, strength reduction, function calling sequence 59649 handling, and condition code support, in addition to generalizing 59650 the code for frame pointer elimination and delay slot scheduling. 59651 Richard Kenner was also the head maintainer of GCC for several 59652 years. 59653 59654 * Mumit Khan for various contributions to the Cygwin and Mingw32 59655 ports and maintaining binary releases for Microsoft Windows hosts, 59656 and for massive libstdc++ porting work to Cygwin/Mingw32. 59657 59658 * Robin Kirkham for cpu32 support. 59659 59660 * Mark Klein for PA improvements. 59661 59662 * Thomas Koenig for various bug fixes. 59663 59664 * Bruce Korb for the new and improved fixincludes code. 59665 59666 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3 59667 effort. 59668 59669 * Maxim Kuvyrkov for contributions to the instruction scheduler, the 59670 Android and m68k/Coldfire ports, and optimizations. 59671 59672 * Charles LaBrec contributed the support for the Integrated Solutions 59673 68020 system. 59674 59675 * Asher Langton and Mike Kumbera for contributing Cray pointer 59676 support to GNU Fortran, and for other GNU Fortran improvements. 59677 59678 * Jeff Law for his direction via the steering committee, coordinating 59679 the entire egcs project and GCC 2.95, rolling out snapshots and 59680 releases, handling merges from GCC2, reviewing tons of patches that 59681 might have fallen through the cracks else, and random but extensive 59682 hacking. 59683 59684 * Walter Lee for work on the TILE-Gx and TILEPro ports. 59685 59686 * Marc Lehmann for his direction via the steering committee and 59687 helping with analysis and improvements of x86 performance. 59688 59689 * Victor Leikehman for work on GNU Fortran. 59690 59691 * Ted Lemon wrote parts of the RTL reader and printer. 59692 59693 * Kriang Lerdsuwanakij for C++ improvements including template as 59694 template parameter support, and many C++ fixes. 59695 59696 * Warren Levy for tremendous work on libgcj (Java Runtime Library) 59697 and random work on the Java front end. 59698 59699 * Alain Lichnewsky ported GCC to the MIPS CPU. 59700 59701 * Oskar Liljeblad for hacking on AWT and his many Java bug reports 59702 and patches. 59703 59704 * Robert Lipe for OpenServer support, new testsuites, testing, etc. 59705 59706 * Chen Liqin for various S+core related fixes/improvement, and for 59707 maintaining the S+core port. 59708 59709 * Martin Liska for his work on identical code folding, the 59710 sanitizers, HSA, general bug fixing and for running automated 59711 regression testing of GCC and reporting numerous bugs. 59712 59713 * Weiwen Liu for testing and various bug fixes. 59714 59715 * Manuel Lo'pez-Iba'n~ez for improving '-Wconversion' and many other 59716 diagnostics fixes and improvements. 59717 59718 * Dave Love for his ongoing work with the Fortran front end and 59719 runtime libraries. 59720 59721 * Martin von Lo"wis for internal consistency checking infrastructure, 59722 various C++ improvements including namespace support, and tons of 59723 assistance with libstdc++/compiler merges. 59724 59725 * H.J. Lu for his previous contributions to the steering committee, 59726 many x86 bug reports, prototype patches, and keeping the GNU/Linux 59727 ports working. 59728 59729 * Greg McGary for random fixes and (someday) bounded pointers. 59730 59731 * Andrew MacLeod for his ongoing work in building a real EH system, 59732 various code generation improvements, work on the global optimizer, 59733 etc. 59734 59735 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC 59736 hacking improvements to compile-time performance, overall knowledge 59737 and direction in the area of instruction scheduling, design and 59738 implementation of the automaton based instruction scheduler and 59739 design and implementation of the integrated and local register 59740 allocators. 59741 59742 * David Malcolm for his work on improving GCC diagnostics, JIT, 59743 self-tests and unit testing. 59744 59745 * Bob Manson for his behind the scenes work on dejagnu. 59746 59747 * John Marino for contributing the DragonFly BSD port. 59748 59749 * Philip Martin for lots of libstdc++ string and vector iterator 59750 fixes and improvements, and string clean up and testsuites. 59751 59752 * Michael Matz for his work on dominance tree discovery, the x86-64 59753 port, link-time optimization framework and general optimization 59754 improvements. 59755 59756 * All of the Mauve project contributors for Java test code. 59757 59758 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements. 59759 59760 * Adam Megacz for his work on the Microsoft Windows port of GCJ. 59761 59762 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS, 59763 powerpc, haifa, ECOFF debug support, and other assorted hacking. 59764 59765 * Jason Merrill for his direction via the steering committee and 59766 leading the G++ effort. 59767 59768 * Martin Michlmayr for testing GCC on several architectures using the 59769 entire Debian archive. 59770 59771 * David Miller for his direction via the steering committee, lots of 59772 SPARC work, improvements in jump.c and interfacing with the Linux 59773 kernel developers. 59774 59775 * Gary Miller ported GCC to Charles River Data Systems machines. 59776 59777 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning 59778 the entire libstdc++ testsuite namespace-compatible. 59779 59780 * Mark Mitchell for his direction via the steering committee, 59781 mountains of C++ work, load/store hoisting out of loops, alias 59782 analysis improvements, ISO C 'restrict' support, and serving as 59783 release manager from 2000 to 2011. 59784 59785 * Alan Modra for various GNU/Linux bits and testing. 59786 59787 * Toon Moene for his direction via the steering committee, Fortran 59788 maintenance, and his ongoing work to make us make Fortran run fast. 59789 59790 * Jason Molenda for major help in the care and feeding of all the 59791 services on the gcc.gnu.org (formerly egcs.cygnus.com) 59792 machine--mail, web services, ftp services, etc etc. Doing all this 59793 work on scrap paper and the backs of envelopes would have been... 59794 difficult. 59795 59796 * Catherine Moore for fixing various ugly problems we have sent her 59797 way, including the haifa bug which was killing the Alpha & PowerPC 59798 Linux kernels. 59799 59800 * Mike Moreton for his various Java patches. 59801 59802 * David Mosberger-Tang for various Alpha improvements, and for the 59803 initial IA-64 port. 59804 59805 * Stephen Moshier contributed the floating point emulator that 59806 assists in cross-compilation and permits support for floating point 59807 numbers wider than 64 bits and for ISO C99 support. 59808 59809 * Bill Moyer for his behind the scenes work on various issues. 59810 59811 * Philippe De Muyter for his work on the m68k port. 59812 59813 * Joseph S. Myers for his work on the PDP-11 port, format checking 59814 and ISO C99 support, and continuous emphasis on (and contributions 59815 to) documentation. 59816 59817 * Nathan Myers for his work on libstdc++-v3: architecture and 59818 authorship through the first three snapshots, including 59819 implementation of locale infrastructure, string, shadow C headers, 59820 and the initial project documentation (DESIGN, CHECKLIST, and so 59821 forth). Later, more work on MT-safe string and shadow headers. 59822 59823 * Felix Natter for documentation on porting libstdc++. 59824 59825 * Nathanael Nerode for cleaning up the configuration/build process. 59826 59827 * NeXT, Inc. donated the front end that supports the Objective-C 59828 language. 59829 59830 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to the 59831 search engine setup, various documentation fixes and other small 59832 fixes. 59833 59834 * Geoff Noer for his work on getting cygwin native builds working. 59835 59836 * Vegard Nossum for running automated regression testing of GCC and 59837 reporting numerous bugs. 59838 59839 * Diego Novillo for his work on Tree SSA, OpenMP, SPEC performance 59840 tracking web pages, GIMPLE tuples, and assorted fixes. 59841 59842 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64, 59843 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and related 59844 infrastructure improvements. 59845 59846 * Alexandre Oliva for various build infrastructure improvements, 59847 scripts and amazing testing work, including keeping libtool issues 59848 sane and happy. 59849 59850 * Stefan Olsson for work on mt_alloc. 59851 59852 * Melissa O'Neill for various NeXT fixes. 59853 59854 * Rainer Orth for random MIPS work, including improvements to GCC's 59855 o32 ABI support, improvements to dejagnu's MIPS support, Java 59856 configuration clean-ups and porting work, and maintaining the IRIX, 59857 Solaris 2, and Tru64 UNIX ports. 59858 59859 * Steven Pemberton for his contribution of 'enquire' which allowed 59860 GCC to determine various properties of the floating point unit and 59861 generate 'float.h' in older versions of GCC. 59862 59863 * Hartmut Penner for work on the s390 port. 59864 59865 * Paul Petersen wrote the machine description for the Alliant FX/8. 59866 59867 * Alexandre Petit-Bianco for implementing much of the Java compiler 59868 and continued Java maintainership. 59869 59870 * Matthias Pfaller for major improvements to the NS32k port. 59871 59872 * Gerald Pfeifer for his direction via the steering committee, 59873 pointing out lots of problems we need to solve, maintenance of the 59874 web pages, and taking care of documentation maintenance in general. 59875 59876 * Marek Polacek for his work on the C front end, the sanitizers and 59877 general bug fixing. 59878 59879 * Andrew Pinski for processing bug reports by the dozen. 59880 59881 * Ovidiu Predescu for his work on the Objective-C front end and 59882 runtime libraries. 59883 59884 * Jerry Quinn for major performance improvements in C++ formatted 59885 I/O. 59886 59887 * Ken Raeburn for various improvements to checker, MIPS ports and 59888 various cleanups in the compiler. 59889 59890 * Rolf W. Rasmussen for hacking on AWT. 59891 59892 * David Reese of Sun Microsystems contributed to the Solaris on 59893 PowerPC port. 59894 59895 * John Regehr for running automated regression testing of GCC and 59896 reporting numerous bugs. 59897 59898 * Volker Reichelt for running automated regression testing of GCC and 59899 reporting numerous bugs and for keeping up with the problem 59900 reports. 59901 59902 * Joern Rennecke for maintaining the sh port, loop, regmove & reload 59903 hacking and developing and maintaining the Epiphany port. 59904 59905 * Loren J. Rittle for improvements to libstdc++-v3 including the 59906 FreeBSD port, threading fixes, thread-related configury changes, 59907 critical threading documentation, and solutions to really tricky 59908 I/O problems, as well as keeping GCC properly working on FreeBSD 59909 and continuous testing. 59910 59911 * Craig Rodrigues for processing tons of bug reports. 59912 59913 * Ola Ro"nnerup for work on mt_alloc. 59914 59915 * Gavin Romig-Koch for lots of behind the scenes MIPS work. 59916 59917 * David Ronis inspired and encouraged Craig to rewrite the G77 59918 documentation in texinfo format by contributing a first pass at a 59919 translation of the old 'g77-0.5.16/f/DOC' file. 59920 59921 * Ken Rose for fixes to GCC's delay slot filling code. 59922 59923 * Ira Rosen for her contributions to the auto-vectorizer. 59924 59925 * Paul Rubin wrote most of the preprocessor. 59926 59927 * Pe'tur Runo'lfsson for major performance improvements in C++ 59928 formatted I/O and large file support in C++ filebuf. 59929 59930 * Chip Salzenberg for libstdc++ patches and improvements to locales, 59931 traits, Makefiles, libio, libtool hackery, and "long long" support. 59932 59933 * Juha Sarlin for improvements to the H8 code generator. 59934 59935 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series 59936 300. 59937 59938 * Roger Sayle for improvements to constant folding and GCC's RTL 59939 optimizers as well as for fixing numerous bugs. 59940 59941 * Bradley Schatz for his work on the GCJ FAQ. 59942 59943 * Peter Schauer wrote the code to allow debugging to work on the 59944 Alpha. 59945 59946 * William Schelter did most of the work on the Intel 80386 support. 59947 59948 * Tobias Schlu"ter for work on GNU Fortran. 59949 59950 * Bernd Schmidt for various code generation improvements and major 59951 work in the reload pass, serving as release manager for GCC 2.95.3, 59952 and work on the Blackfin and C6X ports. 59953 59954 * Peter Schmid for constant testing of libstdc++--especially 59955 application testing, going above and beyond what was requested for 59956 the release criteria--and libstdc++ header file tweaks. 59957 59958 * Jason Schroeder for jcf-dump patches. 59959 59960 * Andreas Schwab for his work on the m68k port. 59961 59962 * Lars Segerlund for work on GNU Fortran. 59963 59964 * Dodji Seketeli for numerous C++ bug fixes and debug info 59965 improvements. 59966 59967 * Tim Shen for major work on '<regex>'. 59968 59969 * Joel Sherrill for his direction via the steering committee, RTEMS 59970 contributions and RTEMS testing. 59971 59972 * Nathan Sidwell for many C++ fixes/improvements. 59973 59974 * Jeffrey Siegal for helping RMS with the original design of GCC, 59975 some code which handles the parse tree and RTL data structures, 59976 constant folding and help with the original VAX & m68k ports. 59977 59978 * Kenny Simpson for prompting libstdc++ fixes due to defect reports 59979 from the LWG (thereby keeping GCC in line with updates from the 59980 ISO). 59981 59982 * Franz Sirl for his ongoing work with making the PPC port stable for 59983 GNU/Linux. 59984 59985 * Andrey Slepuhin for assorted AIX hacking. 59986 59987 * Trevor Smigiel for contributing the SPU port. 59988 59989 * Christopher Smith did the port for Convex machines. 59990 59991 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports. 59992 Retired from GCC maintainership August 2010, having mentored two 59993 new maintainers into the role. 59994 59995 * Randy Smith finished the Sun FPA support. 59996 59997 * Ed Smith-Rowland for his continuous work on libstdc++-v3, special 59998 functions, '<random>', and various improvements to C++11 features. 59999 60000 * Scott Snyder for queue, iterator, istream, and string fixes and 60001 libstdc++ testsuite entries. Also for providing the patch to G77 60002 to add rudimentary support for 'INTEGER*1', 'INTEGER*2', and 60003 'LOGICAL*1'. 60004 60005 * Zdenek Sojka for running automated regression testing of GCC and 60006 reporting numerous bugs. 60007 60008 * Arseny Solokha for running automated regression testing of GCC and 60009 reporting numerous bugs. 60010 60011 * Jayant Sonar for contributing the CR16 port. 60012 60013 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique. 60014 60015 * Richard Stallman, for writing the original GCC and launching the 60016 GNU project. 60017 60018 * Jan Stein of the Chalmers Computer Society provided support for 60019 Genix, as well as part of the 32000 machine description. 60020 60021 * Gerhard Steinmetz for running automated regression testing of GCC 60022 and reporting numerous bugs. 60023 60024 * Nigel Stephens for various mips16 related fixes/improvements. 60025 60026 * Jonathan Stone wrote the machine description for the Pyramid 60027 computer. 60028 60029 * Graham Stott for various infrastructure improvements. 60030 60031 * John Stracke for his Java HTTP protocol fixes. 60032 60033 * Mike Stump for his Elxsi port, G++ contributions over the years and 60034 more recently his vxworks contributions 60035 60036 * Jeff Sturm for Java porting help, bug fixes, and encouragement. 60037 60038 * Zhendong Su for running automated regression testing of GCC and 60039 reporting numerous bugs. 60040 60041 * Chengnian Sun for running automated regression testing of GCC and 60042 reporting numerous bugs. 60043 60044 * Shigeya Suzuki for this fixes for the bsdi platforms. 60045 60046 * Ian Lance Taylor for the Go frontend, the initial mips16 and mips64 60047 support, general configury hacking, fixincludes, etc. 60048 60049 * Holger Teutsch provided the support for the Clipper CPU. 60050 60051 * Gary Thomas for his ongoing work to make the PPC work for 60052 GNU/Linux. 60053 60054 * Paul Thomas for contributions to GNU Fortran. 60055 60056 * Philipp Thomas for random bug fixes throughout the compiler 60057 60058 * Jason Thorpe for thread support in libstdc++ on NetBSD. 60059 60060 * Kresten Krab Thorup wrote the run time support for the Objective-C 60061 language and the fantastic Java bytecode interpreter. 60062 60063 * Michael Tiemann for random bug fixes, the first instruction 60064 scheduler, initial C++ support, function integration, NS32k, SPARC 60065 and M88k machine description work, delay slot scheduling. 60066 60067 * Andreas Tobler for his work porting libgcj to Darwin. 60068 60069 * Teemu Torma for thread safe exception handling support. 60070 60071 * Leonard Tower wrote parts of the parser, RTL generator, and RTL 60072 definitions, and of the VAX machine description. 60073 60074 * Daniel Towner and Hariharan Sandanagobalane contributed and 60075 maintain the picoChip port. 60076 60077 * Tom Tromey for internationalization support and for his many Java 60078 contributions and libgcj maintainership. 60079 60080 * Lassi Tuura for improvements to config.guess to determine HP 60081 processor types. 60082 60083 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes. 60084 60085 * Andy Vaught for the design and initial implementation of the GNU 60086 Fortran front end. 60087 60088 * Brent Verner for work with the libstdc++ cshadow files and their 60089 associated configure steps. 60090 60091 * Todd Vierling for contributions for NetBSD ports. 60092 60093 * Andrew Waterman for contributing the RISC-V port, as well as 60094 maintaining it. 60095 60096 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML 60097 guidance and maintaining libstdc++. 60098 60099 * Dean Wakerley for converting the install documentation from HTML to 60100 texinfo in time for GCC 3.0. 60101 60102 * Krister Walfridsson for random bug fixes. 60103 60104 * Feng Wang for contributions to GNU Fortran. 60105 60106 * Stephen M. Webb for time and effort on making libstdc++ shadow 60107 files work with the tricky Solaris 8+ headers, and for pushing the 60108 build-time header tree. Also, for starting and driving the 60109 '<regex>' effort. 60110 60111 * John Wehle for various improvements for the x86 code generator, 60112 related infrastructure improvements to help x86 code generation, 60113 value range propagation and other work, WE32k port. 60114 60115 * Ulrich Weigand for work on the s390 port. 60116 60117 * Janus Weil for contributions to GNU Fortran. 60118 60119 * Zack Weinberg for major work on cpplib and various other bug fixes. 60120 60121 * Matt Welsh for help with Linux Threads support in GCJ. 60122 60123 * Urban Widmark for help fixing java.io. 60124 60125 * Mark Wielaard for new Java library code and his work integrating 60126 with Classpath. 60127 60128 * Dale Wiles helped port GCC to the Tahoe. 60129 60130 * Bob Wilson from Tensilica, Inc. for the Xtensa port. 60131 60132 * Jim Wilson for his direction via the steering committee, tackling 60133 hard problems in various places that nobody else wanted to work on, 60134 strength reduction and other loop optimizations. 60135 60136 * Paul Woegerer and Tal Agmon for the CRX port. 60137 60138 * Carlo Wood for various fixes. 60139 60140 * Tom Wood for work on the m88k port. 60141 60142 * Chung-Ju Wu for his work on the Andes NDS32 port. 60143 60144 * Canqun Yang for work on GNU Fortran. 60145 60146 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine 60147 description for the Tron architecture (specifically, the Gmicro). 60148 60149 * Kevin Zachmann helped port GCC to the Tahoe. 60150 60151 * Ayal Zaks for Swing Modulo Scheduling (SMS). 60152 60153 * Qirun Zhang for running automated regression testing of GCC and 60154 reporting numerous bugs. 60155 60156 * Xiaoqiang Zhang for work on GNU Fortran. 60157 60158 * Gilles Zunino for help porting Java to Irix. 60159 60160 The following people are recognized for their contributions to GNAT, 60161the Ada front end of GCC: 60162 * Bernard Banner 60163 60164 * Romain Berrendonner 60165 60166 * Geert Bosch 60167 60168 * Emmanuel Briot 60169 60170 * Joel Brobecker 60171 60172 * Ben Brosgol 60173 60174 * Vincent Celier 60175 60176 * Arnaud Charlet 60177 60178 * Chien Chieng 60179 60180 * Cyrille Comar 60181 60182 * Cyrille Crozes 60183 60184 * Robert Dewar 60185 60186 * Gary Dismukes 60187 60188 * Robert Duff 60189 60190 * Ed Falis 60191 60192 * Ramon Fernandez 60193 60194 * Sam Figueroa 60195 60196 * Vasiliy Fofanov 60197 60198 * Michael Friess 60199 60200 * Franco Gasperoni 60201 60202 * Ted Giering 60203 60204 * Matthew Gingell 60205 60206 * Laurent Guerby 60207 60208 * Jerome Guitton 60209 60210 * Olivier Hainque 60211 60212 * Jerome Hugues 60213 60214 * Hristian Kirtchev 60215 60216 * Jerome Lambourg 60217 60218 * Bruno Leclerc 60219 60220 * Albert Lee 60221 60222 * Sean McNeil 60223 60224 * Javier Miranda 60225 60226 * Laurent Nana 60227 60228 * Pascal Obry 60229 60230 * Dong-Ik Oh 60231 60232 * Laurent Pautet 60233 60234 * Brett Porter 60235 60236 * Thomas Quinot 60237 60238 * Nicolas Roche 60239 60240 * Pat Rogers 60241 60242 * Jose Ruiz 60243 60244 * Douglas Rupp 60245 60246 * Sergey Rybin 60247 60248 * Gail Schenker 60249 60250 * Ed Schonberg 60251 60252 * Nicolas Setton 60253 60254 * Samuel Tardieu 60255 60256 The following people are recognized for their contributions of new 60257features, bug reports, testing and integration of classpath/libgcj for 60258GCC version 4.1: 60259 * Lillian Angel for 'JTree' implementation and lots Free Swing 60260 additions and bug fixes. 60261 60262 * Wolfgang Baer for 'GapContent' bug fixes. 60263 60264 * Anthony Balkissoon for 'JList', Free Swing 1.5 updates and mouse 60265 event fixes, lots of Free Swing work including 'JTable' editing. 60266 60267 * Stuart Ballard for RMI constant fixes. 60268 60269 * Goffredo Baroncelli for 'HTTPURLConnection' fixes. 60270 60271 * Gary Benson for 'MessageFormat' fixes. 60272 60273 * Daniel Bonniot for 'Serialization' fixes. 60274 60275 * Chris Burdess for lots of gnu.xml and http protocol fixes, 'StAX' 60276 and 'DOM xml:id' support. 60277 60278 * Ka-Hing Cheung for 'TreePath' and 'TreeSelection' fixes. 60279 60280 * Archie Cobbs for build fixes, VM interface updates, 60281 'URLClassLoader' updates. 60282 60283 * Kelley Cook for build fixes. 60284 60285 * Martin Cordova for Suggestions for better 'SocketTimeoutException'. 60286 60287 * David Daney for 'BitSet' bug fixes, 'HttpURLConnection' rewrite and 60288 improvements. 60289 60290 * Thomas Fitzsimmons for lots of upgrades to the gtk+ AWT and Cairo 60291 2D support. Lots of imageio framework additions, lots of AWT and 60292 Free Swing bug fixes. 60293 60294 * Jeroen Frijters for 'ClassLoader' and nio cleanups, serialization 60295 fixes, better 'Proxy' support, bug fixes and IKVM integration. 60296 60297 * Santiago Gala for 'AccessControlContext' fixes. 60298 60299 * Nicolas Geoffray for 'VMClassLoader' and 'AccessController' 60300 improvements. 60301 60302 * David Gilbert for 'basic' and 'metal' icon and plaf support and 60303 lots of documenting, Lots of Free Swing and metal theme additions. 60304 'MetalIconFactory' implementation. 60305 60306 * Anthony Green for 'MIDI' framework, 'ALSA' and 'DSSI' providers. 60307 60308 * Andrew Haley for 'Serialization' and 'URLClassLoader' fixes, gcj 60309 build speedups. 60310 60311 * Kim Ho for 'JFileChooser' implementation. 60312 60313 * Andrew John Hughes for 'Locale' and net fixes, URI RFC2986 updates, 60314 'Serialization' fixes, 'Properties' XML support and generic branch 60315 work, VMIntegration guide update. 60316 60317 * Bastiaan Huisman for 'TimeZone' bug fixing. 60318 60319 * Andreas Jaeger for mprec updates. 60320 60321 * Paul Jenner for better '-Werror' support. 60322 60323 * Ito Kazumitsu for 'NetworkInterface' implementation and updates. 60324 60325 * Roman Kennke for 'BoxLayout', 'GrayFilter' and 'SplitPane', plus 60326 bug fixes all over. Lots of Free Swing work including styled text. 60327 60328 * Simon Kitching for 'String' cleanups and optimization suggestions. 60329 60330 * Michael Koch for configuration fixes, 'Locale' updates, bug and 60331 build fixes. 60332 60333 * Guilhem Lavaux for configuration, thread and channel fixes and 60334 Kaffe integration. JCL native 'Pointer' updates. Logger bug 60335 fixes. 60336 60337 * David Lichteblau for JCL support library global/local reference 60338 cleanups. 60339 60340 * Aaron Luchko for JDWP updates and documentation fixes. 60341 60342 * Ziga Mahkovec for 'Graphics2D' upgraded to Cairo 0.5 and new regex 60343 features. 60344 60345 * Sven de Marothy for BMP imageio support, CSS and 'TextLayout' 60346 fixes. 'GtkImage' rewrite, 2D, awt, free swing and date/time fixes 60347 and implementing the Qt4 peers. 60348 60349 * Casey Marshall for crypto algorithm fixes, 'FileChannel' lock, 60350 'SystemLogger' and 'FileHandler' rotate implementations, NIO 60351 'FileChannel.map' support, security and policy updates. 60352 60353 * Bryce McKinlay for RMI work. 60354 60355 * Audrius Meskauskas for lots of Free Corba, RMI and HTML work plus 60356 testing and documenting. 60357 60358 * Kalle Olavi Niemitalo for build fixes. 60359 60360 * Rainer Orth for build fixes. 60361 60362 * Andrew Overholt for 'File' locking fixes. 60363 60364 * Ingo Proetel for 'Image', 'Logger' and 'URLClassLoader' updates. 60365 60366 * Olga Rodimina for 'MenuSelectionManager' implementation. 60367 60368 * Jan Roehrich for 'BasicTreeUI' and 'JTree' fixes. 60369 60370 * Julian Scheid for documentation updates and gjdoc support. 60371 60372 * Christian Schlichtherle for zip fixes and cleanups. 60373 60374 * Robert Schuster for documentation updates and beans fixes, 60375 'TreeNode' enumerations and 'ActionCommand' and various fixes, XML 60376 and URL, AWT and Free Swing bug fixes. 60377 60378 * Keith Seitz for lots of JDWP work. 60379 60380 * Christian Thalinger for 64-bit cleanups, Configuration and VM 60381 interface fixes and 'CACAO' integration, 'fdlibm' updates. 60382 60383 * Gael Thomas for 'VMClassLoader' boot packages support suggestions. 60384 60385 * Andreas Tobler for Darwin and Solaris testing and fixing, 'Qt4' 60386 support for Darwin/OS X, 'Graphics2D' support, 'gtk+' updates. 60387 60388 * Dalibor Topic for better 'DEBUG' support, build cleanups and Kaffe 60389 integration. 'Qt4' build infrastructure, 'SHA1PRNG' and 60390 'GdkPixbugDecoder' updates. 60391 60392 * Tom Tromey for Eclipse integration, generics work, lots of bug 60393 fixes and gcj integration including coordinating The Big Merge. 60394 60395 * Mark Wielaard for bug fixes, packaging and release management, 60396 'Clipboard' implementation, system call interrupts and network 60397 timeouts and 'GdkPixpufDecoder' fixes. 60398 60399 In addition to the above, all of which also contributed time and energy 60400in testing GCC, we would like to thank the following for their 60401contributions to testing: 60402 60403 * Michael Abd-El-Malek 60404 60405 * Thomas Arend 60406 60407 * Bonzo Armstrong 60408 60409 * Steven Ashe 60410 60411 * Chris Baldwin 60412 60413 * David Billinghurst 60414 60415 * Jim Blandy 60416 60417 * Stephane Bortzmeyer 60418 60419 * Horst von Brand 60420 60421 * Frank Braun 60422 60423 * Rodney Brown 60424 60425 * Sidney Cadot 60426 60427 * Bradford Castalia 60428 60429 * Robert Clark 60430 60431 * Jonathan Corbet 60432 60433 * Ralph Doncaster 60434 60435 * Richard Emberson 60436 60437 * Levente Farkas 60438 60439 * Graham Fawcett 60440 60441 * Mark Fernyhough 60442 60443 * Robert A. French 60444 60445 * Jo"rgen Freyh 60446 60447 * Mark K. Gardner 60448 60449 * Charles-Antoine Gauthier 60450 60451 * Yung Shing Gene 60452 60453 * David Gilbert 60454 60455 * Simon Gornall 60456 60457 * Fred Gray 60458 60459 * John Griffin 60460 60461 * Patrik Hagglund 60462 60463 * Phil Hargett 60464 60465 * Amancio Hasty 60466 60467 * Takafumi Hayashi 60468 60469 * Bryan W. Headley 60470 60471 * Kevin B. Hendricks 60472 60473 * Joep Jansen 60474 60475 * Christian Joensson 60476 60477 * Michel Kern 60478 60479 * David Kidd 60480 60481 * Tobias Kuipers 60482 60483 * Anand Krishnaswamy 60484 60485 * A. O. V. Le Blanc 60486 60487 * llewelly 60488 60489 * Damon Love 60490 60491 * Brad Lucier 60492 60493 * Matthias Klose 60494 60495 * Martin Knoblauch 60496 60497 * Rick Lutowski 60498 60499 * Jesse Macnish 60500 60501 * Stefan Morrell 60502 60503 * Anon A. Mous 60504 60505 * Matthias Mueller 60506 60507 * Pekka Nikander 60508 60509 * Rick Niles 60510 60511 * Jon Olson 60512 60513 * Magnus Persson 60514 60515 * Chris Pollard 60516 60517 * Richard Polton 60518 60519 * Derk Reefman 60520 60521 * David Rees 60522 60523 * Paul Reilly 60524 60525 * Tom Reilly 60526 60527 * Torsten Rueger 60528 60529 * Danny Sadinoff 60530 60531 * Marc Schifer 60532 60533 * Erik Schnetter 60534 60535 * Wayne K. Schroll 60536 60537 * David Schuler 60538 60539 * Vin Shelton 60540 60541 * Tim Souder 60542 60543 * Adam Sulmicki 60544 60545 * Bill Thorson 60546 60547 * George Talbot 60548 60549 * Pedro A. M. Vazquez 60550 60551 * Gregory Warnes 60552 60553 * Ian Watson 60554 60555 * David E. Young 60556 60557 * And many others 60558 60559 And finally we'd like to thank everyone who uses the compiler, provides 60560feedback and generally reminds us why we're doing this work in the first 60561place. 60562 60563 60564File: gcc.info, Node: Option Index, Next: Keyword Index, Prev: Contributors, Up: Top 60565 60566Option Index 60567************ 60568 60569GCC's command line options are indexed here without any initial '-' or 60570'--'. Where an option has both positive and negative forms (such as 60571'-fOPTION' and '-fno-OPTION'), relevant entries in the manual are 60572indexed under the most appropriate form; it may sometimes be useful to 60573look up both forms. 60574 60575[index] 60576* Menu: 60577 60578* ###: Overall Options. (line 214) 60579* 80387: x86 Options. (line 532) 60580* A: Preprocessor Options. 60581 (line 334) 60582* allowable_client: Darwin Options. (line 196) 60583* all_load: Darwin Options. (line 110) 60584* analyzer: Static Analyzer Options. 60585 (line 7) 60586* ansi: Standards. (line 13) 60587* ansi <1>: C Dialect Options. (line 11) 60588* ansi <2>: Other Builtins. (line 31) 60589* ansi <3>: Non-bugs. (line 107) 60590* arch_errors_fatal: Darwin Options. (line 114) 60591* aux-info: C Dialect Options. (line 229) 60592* B: Directory Options. (line 122) 60593* Bdynamic: VxWorks Options. (line 22) 60594* bind_at_load: Darwin Options. (line 118) 60595* Bstatic: VxWorks Options. (line 22) 60596* bundle: Darwin Options. (line 123) 60597* bundle_loader: Darwin Options. (line 127) 60598* c: Overall Options. (line 169) 60599* C: Preprocessor Options. 60600 (line 343) 60601* c <1>: Link Options. (line 20) 60602* CC: Preprocessor Options. 60603 (line 355) 60604* client_name: Darwin Options. (line 196) 60605* compatibility_version: Darwin Options. (line 196) 60606* coverage: Instrumentation Options. 60607 (line 50) 60608* current_version: Darwin Options. (line 196) 60609* D: Preprocessor Options. 60610 (line 19) 60611* d: Preprocessor Options. 60612 (line 409) 60613* d <1>: Developer Options. (line 52) 60614* da: Developer Options. (line 246) 60615* dA: Developer Options. (line 249) 60616* dD: Preprocessor Options. 60617 (line 433) 60618* dD <1>: Developer Options. (line 253) 60619* dead_strip: Darwin Options. (line 196) 60620* dependency-file: Darwin Options. (line 196) 60621* dH: Developer Options. (line 257) 60622* dI: Preprocessor Options. 60623 (line 443) 60624* dM: Preprocessor Options. 60625 (line 418) 60626* dN: Preprocessor Options. 60627 (line 439) 60628* dp: Developer Options. (line 260) 60629* dP: Developer Options. (line 265) 60630* dU: Preprocessor Options. 60631 (line 447) 60632* dump-analyzer-exploded-nodes: Static Analyzer Options. 60633 (line 290) 60634* dump-analyzer-exploded-nodes-2: Static Analyzer Options. 60635 (line 294) 60636* dump-analyzer-exploded-nodes-3: Static Analyzer Options. 60637 (line 298) 60638* dumpfullversion: Developer Options. (line 1025) 60639* dumpmachine: Developer Options. (line 1013) 60640* dumpspecs: Developer Options. (line 1030) 60641* dumpversion: Developer Options. (line 1017) 60642* dx: Developer Options. (line 269) 60643* dylib_file: Darwin Options. (line 196) 60644* dylinker_install_name: Darwin Options. (line 196) 60645* dynamic: Darwin Options. (line 196) 60646* dynamiclib: Darwin Options. (line 131) 60647* E: Overall Options. (line 190) 60648* E <1>: Link Options. (line 20) 60649* e: Link Options. (line 169) 60650* EB: ARC Options. (line 597) 60651* EB <1>: C-SKY Options. (line 29) 60652* EB <2>: MIPS Options. (line 7) 60653* EL: ARC Options. (line 606) 60654* EL <1>: C-SKY Options. (line 31) 60655* EL <2>: MIPS Options. (line 10) 60656* entry: Link Options. (line 169) 60657* exported_symbols_list: Darwin Options. (line 196) 60658* F: Darwin Options. (line 31) 60659* fabi-compat-version: C++ Dialect Options. 60660 (line 88) 60661* fabi-version: C++ Dialect Options. 60662 (line 24) 60663* faccess-control: C++ Dialect Options. 60664 (line 104) 60665* fada-spec-parent: Overall Options. (line 397) 60666* faggressive-loop-optimizations: Optimize Options. (line 546) 60667* falign-functions: Optimize Options. (line 1668) 60668* falign-jumps: Optimize Options. (line 1750) 60669* falign-labels: Optimize Options. (line 1709) 60670* falign-loops: Optimize Options. (line 1729) 60671* faligned-new: C++ Dialect Options. 60672 (line 108) 60673* fallow-parameterless-variadic-functions: C Dialect Options. 60674 (line 245) 60675* fallow-store-data-races: Optimize Options. (line 1770) 60676* fanalyzer: Static Analyzer Options. 60677 (line 7) 60678* fanalyzer-call-summaries: Static Analyzer Options. 60679 (line 174) 60680* fanalyzer-checker: Static Analyzer Options. 60681 (line 183) 60682* fanalyzer-fine-grained: Static Analyzer Options. 60683 (line 191) 60684* fanalyzer-show-duplicate-count: Static Analyzer Options. 60685 (line 201) 60686* fanalyzer-state-merge: Static Analyzer Options. 60687 (line 208) 60688* fanalyzer-state-purge: Static Analyzer Options. 60689 (line 216) 60690* fanalyzer-transitivity: Static Analyzer Options. 60691 (line 227) 60692* fasan-shadow-offset: Instrumentation Options. 60693 (line 456) 60694* fasm: C Dialect Options. (line 252) 60695* fassociative-math: Optimize Options. (line 2283) 60696* fasynchronous-unwind-tables: Code Gen Options. (line 156) 60697* fauto-inc-dec: Optimize Options. (line 568) 60698* fauto-profile: Optimize Options. (line 2158) 60699* fbranch-count-reg: Optimize Options. (line 420) 60700* fbranch-probabilities: Optimize Options. (line 2429) 60701* fbuiltin: C Dialect Options. (line 266) 60702* fcall-saved: Code Gen Options. (line 448) 60703* fcall-used: Code Gen Options. (line 434) 60704* fcaller-saves: Optimize Options. (line 926) 60705* fcallgraph-info: Developer Options. (line 34) 60706* fcf-protection: Instrumentation Options. 60707 (line 529) 60708* fchar8_t: C++ Dialect Options. 60709 (line 118) 60710* fcheck-new: C++ Dialect Options. 60711 (line 161) 60712* fchecking: Developer Options. (line 705) 60713* fcode-hoisting: Optimize Options. (line 967) 60714* fcombine-stack-adjustments: Optimize Options. (line 938) 60715* fcommon: Code Gen Options. (line 231) 60716* fcommon <1>: Common Variable Attributes. 60717 (line 176) 60718* fcompare-debug: Developer Options. (line 799) 60719* fcompare-debug-second: Developer Options. (line 825) 60720* fcompare-elim: Optimize Options. (line 2088) 60721* fconcepts: C++ Dialect Options. 60722 (line 172) 60723* fconcepts-ts: C++ Dialect Options. 60724 (line 172) 60725* fcond-mismatch: C Dialect Options. (line 396) 60726* fconserve-stack: Optimize Options. (line 957) 60727* fconstant-string-class: Objective-C and Objective-C++ Dialect Options. 60728 (line 30) 60729* fconstexpr-cache-depth: C++ Dialect Options. 60730 (line 188) 60731* fconstexpr-depth: C++ Dialect Options. 60732 (line 182) 60733* fconstexpr-loop-limit: C++ Dialect Options. 60734 (line 198) 60735* fconstexpr-ops-limit: C++ Dialect Options. 60736 (line 203) 60737* fcoroutines: C++ Dialect Options. 60738 (line 212) 60739* fcprop-registers: Optimize Options. (line 2100) 60740* fcrossjumping: Optimize Options. (line 561) 60741* fcse-follow-jumps: Optimize Options. (line 480) 60742* fcse-skip-blocks: Optimize Options. (line 489) 60743* fcx-fortran-rules: Optimize Options. (line 2416) 60744* fcx-limited-range: Optimize Options. (line 2404) 60745* fdata-sections: Optimize Options. (line 2567) 60746* fdbg-cnt: Developer Options. (line 935) 60747* fdbg-cnt-list: Developer Options. (line 932) 60748* fdce: Optimize Options. (line 574) 60749* fdebug-cpp: Preprocessor Options. 60750 (line 454) 60751* fdebug-prefix-map: Debugging Options. (line 141) 60752* fdebug-types-section: Debugging Options. (line 192) 60753* fdeclone-ctor-dtor: Optimize Options. (line 597) 60754* fdefer-pop: Optimize Options. (line 221) 60755* fdelayed-branch: Optimize Options. (line 750) 60756* fdelete-dead-exceptions: Code Gen Options. (line 141) 60757* fdelete-null-pointer-checks: Optimize Options. (line 608) 60758* fdevirtualize: Optimize Options. (line 629) 60759* fdevirtualize-at-ltrans: Optimize Options. (line 646) 60760* fdevirtualize-speculatively: Optimize Options. (line 636) 60761* fdiagnostics-color: Diagnostic Message Formatting Options. 60762 (line 40) 60763* fdiagnostics-format: Diagnostic Message Formatting Options. 60764 (line 398) 60765* fdiagnostics-generate-patch: Diagnostic Message Formatting Options. 60766 (line 240) 60767* fdiagnostics-minimum-margin-width: Diagnostic Message Formatting Options. 60768 (line 209) 60769* fdiagnostics-parseable-fixits: Diagnostic Message Formatting Options. 60770 (line 213) 60771* fdiagnostics-path-format: Diagnostic Message Formatting Options. 60772 (line 290) 60773* fdiagnostics-show-caret: Diagnostic Message Formatting Options. 60774 (line 173) 60775* fdiagnostics-show-cwe: Diagnostic Message Formatting Options. 60776 (line 195) 60777* fdiagnostics-show-labels: Diagnostic Message Formatting Options. 60778 (line 182) 60779* fdiagnostics-show-line-numbers: Diagnostic Message Formatting Options. 60780 (line 204) 60781* fdiagnostics-show-location: Diagnostic Message Formatting Options. 60782 (line 25) 60783* fdiagnostics-show-option: Diagnostic Message Formatting Options. 60784 (line 167) 60785* fdiagnostics-show-path-depths: Diagnostic Message Formatting Options. 60786 (line 381) 60787* fdiagnostics-show-template-tree: Diagnostic Message Formatting Options. 60788 (line 258) 60789* fdiagnostics-urls: Diagnostic Message Formatting Options. 60790 (line 129) 60791* fdirectives-only: Preprocessor Options. 60792 (line 202) 60793* fdisable-: Developer Options. (line 636) 60794* fdollars-in-identifiers: Preprocessor Options. 60795 (line 223) 60796* fdollars-in-identifiers <1>: Interoperation. (line 141) 60797* fdpic: SH Options. (line 388) 60798* fdse: Optimize Options. (line 578) 60799* fdump-ada-spec: Overall Options. (line 392) 60800* fdump-analyzer: Static Analyzer Options. 60801 (line 272) 60802* fdump-analyzer-callgraph: Static Analyzer Options. 60803 (line 281) 60804* fdump-analyzer-exploded-graph: Static Analyzer Options. 60805 (line 285) 60806* fdump-analyzer-state-purge: Static Analyzer Options. 60807 (line 303) 60808* fdump-analyzer-stderr: Static Analyzer Options. 60809 (line 277) 60810* fdump-analyzer-supergraph: Static Analyzer Options. 60811 (line 309) 60812* fdump-debug: Developer Options. (line 273) 60813* fdump-earlydebug: Developer Options. (line 277) 60814* fdump-final-insns: Developer Options. (line 793) 60815* fdump-go-spec: Overall Options. (line 401) 60816* fdump-ipa: Developer Options. (line 303) 60817* fdump-lang: Developer Options. (line 335) 60818* fdump-lang-all: Developer Options. (line 335) 60819* fdump-noaddr: Developer Options. (line 281) 60820* fdump-passes: Developer Options. (line 353) 60821* fdump-rtl-alignments: Developer Options. (line 65) 60822* fdump-rtl-all: Developer Options. (line 246) 60823* fdump-rtl-asmcons: Developer Options. (line 68) 60824* fdump-rtl-auto_inc_dec: Developer Options. (line 72) 60825* fdump-rtl-barriers: Developer Options. (line 76) 60826* fdump-rtl-bbpart: Developer Options. (line 79) 60827* fdump-rtl-bbro: Developer Options. (line 82) 60828* fdump-rtl-btl2: Developer Options. (line 86) 60829* fdump-rtl-btl2 <1>: Developer Options. (line 86) 60830* fdump-rtl-bypass: Developer Options. (line 90) 60831* fdump-rtl-ce1: Developer Options. (line 101) 60832* fdump-rtl-ce2: Developer Options. (line 101) 60833* fdump-rtl-ce3: Developer Options. (line 101) 60834* fdump-rtl-combine: Developer Options. (line 93) 60835* fdump-rtl-compgotos: Developer Options. (line 96) 60836* fdump-rtl-cprop_hardreg: Developer Options. (line 105) 60837* fdump-rtl-csa: Developer Options. (line 108) 60838* fdump-rtl-cse1: Developer Options. (line 112) 60839* fdump-rtl-cse2: Developer Options. (line 112) 60840* fdump-rtl-dbr: Developer Options. (line 119) 60841* fdump-rtl-dce: Developer Options. (line 116) 60842* fdump-rtl-dce1: Developer Options. (line 123) 60843* fdump-rtl-dce2: Developer Options. (line 123) 60844* fdump-rtl-dfinish: Developer Options. (line 242) 60845* fdump-rtl-dfinit: Developer Options. (line 242) 60846* fdump-rtl-eh: Developer Options. (line 127) 60847* fdump-rtl-eh_ranges: Developer Options. (line 130) 60848* fdump-rtl-expand: Developer Options. (line 133) 60849* fdump-rtl-fwprop1: Developer Options. (line 137) 60850* fdump-rtl-fwprop2: Developer Options. (line 137) 60851* fdump-rtl-gcse1: Developer Options. (line 142) 60852* fdump-rtl-gcse2: Developer Options. (line 142) 60853* fdump-rtl-init-regs: Developer Options. (line 146) 60854* fdump-rtl-initvals: Developer Options. (line 149) 60855* fdump-rtl-into_cfglayout: Developer Options. (line 152) 60856* fdump-rtl-ira: Developer Options. (line 155) 60857* fdump-rtl-jump: Developer Options. (line 158) 60858* fdump-rtl-loop2: Developer Options. (line 161) 60859* fdump-rtl-mach: Developer Options. (line 165) 60860* fdump-rtl-mode_sw: Developer Options. (line 169) 60861* fdump-rtl-outof_cfglayout: Developer Options. (line 175) 60862* fdump-rtl-PASS: Developer Options. (line 52) 60863* fdump-rtl-peephole2: Developer Options. (line 178) 60864* fdump-rtl-postreload: Developer Options. (line 181) 60865* fdump-rtl-pro_and_epilogue: Developer Options. (line 184) 60866* fdump-rtl-ree: Developer Options. (line 192) 60867* fdump-rtl-regclass: Developer Options. (line 242) 60868* fdump-rtl-rnreg: Developer Options. (line 172) 60869* fdump-rtl-sched1: Developer Options. (line 188) 60870* fdump-rtl-sched2: Developer Options. (line 188) 60871* fdump-rtl-seqabstr: Developer Options. (line 195) 60872* fdump-rtl-shorten: Developer Options. (line 198) 60873* fdump-rtl-sibling: Developer Options. (line 201) 60874* fdump-rtl-sms: Developer Options. (line 212) 60875* fdump-rtl-split1: Developer Options. (line 208) 60876* fdump-rtl-split2: Developer Options. (line 208) 60877* fdump-rtl-split3: Developer Options. (line 208) 60878* fdump-rtl-split4: Developer Options. (line 208) 60879* fdump-rtl-split5: Developer Options. (line 208) 60880* fdump-rtl-stack: Developer Options. (line 216) 60881* fdump-rtl-subreg1: Developer Options. (line 222) 60882* fdump-rtl-subreg2: Developer Options. (line 222) 60883* fdump-rtl-subregs_of_mode_finish: Developer Options. (line 242) 60884* fdump-rtl-subregs_of_mode_init: Developer Options. (line 242) 60885* fdump-rtl-unshare: Developer Options. (line 226) 60886* fdump-rtl-vartrack: Developer Options. (line 229) 60887* fdump-rtl-vregs: Developer Options. (line 232) 60888* fdump-rtl-web: Developer Options. (line 235) 60889* fdump-statistics: Developer Options. (line 357) 60890* fdump-tree: Developer Options. (line 370) 60891* fdump-tree-all: Developer Options. (line 370) 60892* fdump-unnumbered: Developer Options. (line 291) 60893* fdump-unnumbered-links: Developer Options. (line 297) 60894* fdwarf2-cfi-asm: Debugging Options. (line 397) 60895* fearly-inlining: Optimize Options. (line 320) 60896* felide-constructors: C++ Dialect Options. 60897 (line 215) 60898* felide-type: Diagnostic Message Formatting Options. 60899 (line 278) 60900* feliminate-unused-debug-symbols: Debugging Options. (line 121) 60901* feliminate-unused-debug-types: Debugging Options. (line 401) 60902* femit-class-debug-always: Debugging Options. (line 126) 60903* femit-struct-debug-baseonly: Debugging Options. (line 328) 60904* femit-struct-debug-detailed: Debugging Options. (line 355) 60905* femit-struct-debug-reduced: Debugging Options. (line 341) 60906* fenable-: Developer Options. (line 636) 60907* fenforce-eh-specs: C++ Dialect Options. 60908 (line 226) 60909* fexceptions: Code Gen Options. (line 119) 60910* fexcess-precision: Optimize Options. (line 2209) 60911* fexec-charset: Preprocessor Options. 60912 (line 270) 60913* fexpensive-optimizations: Optimize Options. (line 653) 60914* fext-numeric-literals: C++ Dialect Options. 60915 (line 478) 60916* fextended-identifiers: Preprocessor Options. 60917 (line 226) 60918* fextern-tls-init: C++ Dialect Options. 60919 (line 236) 60920* ffast-math: Optimize Options. (line 2233) 60921* ffat-lto-objects: Optimize Options. (line 2065) 60922* ffile-prefix-map: Overall Options. (line 372) 60923* ffinite-loops: Optimize Options. (line 1207) 60924* ffinite-math-only: Optimize Options. (line 2310) 60925* ffix-and-continue: Darwin Options. (line 104) 60926* ffixed: Code Gen Options. (line 422) 60927* ffloat-store: Optimize Options. (line 2195) 60928* ffloat-store <1>: Disappointments. (line 77) 60929* fforward-propagate: Optimize Options. (line 228) 60930* ffp-contract: Optimize Options. (line 237) 60931* ffp-int-builtin-inexact: Optimize Options. (line 2382) 60932* ffreestanding: Standards. (line 99) 60933* ffreestanding <1>: C Dialect Options. (line 314) 60934* ffreestanding <2>: Warning Options. (line 412) 60935* ffreestanding <3>: Common Function Attributes. 60936 (line 421) 60937* ffunction-cse: Optimize Options. (line 434) 60938* ffunction-sections: Optimize Options. (line 2567) 60939* fgcse: Optimize Options. (line 503) 60940* fgcse-after-reload: Optimize Options. (line 539) 60941* fgcse-las: Optimize Options. (line 532) 60942* fgcse-lm: Optimize Options. (line 514) 60943* fgcse-sm: Optimize Options. (line 523) 60944* fgimple: C Dialect Options. (line 300) 60945* fgnu-keywords: C++ Dialect Options. 60946 (line 256) 60947* fgnu-runtime: Objective-C and Objective-C++ Dialect Options. 60948 (line 39) 60949* fgnu-tm: C Dialect Options. (line 353) 60950* fgnu-unique: Code Gen Options. (line 162) 60951* fgnu89-inline: C Dialect Options. (line 190) 60952* fgraphite-identity: Optimize Options. (line 1249) 60953* fguess-branch-probability: Optimize Options. (line 1548) 60954* fhoist-adjacent-loads: Optimize Options. (line 997) 60955* fhosted: C Dialect Options. (line 306) 60956* fident: Code Gen Options. (line 252) 60957* fif-conversion: Optimize Options. (line 582) 60958* fif-conversion2: Optimize Options. (line 591) 60959* fiji: AMD GCN Options. (line 13) 60960* filelist: Darwin Options. (line 196) 60961* fimplement-inlines: C++ Dialect Options. 60962 (line 276) 60963* fimplicit-inline-templates: C++ Dialect Options. 60964 (line 270) 60965* fimplicit-templates: C++ Dialect Options. 60966 (line 262) 60967* findirect-data: Darwin Options. (line 104) 60968* findirect-inlining: Optimize Options. (line 292) 60969* finhibit-size-directive: Code Gen Options. (line 255) 60970* finline: Optimize Options. (line 275) 60971* finline-functions: Optimize Options. (line 300) 60972* finline-functions-called-once: Optimize Options. (line 312) 60973* finline-limit: Optimize Options. (line 336) 60974* finline-small-functions: Optimize Options. (line 283) 60975* finput-charset: Preprocessor Options. 60976 (line 285) 60977* finstrument-functions: Instrumentation Options. 60978 (line 732) 60979* finstrument-functions <1>: Common Function Attributes. 60980 (line 696) 60981* finstrument-functions-exclude-file-list: Instrumentation Options. 60982 (line 768) 60983* finstrument-functions-exclude-function-list: Instrumentation Options. 60984 (line 789) 60985* fipa-bit-cp: Optimize Options. (line 1057) 60986* fipa-cp: Optimize Options. (line 1038) 60987* fipa-cp-clone: Optimize Options. (line 1047) 60988* fipa-icf: Optimize Options. (line 1067) 60989* fipa-profile: Optimize Options. (line 1030) 60990* fipa-pta: Optimize Options. (line 1024) 60991* fipa-pure-const: Optimize Options. (line 1008) 60992* fipa-ra: Optimize Options. (line 944) 60993* fipa-reference: Optimize Options. (line 1012) 60994* fipa-reference-addressable: Optimize Options. (line 1016) 60995* fipa-sra: Optimize Options. (line 329) 60996* fipa-stack-alignment: Optimize Options. (line 1020) 60997* fipa-vrp: Optimize Options. (line 1062) 60998* fira-algorithm: Optimize Options. (line 687) 60999* fira-hoist-pressure: Optimize Options. (line 716) 61000* fira-loop-pressure: Optimize Options. (line 723) 61001* fira-region: Optimize Options. (line 695) 61002* fira-share-save-slots: Optimize Options. (line 731) 61003* fira-share-spill-slots: Optimize Options. (line 737) 61004* fira-verbose: Developer Options. (line 862) 61005* fisolate-erroneous-paths-attribute: Optimize Options. (line 1149) 61006* fisolate-erroneous-paths-dereference: Optimize Options. (line 1141) 61007* fivar-visibility: Objective-C and Objective-C++ Dialect Options. 61008 (line 161) 61009* fivopts: Optimize Options. (line 1374) 61010* fjump-tables: Code Gen Options. (line 414) 61011* fkeep-inline-dllexport: Optimize Options. (line 361) 61012* fkeep-inline-functions: Optimize Options. (line 367) 61013* fkeep-inline-functions <1>: Inline. (line 51) 61014* fkeep-static-consts: Optimize Options. (line 378) 61015* fkeep-static-functions: Optimize Options. (line 374) 61016* flat_namespace: Darwin Options. (line 196) 61017* flax-vector-conversions: C Dialect Options. (line 401) 61018* fleading-underscore: Code Gen Options. (line 478) 61019* flifetime-dse: Optimize Options. (line 667) 61020* flinker-output: Link Options. (line 25) 61021* flive-patching: Optimize Options. (line 1081) 61022* flive-range-shrinkage: Optimize Options. (line 682) 61023* flocal-ivars: Objective-C and Objective-C++ Dialect Options. 61024 (line 152) 61025* floop-block: Optimize Options. (line 1243) 61026* floop-interchange: Optimize Options. (line 1327) 61027* floop-nest-optimize: Optimize Options. (line 1257) 61028* floop-parallelize-all: Optimize Options. (line 1263) 61029* floop-strip-mine: Optimize Options. (line 1243) 61030* floop-unroll-and-jam: Optimize Options. (line 1344) 61031* flra-remat: Optimize Options. (line 743) 61032* flto: Optimize Options. (line 1817) 61033* flto-compression-level: Optimize Options. (line 2039) 61034* flto-partition: Optimize Options. (line 2025) 61035* flto-report: Developer Options. (line 868) 61036* flto-report-wpa: Developer Options. (line 876) 61037* fmacro-prefix-map: Preprocessor Options. 61038 (line 261) 61039* fmath-errno: Optimize Options. (line 2247) 61040* fmax-errors: Warning Options. (line 18) 61041* fmax-include-depth: Preprocessor Options. 61042 (line 235) 61043* fmem-report: Developer Options. (line 880) 61044* fmem-report-wpa: Developer Options. (line 884) 61045* fmerge-all-constants: Optimize Options. (line 397) 61046* fmerge-constants: Optimize Options. (line 387) 61047* fmerge-debug-strings: Debugging Options. (line 134) 61048* fmessage-length: Diagnostic Message Formatting Options. 61049 (line 14) 61050* fmodulo-sched: Optimize Options. (line 408) 61051* fmodulo-sched-allow-regmoves: Optimize Options. (line 413) 61052* fmove-loop-invariants: Optimize Options. (line 2527) 61053* fms-extensions: C Dialect Options. (line 368) 61054* fms-extensions <1>: C++ Dialect Options. 61055 (line 281) 61056* fms-extensions <2>: Unnamed Fields. (line 36) 61057* fnew-inheriting-ctors: C++ Dialect Options. 61058 (line 286) 61059* fnew-ttp-matching: C++ Dialect Options. 61060 (line 292) 61061* fnext-runtime: Objective-C and Objective-C++ Dialect Options. 61062 (line 43) 61063* fnil-receivers: Objective-C and Objective-C++ Dialect Options. 61064 (line 49) 61065* fno-access-control: C++ Dialect Options. 61066 (line 104) 61067* fno-allocation-dce: Optimize Options. (line 1767) 61068* fno-analyzer: Static Analyzer Options. 61069 (line 7) 61070* fno-analyzer-call-summaries: Static Analyzer Options. 61071 (line 174) 61072* fno-analyzer-fine-grained: Static Analyzer Options. 61073 (line 191) 61074* fno-analyzer-show-duplicate-count: Static Analyzer Options. 61075 (line 201) 61076* fno-analyzer-state-merge: Static Analyzer Options. 61077 (line 208) 61078* fno-analyzer-state-purge: Static Analyzer Options. 61079 (line 216) 61080* fno-analyzer-transitivity: Static Analyzer Options. 61081 (line 227) 61082* fno-asm: C Dialect Options. (line 252) 61083* fno-branch-count-reg: Optimize Options. (line 420) 61084* fno-builtin: C Dialect Options. (line 266) 61085* fno-builtin <1>: Warning Options. (line 412) 61086* fno-builtin <2>: Common Function Attributes. 61087 (line 421) 61088* fno-builtin <3>: Other Builtins. (line 21) 61089* fno-canonical-system-headers: Preprocessor Options. 61090 (line 231) 61091* fno-char8_t: C++ Dialect Options. 61092 (line 118) 61093* fno-checking: Developer Options. (line 705) 61094* fno-common: Code Gen Options. (line 231) 61095* fno-common <1>: Common Variable Attributes. 61096 (line 176) 61097* fno-compare-debug: Developer Options. (line 799) 61098* fno-debug-types-section: Debugging Options. (line 192) 61099* fno-default-inline: Inline. (line 68) 61100* fno-defer-pop: Optimize Options. (line 221) 61101* fno-diagnostics-show-caret: Diagnostic Message Formatting Options. 61102 (line 173) 61103* fno-diagnostics-show-cwe: Diagnostic Message Formatting Options. 61104 (line 195) 61105* fno-diagnostics-show-labels: Diagnostic Message Formatting Options. 61106 (line 182) 61107* fno-diagnostics-show-line-numbers: Diagnostic Message Formatting Options. 61108 (line 204) 61109* fno-diagnostics-show-option: Diagnostic Message Formatting Options. 61110 (line 167) 61111* fno-dwarf2-cfi-asm: Debugging Options. (line 397) 61112* fno-elide-constructors: C++ Dialect Options. 61113 (line 215) 61114* fno-elide-type: Diagnostic Message Formatting Options. 61115 (line 278) 61116* fno-eliminate-unused-debug-symbols: Debugging Options. (line 121) 61117* fno-eliminate-unused-debug-types: Debugging Options. (line 401) 61118* fno-enforce-eh-specs: C++ Dialect Options. 61119 (line 226) 61120* fno-ext-numeric-literals: C++ Dialect Options. 61121 (line 478) 61122* fno-extern-tls-init: C++ Dialect Options. 61123 (line 236) 61124* fno-finite-loops: Optimize Options. (line 1207) 61125* fno-fp-int-builtin-inexact: Optimize Options. (line 2382) 61126* fno-function-cse: Optimize Options. (line 434) 61127* fno-gnu-keywords: C++ Dialect Options. 61128 (line 256) 61129* fno-gnu-unique: Code Gen Options. (line 162) 61130* fno-guess-branch-probability: Optimize Options. (line 1548) 61131* fno-ident: Code Gen Options. (line 252) 61132* fno-implement-inlines: C++ Dialect Options. 61133 (line 276) 61134* fno-implement-inlines <1>: C++ Interface. (line 66) 61135* fno-implicit-inline-templates: C++ Dialect Options. 61136 (line 270) 61137* fno-implicit-templates: C++ Dialect Options. 61138 (line 262) 61139* fno-implicit-templates <1>: Template Instantiation. 61140 (line 94) 61141* fno-inline: Optimize Options. (line 275) 61142* fno-ira-share-save-slots: Optimize Options. (line 731) 61143* fno-ira-share-spill-slots: Optimize Options. (line 737) 61144* fno-jump-tables: Code Gen Options. (line 414) 61145* fno-keep-inline-dllexport: Optimize Options. (line 361) 61146* fno-lifetime-dse: Optimize Options. (line 667) 61147* fno-local-ivars: Objective-C and Objective-C++ Dialect Options. 61148 (line 152) 61149* fno-math-errno: Optimize Options. (line 2247) 61150* fno-merge-debug-strings: Debugging Options. (line 134) 61151* fno-nil-receivers: Objective-C and Objective-C++ Dialect Options. 61152 (line 49) 61153* fno-nonansi-builtins: C++ Dialect Options. 61154 (line 299) 61155* fno-operator-names: C++ Dialect Options. 61156 (line 315) 61157* fno-optional-diags: C++ Dialect Options. 61158 (line 319) 61159* fno-peephole: Optimize Options. (line 1539) 61160* fno-peephole2: Optimize Options. (line 1539) 61161* fno-plt: Code Gen Options. (line 396) 61162* fno-pretty-templates: C++ Dialect Options. 61163 (line 329) 61164* fno-printf-return-value: Optimize Options. (line 1516) 61165* fno-rtti: C++ Dialect Options. 61166 (line 341) 61167* fno-sanitize-recover: Instrumentation Options. 61168 (line 465) 61169* fno-sanitize=all: Instrumentation Options. 61170 (line 450) 61171* fno-sched-interblock: Optimize Options. (line 776) 61172* fno-sched-spec: Optimize Options. (line 781) 61173* fno-set-stack-executable: x86 Windows Options. 61174 (line 46) 61175* fno-show-column: Diagnostic Message Formatting Options. 61176 (line 393) 61177* fno-signed-bitfields: C Dialect Options. (line 434) 61178* fno-signed-zeros: Optimize Options. (line 2322) 61179* fno-stack-limit: Instrumentation Options. 61180 (line 644) 61181* fno-threadsafe-statics: C++ Dialect Options. 61182 (line 396) 61183* fno-toplevel-reorder: Optimize Options. (line 1782) 61184* fno-trapping-math: Optimize Options. (line 2332) 61185* fno-unsigned-bitfields: C Dialect Options. (line 434) 61186* fno-use-cxa-get-exception-ptr: C++ Dialect Options. 61187 (line 409) 61188* fno-var-tracking-assignments: Debugging Options. (line 161) 61189* fno-var-tracking-assignments-toggle: Developer Options. (line 846) 61190* fno-weak: C++ Dialect Options. 61191 (line 471) 61192* fno-working-directory: Preprocessor Options. 61193 (line 320) 61194* fno-writable-relocated-rdata: x86 Windows Options. 61195 (line 53) 61196* fno-zero-initialized-in-bss: Optimize Options. (line 445) 61197* fnon-call-exceptions: Code Gen Options. (line 133) 61198* fnonansi-builtins: C++ Dialect Options. 61199 (line 299) 61200* fnothrow-opt: C++ Dialect Options. 61201 (line 304) 61202* fobjc-abi-version: Objective-C and Objective-C++ Dialect Options. 61203 (line 56) 61204* fobjc-call-cxx-cdtors: Objective-C and Objective-C++ Dialect Options. 61205 (line 67) 61206* fobjc-direct-dispatch: Objective-C and Objective-C++ Dialect Options. 61207 (line 92) 61208* fobjc-exceptions: Objective-C and Objective-C++ Dialect Options. 61209 (line 96) 61210* fobjc-gc: Objective-C and Objective-C++ Dialect Options. 61211 (line 104) 61212* fobjc-nilcheck: Objective-C and Objective-C++ Dialect Options. 61213 (line 110) 61214* fobjc-std: Objective-C and Objective-C++ Dialect Options. 61215 (line 119) 61216* fomit-frame-pointer: Optimize Options. (line 248) 61217* fopenacc: C Dialect Options. (line 325) 61218* fopenacc-dim: C Dialect Options. (line 334) 61219* fopenmp: C Dialect Options. (line 340) 61220* fopenmp-simd: C Dialect Options. (line 349) 61221* foperator-names: C++ Dialect Options. 61222 (line 315) 61223* fopt-info: Developer Options. (line 476) 61224* foptimize-sibling-calls: Optimize Options. (line 263) 61225* foptimize-strlen: Optimize Options. (line 268) 61226* foptional-diags: C++ Dialect Options. 61227 (line 319) 61228* force_cpusubtype_ALL: Darwin Options. (line 135) 61229* force_flat_namespace: Darwin Options. (line 196) 61230* fpack-struct: Code Gen Options. (line 465) 61231* fpartial-inlining: Optimize Options. (line 1491) 61232* fpatchable-function-entry: Instrumentation Options. 61233 (line 801) 61234* fpcc-struct-return: Code Gen Options. (line 175) 61235* fpcc-struct-return <1>: Incompatibilities. (line 170) 61236* fpch-deps: Preprocessor Options. 61237 (line 295) 61238* fpch-preprocess: Preprocessor Options. 61239 (line 303) 61240* fpeel-loops: Optimize Options. (line 2519) 61241* fpeephole: Optimize Options. (line 1539) 61242* fpeephole2: Optimize Options. (line 1539) 61243* fpermissive: C++ Dialect Options. 61244 (line 324) 61245* fpermitted-flt-eval-methods: C Dialect Options. (line 207) 61246* fpermitted-flt-eval-methods=c11: C Dialect Options. (line 207) 61247* fpermitted-flt-eval-methods=ts-18661-3: C Dialect Options. (line 207) 61248* fpic: Code Gen Options. (line 353) 61249* fPIC: Code Gen Options. (line 374) 61250* fpie: Code Gen Options. (line 387) 61251* fPIE: Code Gen Options. (line 387) 61252* fplan9-extensions: C Dialect Options. (line 386) 61253* fplan9-extensions <1>: Unnamed Fields. (line 43) 61254* fplt: Code Gen Options. (line 396) 61255* fplugin: Overall Options. (line 381) 61256* fplugin-arg: Overall Options. (line 388) 61257* fpost-ipa-mem-report: Developer Options. (line 889) 61258* fpre-ipa-mem-report: Developer Options. (line 888) 61259* fpredictive-commoning: Optimize Options. (line 1498) 61260* fprefetch-loop-arrays: Optimize Options. (line 1506) 61261* fpreprocessed: Preprocessor Options. 61262 (line 189) 61263* fpretty-templates: C++ Dialect Options. 61264 (line 329) 61265* fprintf-return-value: Optimize Options. (line 1516) 61266* fprofile-abs-path: Instrumentation Options. 61267 (line 106) 61268* fprofile-arcs: Instrumentation Options. 61269 (line 30) 61270* fprofile-arcs <1>: Other Builtins. (line 563) 61271* fprofile-correction: Optimize Options. (line 2107) 61272* fprofile-dir: Instrumentation Options. 61273 (line 112) 61274* fprofile-exclude-files: Instrumentation Options. 61275 (line 200) 61276* fprofile-filter-files: Instrumentation Options. 61277 (line 192) 61278* fprofile-generate: Instrumentation Options. 61279 (line 137) 61280* fprofile-note: Instrumentation Options. 61281 (line 154) 61282* fprofile-partial-training: Optimize Options. (line 2116) 61283* fprofile-prefix-path: Instrumentation Options. 61284 (line 160) 61285* fprofile-reorder-functions: Optimize Options. (line 2459) 61286* fprofile-report: Developer Options. (line 893) 61287* fprofile-reproducible: Instrumentation Options. 61288 (line 208) 61289* fprofile-update: Instrumentation Options. 61290 (line 175) 61291* fprofile-use: Optimize Options. (line 2130) 61292* fprofile-values: Optimize Options. (line 2449) 61293* fpu: RX Options. (line 17) 61294* frandom-seed: Developer Options. (line 710) 61295* freciprocal-math: Optimize Options. (line 2300) 61296* frecord-gcc-switches: Code Gen Options. (line 341) 61297* free: Optimize Options. (line 659) 61298* freg-struct-return: Code Gen Options. (line 193) 61299* frename-registers: Optimize Options. (line 2478) 61300* freorder-blocks: Optimize Options. (line 1569) 61301* freorder-blocks-algorithm: Optimize Options. (line 1575) 61302* freorder-blocks-and-partition: Optimize Options. (line 1586) 61303* freorder-functions: Optimize Options. (line 1603) 61304* freplace-objc-classes: Objective-C and Objective-C++ Dialect Options. 61305 (line 130) 61306* freport-bug: Developer Options. (line 287) 61307* frerun-cse-after-loop: Optimize Options. (line 497) 61308* freschedule-modulo-scheduled-loops: Optimize Options. (line 875) 61309* frounding-math: Optimize Options. (line 2347) 61310* frtti: C++ Dialect Options. 61311 (line 341) 61312* fsanitize-address-use-after-scope: Instrumentation Options. 61313 (line 501) 61314* fsanitize-coverage=trace-cmp: Instrumentation Options. 61315 (line 516) 61316* fsanitize-coverage=trace-pc: Instrumentation Options. 61317 (line 512) 61318* fsanitize-recover: Instrumentation Options. 61319 (line 465) 61320* fsanitize-sections: Instrumentation Options. 61321 (line 461) 61322* fsanitize-undefined-trap-on-error: Instrumentation Options. 61323 (line 505) 61324* fsanitize=address: Instrumentation Options. 61325 (line 236) 61326* fsanitize=alignment: Instrumentation Options. 61327 (line 376) 61328* fsanitize=bool: Instrumentation Options. 61329 (line 414) 61330* fsanitize=bounds: Instrumentation Options. 61331 (line 363) 61332* fsanitize=bounds-strict: Instrumentation Options. 61333 (line 369) 61334* fsanitize=builtin: Instrumentation Options. 61335 (line 438) 61336* fsanitize=enum: Instrumentation Options. 61337 (line 419) 61338* fsanitize=float-cast-overflow: Instrumentation Options. 61339 (line 394) 61340* fsanitize=float-divide-by-zero: Instrumentation Options. 61341 (line 388) 61342* fsanitize=integer-divide-by-zero: Instrumentation Options. 61343 (line 326) 61344* fsanitize=kernel-address: Instrumentation Options. 61345 (line 250) 61346* fsanitize=leak: Instrumentation Options. 61347 (line 288) 61348* fsanitize=nonnull-attribute: Instrumentation Options. 61349 (line 402) 61350* fsanitize=null: Instrumentation Options. 61351 (line 340) 61352* fsanitize=object-size: Instrumentation Options. 61353 (line 383) 61354* fsanitize=pointer-compare: Instrumentation Options. 61355 (line 254) 61356* fsanitize=pointer-overflow: Instrumentation Options. 61357 (line 432) 61358* fsanitize=pointer-subtract: Instrumentation Options. 61359 (line 264) 61360* fsanitize=return: Instrumentation Options. 61361 (line 348) 61362* fsanitize=returns-nonnull-attribute: Instrumentation Options. 61363 (line 408) 61364* fsanitize=shift: Instrumentation Options. 61365 (line 306) 61366* fsanitize=shift-base: Instrumentation Options. 61367 (line 319) 61368* fsanitize=shift-exponent: Instrumentation Options. 61369 (line 314) 61370* fsanitize=signed-integer-overflow: Instrumentation Options. 61371 (line 354) 61372* fsanitize=thread: Instrumentation Options. 61373 (line 274) 61374* fsanitize=undefined: Instrumentation Options. 61375 (line 298) 61376* fsanitize=unreachable: Instrumentation Options. 61377 (line 330) 61378* fsanitize=vla-bound: Instrumentation Options. 61379 (line 336) 61380* fsanitize=vptr: Instrumentation Options. 61381 (line 425) 61382* fsave-optimization-record: Developer Options. (line 582) 61383* fsched-critical-path-heuristic: Optimize Options. (line 841) 61384* fsched-dep-count-heuristic: Optimize Options. (line 868) 61385* fsched-group-heuristic: Optimize Options. (line 835) 61386* fsched-interblock: Optimize Options. (line 776) 61387* fsched-last-insn-heuristic: Optimize Options. (line 861) 61388* fsched-pressure: Optimize Options. (line 786) 61389* fsched-rank-heuristic: Optimize Options. (line 854) 61390* fsched-spec: Optimize Options. (line 781) 61391* fsched-spec-insn-heuristic: Optimize Options. (line 847) 61392* fsched-spec-load: Optimize Options. (line 795) 61393* fsched-spec-load-dangerous: Optimize Options. (line 800) 61394* fsched-stalled-insns: Optimize Options. (line 806) 61395* fsched-stalled-insns-dep: Optimize Options. (line 816) 61396* fsched-verbose: Developer Options. (line 622) 61397* fsched2-use-superblocks: Optimize Options. (line 825) 61398* fschedule-fusion: Optimize Options. (line 2488) 61399* fschedule-insns: Optimize Options. (line 757) 61400* fschedule-insns2: Optimize Options. (line 767) 61401* fsection-anchors: Optimize Options. (line 2600) 61402* fsel-sched-pipelining: Optimize Options. (line 888) 61403* fsel-sched-pipelining-outer-loops: Optimize Options. (line 893) 61404* fselective-scheduling: Optimize Options. (line 880) 61405* fselective-scheduling2: Optimize Options. (line 884) 61406* fsemantic-interposition: Optimize Options. (line 898) 61407* fset-stack-executable: x86 Windows Options. 61408 (line 46) 61409* fshort-enums: Code Gen Options. (line 211) 61410* fshort-enums <1>: Structures unions enumerations and bit-fields implementation. 61411 (line 48) 61412* fshort-enums <2>: Common Type Attributes. 61413 (line 288) 61414* fshort-enums <3>: Non-bugs. (line 42) 61415* fshort-wchar: Code Gen Options. (line 221) 61416* fshow-column: Diagnostic Message Formatting Options. 61417 (line 393) 61418* fshrink-wrap: Optimize Options. (line 915) 61419* fshrink-wrap-separate: Optimize Options. (line 920) 61420* fsignaling-nans: Optimize Options. (line 2367) 61421* fsigned-bitfields: C Dialect Options. (line 434) 61422* fsigned-bitfields <1>: Non-bugs. (line 57) 61423* fsigned-char: C Dialect Options. (line 424) 61424* fsigned-char <1>: Characters implementation. 61425 (line 31) 61426* fsigned-zeros: Optimize Options. (line 2322) 61427* fsimd-cost-model: Optimize Options. (line 1448) 61428* fsingle-precision-constant: Optimize Options. (line 2400) 61429* fsized-deallocation: C++ Dialect Options. 61430 (line 356) 61431* fsplit-ivs-in-unroller: Optimize Options. (line 1469) 61432* fsplit-loops: Optimize Options. (line 2531) 61433* fsplit-paths: Optimize Options. (line 1464) 61434* fsplit-stack: Instrumentation Options. 61435 (line 661) 61436* fsplit-stack <1>: Common Function Attributes. 61437 (line 746) 61438* fsplit-wide-types: Optimize Options. (line 466) 61439* fsplit-wide-types-early: Optimize Options. (line 474) 61440* fssa-backprop: Optimize Options. (line 1173) 61441* fssa-phiopt: Optimize Options. (line 1179) 61442* fsso-struct: C Dialect Options. (line 440) 61443* fstack-check: Instrumentation Options. 61444 (line 587) 61445* fstack-clash-protection: Instrumentation Options. 61446 (line 629) 61447* fstack-limit-register: Instrumentation Options. 61448 (line 644) 61449* fstack-limit-symbol: Instrumentation Options. 61450 (line 644) 61451* fstack-protector: Instrumentation Options. 61452 (line 562) 61453* fstack-protector-all: Instrumentation Options. 61454 (line 573) 61455* fstack-protector-explicit: Instrumentation Options. 61456 (line 583) 61457* fstack-protector-strong: Instrumentation Options. 61458 (line 576) 61459* fstack-usage: Developer Options. (line 897) 61460* fstack_reuse: Code Gen Options. (line 15) 61461* fstats: Developer Options. (line 926) 61462* fstdarg-opt: Optimize Options. (line 2596) 61463* fstore-merging: Optimize Options. (line 1398) 61464* fstrict-aliasing: Optimize Options. (line 1618) 61465* fstrict-enums: C++ Dialect Options. 61466 (line 366) 61467* fstrict-overflow: Code Gen Options. (line 115) 61468* fstrict-volatile-bitfields: Code Gen Options. (line 589) 61469* fstrong-eval-order: C++ Dialect Options. 61470 (line 375) 61471* fsync-libcalls: Code Gen Options. (line 621) 61472* fsyntax-only: Warning Options. (line 14) 61473* ftabstop: Preprocessor Options. 61474 (line 238) 61475* ftemplate-backtrace-limit: C++ Dialect Options. 61476 (line 383) 61477* ftemplate-depth: C++ Dialect Options. 61478 (line 387) 61479* ftest-coverage: Instrumentation Options. 61480 (line 97) 61481* fthread-jumps: Optimize Options. (line 457) 61482* fthreadsafe-statics: C++ Dialect Options. 61483 (line 396) 61484* ftime-report: Developer Options. (line 854) 61485* ftime-report-details: Developer Options. (line 858) 61486* ftls-model: Code Gen Options. (line 489) 61487* ftoplevel-reorder: Optimize Options. (line 1782) 61488* ftracer: Optimize Options. (line 2496) 61489* ftrack-macro-expansion: Preprocessor Options. 61490 (line 244) 61491* ftrampolines: Code Gen Options. (line 500) 61492* ftrapping-math: Optimize Options. (line 2332) 61493* ftrapv: Code Gen Options. (line 91) 61494* ftree-bit-ccp: Optimize Options. (line 1161) 61495* ftree-builtin-call-dce: Optimize Options. (line 1201) 61496* ftree-ccp: Optimize Options. (line 1168) 61497* ftree-ch: Optimize Options. (line 1230) 61498* ftree-coalesce-vars: Optimize Options. (line 1269) 61499* ftree-copy-prop: Optimize Options. (line 1003) 61500* ftree-dce: Optimize Options. (line 1197) 61501* ftree-dominator-opts: Optimize Options. (line 1216) 61502* ftree-dse: Optimize Options. (line 1223) 61503* ftree-forwprop: Optimize Options. (line 982) 61504* ftree-fre: Optimize Options. (line 986) 61505* ftree-loop-distribute-patterns: Optimize Options. (line 1305) 61506* ftree-loop-distribution: Optimize Options. (line 1286) 61507* ftree-loop-if-convert: Optimize Options. (line 1279) 61508* ftree-loop-im: Optimize Options. (line 1350) 61509* ftree-loop-ivcanon: Optimize Options. (line 1359) 61510* ftree-loop-linear: Optimize Options. (line 1243) 61511* ftree-loop-optimize: Optimize Options. (line 1237) 61512* ftree-loop-vectorize: Optimize Options. (line 1424) 61513* ftree-parallelize-loops: Optimize Options. (line 1379) 61514* ftree-partial-pre: Optimize Options. (line 978) 61515* ftree-phiprop: Optimize Options. (line 993) 61516* ftree-pre: Optimize Options. (line 974) 61517* ftree-pta: Optimize Options. (line 1388) 61518* ftree-reassoc: Optimize Options. (line 963) 61519* ftree-scev-cprop: Optimize Options. (line 1365) 61520* ftree-sink: Optimize Options. (line 1157) 61521* ftree-slp-vectorize: Optimize Options. (line 1429) 61522* ftree-slsr: Optimize Options. (line 1413) 61523* ftree-sra: Optimize Options. (line 1392) 61524* ftree-switch-conversion: Optimize Options. (line 1184) 61525* ftree-tail-merge: Optimize Options. (line 1189) 61526* ftree-ter: Optimize Options. (line 1405) 61527* ftree-vectorize: Optimize Options. (line 1419) 61528* ftree-vrp: Optimize Options. (line 1455) 61529* funconstrained-commons: Optimize Options. (line 555) 61530* funit-at-a-time: Optimize Options. (line 1775) 61531* funroll-all-loops: Optimize Options. (line 2513) 61532* funroll-loops: Optimize Options. (line 2503) 61533* funsafe-math-optimizations: Optimize Options. (line 2265) 61534* funsigned-bitfields: C Dialect Options. (line 434) 61535* funsigned-bitfields <1>: Structures unions enumerations and bit-fields implementation. 61536 (line 17) 61537* funsigned-bitfields <2>: Non-bugs. (line 57) 61538* funsigned-char: C Dialect Options. (line 406) 61539* funsigned-char <1>: Characters implementation. 61540 (line 31) 61541* funswitch-loops: Optimize Options. (line 2537) 61542* funwind-tables: Code Gen Options. (line 149) 61543* fuse-cxa-atexit: C++ Dialect Options. 61544 (line 402) 61545* fuse-cxa-get-exception-ptr: C++ Dialect Options. 61546 (line 409) 61547* fuse-ld=bfd: Link Options. (line 74) 61548* fuse-ld=gold: Link Options. (line 77) 61549* fuse-ld=lld: Link Options. (line 80) 61550* fuse-linker-plugin: Optimize Options. (line 2047) 61551* fvar-tracking: Debugging Options. (line 151) 61552* fvar-tracking-assignments: Debugging Options. (line 161) 61553* fvar-tracking-assignments-toggle: Developer Options. (line 846) 61554* fvariable-expansion-in-unroller: Optimize Options. (line 1483) 61555* fvect-cost-model: Optimize Options. (line 1434) 61556* fverbose-asm: Code Gen Options. (line 262) 61557* fversion-loops-for-strides: Optimize Options. (line 2544) 61558* fvisibility: Code Gen Options. (line 524) 61559* fvisibility-inlines-hidden: C++ Dialect Options. 61560 (line 414) 61561* fvisibility-ms-compat: C++ Dialect Options. 61562 (line 442) 61563* fvpt: Optimize Options. (line 2466) 61564* fvtable-verify: Instrumentation Options. 61565 (line 679) 61566* fvtv-counts: Instrumentation Options. 61567 (line 715) 61568* fvtv-debug: Instrumentation Options. 61569 (line 702) 61570* fweak: C++ Dialect Options. 61571 (line 471) 61572* fweb: Optimize Options. (line 1795) 61573* fwhole-program: Optimize Options. (line 1806) 61574* fwide-exec-charset: Preprocessor Options. 61575 (line 275) 61576* fworking-directory: Preprocessor Options. 61577 (line 320) 61578* fwrapv: Code Gen Options. (line 99) 61579* fwrapv-pointer: Code Gen Options. (line 109) 61580* fwritable-relocated-rdata: x86 Windows Options. 61581 (line 53) 61582* fzero-initialized-in-bss: Optimize Options. (line 445) 61583* fzero-link: Objective-C and Objective-C++ Dialect Options. 61584 (line 140) 61585* g: Debugging Options. (line 25) 61586* G: ARC Options. (line 416) 61587* G <1>: M32R/D Options. (line 57) 61588* G <2>: MIPS Options. (line 460) 61589* G <3>: Nios II Options. (line 9) 61590* G <4>: RS/6000 and PowerPC Options. 61591 (line 713) 61592* G <5>: System V Options. (line 10) 61593* gas-loc-support: Debugging Options. (line 221) 61594* gas-locview-support: Debugging Options. (line 237) 61595* gcolumn-info: Debugging Options. (line 249) 61596* gdescribe-dies: Debugging Options. (line 179) 61597* gdwarf: Debugging Options. (line 45) 61598* gen-decls: Objective-C and Objective-C++ Dialect Options. 61599 (line 166) 61600* gfull: Darwin Options. (line 69) 61601* ggdb: Debugging Options. (line 38) 61602* ggnu-pubnames: Debugging Options. (line 187) 61603* ginline-points: Debugging Options. (line 308) 61604* ginternal-reset-location-views: Debugging Options. (line 297) 61605* gno-as-loc-support: Debugging Options. (line 233) 61606* gno-column-info: Debugging Options. (line 249) 61607* gno-inline-points: Debugging Options. (line 308) 61608* gno-internal-reset-location-views: Debugging Options. (line 297) 61609* gno-record-gcc-switches: Debugging Options. (line 202) 61610* gno-statement-frontiers: Debugging Options. (line 254) 61611* gno-strict-dwarf: Debugging Options. (line 217) 61612* gno-variable-location-views: Debugging Options. (line 265) 61613* gpubnames: Debugging Options. (line 184) 61614* grecord-gcc-switches: Debugging Options. (line 202) 61615* gsplit-dwarf: Debugging Options. (line 172) 61616* gstabs: Debugging Options. (line 63) 61617* gstabs+: Debugging Options. (line 71) 61618* gstatement-frontiers: Debugging Options. (line 254) 61619* gstrict-dwarf: Debugging Options. (line 211) 61620* gtoggle: Developer Options. (line 838) 61621* gused: Darwin Options. (line 64) 61622* gvariable-location-views: Debugging Options. (line 265) 61623* gvariable-location-views=incompat5: Debugging Options. (line 265) 61624* gvms: Debugging Options. (line 90) 61625* gxcoff: Debugging Options. (line 77) 61626* gxcoff+: Debugging Options. (line 82) 61627* gz: Debugging Options. (line 317) 61628* H: Preprocessor Options. 61629 (line 402) 61630* headerpad_max_install_names: Darwin Options. (line 196) 61631* help: Overall Options. (line 220) 61632* I: Directory Options. (line 13) 61633* I-: Directory Options. (line 65) 61634* idirafter: Directory Options. (line 13) 61635* iframework: Darwin Options. (line 57) 61636* imacros: Preprocessor Options. 61637 (line 57) 61638* image_base: Darwin Options. (line 196) 61639* imultilib: Directory Options. (line 98) 61640* include: Preprocessor Options. 61641 (line 46) 61642* init: Darwin Options. (line 196) 61643* install_name: Darwin Options. (line 196) 61644* iplugindir=: Directory Options. (line 113) 61645* iprefix: Directory Options. (line 80) 61646* iquote: Directory Options. (line 13) 61647* isysroot: Directory Options. (line 92) 61648* isystem: Directory Options. (line 13) 61649* iwithprefix: Directory Options. (line 86) 61650* iwithprefixbefore: Directory Options. (line 86) 61651* keep_private_externs: Darwin Options. (line 196) 61652* l: Link Options. (line 84) 61653* L: Directory Options. (line 118) 61654* lobjc: Link Options. (line 110) 61655* M: Preprocessor Options. 61656 (line 77) 61657* m: RS/6000 and PowerPC Options. 61658 (line 520) 61659* m1: SH Options. (line 9) 61660* m10: PDP-11 Options. (line 29) 61661* m128bit-long-double: x86 Options. (line 584) 61662* m16: x86 Options. (line 1425) 61663* m16-bit: CRIS Options. (line 64) 61664* m16-bit <1>: NDS32 Options. (line 51) 61665* m1reg-: Adapteva Epiphany Options. 61666 (line 131) 61667* m2: SH Options. (line 12) 61668* m210: MCore Options. (line 43) 61669* m2a: SH Options. (line 30) 61670* m2a-nofpu: SH Options. (line 18) 61671* m2a-single: SH Options. (line 26) 61672* m2a-single-only: SH Options. (line 22) 61673* m3: SH Options. (line 34) 61674* m31: S/390 and zSeries Options. 61675 (line 86) 61676* m32: Nvidia PTX Options. (line 10) 61677* m32 <1>: RS/6000 and PowerPC Options. 61678 (line 244) 61679* m32 <2>: SPARC Options. (line 315) 61680* m32 <3>: TILE-Gx Options. (line 23) 61681* m32 <4>: TILEPro Options. (line 13) 61682* m32 <5>: x86 Options. (line 1425) 61683* m32-bit: CRIS Options. (line 64) 61684* m32bit-doubles: RL78 Options. (line 73) 61685* m32bit-doubles <1>: RX Options. (line 10) 61686* m32r: M32R/D Options. (line 15) 61687* m32r2: M32R/D Options. (line 9) 61688* m32rx: M32R/D Options. (line 12) 61689* m340: MCore Options. (line 43) 61690* m3dnow: x86 Options. (line 807) 61691* m3dnowa: x86 Options. (line 808) 61692* m3e: SH Options. (line 37) 61693* m4: SH Options. (line 51) 61694* m4-100: SH Options. (line 54) 61695* m4-100-nofpu: SH Options. (line 57) 61696* m4-100-single: SH Options. (line 61) 61697* m4-100-single-only: SH Options. (line 65) 61698* m4-200: SH Options. (line 69) 61699* m4-200-nofpu: SH Options. (line 72) 61700* m4-200-single: SH Options. (line 76) 61701* m4-200-single-only: SH Options. (line 80) 61702* m4-300: SH Options. (line 84) 61703* m4-300-nofpu: SH Options. (line 87) 61704* m4-300-single: SH Options. (line 91) 61705* m4-300-single-only: SH Options. (line 95) 61706* m4-340: SH Options. (line 99) 61707* m4-500: SH Options. (line 102) 61708* m4-nofpu: SH Options. (line 40) 61709* m4-single: SH Options. (line 47) 61710* m4-single-only: SH Options. (line 43) 61711* m40: PDP-11 Options. (line 23) 61712* m45: PDP-11 Options. (line 26) 61713* m4a: SH Options. (line 118) 61714* m4a-nofpu: SH Options. (line 106) 61715* m4a-single: SH Options. (line 114) 61716* m4a-single-only: SH Options. (line 110) 61717* m4al: SH Options. (line 121) 61718* m4byte-functions: MCore Options. (line 27) 61719* m5200: M680x0 Options. (line 144) 61720* m5206e: M680x0 Options. (line 153) 61721* m528x: M680x0 Options. (line 157) 61722* m5307: M680x0 Options. (line 161) 61723* m5407: M680x0 Options. (line 165) 61724* m64: Nvidia PTX Options. (line 10) 61725* m64 <1>: RS/6000 and PowerPC Options. 61726 (line 244) 61727* m64 <2>: S/390 and zSeries Options. 61728 (line 86) 61729* m64 <3>: SPARC Options. (line 315) 61730* m64 <4>: TILE-Gx Options. (line 23) 61731* m64 <5>: x86 Options. (line 1425) 61732* m64bit-doubles: RL78 Options. (line 73) 61733* m64bit-doubles <1>: RX Options. (line 10) 61734* m68000: M680x0 Options. (line 93) 61735* m68010: M680x0 Options. (line 101) 61736* m68020: M680x0 Options. (line 107) 61737* m68020-40: M680x0 Options. (line 175) 61738* m68020-60: M680x0 Options. (line 184) 61739* m68030: M680x0 Options. (line 112) 61740* m68040: M680x0 Options. (line 117) 61741* m68060: M680x0 Options. (line 126) 61742* m68881: M680x0 Options. (line 194) 61743* m8-bit: CRIS Options. (line 64) 61744* m8bit-idiv: x86 Options. (line 1346) 61745* m8byte-align: V850 Options. (line 170) 61746* m96bit-long-double: x86 Options. (line 584) 61747* mA6: ARC Options. (line 23) 61748* mA7: ARC Options. (line 30) 61749* mabi: AArch64 Options. (line 9) 61750* mabi <1>: ARM Options. (line 9) 61751* mabi <2>: PRU Options. (line 28) 61752* mabi <3>: RISC-V Options. (line 17) 61753* mabi <4>: RS/6000 and PowerPC Options. 61754 (line 551) 61755* mabi <5>: x86 Options. (line 1039) 61756* mabi=32: MIPS Options. (line 156) 61757* mabi=64: MIPS Options. (line 156) 61758* mabi=eabi: MIPS Options. (line 156) 61759* mabi=elfv1: RS/6000 and PowerPC Options. 61760 (line 573) 61761* mabi=elfv2: RS/6000 and PowerPC Options. 61762 (line 579) 61763* mabi=gnu: MMIX Options. (line 20) 61764* mabi=ibmlongdouble: RS/6000 and PowerPC Options. 61765 (line 557) 61766* mabi=ieeelongdouble: RS/6000 and PowerPC Options. 61767 (line 565) 61768* mabi=mmixware: MMIX Options. (line 20) 61769* mabi=n32: MIPS Options. (line 156) 61770* mabi=o64: MIPS Options. (line 156) 61771* mabicalls: MIPS Options. (line 192) 61772* mabm: x86 Options. (line 810) 61773* mabort-on-noreturn: ARM Options. (line 750) 61774* mabs=2008: MIPS Options. (line 300) 61775* mabs=legacy: MIPS Options. (line 300) 61776* mabsdata: AVR Options. (line 163) 61777* mabsdiff: MeP Options. (line 7) 61778* mac0: PDP-11 Options. (line 16) 61779* macc-4: FRV Options. (line 139) 61780* macc-8: FRV Options. (line 143) 61781* maccumulate-args: AVR Options. (line 170) 61782* maccumulate-outgoing-args: SH Options. (line 314) 61783* maccumulate-outgoing-args <1>: x86 Options. (line 1083) 61784* maddress-mode=long: x86 Options. (line 1475) 61785* maddress-mode=short: x86 Options. (line 1480) 61786* mads: RS/6000 and PowerPC Options. 61787 (line 613) 61788* madx: x86 Options. (line 811) 61789* maes: x86 Options. (line 788) 61790* maix-struct-return: RS/6000 and PowerPC Options. 61791 (line 544) 61792* maix32: RS/6000 and PowerPC Options. 61793 (line 282) 61794* maix64: RS/6000 and PowerPC Options. 61795 (line 282) 61796* malign-300: H8/300 Options. (line 41) 61797* malign-call: ARC Options. (line 435) 61798* malign-data: RISC-V Options. (line 127) 61799* malign-data <1>: x86 Options. (line 624) 61800* malign-double: x86 Options. (line 569) 61801* malign-int: M680x0 Options. (line 261) 61802* malign-labels: FRV Options. (line 128) 61803* malign-loops: M32R/D Options. (line 73) 61804* malign-natural: RS/6000 and PowerPC Options. 61805 (line 320) 61806* malign-power: RS/6000 and PowerPC Options. 61807 (line 320) 61808* malign-stringops: x86 Options. (line 1219) 61809* mall-opts: MeP Options. (line 11) 61810* malloc-cc: FRV Options. (line 31) 61811* mallow-string-insns: RX Options. (line 150) 61812* mallregs: RL78 Options. (line 66) 61813* maltivec: RS/6000 and PowerPC Options. 61814 (line 136) 61815* mam33: MN10300 Options. (line 17) 61816* mam33-2: MN10300 Options. (line 24) 61817* mam34: MN10300 Options. (line 27) 61818* manchor: C-SKY Options. (line 110) 61819* mandroid: GNU/Linux Options. (line 26) 61820* mannotate-align: ARC Options. (line 382) 61821* mapcs: ARM Options. (line 21) 61822* mapcs-frame: ARM Options. (line 13) 61823* mapp-regs: SPARC Options. (line 10) 61824* mapp-regs <1>: V850 Options. (line 181) 61825* mARC600: ARC Options. (line 23) 61826* mARC601: ARC Options. (line 27) 61827* mARC700: ARC Options. (line 30) 61828* march: AArch64 Options. (line 165) 61829* march <1>: AMD GCN Options. (line 9) 61830* march <2>: ARM Options. (line 80) 61831* march <3>: C6X Options. (line 7) 61832* march <4>: CRIS Options. (line 10) 61833* march <5>: HPPA Options. (line 9) 61834* march <6>: HPPA Options. (line 162) 61835* march <7>: M680x0 Options. (line 12) 61836* march <8>: MIPS Options. (line 14) 61837* march <9>: NDS32 Options. (line 64) 61838* march <10>: Nios II Options. (line 94) 61839* march <11>: Nvidia PTX Options. (line 13) 61840* march <12>: RISC-V Options. (line 54) 61841* march <13>: S/390 and zSeries Options. 61842 (line 148) 61843* march <14>: x86 Options. (line 9) 61844* march=: C-SKY Options. (line 9) 61845* marclinux: ARC Options. (line 388) 61846* marclinux_prof: ARC Options. (line 395) 61847* margonaut: ARC Options. (line 593) 61848* marm: ARM Options. (line 822) 61849* mas100-syntax: RX Options. (line 76) 61850* masm-hex: MSP430 Options. (line 9) 61851* masm-syntax-unified: ARM Options. (line 921) 61852* masm=DIALECT: x86 Options. (line 518) 61853* matomic: ARC Options. (line 155) 61854* matomic-model=MODEL: SH Options. (line 193) 61855* mauto-litpools: Xtensa Options. (line 60) 61856* mauto-modify-reg: ARC Options. (line 438) 61857* mauto-pic: IA-64 Options. (line 50) 61858* maverage: MeP Options. (line 16) 61859* mavoid-indexed-addresses: RS/6000 and PowerPC Options. 61860 (line 359) 61861* mavx: x86 Options. (line 776) 61862* mavx2: x86 Options. (line 777) 61863* mavx256-split-unaligned-load: x86 Options. (line 1354) 61864* mavx256-split-unaligned-store: x86 Options. (line 1354) 61865* mavx5124fmaps: x86 Options. (line 838) 61866* mavx5124vnniw: x86 Options. (line 840) 61867* mavx512bf16: x86 Options. (line 827) 61868* mavx512bitalg: x86 Options. (line 832) 61869* mavx512bw: x86 Options. (line 783) 61870* mavx512cd: x86 Options. (line 781) 61871* mavx512dq: x86 Options. (line 784) 61872* mavx512er: x86 Options. (line 780) 61873* mavx512f: x86 Options. (line 778) 61874* mavx512ifma: x86 Options. (line 785) 61875* mavx512pf: x86 Options. (line 779) 61876* mavx512vbmi: x86 Options. (line 786) 61877* mavx512vbmi2: x86 Options. (line 826) 61878* mavx512vl: x86 Options. (line 782) 61879* mavx512vnni: x86 Options. (line 839) 61880* mavx512vp2intersect: x86 Options. (line 837) 61881* mavx512vpopcntdq: x86 Options. (line 836) 61882* max-vect-align: Adapteva Epiphany Options. 61883 (line 119) 61884* mb: SH Options. (line 126) 61885* mbackchain: S/390 and zSeries Options. 61886 (line 35) 61887* mbarrel-shift-enabled: LM32 Options. (line 9) 61888* mbarrel-shifter: ARC Options. (line 10) 61889* mbarrel_shifter: ARC Options. (line 613) 61890* mbase-addresses: MMIX Options. (line 53) 61891* mbased=: MeP Options. (line 20) 61892* mbbit-peephole: ARC Options. (line 441) 61893* mbe8: ARM Options. (line 72) 61894* mbig: RS/6000 and PowerPC Options. 61895 (line 439) 61896* mbig-endian: AArch64 Options. (line 20) 61897* mbig-endian <1>: ARC Options. (line 596) 61898* mbig-endian <2>: ARM Options. (line 67) 61899* mbig-endian <3>: C6X Options. (line 13) 61900* mbig-endian <4>: C-SKY Options. (line 28) 61901* mbig-endian <5>: eBPF Options. (line 22) 61902* mbig-endian <6>: IA-64 Options. (line 9) 61903* mbig-endian <7>: MCore Options. (line 39) 61904* mbig-endian <8>: MicroBlaze Options. (line 56) 61905* mbig-endian <9>: NDS32 Options. (line 9) 61906* mbig-endian <10>: RS/6000 and PowerPC Options. 61907 (line 439) 61908* mbig-endian <11>: TILE-Gx Options. (line 29) 61909* mbig-endian-data: RX Options. (line 42) 61910* mbig-switch: V850 Options. (line 176) 61911* mbigtable: SH Options. (line 141) 61912* mbionic: GNU/Linux Options. (line 22) 61913* mbit-align: RS/6000 and PowerPC Options. 61914 (line 391) 61915* mbit-ops: CR16 Options. (line 25) 61916* mbitfield: M680x0 Options. (line 231) 61917* mbitops: MeP Options. (line 26) 61918* mbitops <1>: SH Options. (line 145) 61919* mblock-compare-inline-limit: RS/6000 and PowerPC Options. 61920 (line 693) 61921* mblock-compare-inline-loop-limit: RS/6000 and PowerPC Options. 61922 (line 699) 61923* mblock-move-inline-limit: RS/6000 and PowerPC Options. 61924 (line 687) 61925* mbmi: x86 Options. (line 812) 61926* mbmi2: x86 Options. (line 813) 61927* mboard: OpenRISC Options. (line 9) 61928* mbranch-cost: Adapteva Epiphany Options. 61929 (line 18) 61930* mbranch-cost <1>: AVR Options. (line 185) 61931* mbranch-cost <2>: MIPS Options. (line 785) 61932* mbranch-cost <3>: RISC-V Options. (line 9) 61933* mbranch-cost=: C-SKY Options. (line 143) 61934* mbranch-cost=NUM: SH Options. (line 334) 61935* mbranch-cost=NUMBER: M32R/D Options. (line 82) 61936* mbranch-index: ARC Options. (line 329) 61937* mbranch-likely: MIPS Options. (line 792) 61938* mbranch-predict: MMIX Options. (line 48) 61939* mbranch-protection: AArch64 Options. (line 296) 61940* mbss-plt: RS/6000 and PowerPC Options. 61941 (line 160) 61942* mbuild-constants: DEC Alpha Options. (line 141) 61943* mbwx: DEC Alpha Options. (line 163) 61944* mbypass-cache: Nios II Options. (line 103) 61945* mc68000: M680x0 Options. (line 93) 61946* mc68020: M680x0 Options. (line 107) 61947* mc=: MeP Options. (line 31) 61948* mcache: C-SKY Options. (line 77) 61949* mcache-block-size: NDS32 Options. (line 60) 61950* mcache-volatile: Nios II Options. (line 109) 61951* mcall-eabi: RS/6000 and PowerPC Options. 61952 (line 514) 61953* mcall-freebsd: RS/6000 and PowerPC Options. 61954 (line 528) 61955* mcall-linux: RS/6000 and PowerPC Options. 61956 (line 524) 61957* mcall-ms2sysv-xlogues: x86 Options. (line 1059) 61958* mcall-netbsd: RS/6000 and PowerPC Options. 61959 (line 532) 61960* mcall-netbsd <1>: RS/6000 and PowerPC Options. 61961 (line 536) 61962* mcall-prologues: AVR Options. (line 190) 61963* mcall-sysv: RS/6000 and PowerPC Options. 61964 (line 506) 61965* mcall-sysv-eabi: RS/6000 and PowerPC Options. 61966 (line 514) 61967* mcall-sysv-noeabi: RS/6000 and PowerPC Options. 61968 (line 517) 61969* mcallee-super-interworking: ARM Options. (line 851) 61970* mcaller-copies: HPPA Options. (line 23) 61971* mcaller-super-interworking: ARM Options. (line 858) 61972* mcallgraph-data: MCore Options. (line 31) 61973* mcase-vector-pcrel: ARC Options. (line 450) 61974* mcbcond: SPARC Options. (line 260) 61975* mcbranch-force-delay-slot: SH Options. (line 349) 61976* mcc-init: CRIS Options. (line 42) 61977* mccrt: C-SKY Options. (line 139) 61978* mcfv4e: M680x0 Options. (line 169) 61979* mcheck-zero-division: MIPS Options. (line 570) 61980* mcix: DEC Alpha Options. (line 163) 61981* mcld: x86 Options. (line 892) 61982* mcldemote: x86 Options. (line 841) 61983* mclear-hwcap: Solaris 2 Options. (line 9) 61984* mclflushopt: x86 Options. (line 790) 61985* mclip: MeP Options. (line 35) 61986* mclwb: x86 Options. (line 791) 61987* mclzero: x86 Options. (line 824) 61988* mcmodel: NDS32 Options. (line 67) 61989* mcmodel <1>: SPARC Options. (line 320) 61990* mcmodel=kernel: x86 Options. (line 1459) 61991* mcmodel=large: AArch64 Options. (line 45) 61992* mcmodel=large <1>: RS/6000 and PowerPC Options. 61993 (line 130) 61994* mcmodel=large <2>: TILE-Gx Options. (line 14) 61995* mcmodel=large <3>: x86 Options. (line 1471) 61996* mcmodel=medany: RISC-V Options. (line 105) 61997* mcmodel=medium: RS/6000 and PowerPC Options. 61998 (line 125) 61999* mcmodel=medium <1>: x86 Options. (line 1464) 62000* mcmodel=medlow: RISC-V Options. (line 98) 62001* mcmodel=small: AArch64 Options. (line 39) 62002* mcmodel=small <1>: RS/6000 and PowerPC Options. 62003 (line 121) 62004* mcmodel=small <2>: TILE-Gx Options. (line 9) 62005* mcmodel=small <3>: x86 Options. (line 1453) 62006* mcmodel=tiny: AArch64 Options. (line 34) 62007* mcmov: NDS32 Options. (line 21) 62008* mcmov <1>: OpenRISC Options. (line 45) 62009* mcmove: Adapteva Epiphany Options. 62010 (line 23) 62011* mcmpb: RS/6000 and PowerPC Options. 62012 (line 25) 62013* mcmse: ARM Options. (line 950) 62014* mcode-density: ARC Options. (line 163) 62015* mcode-density-frame: ARC Options. (line 511) 62016* mcode-readable: MIPS Options. (line 530) 62017* mcode-region: MSP430 Options. (line 135) 62018* mcompact-branches=always: MIPS Options. (line 804) 62019* mcompact-branches=never: MIPS Options. (line 804) 62020* mcompact-branches=optimal: MIPS Options. (line 804) 62021* mcompact-casesi: ARC Options. (line 454) 62022* mcompat-align-parm: RS/6000 and PowerPC Options. 62023 (line 898) 62024* mcompress: FT32 Options. (line 26) 62025* mcond-exec: FRV Options. (line 187) 62026* mcond-move: FRV Options. (line 159) 62027* mconfig=: MeP Options. (line 39) 62028* mconsole: x86 Windows Options. 62029 (line 9) 62030* mconst-align: CRIS Options. (line 55) 62031* mconst16: Xtensa Options. (line 10) 62032* mconstant-gp: IA-64 Options. (line 46) 62033* mconstpool: C-SKY Options. (line 127) 62034* mcop: MeP Options. (line 48) 62035* mcop32: MeP Options. (line 53) 62036* mcop64: MeP Options. (line 56) 62037* mcorea: Blackfin Options. (line 154) 62038* mcoreb: Blackfin Options. (line 161) 62039* mcp: C-SKY Options. (line 74) 62040* mcpu: AArch64 Options. (line 243) 62041* mcpu <1>: ARC Options. (line 18) 62042* mcpu <2>: ARM Options. (line 620) 62043* mcpu <3>: CRIS Options. (line 10) 62044* mcpu <4>: DEC Alpha Options. (line 215) 62045* mcpu <5>: FRV Options. (line 258) 62046* mcpu <6>: M680x0 Options. (line 28) 62047* mcpu <7>: picoChip Options. (line 9) 62048* mcpu <8>: RL78 Options. (line 32) 62049* mcpu <9>: RS/6000 and PowerPC Options. 62050 (line 62) 62051* mcpu <10>: RX Options. (line 30) 62052* mcpu <11>: SPARC Options. (line 115) 62053* mcpu <12>: TILE-Gx Options. (line 18) 62054* mcpu <13>: TILEPro Options. (line 9) 62055* mcpu <14>: Visium Options. (line 33) 62056* mcpu <15>: x86 Options. (line 464) 62057* mcpu32: M680x0 Options. (line 135) 62058* mcpu=: Blackfin Options. (line 7) 62059* mcpu= <1>: C-SKY Options. (line 14) 62060* mcpu= <2>: M32C Options. (line 7) 62061* mcpu= <3>: MicroBlaze Options. (line 20) 62062* mcpu= <4>: MSP430 Options. (line 72) 62063* mcr16c: CR16 Options. (line 14) 62064* mcr16cplus: CR16 Options. (line 14) 62065* mcrc: MIPS Options. (line 416) 62066* mcrc32: x86 Options. (line 960) 62067* mcrypto: RS/6000 and PowerPC Options. 62068 (line 177) 62069* mcsync-anomaly: Blackfin Options. (line 57) 62070* mcsync-anomaly <1>: Blackfin Options. (line 63) 62071* mctor-dtor: NDS32 Options. (line 81) 62072* mcustom-fpu-cfg: Nios II Options. (line 259) 62073* mcustom-INSN: Nios II Options. (line 139) 62074* mcx16: x86 Options. (line 933) 62075* MD: Preprocessor Options. 62076 (line 169) 62077* mdalign: SH Options. (line 132) 62078* mdata-align: CRIS Options. (line 55) 62079* mdata-model: CR16 Options. (line 28) 62080* mdata-region: MSP430 Options. (line 135) 62081* mdc: MeP Options. (line 62) 62082* mdebug: M32R/D Options. (line 69) 62083* mdebug <1>: S/390 and zSeries Options. 62084 (line 144) 62085* mdebug <2>: Visium Options. (line 7) 62086* mdebug-main=PREFIX: VMS Options. (line 13) 62087* mdec-asm: PDP-11 Options. (line 46) 62088* mdisable-callt: V850 Options. (line 92) 62089* mdisable-fpregs: HPPA Options. (line 34) 62090* mdisable-indexing: HPPA Options. (line 40) 62091* mdiv: C-SKY Options. (line 93) 62092* mdiv <1>: M680x0 Options. (line 206) 62093* mdiv <2>: MCore Options. (line 15) 62094* mdiv <3>: MeP Options. (line 65) 62095* mdiv <4>: RISC-V Options. (line 49) 62096* mdiv-rem: ARC Options. (line 160) 62097* mdiv=STRATEGY: SH Options. (line 284) 62098* mdivide-breaks: MIPS Options. (line 576) 62099* mdivide-enabled: LM32 Options. (line 12) 62100* mdivide-traps: MIPS Options. (line 576) 62101* mdivsi3_libfunc=NAME: SH Options. (line 320) 62102* mdll: x86 Windows Options. 62103 (line 16) 62104* mdlmzb: RS/6000 and PowerPC Options. 62105 (line 384) 62106* mdmx: MIPS Options. (line 376) 62107* mdouble: AVR Options. (line 195) 62108* mdouble <1>: FRV Options. (line 48) 62109* mdouble-float: C-SKY Options. (line 42) 62110* mdouble-float <1>: MIPS Options. (line 288) 62111* mdouble-float <2>: OpenRISC Options. (line 33) 62112* mdpfp: ARC Options. (line 99) 62113* mdpfp-compact: ARC Options. (line 100) 62114* mdpfp-fast: ARC Options. (line 104) 62115* mdpfp_compact: ARC Options. (line 616) 62116* mdpfp_fast: ARC Options. (line 619) 62117* mdsp: C-SKY Options. (line 86) 62118* mdsp <1>: MIPS Options. (line 353) 62119* mdsp-packa: ARC Options. (line 335) 62120* mdspr2: MIPS Options. (line 359) 62121* mdsp_packa: ARC Options. (line 622) 62122* mdump-tune-features: x86 Options. (line 874) 62123* mdvbf: ARC Options. (line 340) 62124* mdwarf2-asm: IA-64 Options. (line 94) 62125* mdword: FRV Options. (line 40) 62126* mdword <1>: FRV Options. (line 44) 62127* mdynamic-no-pic: RS/6000 and PowerPC Options. 62128 (line 444) 62129* mea: ARC Options. (line 112) 62130* mEA: ARC Options. (line 625) 62131* meabi: RS/6000 and PowerPC Options. 62132 (line 632) 62133* mearly-cbranchsi: ARC Options. (line 476) 62134* mearly-stop-bits: IA-64 Options. (line 100) 62135* meb: MeP Options. (line 68) 62136* meb <1>: Moxie Options. (line 7) 62137* meb <2>: Nios II Options. (line 90) 62138* meb <3>: Score Options. (line 9) 62139* medsp: C-SKY Options. (line 87) 62140* mel: MeP Options. (line 71) 62141* mel <1>: Moxie Options. (line 11) 62142* mel <2>: Nios II Options. (line 90) 62143* mel <3>: Score Options. (line 12) 62144* melf: CRIS Options. (line 87) 62145* melf <1>: MMIX Options. (line 43) 62146* melrw: C-SKY Options. (line 60) 62147* memb: RS/6000 and PowerPC Options. 62148 (line 627) 62149* membedded-data: MIPS Options. (line 517) 62150* memregs=: M32C Options. (line 21) 62151* menqcmd: x86 Options. (line 835) 62152* mep: V850 Options. (line 16) 62153* mepsilon: MMIX Options. (line 15) 62154* mesa: S/390 and zSeries Options. 62155 (line 94) 62156* metrax100: CRIS Options. (line 27) 62157* metrax4: CRIS Options. (line 27) 62158* meva: MIPS Options. (line 403) 62159* mexpand-adddi: ARC Options. (line 479) 62160* mexplicit-relocs: DEC Alpha Options. (line 176) 62161* mexplicit-relocs <1>: MIPS Options. (line 561) 62162* mexr: H8/300 Options. (line 28) 62163* mexr <1>: H8/300 Options. (line 33) 62164* mext-perf: NDS32 Options. (line 27) 62165* mext-perf2: NDS32 Options. (line 33) 62166* mext-string: NDS32 Options. (line 39) 62167* mextern-sdata: MIPS Options. (line 480) 62168* MF: Preprocessor Options. 62169 (line 111) 62170* mf16c: x86 Options. (line 795) 62171* mfancy-math-387: x86 Options. (line 559) 62172* mfast-fp: Blackfin Options. (line 130) 62173* mfast-indirect-calls: HPPA Options. (line 52) 62174* mfast-sw-div: Nios II Options. (line 115) 62175* mfaster-structs: SPARC Options. (line 91) 62176* mfdiv: RISC-V Options. (line 42) 62177* mfdivdu: C-SKY Options. (line 48) 62178* mfdpic: ARM Options. (line 965) 62179* mfdpic <1>: FRV Options. (line 72) 62180* mfentry: x86 Options. (line 1293) 62181* mfentry-name: x86 Options. (line 1324) 62182* mfentry-section: x86 Options. (line 1328) 62183* mfix: DEC Alpha Options. (line 163) 62184* mfix-24k: MIPS Options. (line 641) 62185* mfix-and-continue: Darwin Options. (line 104) 62186* mfix-at697f: SPARC Options. (line 294) 62187* mfix-cmse-cve-2021-35465: ARM Options. (line 956) 62188* mfix-cortex-a53-835769: AArch64 Options. (line 105) 62189* mfix-cortex-a53-843419: AArch64 Options. (line 112) 62190* mfix-cortex-m3-ldrd: ARM Options. (line 892) 62191* mfix-gr712rc: SPARC Options. (line 307) 62192* mfix-r10000: MIPS Options. (line 663) 62193* mfix-r4000: MIPS Options. (line 647) 62194* mfix-r4400: MIPS Options. (line 657) 62195* mfix-r5900: MIPS Options. (line 674) 62196* mfix-rm7000: MIPS Options. (line 684) 62197* mfix-sb1: MIPS Options. (line 709) 62198* mfix-ut699: SPARC Options. (line 299) 62199* mfix-ut700: SPARC Options. (line 303) 62200* mfix-vr4120: MIPS Options. (line 689) 62201* mfix-vr4130: MIPS Options. (line 702) 62202* mfixed-cc: FRV Options. (line 35) 62203* mfixed-range: HPPA Options. (line 59) 62204* mfixed-range <1>: IA-64 Options. (line 105) 62205* mfixed-range <2>: SH Options. (line 327) 62206* mflat: SPARC Options. (line 22) 62207* mflip-mips16: MIPS Options. (line 128) 62208* mflip-thumb: ARM Options. (line 834) 62209* mfloat-abi: ARM Options. (line 41) 62210* mfloat-ieee: DEC Alpha Options. (line 171) 62211* mfloat-vax: DEC Alpha Options. (line 171) 62212* mfloat128: RS/6000 and PowerPC Options. 62213 (line 214) 62214* mfloat128-hardware: RS/6000 and PowerPC Options. 62215 (line 235) 62216* mflush-func: MIPS Options. (line 776) 62217* mflush-func=NAME: M32R/D Options. (line 93) 62218* mflush-trap=NUMBER: M32R/D Options. (line 86) 62219* mfma: x86 Options. (line 796) 62220* mfma4: x86 Options. (line 799) 62221* mfmaf: SPARC Options. (line 267) 62222* mfmovd: SH Options. (line 148) 62223* mforce-indirect-call: x86 Options. (line 1048) 62224* mforce-no-pic: Xtensa Options. (line 41) 62225* mfp-exceptions: MIPS Options. (line 824) 62226* mfp-mode: Adapteva Epiphany Options. 62227 (line 71) 62228* mfp-reg: DEC Alpha Options. (line 25) 62229* mfp-ret-in-387: x86 Options. (line 549) 62230* mfp-rounding-mode: DEC Alpha Options. (line 85) 62231* mfp-trap-mode: DEC Alpha Options. (line 63) 62232* mfp16-format: ARM Options. (line 728) 62233* mfp32: MIPS Options. (line 258) 62234* mfp64: MIPS Options. (line 261) 62235* mfpmath: Optimize Options. (line 2226) 62236* mfpmath <1>: x86 Options. (line 467) 62237* mfpr-32: FRV Options. (line 15) 62238* mfpr-64: FRV Options. (line 19) 62239* mfprnd: RS/6000 and PowerPC Options. 62240 (line 25) 62241* mfpu: ARC Options. (line 231) 62242* mfpu <1>: ARM Options. (line 700) 62243* mfpu <2>: PDP-11 Options. (line 9) 62244* mfpu <3>: SPARC Options. (line 34) 62245* mfpu <4>: Visium Options. (line 19) 62246* mfpu=: C-SKY Options. (line 53) 62247* mfpxx: MIPS Options. (line 264) 62248* mfract-convert-truncate: AVR Options. (line 286) 62249* mframe-header-opt: MIPS Options. (line 885) 62250* mfriz: RS/6000 and PowerPC Options. 62251 (line 869) 62252* mfsca: SH Options. (line 365) 62253* mfsgsbase: x86 Options. (line 792) 62254* mfsmuld: SPARC Options. (line 274) 62255* mfsrra: SH Options. (line 374) 62256* mft32b: FT32 Options. (line 23) 62257* mfull-regs: NDS32 Options. (line 18) 62258* mfull-toc: RS/6000 and PowerPC Options. 62259 (line 255) 62260* mfunction-return: x86 Options. (line 1396) 62261* mfused-madd: IA-64 Options. (line 88) 62262* mfused-madd <1>: MIPS Options. (line 624) 62263* mfused-madd <2>: RS/6000 and PowerPC Options. 62264 (line 368) 62265* mfused-madd <3>: S/390 and zSeries Options. 62266 (line 182) 62267* mfused-madd <4>: SH Options. (line 356) 62268* mfused-madd <5>: Xtensa Options. (line 19) 62269* mfxsr: x86 Options. (line 815) 62270* MG: Preprocessor Options. 62271 (line 122) 62272* mg: VAX Options. (line 17) 62273* mg10: RL78 Options. (line 62) 62274* mg13: RL78 Options. (line 62) 62275* mg14: RL78 Options. (line 62) 62276* mgas: HPPA Options. (line 75) 62277* mgas-isr-prologues: AVR Options. (line 203) 62278* mgcc-abi: V850 Options. (line 148) 62279* mgeneral-regs-only: AArch64 Options. (line 24) 62280* mgeneral-regs-only <1>: ARM Options. (line 57) 62281* mgeneral-regs-only <2>: x86 Options. (line 1372) 62282* mgfni: x86 Options. (line 828) 62283* mghs: V850 Options. (line 127) 62284* mginv: MIPS Options. (line 421) 62285* mglibc: GNU/Linux Options. (line 9) 62286* mgnu: VAX Options. (line 13) 62287* mgnu-as: IA-64 Options. (line 18) 62288* mgnu-asm: PDP-11 Options. (line 49) 62289* mgnu-attribute: RS/6000 and PowerPC Options. 62290 (line 586) 62291* mgnu-ld: HPPA Options. (line 111) 62292* mgnu-ld <1>: IA-64 Options. (line 23) 62293* mgomp: Nvidia PTX Options. (line 53) 62294* mgotplt: CRIS Options. (line 81) 62295* mgp32: MIPS Options. (line 252) 62296* mgp64: MIPS Options. (line 255) 62297* mgpopt: MIPS Options. (line 502) 62298* mgpopt <1>: Nios II Options. (line 16) 62299* mgpr-32: FRV Options. (line 7) 62300* mgpr-64: FRV Options. (line 11) 62301* mgprel-ro: FRV Options. (line 99) 62302* mgprel-sec: Nios II Options. (line 65) 62303* mh: H8/300 Options. (line 14) 62304* mhal: Nios II Options. (line 304) 62305* mhalf-reg-file: Adapteva Epiphany Options. 62306 (line 9) 62307* mhard-dfp: RS/6000 and PowerPC Options. 62308 (line 25) 62309* mhard-dfp <1>: S/390 and zSeries Options. 62310 (line 20) 62311* mhard-div: OpenRISC Options. (line 19) 62312* mhard-float: C-SKY Options. (line 35) 62313* mhard-float <1>: FRV Options. (line 23) 62314* mhard-float <2>: M680x0 Options. (line 194) 62315* mhard-float <3>: MicroBlaze Options. (line 10) 62316* mhard-float <4>: MIPS Options. (line 267) 62317* mhard-float <5>: OpenRISC Options. (line 29) 62318* mhard-float <6>: RS/6000 and PowerPC Options. 62319 (line 332) 62320* mhard-float <7>: S/390 and zSeries Options. 62321 (line 11) 62322* mhard-float <8>: SPARC Options. (line 34) 62323* mhard-float <9>: V850 Options. (line 113) 62324* mhard-float <10>: Visium Options. (line 19) 62325* mhard-float <11>: x86 Options. (line 532) 62326* mhard-mul: OpenRISC Options. (line 24) 62327* mhard-quad-float: SPARC Options. (line 55) 62328* mharden-sls: AArch64 Options. (line 309) 62329* mhardlit: MCore Options. (line 10) 62330* mhigh-registers: C-SKY Options. (line 104) 62331* mhle: x86 Options. (line 821) 62332* mhotpatch: S/390 and zSeries Options. 62333 (line 217) 62334* mhp-ld: HPPA Options. (line 123) 62335* mhtm: RS/6000 and PowerPC Options. 62336 (line 183) 62337* mhtm <1>: S/390 and zSeries Options. 62338 (line 104) 62339* mhw-div: Nios II Options. (line 124) 62340* mhw-mul: Nios II Options. (line 124) 62341* mhw-mulx: Nios II Options. (line 124) 62342* mhwmult=: MSP430 Options. (line 93) 62343* miamcu: x86 Options. (line 1425) 62344* micplb: Blackfin Options. (line 175) 62345* mid-shared-library: Blackfin Options. (line 78) 62346* mid-shared-library <1>: Blackfin Options. (line 85) 62347* mieee: DEC Alpha Options. (line 39) 62348* mieee <1>: SH Options. (line 165) 62349* mieee-conformant: DEC Alpha Options. (line 134) 62350* mieee-fp: x86 Options. (line 526) 62351* mieee-with-inexact: DEC Alpha Options. (line 52) 62352* milp32: IA-64 Options. (line 121) 62353* mimadd: MIPS Options. (line 617) 62354* mimpure-text: Solaris 2 Options. (line 15) 62355* mincoming-stack-boundary: x86 Options. (line 742) 62356* mindexed-loads: ARC Options. (line 483) 62357* mindirect-branch: x86 Options. (line 1377) 62358* mindirect-branch-register: x86 Options. (line 1415) 62359* minline-all-stringops: x86 Options. (line 1224) 62360* minline-float-divide-max-throughput: IA-64 Options. (line 58) 62361* minline-float-divide-min-latency: IA-64 Options. (line 54) 62362* minline-ic_invalidate: SH Options. (line 174) 62363* minline-int-divide: IA-64 Options. (line 73) 62364* minline-int-divide-max-throughput: IA-64 Options. (line 69) 62365* minline-int-divide-min-latency: IA-64 Options. (line 65) 62366* minline-plt: Blackfin Options. (line 135) 62367* minline-plt <1>: FRV Options. (line 81) 62368* minline-sqrt-max-throughput: IA-64 Options. (line 80) 62369* minline-sqrt-min-latency: IA-64 Options. (line 76) 62370* minline-stringops-dynamically: x86 Options. (line 1232) 62371* minrt: MSP430 Options. (line 115) 62372* minrt <1>: PRU Options. (line 9) 62373* minsert-sched-nops: RS/6000 and PowerPC Options. 62374 (line 484) 62375* minstrument-return: x86 Options. (line 1312) 62376* mint-register: RX Options. (line 100) 62377* mint16: PDP-11 Options. (line 33) 62378* mint32: CR16 Options. (line 22) 62379* mint32 <1>: H8/300 Options. (line 38) 62380* mint32 <2>: PDP-11 Options. (line 37) 62381* mint8: AVR Options. (line 213) 62382* minterlink-compressed: MIPS Options. (line 135) 62383* minterlink-mips16: MIPS Options. (line 147) 62384* mio-volatile: MeP Options. (line 74) 62385* mips1: MIPS Options. (line 80) 62386* mips16: MIPS Options. (line 120) 62387* mips2: MIPS Options. (line 83) 62388* mips3: MIPS Options. (line 86) 62389* mips32: MIPS Options. (line 92) 62390* mips32r3: MIPS Options. (line 95) 62391* mips32r5: MIPS Options. (line 98) 62392* mips32r6: MIPS Options. (line 101) 62393* mips3d: MIPS Options. (line 382) 62394* mips4: MIPS Options. (line 89) 62395* mips64: MIPS Options. (line 104) 62396* mips64r2: MIPS Options. (line 107) 62397* mips64r3: MIPS Options. (line 110) 62398* mips64r5: MIPS Options. (line 113) 62399* mips64r6: MIPS Options. (line 116) 62400* mirq-ctrl-saved: ARC Options. (line 296) 62401* misel: RS/6000 and PowerPC Options. 62402 (line 166) 62403* misize: ARC Options. (line 379) 62404* misize <1>: SH Options. (line 186) 62405* misr-vector-size: NDS32 Options. (line 57) 62406* missue-rate=NUMBER: M32R/D Options. (line 79) 62407* mistack: C-SKY Options. (line 65) 62408* mivc2: MeP Options. (line 59) 62409* mjli-alawys: ARC Options. (line 14) 62410* mjsr: RX Options. (line 169) 62411* mjump-in-delay: HPPA Options. (line 30) 62412* mkernel: Darwin Options. (line 82) 62413* mkernel <1>: eBPF Options. (line 13) 62414* mknuthdiv: MMIX Options. (line 32) 62415* ml: MeP Options. (line 78) 62416* ml <1>: SH Options. (line 129) 62417* mlarge: MSP430 Options. (line 82) 62418* mlarge-data: DEC Alpha Options. (line 187) 62419* mlarge-data-threshold: x86 Options. (line 631) 62420* mlarge-text: DEC Alpha Options. (line 205) 62421* mleadz: MeP Options. (line 81) 62422* mleaf-id-shared-library: Blackfin Options. (line 89) 62423* mleaf-id-shared-library <1>: Blackfin Options. (line 95) 62424* mlibfuncs: MMIX Options. (line 10) 62425* mlibrary-pic: FRV Options. (line 135) 62426* mlinked-fp: FRV Options. (line 116) 62427* mlinker-opt: HPPA Options. (line 85) 62428* mlinux: CRIS Options. (line 91) 62429* mlittle: RS/6000 and PowerPC Options. 62430 (line 433) 62431* mlittle-endian: AArch64 Options. (line 30) 62432* mlittle-endian <1>: ARC Options. (line 605) 62433* mlittle-endian <2>: ARM Options. (line 63) 62434* mlittle-endian <3>: C6X Options. (line 16) 62435* mlittle-endian <4>: C-SKY Options. (line 30) 62436* mlittle-endian <5>: eBPF Options. (line 25) 62437* mlittle-endian <6>: IA-64 Options. (line 13) 62438* mlittle-endian <7>: MCore Options. (line 39) 62439* mlittle-endian <8>: MicroBlaze Options. (line 59) 62440* mlittle-endian <9>: NDS32 Options. (line 12) 62441* mlittle-endian <10>: RS/6000 and PowerPC Options. 62442 (line 433) 62443* mlittle-endian <11>: TILE-Gx Options. (line 29) 62444* mlittle-endian-data: RX Options. (line 42) 62445* mliw: MN10300 Options. (line 54) 62446* mll64: ARC Options. (line 167) 62447* mllsc: MIPS Options. (line 339) 62448* mload-store-pairs: MIPS Options. (line 590) 62449* mlocal-sdata: MIPS Options. (line 468) 62450* mlock: ARC Options. (line 345) 62451* mlong-calls: Adapteva Epiphany Options. 62452 (line 55) 62453* mlong-calls <1>: ARC Options. (line 404) 62454* mlong-calls <2>: ARM Options. (line 755) 62455* mlong-calls <3>: Blackfin Options. (line 118) 62456* mlong-calls <4>: FRV Options. (line 122) 62457* mlong-calls <5>: HPPA Options. (line 136) 62458* mlong-calls <6>: MIPS Options. (line 603) 62459* mlong-calls <7>: V850 Options. (line 10) 62460* mlong-double: AVR Options. (line 195) 62461* mlong-double-128: S/390 and zSeries Options. 62462 (line 29) 62463* mlong-double-128 <1>: x86 Options. (line 610) 62464* mlong-double-64: S/390 and zSeries Options. 62465 (line 29) 62466* mlong-double-64 <1>: x86 Options. (line 610) 62467* mlong-double-80: x86 Options. (line 610) 62468* mlong-jump-table-offsets: M680x0 Options. (line 339) 62469* mlong-jumps: V850 Options. (line 108) 62470* mlong-load-store: HPPA Options. (line 66) 62471* mlong32: MIPS Options. (line 443) 62472* mlong64: MIPS Options. (line 438) 62473* mlongcall: RS/6000 and PowerPC Options. 62474 (line 727) 62475* mlongcalls: Xtensa Options. (line 87) 62476* mloongson-ext: MIPS Options. (line 430) 62477* mloongson-ext2: MIPS Options. (line 434) 62478* mloongson-mmi: MIPS Options. (line 425) 62479* mloop: PRU Options. (line 25) 62480* mloop <1>: V850 Options. (line 121) 62481* mlow-precision-div: AArch64 Options. (line 135) 62482* mlow-precision-recip-sqrt: AArch64 Options. (line 118) 62483* mlow-precision-sqrt: AArch64 Options. (line 126) 62484* mlow64k: Blackfin Options. (line 67) 62485* mlp64: IA-64 Options. (line 121) 62486* mlpc-width: ARC Options. (line 313) 62487* mlra: ARC Options. (line 488) 62488* mlra <1>: FT32 Options. (line 16) 62489* mlra <2>: PDP-11 Options. (line 52) 62490* mlra <3>: SPARC Options. (line 111) 62491* mlra-priority-compact: ARC Options. (line 496) 62492* mlra-priority-noncompact: ARC Options. (line 499) 62493* mlra-priority-none: ARC Options. (line 493) 62494* mlwp: x86 Options. (line 806) 62495* mlxc1-sxc1: MIPS Options. (line 895) 62496* mlzcnt: x86 Options. (line 814) 62497* MM: Preprocessor Options. 62498 (line 102) 62499* mm: MeP Options. (line 84) 62500* mmac: CR16 Options. (line 9) 62501* mmac <1>: Score Options. (line 21) 62502* mmac-24: ARC Options. (line 354) 62503* mmac-d16: ARC Options. (line 350) 62504* mmac_24: ARC Options. (line 628) 62505* mmac_d16: ARC Options. (line 631) 62506* mmad: MIPS Options. (line 612) 62507* mmadd4: MIPS Options. (line 900) 62508* mmain-is-OS_task: AVR Options. (line 219) 62509* mmainkernel: Nvidia PTX Options. (line 18) 62510* mmalloc64: VMS Options. (line 17) 62511* mmanual-endbr: x86 Options. (line 1053) 62512* mmax: DEC Alpha Options. (line 163) 62513* mmax-constant-size: RX Options. (line 82) 62514* mmax-stack-frame: CRIS Options. (line 23) 62515* mmcount-ra-address: MIPS Options. (line 872) 62516* mmcu: AVR Options. (line 9) 62517* mmcu <1>: MIPS Options. (line 399) 62518* mmcu <2>: PRU Options. (line 17) 62519* mmcu=: MSP430 Options. (line 14) 62520* MMD: Preprocessor Options. 62521 (line 185) 62522* mmedia: FRV Options. (line 56) 62523* mmedium-calls: ARC Options. (line 408) 62524* mmemcpy: MicroBlaze Options. (line 13) 62525* mmemcpy <1>: MIPS Options. (line 597) 62526* mmemcpy-strategy=STRATEGY: x86 Options. (line 1254) 62527* mmemory-latency: DEC Alpha Options. (line 268) 62528* mmemory-model: SPARC Options. (line 348) 62529* mmemset-strategy=STRATEGY: x86 Options. (line 1266) 62530* mmfcrf: RS/6000 and PowerPC Options. 62531 (line 25) 62532* mmicromips: MIPS Options. (line 387) 62533* mmillicode: ARC Options. (line 502) 62534* mminimal-toc: RS/6000 and PowerPC Options. 62535 (line 255) 62536* mminmax: MeP Options. (line 87) 62537* mmixed-code: ARC Options. (line 516) 62538* mmma: RS/6000 and PowerPC Options. 62539 (line 943) 62540* mmmx: x86 Options. (line 767) 62541* mmodel=large: M32R/D Options. (line 33) 62542* mmodel=medium: M32R/D Options. (line 27) 62543* mmodel=small: M32R/D Options. (line 18) 62544* mmovbe: x86 Options. (line 952) 62545* mmovdir64b: x86 Options. (line 834) 62546* mmovdiri: x86 Options. (line 833) 62547* mmp: C-SKY Options. (line 71) 62548* mmpy: ARC Options. (line 117) 62549* mmpy-option: ARC Options. (line 173) 62550* mms-bitfields: x86 Options. (line 1099) 62551* mmt: MIPS Options. (line 395) 62552* mmul: RL78 Options. (line 15) 62553* mmul-bug-workaround: CRIS Options. (line 32) 62554* mmul.x: Moxie Options. (line 14) 62555* mmul32x16: ARC Options. (line 121) 62556* mmul64: ARC Options. (line 124) 62557* mmuladd: FRV Options. (line 64) 62558* mmulhw: RS/6000 and PowerPC Options. 62559 (line 377) 62560* mmult: MeP Options. (line 90) 62561* mmult-bug: MN10300 Options. (line 9) 62562* mmultcost: ARC Options. (line 578) 62563* mmulti-cond-exec: FRV Options. (line 215) 62564* mmulticore: Blackfin Options. (line 139) 62565* mmultiple: RS/6000 and PowerPC Options. 62566 (line 338) 62567* mmultiple-stld: C-SKY Options. (line 121) 62568* mmusl: GNU/Linux Options. (line 18) 62569* mmvcle: S/390 and zSeries Options. 62570 (line 138) 62571* mmvme: RS/6000 and PowerPC Options. 62572 (line 608) 62573* mmwaitx: x86 Options. (line 823) 62574* mn: H8/300 Options. (line 20) 62575* mn-flash: AVR Options. (line 224) 62576* mnan=2008: MIPS Options. (line 320) 62577* mnan=legacy: MIPS Options. (line 320) 62578* mneon-for-64bits: ARM Options. (line 912) 62579* mnested-cond-exec: FRV Options. (line 230) 62580* mnewlib: OpenRISC Options. (line 13) 62581* mnhwloop: Score Options. (line 15) 62582* mno-16-bit: NDS32 Options. (line 54) 62583* mno-4byte-functions: MCore Options. (line 27) 62584* mno-8byte-align: V850 Options. (line 170) 62585* mno-abicalls: MIPS Options. (line 192) 62586* mno-ac0: PDP-11 Options. (line 20) 62587* mno-align-double: x86 Options. (line 569) 62588* mno-align-int: M680x0 Options. (line 261) 62589* mno-align-loops: M32R/D Options. (line 76) 62590* mno-align-stringops: x86 Options. (line 1219) 62591* mno-allow-string-insns: RX Options. (line 150) 62592* mno-altivec: RS/6000 and PowerPC Options. 62593 (line 136) 62594* mno-am33: MN10300 Options. (line 20) 62595* mno-app-regs: SPARC Options. (line 10) 62596* mno-app-regs <1>: V850 Options. (line 185) 62597* mno-as100-syntax: RX Options. (line 76) 62598* mno-auto-litpools: Xtensa Options. (line 60) 62599* mno-avoid-indexed-addresses: RS/6000 and PowerPC Options. 62600 (line 359) 62601* mno-backchain: S/390 and zSeries Options. 62602 (line 35) 62603* mno-base-addresses: MMIX Options. (line 53) 62604* mno-bit-align: RS/6000 and PowerPC Options. 62605 (line 391) 62606* mno-bitfield: M680x0 Options. (line 227) 62607* mno-branch-likely: MIPS Options. (line 792) 62608* mno-branch-predict: MMIX Options. (line 48) 62609* mno-brcc: ARC Options. (line 444) 62610* mno-bwx: DEC Alpha Options. (line 163) 62611* mno-bypass-cache: Nios II Options. (line 103) 62612* mno-cache-volatile: Nios II Options. (line 109) 62613* mno-call-ms2sysv-xlogues: x86 Options. (line 1059) 62614* mno-callgraph-data: MCore Options. (line 31) 62615* mno-cbcond: SPARC Options. (line 260) 62616* mno-check-zero-division: MIPS Options. (line 570) 62617* mno-cix: DEC Alpha Options. (line 163) 62618* mno-clearbss: MicroBlaze Options. (line 16) 62619* mno-cmov: NDS32 Options. (line 24) 62620* mno-cmpb: RS/6000 and PowerPC Options. 62621 (line 25) 62622* mno-cond-exec: ARC Options. (line 458) 62623* mno-cond-exec <1>: FRV Options. (line 194) 62624* mno-cond-move: FRV Options. (line 166) 62625* mno-const-align: CRIS Options. (line 55) 62626* mno-const16: Xtensa Options. (line 10) 62627* mno-crc: MIPS Options. (line 416) 62628* mno-crt0: MN10300 Options. (line 43) 62629* mno-crt0 <1>: Moxie Options. (line 18) 62630* mno-crypto: RS/6000 and PowerPC Options. 62631 (line 177) 62632* mno-csync-anomaly: Blackfin Options. (line 63) 62633* mno-custom-INSN: Nios II Options. (line 139) 62634* mno-data-align: CRIS Options. (line 55) 62635* mno-debug: S/390 and zSeries Options. 62636 (line 144) 62637* mno-default: x86 Options. (line 888) 62638* mno-disable-callt: V850 Options. (line 92) 62639* mno-div: M680x0 Options. (line 206) 62640* mno-div <1>: MCore Options. (line 15) 62641* mno-dlmzb: RS/6000 and PowerPC Options. 62642 (line 384) 62643* mno-double: FRV Options. (line 52) 62644* mno-dpfp-lrsr: ARC Options. (line 108) 62645* mno-dsp: MIPS Options. (line 353) 62646* mno-dspr2: MIPS Options. (line 359) 62647* mno-dwarf2-asm: IA-64 Options. (line 94) 62648* mno-dword: FRV Options. (line 44) 62649* mno-eabi: RS/6000 and PowerPC Options. 62650 (line 632) 62651* mno-early-stop-bits: IA-64 Options. (line 100) 62652* mno-eflags: FRV Options. (line 155) 62653* mno-embedded-data: MIPS Options. (line 517) 62654* mno-ep: V850 Options. (line 16) 62655* mno-epsilon: MMIX Options. (line 15) 62656* mno-eva: MIPS Options. (line 403) 62657* mno-explicit-relocs: DEC Alpha Options. (line 176) 62658* mno-explicit-relocs <1>: MIPS Options. (line 561) 62659* mno-exr: H8/300 Options. (line 33) 62660* mno-ext-perf: NDS32 Options. (line 30) 62661* mno-ext-perf2: NDS32 Options. (line 36) 62662* mno-ext-string: NDS32 Options. (line 42) 62663* mno-extern-sdata: MIPS Options. (line 480) 62664* mno-fancy-math-387: x86 Options. (line 559) 62665* mno-fast-sw-div: Nios II Options. (line 115) 62666* mno-faster-structs: SPARC Options. (line 91) 62667* mno-fdpic: ARM Options. (line 965) 62668* mno-fix: DEC Alpha Options. (line 163) 62669* mno-fix-24k: MIPS Options. (line 641) 62670* mno-fix-cortex-a53-835769: AArch64 Options. (line 105) 62671* mno-fix-cortex-a53-843419: AArch64 Options. (line 112) 62672* mno-fix-r10000: MIPS Options. (line 663) 62673* mno-fix-r4000: MIPS Options. (line 647) 62674* mno-fix-r4400: MIPS Options. (line 657) 62675* mno-flat: SPARC Options. (line 22) 62676* mno-float: MIPS Options. (line 274) 62677* mno-float128: RS/6000 and PowerPC Options. 62678 (line 214) 62679* mno-float128-hardware: RS/6000 and PowerPC Options. 62680 (line 235) 62681* mno-flush-func: M32R/D Options. (line 98) 62682* mno-flush-trap: M32R/D Options. (line 90) 62683* mno-fmaf: SPARC Options. (line 267) 62684* mno-fp-in-toc: RS/6000 and PowerPC Options. 62685 (line 255) 62686* mno-fp-regs: DEC Alpha Options. (line 25) 62687* mno-fp-ret-in-387: x86 Options. (line 549) 62688* mno-fprnd: RS/6000 and PowerPC Options. 62689 (line 25) 62690* mno-fpu: SPARC Options. (line 39) 62691* mno-fpu <1>: Visium Options. (line 24) 62692* mno-fsca: SH Options. (line 365) 62693* mno-fsmuld: SPARC Options. (line 274) 62694* mno-fsrra: SH Options. (line 374) 62695* mno-fused-madd: IA-64 Options. (line 88) 62696* mno-fused-madd <1>: MIPS Options. (line 624) 62697* mno-fused-madd <2>: RS/6000 and PowerPC Options. 62698 (line 368) 62699* mno-fused-madd <3>: S/390 and zSeries Options. 62700 (line 182) 62701* mno-fused-madd <4>: SH Options. (line 356) 62702* mno-fused-madd <5>: Xtensa Options. (line 19) 62703* mno-ginv: MIPS Options. (line 421) 62704* mno-gnu-as: IA-64 Options. (line 18) 62705* mno-gnu-attribute: RS/6000 and PowerPC Options. 62706 (line 586) 62707* mno-gnu-ld: IA-64 Options. (line 23) 62708* mno-gotplt: CRIS Options. (line 81) 62709* mno-gpopt: MIPS Options. (line 502) 62710* mno-gpopt <1>: Nios II Options. (line 16) 62711* mno-hard-dfp: RS/6000 and PowerPC Options. 62712 (line 25) 62713* mno-hard-dfp <1>: S/390 and zSeries Options. 62714 (line 20) 62715* mno-hardlit: MCore Options. (line 10) 62716* mno-htm: RS/6000 and PowerPC Options. 62717 (line 183) 62718* mno-htm <1>: S/390 and zSeries Options. 62719 (line 104) 62720* mno-hw-div: Nios II Options. (line 124) 62721* mno-hw-mul: Nios II Options. (line 124) 62722* mno-hw-mulx: Nios II Options. (line 124) 62723* mno-id-shared-library: Blackfin Options. (line 85) 62724* mno-ieee: SH Options. (line 165) 62725* mno-ieee-fp: x86 Options. (line 526) 62726* mno-imadd: MIPS Options. (line 617) 62727* mno-inline-float-divide: IA-64 Options. (line 62) 62728* mno-inline-int-divide: IA-64 Options. (line 73) 62729* mno-inline-sqrt: IA-64 Options. (line 84) 62730* mno-int16: PDP-11 Options. (line 37) 62731* mno-int32: PDP-11 Options. (line 33) 62732* mno-interlink-compressed: MIPS Options. (line 135) 62733* mno-interlink-mips16: MIPS Options. (line 147) 62734* mno-interrupts: AVR Options. (line 227) 62735* mno-isel: RS/6000 and PowerPC Options. 62736 (line 166) 62737* mno-jsr: RX Options. (line 169) 62738* mno-knuthdiv: MMIX Options. (line 32) 62739* mno-leaf-id-shared-library: Blackfin Options. (line 95) 62740* mno-libfuncs: MMIX Options. (line 10) 62741* mno-liw: MN10300 Options. (line 59) 62742* mno-llsc: MIPS Options. (line 339) 62743* mno-load-store-pairs: MIPS Options. (line 590) 62744* mno-local-sdata: MIPS Options. (line 468) 62745* mno-long-calls: ARM Options. (line 755) 62746* mno-long-calls <1>: Blackfin Options. (line 118) 62747* mno-long-calls <2>: HPPA Options. (line 136) 62748* mno-long-calls <3>: MIPS Options. (line 603) 62749* mno-long-calls <4>: V850 Options. (line 10) 62750* mno-long-jumps: V850 Options. (line 108) 62751* mno-longcall: RS/6000 and PowerPC Options. 62752 (line 727) 62753* mno-longcalls: Xtensa Options. (line 87) 62754* mno-loongson-ext: MIPS Options. (line 430) 62755* mno-loongson-ext2: MIPS Options. (line 434) 62756* mno-loongson-mmi: MIPS Options. (line 425) 62757* mno-low-precision-div: AArch64 Options. (line 135) 62758* mno-low-precision-recip-sqrt: AArch64 Options. (line 118) 62759* mno-low-precision-sqrt: AArch64 Options. (line 126) 62760* mno-low64k: Blackfin Options. (line 71) 62761* mno-lra: SPARC Options. (line 111) 62762* mno-lsim: FR30 Options. (line 14) 62763* mno-lsim <1>: MCore Options. (line 46) 62764* mno-mad: MIPS Options. (line 612) 62765* mno-max: DEC Alpha Options. (line 163) 62766* mno-mcount-ra-address: MIPS Options. (line 872) 62767* mno-mcu: MIPS Options. (line 399) 62768* mno-mdmx: MIPS Options. (line 376) 62769* mno-media: FRV Options. (line 60) 62770* mno-memcpy: MIPS Options. (line 597) 62771* mno-mfcrf: RS/6000 and PowerPC Options. 62772 (line 25) 62773* mno-mips16: MIPS Options. (line 120) 62774* mno-mips3d: MIPS Options. (line 382) 62775* mno-mma: RS/6000 and PowerPC Options. 62776 (line 943) 62777* mno-mmicromips: MIPS Options. (line 387) 62778* mno-mpy: ARC Options. (line 117) 62779* mno-ms-bitfields: x86 Options. (line 1099) 62780* mno-mt: MIPS Options. (line 395) 62781* mno-mul-bug-workaround: CRIS Options. (line 32) 62782* mno-muladd: FRV Options. (line 68) 62783* mno-mulhw: RS/6000 and PowerPC Options. 62784 (line 377) 62785* mno-mult-bug: MN10300 Options. (line 13) 62786* mno-multi-cond-exec: FRV Options. (line 223) 62787* mno-multiple: RS/6000 and PowerPC Options. 62788 (line 338) 62789* mno-mvcle: S/390 and zSeries Options. 62790 (line 138) 62791* mno-nested-cond-exec: FRV Options. (line 237) 62792* mno-odd-spreg: MIPS Options. (line 293) 62793* mno-omit-leaf-frame-pointer: AArch64 Options. (line 58) 62794* mno-optimize-membar: FRV Options. (line 249) 62795* mno-opts: MeP Options. (line 93) 62796* mno-pack: FRV Options. (line 151) 62797* mno-packed-stack: S/390 and zSeries Options. 62798 (line 54) 62799* mno-paired-single: MIPS Options. (line 370) 62800* mno-pc-relative-literal-loads: AArch64 Options. (line 282) 62801* mno-pcrel: RS/6000 and PowerPC Options. 62802 (line 931) 62803* mno-pic: IA-64 Options. (line 26) 62804* mno-pid: RX Options. (line 117) 62805* mno-plt: MIPS Options. (line 219) 62806* mno-pltseq: RS/6000 and PowerPC Options. 62807 (line 764) 62808* mno-popc: SPARC Options. (line 281) 62809* mno-popcntb: RS/6000 and PowerPC Options. 62810 (line 25) 62811* mno-popcntd: RS/6000 and PowerPC Options. 62812 (line 25) 62813* mno-postinc: Adapteva Epiphany Options. 62814 (line 109) 62815* mno-postmodify: Adapteva Epiphany Options. 62816 (line 109) 62817* mno-power8-fusion: RS/6000 and PowerPC Options. 62818 (line 189) 62819* mno-power8-vector: RS/6000 and PowerPC Options. 62820 (line 195) 62821* mno-powerpc-gfxopt: RS/6000 and PowerPC Options. 62822 (line 25) 62823* mno-powerpc-gpopt: RS/6000 and PowerPC Options. 62824 (line 25) 62825* mno-powerpc64: RS/6000 and PowerPC Options. 62826 (line 25) 62827* mno-prefixed: RS/6000 and PowerPC Options. 62828 (line 937) 62829* mno-prolog-function: V850 Options. (line 23) 62830* mno-prologue-epilogue: CRIS Options. (line 71) 62831* mno-prototype: RS/6000 and PowerPC Options. 62832 (line 592) 62833* mno-push-args: x86 Options. (line 1076) 62834* mno-quad-memory: RS/6000 and PowerPC Options. 62835 (line 202) 62836* mno-quad-memory-atomic: RS/6000 and PowerPC Options. 62837 (line 208) 62838* mno-readonly-in-sdata: RS/6000 and PowerPC Options. 62839 (line 683) 62840* mno-red-zone: x86 Options. (line 1445) 62841* mno-register-names: IA-64 Options. (line 37) 62842* mno-regnames: RS/6000 and PowerPC Options. 62843 (line 721) 62844* mno-relax: PRU Options. (line 21) 62845* mno-relax <1>: V850 Options. (line 103) 62846* mno-relax-immediate: MCore Options. (line 19) 62847* mno-relocatable: RS/6000 and PowerPC Options. 62848 (line 407) 62849* mno-relocatable-lib: RS/6000 and PowerPC Options. 62850 (line 418) 62851* mno-renesas: SH Options. (line 155) 62852* mno-round-nearest: Adapteva Epiphany Options. 62853 (line 51) 62854* mno-save-mduc-in-interrupts: RL78 Options. (line 79) 62855* mno-scc: FRV Options. (line 180) 62856* mno-sched-ar-data-spec: IA-64 Options. (line 135) 62857* mno-sched-ar-in-data-spec: IA-64 Options. (line 157) 62858* mno-sched-br-data-spec: IA-64 Options. (line 128) 62859* mno-sched-br-in-data-spec: IA-64 Options. (line 150) 62860* mno-sched-control-spec: IA-64 Options. (line 142) 62861* mno-sched-count-spec-in-critical-path: IA-64 Options. (line 185) 62862* mno-sched-in-control-spec: IA-64 Options. (line 164) 62863* mno-sched-prefer-non-control-spec-insns: IA-64 Options. (line 178) 62864* mno-sched-prefer-non-data-spec-insns: IA-64 Options. (line 171) 62865* mno-sched-prolog: ARM Options. (line 32) 62866* mno-sdata: ARC Options. (line 422) 62867* mno-sdata <1>: IA-64 Options. (line 42) 62868* mno-sdata <2>: RS/6000 and PowerPC Options. 62869 (line 678) 62870* mno-sep-data: Blackfin Options. (line 113) 62871* mno-serialize-volatile: Xtensa Options. (line 35) 62872* mno-setlb: MN10300 Options. (line 69) 62873* mno-short: M680x0 Options. (line 222) 62874* mno-side-effects: CRIS Options. (line 46) 62875* mno-sim: RX Options. (line 71) 62876* mno-single-exit: MMIX Options. (line 65) 62877* mno-slow-bytes: MCore Options. (line 35) 62878* mno-small-exec: S/390 and zSeries Options. 62879 (line 79) 62880* mno-smartmips: MIPS Options. (line 366) 62881* mno-soft-cmpsf: Adapteva Epiphany Options. 62882 (line 29) 62883* mno-soft-float: DEC Alpha Options. (line 10) 62884* mno-space-regs: HPPA Options. (line 45) 62885* mno-specld-anomaly: Blackfin Options. (line 53) 62886* mno-split-addresses: MIPS Options. (line 555) 62887* mno-split-lohi: Adapteva Epiphany Options. 62888 (line 109) 62889* mno-stack-align: CRIS Options. (line 55) 62890* mno-stack-bias: SPARC Options. (line 372) 62891* mno-std-struct-return: SPARC Options. (line 102) 62892* mno-strict-align: AArch64 Options. (line 52) 62893* mno-strict-align <1>: M680x0 Options. (line 280) 62894* mno-strict-align <2>: RS/6000 and PowerPC Options. 62895 (line 402) 62896* mno-subxc: SPARC Options. (line 288) 62897* mno-sum-in-toc: RS/6000 and PowerPC Options. 62898 (line 255) 62899* mno-sym32: MIPS Options. (line 453) 62900* mno-target-align: Xtensa Options. (line 74) 62901* mno-text-section-literals: Xtensa Options. (line 47) 62902* mno-tls-markers: RS/6000 and PowerPC Options. 62903 (line 776) 62904* mno-toc: RS/6000 and PowerPC Options. 62905 (line 427) 62906* mno-toplevel-symbols: MMIX Options. (line 39) 62907* mno-tpf-trace: S/390 and zSeries Options. 62908 (line 168) 62909* mno-tpf-trace-skip: S/390 and zSeries Options. 62910 (line 174) 62911* mno-unaligned-access: ARM Options. (line 899) 62912* mno-unaligned-doubles: SPARC Options. (line 73) 62913* mno-uninit-const-in-rodata: MIPS Options. (line 525) 62914* mno-update: RS/6000 and PowerPC Options. 62915 (line 349) 62916* mno-user-mode: SPARC Options. (line 85) 62917* mno-usermode: SH Options. (line 274) 62918* mno-v3push: NDS32 Options. (line 48) 62919* mno-v8plus: SPARC Options. (line 214) 62920* mno-vect-double: Adapteva Epiphany Options. 62921 (line 115) 62922* mno-virt: MIPS Options. (line 407) 62923* mno-vis: SPARC Options. (line 221) 62924* mno-vis2: SPARC Options. (line 227) 62925* mno-vis3: SPARC Options. (line 235) 62926* mno-vis4: SPARC Options. (line 243) 62927* mno-vis4b: SPARC Options. (line 251) 62928* mno-vliw-branch: FRV Options. (line 208) 62929* mno-volatile-asm-stop: IA-64 Options. (line 32) 62930* mno-volatile-cache: ARC Options. (line 431) 62931* mno-vrsave: RS/6000 and PowerPC Options. 62932 (line 152) 62933* mno-vsx: RS/6000 and PowerPC Options. 62934 (line 171) 62935* mno-vx: S/390 and zSeries Options. 62936 (line 112) 62937* mno-warn-devices-csv: MSP430 Options. (line 153) 62938* mno-warn-mcu: MSP430 Options. (line 65) 62939* mno-warn-multiple-fast-interrupts: RX Options. (line 143) 62940* mno-wide-bitfields: MCore Options. (line 23) 62941* mno-xgot: M680x0 Options. (line 312) 62942* mno-xgot <1>: MIPS Options. (line 229) 62943* mno-xl-compat: RS/6000 and PowerPC Options. 62944 (line 290) 62945* mno-xpa: MIPS Options. (line 411) 62946* mno-zdcbranch: SH Options. (line 341) 62947* mno-zero-extend: MMIX Options. (line 26) 62948* mno-zvector: S/390 and zSeries Options. 62949 (line 123) 62950* mnobitfield: M680x0 Options. (line 227) 62951* mnodiv: FT32 Options. (line 20) 62952* mnomacsave: SH Options. (line 160) 62953* mnop-fun-dllimport: x86 Windows Options. 62954 (line 22) 62955* mnop-mcount: x86 Options. (line 1306) 62956* mnopm: FT32 Options. (line 29) 62957* mnops: Adapteva Epiphany Options. 62958 (line 26) 62959* mnorm: ARC Options. (line 128) 62960* modd-spreg: MIPS Options. (line 293) 62961* momit-leaf-frame-pointer: AArch64 Options. (line 58) 62962* momit-leaf-frame-pointer <1>: Blackfin Options. (line 43) 62963* momit-leaf-frame-pointer <2>: x86 Options. (line 1270) 62964* mone-byte-bool: Darwin Options. (line 90) 62965* moptimize: Nvidia PTX Options. (line 22) 62966* moptimize-membar: FRV Options. (line 244) 62967* moptimize-membar <1>: FRV Options. (line 249) 62968* moverride: AArch64 Options. (line 269) 62969* MP: Preprocessor Options. 62970 (line 132) 62971* mpa-risc-1-0: HPPA Options. (line 19) 62972* mpa-risc-1-1: HPPA Options. (line 19) 62973* mpa-risc-2-0: HPPA Options. (line 19) 62974* mpack: FRV Options. (line 147) 62975* mpacked-stack: S/390 and zSeries Options. 62976 (line 54) 62977* mpadstruct: SH Options. (line 189) 62978* mpaired-single: MIPS Options. (line 370) 62979* mpc-relative-literal-loads: AArch64 Options. (line 282) 62980* mpc32: x86 Options. (line 691) 62981* mpc64: x86 Options. (line 691) 62982* mpc80: x86 Options. (line 691) 62983* mpclmul: x86 Options. (line 789) 62984* mpconfig: x86 Options. (line 797) 62985* mpcrel: M680x0 Options. (line 272) 62986* mpcrel <1>: RS/6000 and PowerPC Options. 62987 (line 931) 62988* mpdebug: CRIS Options. (line 36) 62989* mpe: RS/6000 and PowerPC Options. 62990 (line 309) 62991* mpe-aligned-commons: x86 Windows Options. 62992 (line 59) 62993* mpic-data-is-text-relative: ARM Options. (line 792) 62994* mpic-data-is-text-relative <1>: MicroBlaze Options. (line 70) 62995* mpic-register: ARM Options. (line 785) 62996* mpid: RX Options. (line 117) 62997* mpku: x86 Options. (line 825) 62998* mplt: MIPS Options. (line 219) 62999* mpltseq: RS/6000 and PowerPC Options. 63000 (line 764) 63001* mpointer-size=SIZE: VMS Options. (line 20) 63002* mpointers-to-nested-functions: RS/6000 and PowerPC Options. 63003 (line 877) 63004* mpoke-function-name: ARM Options. (line 800) 63005* mpopc: SPARC Options. (line 281) 63006* mpopcnt: x86 Options. (line 809) 63007* mpopcntb: RS/6000 and PowerPC Options. 63008 (line 25) 63009* mpopcntd: RS/6000 and PowerPC Options. 63010 (line 25) 63011* mportable-runtime: HPPA Options. (line 71) 63012* mpostinc: Adapteva Epiphany Options. 63013 (line 109) 63014* mpostmodify: Adapteva Epiphany Options. 63015 (line 109) 63016* mpower8-fusion: RS/6000 and PowerPC Options. 63017 (line 189) 63018* mpower8-vector: RS/6000 and PowerPC Options. 63019 (line 195) 63020* mpowerpc-gfxopt: RS/6000 and PowerPC Options. 63021 (line 25) 63022* mpowerpc-gpopt: RS/6000 and PowerPC Options. 63023 (line 25) 63024* mpowerpc64: RS/6000 and PowerPC Options. 63025 (line 25) 63026* mprefer-avx128: x86 Options. (line 912) 63027* mprefer-short-insn-regs: Adapteva Epiphany Options. 63028 (line 13) 63029* mprefer-vector-width: x86 Options. (line 916) 63030* mprefergot: SH Options. (line 268) 63031* mpreferred-stack-boundary: RISC-V Options. (line 73) 63032* mpreferred-stack-boundary <1>: x86 Options. (line 721) 63033* mprefetchwt1: x86 Options. (line 802) 63034* mprefixed: RS/6000 and PowerPC Options. 63035 (line 937) 63036* mpretend-cmove: SH Options. (line 383) 63037* mprfchw: x86 Options. (line 800) 63038* mprint-tune-info: ARM Options. (line 933) 63039* mprioritize-restricted-insns: RS/6000 and PowerPC Options. 63040 (line 456) 63041* mprolog-function: V850 Options. (line 23) 63042* mprologue-epilogue: CRIS Options. (line 71) 63043* mprototype: RS/6000 and PowerPC Options. 63044 (line 592) 63045* mptwrite: x86 Options. (line 793) 63046* mpure-code: ARM Options. (line 943) 63047* mpush-args: x86 Options. (line 1076) 63048* mpushpop: C-SKY Options. (line 114) 63049* MQ: Preprocessor Options. 63050 (line 159) 63051* mq-class: ARC Options. (line 521) 63052* mquad-memory: RS/6000 and PowerPC Options. 63053 (line 202) 63054* mquad-memory-atomic: RS/6000 and PowerPC Options. 63055 (line 208) 63056* mr0rel-sec: Nios II Options. (line 76) 63057* mr10k-cache-barrier: MIPS Options. (line 714) 63058* mRcq: ARC Options. (line 525) 63059* mRcw: ARC Options. (line 529) 63060* mrdpid: x86 Options. (line 801) 63061* mrdrnd: x86 Options. (line 794) 63062* mrdseed: x86 Options. (line 803) 63063* mreadonly-in-sdata: RS/6000 and PowerPC Options. 63064 (line 683) 63065* mrecip: RS/6000 and PowerPC Options. 63066 (line 784) 63067* mrecip <1>: x86 Options. (line 966) 63068* mrecip-precision: RS/6000 and PowerPC Options. 63069 (line 841) 63070* mrecip=opt: RS/6000 and PowerPC Options. 63071 (line 797) 63072* mrecip=opt <1>: x86 Options. (line 988) 63073* mrecord-mcount: x86 Options. (line 1300) 63074* mrecord-return: x86 Options. (line 1320) 63075* mred-zone: x86 Options. (line 1445) 63076* mreduced-regs: NDS32 Options. (line 15) 63077* mregister-names: IA-64 Options. (line 37) 63078* mregnames: RS/6000 and PowerPC Options. 63079 (line 721) 63080* mregparm: x86 Options. (line 661) 63081* mrelax: AVR Options. (line 231) 63082* mrelax <1>: H8/300 Options. (line 9) 63083* mrelax <2>: MN10300 Options. (line 46) 63084* mrelax <3>: MSP430 Options. (line 88) 63085* mrelax <4>: NDS32 Options. (line 84) 63086* mrelax <5>: RX Options. (line 95) 63087* mrelax <6>: SH Options. (line 137) 63088* mrelax <7>: V850 Options. (line 103) 63089* mrelax-immediate: MCore Options. (line 19) 63090* mrelax-pic-calls: MIPS Options. (line 859) 63091* mrelocatable: RS/6000 and PowerPC Options. 63092 (line 407) 63093* mrelocatable-lib: RS/6000 and PowerPC Options. 63094 (line 418) 63095* mrenesas: SH Options. (line 152) 63096* mrepeat: MeP Options. (line 96) 63097* mrestrict-it: ARM Options. (line 927) 63098* mreturn-pointer-on-d0: MN10300 Options. (line 36) 63099* mrf16: ARC Options. (line 324) 63100* mrgf-banked-regs: ARC Options. (line 304) 63101* mrh850-abi: V850 Options. (line 127) 63102* mrl78: RL78 Options. (line 62) 63103* mrmw: AVR Options. (line 245) 63104* mror: OpenRISC Options. (line 49) 63105* mrori: OpenRISC Options. (line 54) 63106* mround-nearest: Adapteva Epiphany Options. 63107 (line 51) 63108* mrtd: M680x0 Options. (line 236) 63109* mrtd <1>: x86 Options. (line 637) 63110* mrtd <2>: x86 Function Attributes. 63111 (line 9) 63112* mrtm: x86 Options. (line 820) 63113* mrtp: VxWorks Options. (line 11) 63114* mrtsc: ARC Options. (line 358) 63115* ms: H8/300 Options. (line 17) 63116* ms <1>: MeP Options. (line 100) 63117* ms2600: H8/300 Options. (line 24) 63118* msahf: x86 Options. (line 942) 63119* msatur: MeP Options. (line 105) 63120* msave-acc-in-interrupts: RX Options. (line 109) 63121* msave-mduc-in-interrupts: RL78 Options. (line 79) 63122* msave-restore: RISC-V Options. (line 87) 63123* msave-toc-indirect: RS/6000 and PowerPC Options. 63124 (line 889) 63125* mscc: FRV Options. (line 173) 63126* msched-ar-data-spec: IA-64 Options. (line 135) 63127* msched-ar-in-data-spec: IA-64 Options. (line 157) 63128* msched-br-data-spec: IA-64 Options. (line 128) 63129* msched-br-in-data-spec: IA-64 Options. (line 150) 63130* msched-control-spec: IA-64 Options. (line 142) 63131* msched-costly-dep: RS/6000 and PowerPC Options. 63132 (line 463) 63133* msched-count-spec-in-critical-path: IA-64 Options. (line 185) 63134* msched-fp-mem-deps-zero-cost: IA-64 Options. (line 202) 63135* msched-in-control-spec: IA-64 Options. (line 164) 63136* msched-max-memory-insns: IA-64 Options. (line 211) 63137* msched-max-memory-insns-hard-limit: IA-64 Options. (line 217) 63138* msched-prefer-non-control-spec-insns: IA-64 Options. (line 178) 63139* msched-prefer-non-data-spec-insns: IA-64 Options. (line 171) 63140* msched-prolog: ARM Options. (line 32) 63141* msched-prolog <1>: C-SKY Options. (line 148) 63142* msched-spec-ldc: IA-64 Options. (line 191) 63143* msched-spec-ldc <1>: IA-64 Options. (line 194) 63144* msched-stop-bits-after-every-cycle: IA-64 Options. (line 198) 63145* mschedule: HPPA Options. (line 78) 63146* mscore5: Score Options. (line 25) 63147* mscore5u: Score Options. (line 28) 63148* mscore7: Score Options. (line 31) 63149* mscore7d: Score Options. (line 35) 63150* msda: V850 Options. (line 40) 63151* msdata: ARC Options. (line 422) 63152* msdata <1>: IA-64 Options. (line 42) 63153* msdata <2>: RS/6000 and PowerPC Options. 63154 (line 665) 63155* msdata=all: C6X Options. (line 30) 63156* msdata=data: RS/6000 and PowerPC Options. 63157 (line 670) 63158* msdata=default: C6X Options. (line 22) 63159* msdata=default <1>: RS/6000 and PowerPC Options. 63160 (line 665) 63161* msdata=eabi: RS/6000 and PowerPC Options. 63162 (line 646) 63163* msdata=none: C6X Options. (line 35) 63164* msdata=none <1>: M32R/D Options. (line 40) 63165* msdata=none <2>: RS/6000 and PowerPC Options. 63166 (line 678) 63167* msdata=sdata: M32R/D Options. (line 49) 63168* msdata=sysv: RS/6000 and PowerPC Options. 63169 (line 656) 63170* msdata=use: M32R/D Options. (line 53) 63171* msdram: Blackfin Options. (line 169) 63172* msdram <1>: MeP Options. (line 110) 63173* msecure-plt: RS/6000 and PowerPC Options. 63174 (line 155) 63175* msecurity: C-SKY Options. (line 80) 63176* msel-sched-dont-check-control-spec: IA-64 Options. (line 207) 63177* msep-data: Blackfin Options. (line 107) 63178* msep-data <1>: Blackfin Options. (line 113) 63179* mserialize-volatile: Xtensa Options. (line 35) 63180* msetlb: MN10300 Options. (line 64) 63181* msext: OpenRISC Options. (line 59) 63182* msfimm: OpenRISC Options. (line 63) 63183* msgx: x86 Options. (line 804) 63184* msha: x86 Options. (line 787) 63185* mshared-library-id: Blackfin Options. (line 100) 63186* mshftimm: OpenRISC Options. (line 68) 63187* mshort: M680x0 Options. (line 216) 63188* mshort-calls: AVR Options. (line 249) 63189* mshstk: x86 Options. (line 956) 63190* mside-effects: CRIS Options. (line 46) 63191* msign-extend-enabled: LM32 Options. (line 18) 63192* msign-return-address: AArch64 Options. (line 288) 63193* msilicon-errata: MSP430 Options. (line 144) 63194* msilicon-errata-warn: MSP430 Options. (line 148) 63195* msim: Blackfin Options. (line 36) 63196* msim <1>: C6X Options. (line 19) 63197* msim <2>: CR16 Options. (line 18) 63198* msim <3>: FT32 Options. (line 9) 63199* msim <4>: M32C Options. (line 13) 63200* msim <5>: MeP Options. (line 114) 63201* msim <6>: MSP430 Options. (line 77) 63202* msim <7>: RL78 Options. (line 7) 63203* msim <8>: RS/6000 and PowerPC Options. 63204 (line 602) 63205* msim <9>: RX Options. (line 71) 63206* msim <10>: Visium Options. (line 13) 63207* msim <11>: Xstormy16 Options. (line 9) 63208* msimd: ARC Options. (line 141) 63209* msimnovec: MeP Options. (line 117) 63210* msingle-exit: MMIX Options. (line 65) 63211* msingle-float: MIPS Options. (line 284) 63212* msingle-pic-base: ARM Options. (line 779) 63213* msingle-pic-base <1>: RS/6000 and PowerPC Options. 63214 (line 450) 63215* msio: HPPA Options. (line 105) 63216* msize-level: ARC Options. (line 533) 63217* mskip-rax-setup: x86 Options. (line 1333) 63218* mslow-bytes: MCore Options. (line 35) 63219* mslow-flash-data: ARM Options. (line 915) 63220* msmall: MSP430 Options. (line 85) 63221* msmall-data: DEC Alpha Options. (line 187) 63222* msmall-data-limit: RISC-V Options. (line 82) 63223* msmall-data-limit <1>: RX Options. (line 47) 63224* msmall-divides: MicroBlaze Options. (line 38) 63225* msmall-exec: S/390 and zSeries Options. 63226 (line 79) 63227* msmall-model: FR30 Options. (line 9) 63228* msmall-text: DEC Alpha Options. (line 205) 63229* msmall16: Adapteva Epiphany Options. 63230 (line 66) 63231* msmallc: Nios II Options. (line 310) 63232* msmart: C-SKY Options. (line 97) 63233* msmartmips: MIPS Options. (line 366) 63234* msoft-cmpsf: Adapteva Epiphany Options. 63235 (line 29) 63236* msoft-div: OpenRISC Options. (line 19) 63237* msoft-float: ARC Options. (line 145) 63238* msoft-float <1>: C-SKY Options. (line 36) 63239* msoft-float <2>: DEC Alpha Options. (line 10) 63240* msoft-float <3>: FRV Options. (line 27) 63241* msoft-float <4>: HPPA Options. (line 91) 63242* msoft-float <5>: M680x0 Options. (line 200) 63243* msoft-float <6>: MicroBlaze Options. (line 7) 63244* msoft-float <7>: MIPS Options. (line 270) 63245* msoft-float <8>: OpenRISC Options. (line 29) 63246* msoft-float <9>: PDP-11 Options. (line 13) 63247* msoft-float <10>: RS/6000 and PowerPC Options. 63248 (line 332) 63249* msoft-float <11>: S/390 and zSeries Options. 63250 (line 11) 63251* msoft-float <12>: SPARC Options. (line 39) 63252* msoft-float <13>: V850 Options. (line 113) 63253* msoft-float <14>: Visium Options. (line 24) 63254* msoft-float <15>: x86 Options. (line 536) 63255* msoft-mul: OpenRISC Options. (line 24) 63256* msoft-quad-float: SPARC Options. (line 59) 63257* msoft-stack: Nvidia PTX Options. (line 26) 63258* msp8: AVR Options. (line 256) 63259* mspace: V850 Options. (line 30) 63260* mspace-regs: HPPA Options. (line 45) 63261* mspecld-anomaly: Blackfin Options. (line 48) 63262* mspecld-anomaly <1>: Blackfin Options. (line 53) 63263* mspfp: ARC Options. (line 132) 63264* mspfp-compact: ARC Options. (line 133) 63265* mspfp-fast: ARC Options. (line 137) 63266* mspfp_compact: ARC Options. (line 634) 63267* mspfp_fast: ARC Options. (line 637) 63268* msplit: PDP-11 Options. (line 40) 63269* msplit-addresses: MIPS Options. (line 555) 63270* msplit-lohi: Adapteva Epiphany Options. 63271 (line 109) 63272* msplit-vecmove-early: Adapteva Epiphany Options. 63273 (line 126) 63274* msse: x86 Options. (line 768) 63275* msse2: x86 Options. (line 769) 63276* msse2avx: x86 Options. (line 1288) 63277* msse3: x86 Options. (line 770) 63278* msse4: x86 Options. (line 772) 63279* msse4.1: x86 Options. (line 774) 63280* msse4.2: x86 Options. (line 775) 63281* msse4a: x86 Options. (line 773) 63282* msseregparm: x86 Options. (line 672) 63283* mssse3: x86 Options. (line 771) 63284* mstack-align: CRIS Options. (line 55) 63285* mstack-bias: SPARC Options. (line 372) 63286* mstack-check-l1: Blackfin Options. (line 74) 63287* mstack-guard: S/390 and zSeries Options. 63288 (line 201) 63289* mstack-increment: MCore Options. (line 50) 63290* mstack-offset: Adapteva Epiphany Options. 63291 (line 37) 63292* mstack-protector-guard: AArch64 Options. (line 64) 63293* mstack-protector-guard <1>: AArch64 Options. (line 79) 63294* mstack-protector-guard <2>: RS/6000 and PowerPC Options. 63295 (line 915) 63296* mstack-protector-guard <3>: x86 Options. (line 1359) 63297* mstack-protector-guard-offset: AArch64 Options. (line 64) 63298* mstack-protector-guard-offset <1>: AArch64 Options. (line 79) 63299* mstack-protector-guard-offset <2>: RS/6000 and PowerPC Options. 63300 (line 915) 63301* mstack-protector-guard-offset <3>: x86 Options. (line 1359) 63302* mstack-protector-guard-reg: AArch64 Options. (line 64) 63303* mstack-protector-guard-reg <1>: AArch64 Options. (line 79) 63304* mstack-protector-guard-reg <2>: RS/6000 and PowerPC Options. 63305 (line 915) 63306* mstack-protector-guard-reg <3>: x86 Options. (line 1359) 63307* mstack-protector-guard-symbol: RS/6000 and PowerPC Options. 63308 (line 915) 63309* mstack-size: AMD GCN Options. (line 23) 63310* mstack-size <1>: C-SKY Options. (line 134) 63311* mstack-size <2>: S/390 and zSeries Options. 63312 (line 201) 63313* mstackrealign: x86 Options. (line 712) 63314* mstd-struct-return: SPARC Options. (line 102) 63315* mstrict-align: AArch64 Options. (line 52) 63316* mstrict-align <1>: M680x0 Options. (line 280) 63317* mstrict-align <2>: RISC-V Options. (line 93) 63318* mstrict-align <3>: RS/6000 and PowerPC Options. 63319 (line 402) 63320* mstrict-X: AVR Options. (line 269) 63321* mstring-compare-inline-limit: RS/6000 and PowerPC Options. 63322 (line 707) 63323* mstringop-strategy=ALG: x86 Options. (line 1236) 63324* mstructure-size-boundary: ARM Options. (line 734) 63325* msubxc: SPARC Options. (line 288) 63326* msv-mode: Visium Options. (line 52) 63327* msve-vector-bits: AArch64 Options. (line 317) 63328* msvr4-struct-return: RS/6000 and PowerPC Options. 63329 (line 547) 63330* mswap: ARC Options. (line 152) 63331* mswape: ARC Options. (line 363) 63332* msym32: MIPS Options. (line 453) 63333* msynci: MIPS Options. (line 845) 63334* msys-crt0: Nios II Options. (line 314) 63335* msys-lib: Nios II Options. (line 318) 63336* MT: Preprocessor Options. 63337 (line 144) 63338* mtarget-align: Xtensa Options. (line 74) 63339* mtas: SH Options. (line 259) 63340* mtbm: x86 Options. (line 822) 63341* mtda: V850 Options. (line 34) 63342* mtelephony: ARC Options. (line 368) 63343* mtext-section-literals: Xtensa Options. (line 47) 63344* mtf: MeP Options. (line 121) 63345* mthread: x86 Windows Options. 63346 (line 26) 63347* mthreads: x86 Options. (line 1091) 63348* mthumb: ARM Options. (line 822) 63349* mthumb-interwork: ARM Options. (line 24) 63350* mtiny-printf: MSP430 Options. (line 122) 63351* mtiny-stack: AVR Options. (line 283) 63352* mtiny=: MeP Options. (line 125) 63353* mTLS: FRV Options. (line 90) 63354* mtls: FRV Options. (line 94) 63355* mtls-dialect: ARM Options. (line 874) 63356* mtls-dialect <1>: x86 Options. (line 1069) 63357* mtls-dialect=desc: AArch64 Options. (line 92) 63358* mtls-dialect=traditional: AArch64 Options. (line 96) 63359* mtls-direct-seg-refs: x86 Options. (line 1278) 63360* mtls-markers: RS/6000 and PowerPC Options. 63361 (line 776) 63362* mtls-size: AArch64 Options. (line 100) 63363* mtls-size <1>: IA-64 Options. (line 112) 63364* mtoc: RS/6000 and PowerPC Options. 63365 (line 427) 63366* mtomcat-stats: FRV Options. (line 254) 63367* mtoplevel-symbols: MMIX Options. (line 39) 63368* mtp: ARM Options. (line 866) 63369* mtp-regno: ARC Options. (line 170) 63370* mtpcs-frame: ARM Options. (line 839) 63371* mtpcs-leaf-frame: ARM Options. (line 845) 63372* mtpf-trace: S/390 and zSeries Options. 63373 (line 168) 63374* mtpf-trace-skip: S/390 and zSeries Options. 63375 (line 174) 63376* mtraceback: RS/6000 and PowerPC Options. 63377 (line 540) 63378* mtrap-precision: DEC Alpha Options. (line 109) 63379* mtrust: C-SKY Options. (line 83) 63380* mtune: AArch64 Options. (line 199) 63381* mtune <1>: AMD GCN Options. (line 10) 63382* mtune <2>: ARC Options. (line 554) 63383* mtune <3>: ARC Options. (line 640) 63384* mtune <4>: ARM Options. (line 571) 63385* mtune <5>: CRIS Options. (line 17) 63386* mtune <6>: DEC Alpha Options. (line 259) 63387* mtune <7>: IA-64 Options. (line 116) 63388* mtune <8>: M680x0 Options. (line 68) 63389* mtune <9>: MIPS Options. (line 66) 63390* mtune <10>: MN10300 Options. (line 30) 63391* mtune <11>: RISC-V Options. (line 59) 63392* mtune <12>: RS/6000 and PowerPC Options. 63393 (line 113) 63394* mtune <13>: S/390 and zSeries Options. 63395 (line 161) 63396* mtune <14>: SPARC Options. (line 199) 63397* mtune <15>: Visium Options. (line 47) 63398* mtune <16>: x86 Options. (line 410) 63399* mtune-ctrl=FEATURE-LIST: x86 Options. (line 879) 63400* muclibc: GNU/Linux Options. (line 14) 63401* muls: Score Options. (line 18) 63402* multcost: ARC Options. (line 645) 63403* multcost=NUMBER: SH Options. (line 281) 63404* multilib-library-pic: FRV Options. (line 110) 63405* multiply-enabled: LM32 Options. (line 15) 63406* multiply_defined: Darwin Options. (line 196) 63407* multiply_defined_unused: Darwin Options. (line 196) 63408* multi_module: Darwin Options. (line 196) 63409* munalign-prob-threshold: ARC Options. (line 582) 63410* munaligned-access: ARM Options. (line 899) 63411* munaligned-doubles: SPARC Options. (line 73) 63412* municode: x86 Windows Options. 63413 (line 30) 63414* muniform-simt: Nvidia PTX Options. (line 38) 63415* muninit-const-in-rodata: MIPS Options. (line 525) 63416* munix: VAX Options. (line 9) 63417* munix-asm: PDP-11 Options. (line 43) 63418* munordered-float: OpenRISC Options. (line 39) 63419* mupdate: RS/6000 and PowerPC Options. 63420 (line 349) 63421* muser-enabled: LM32 Options. (line 21) 63422* muser-mode: SPARC Options. (line 85) 63423* muser-mode <1>: Visium Options. (line 57) 63424* musermode: SH Options. (line 274) 63425* mv3push: NDS32 Options. (line 45) 63426* mv850: V850 Options. (line 49) 63427* mv850e: V850 Options. (line 79) 63428* mv850e1: V850 Options. (line 70) 63429* mv850e2: V850 Options. (line 66) 63430* mv850e2v3: V850 Options. (line 61) 63431* mv850e2v4: V850 Options. (line 57) 63432* mv850e3v5: V850 Options. (line 52) 63433* mv850es: V850 Options. (line 75) 63434* mv8plus: SPARC Options. (line 214) 63435* mvaes: x86 Options. (line 829) 63436* mvdsp: C-SKY Options. (line 88) 63437* mveclibabi: RS/6000 and PowerPC Options. 63438 (line 850) 63439* mveclibabi <1>: x86 Options. (line 1017) 63440* mvect-double: Adapteva Epiphany Options. 63441 (line 115) 63442* mvect8-ret-in-mem: x86 Options. (line 682) 63443* mverbose-cost-dump: AArch64 Options. (line 277) 63444* mverbose-cost-dump <1>: ARM Options. (line 939) 63445* mvirt: MIPS Options. (line 407) 63446* mvis: SPARC Options. (line 221) 63447* mvis2: SPARC Options. (line 227) 63448* mvis3: SPARC Options. (line 235) 63449* mvis4: SPARC Options. (line 243) 63450* mvis4b: SPARC Options. (line 251) 63451* mvliw-branch: FRV Options. (line 201) 63452* mvms-return-codes: VMS Options. (line 9) 63453* mvolatile-asm-stop: IA-64 Options. (line 32) 63454* mvolatile-cache: ARC Options. (line 427) 63455* mvolatile-cache <1>: ARC Options. (line 431) 63456* mvpclmulqdq: x86 Options. (line 831) 63457* mvr4130-align: MIPS Options. (line 834) 63458* mvrsave: RS/6000 and PowerPC Options. 63459 (line 152) 63460* mvsx: RS/6000 and PowerPC Options. 63461 (line 171) 63462* mvx: S/390 and zSeries Options. 63463 (line 112) 63464* mvxworks: RS/6000 and PowerPC Options. 63465 (line 623) 63466* mvzeroupper: x86 Options. (line 906) 63467* mwaitpkg: x86 Options. (line 830) 63468* mwarn-devices-csv: MSP430 Options. (line 153) 63469* mwarn-dynamicstack: S/390 and zSeries Options. 63470 (line 195) 63471* mwarn-framesize: S/390 and zSeries Options. 63472 (line 187) 63473* mwarn-mcu: MSP430 Options. (line 65) 63474* mwarn-multiple-fast-interrupts: RX Options. (line 143) 63475* mwbnoinvd: x86 Options. (line 798) 63476* mwide-bitfields: MCore Options. (line 23) 63477* mwin32: x86 Windows Options. 63478 (line 35) 63479* mwindows: x86 Windows Options. 63480 (line 41) 63481* mword-relocations: ARM Options. (line 885) 63482* mx32: x86 Options. (line 1425) 63483* mxgot: M680x0 Options. (line 312) 63484* mxgot <1>: MIPS Options. (line 229) 63485* mxl-barrel-shift: MicroBlaze Options. (line 32) 63486* mxl-compat: RS/6000 and PowerPC Options. 63487 (line 290) 63488* mxl-float-convert: MicroBlaze Options. (line 50) 63489* mxl-float-sqrt: MicroBlaze Options. (line 53) 63490* mxl-gp-opt: MicroBlaze Options. (line 44) 63491* mxl-multiply-high: MicroBlaze Options. (line 47) 63492* mxl-pattern-compare: MicroBlaze Options. (line 35) 63493* mxl-reorder: MicroBlaze Options. (line 62) 63494* mxl-soft-div: MicroBlaze Options. (line 29) 63495* mxl-soft-mul: MicroBlaze Options. (line 26) 63496* mxl-stack-check: MicroBlaze Options. (line 41) 63497* mxop: x86 Options. (line 805) 63498* mxpa: MIPS Options. (line 411) 63499* mxsave: x86 Options. (line 816) 63500* mxsavec: x86 Options. (line 818) 63501* mxsaveopt: x86 Options. (line 817) 63502* mxsaves: x86 Options. (line 819) 63503* mxy: ARC Options. (line 373) 63504* myellowknife: RS/6000 and PowerPC Options. 63505 (line 618) 63506* mzarch: S/390 and zSeries Options. 63507 (line 94) 63508* mzda: V850 Options. (line 45) 63509* mzdcbranch: SH Options. (line 341) 63510* mzero-extend: MMIX Options. (line 26) 63511* mzvector: S/390 and zSeries Options. 63512 (line 123) 63513* no-80387: x86 Options. (line 536) 63514* no-canonical-prefixes: Directory Options. (line 164) 63515* no-integrated-cpp: Preprocessor Options. 63516 (line 482) 63517* no-pie: Link Options. (line 181) 63518* no-sysroot-suffix: Directory Options. (line 183) 63519* noall_load: Darwin Options. (line 196) 63520* nocpp: MIPS Options. (line 636) 63521* nodefaultlibs: Link Options. (line 119) 63522* nodevicelib: AVR Options. (line 290) 63523* nodevicespecs: AVR Options. (line 293) 63524* nofixprebinding: Darwin Options. (line 196) 63525* nofpu: RX Options. (line 17) 63526* nolibc: Link Options. (line 131) 63527* nolibdld: HPPA Options. (line 188) 63528* nomultidefs: Darwin Options. (line 196) 63529* non-static: VxWorks Options. (line 16) 63530* noprebind: Darwin Options. (line 196) 63531* noseglinkedit: Darwin Options. (line 196) 63532* nostartfiles: Link Options. (line 114) 63533* nostdinc: Directory Options. (line 102) 63534* nostdinc++: C++ Dialect Options. 63535 (line 487) 63536* nostdinc++ <1>: Directory Options. (line 108) 63537* nostdlib: Link Options. (line 143) 63538* no_dead_strip_inits_and_terms: Darwin Options. (line 196) 63539* o: Overall Options. (line 197) 63540* O: Optimize Options. (line 39) 63541* O0: Optimize Options. (line 164) 63542* O1: Optimize Options. (line 39) 63543* O2: Optimize Options. (line 95) 63544* O3: Optimize Options. (line 143) 63545* Ofast: Optimize Options. (line 180) 63546* Og: Optimize Options. (line 188) 63547* Os: Optimize Options. (line 168) 63548* p: Instrumentation Options. 63549 (line 20) 63550* P: Preprocessor Options. 63551 (line 367) 63552* p <1>: Common Function Attributes. 63553 (line 696) 63554* pagezero_size: Darwin Options. (line 196) 63555* param: Optimize Options. (line 2624) 63556* pass-exit-codes: Overall Options. (line 339) 63557* pedantic: Standards. (line 13) 63558* pedantic <1>: Warning Options. (line 86) 63559* pedantic <2>: C Extensions. (line 6) 63560* pedantic <3>: Alternate Keywords. (line 30) 63561* pedantic <4>: Warnings and Errors. 63562 (line 25) 63563* pedantic-errors: Standards. (line 13) 63564* pedantic-errors <1>: Warning Options. (line 129) 63565* pedantic-errors <2>: Non-bugs. (line 216) 63566* pedantic-errors <3>: Warnings and Errors. 63567 (line 25) 63568* pg: Instrumentation Options. 63569 (line 20) 63570* pg <1>: Common Function Attributes. 63571 (line 696) 63572* pie: Link Options. (line 175) 63573* pipe: Overall Options. (line 347) 63574* plt: RISC-V Options. (line 13) 63575* prebind: Darwin Options. (line 196) 63576* prebind_all_twolevel_modules: Darwin Options. (line 196) 63577* print-file-name: Developer Options. (line 946) 63578* print-libgcc-file-name: Developer Options. (line 980) 63579* print-multi-directory: Developer Options. (line 952) 63580* print-multi-lib: Developer Options. (line 957) 63581* print-multi-os-directory: Developer Options. (line 964) 63582* print-multiarch: Developer Options. (line 973) 63583* print-objc-runtime-info: Objective-C and Objective-C++ Dialect Options. 63584 (line 220) 63585* print-prog-name: Developer Options. (line 977) 63586* print-search-dirs: Developer Options. (line 988) 63587* print-sysroot: Developer Options. (line 1001) 63588* print-sysroot-headers-suffix: Developer Options. (line 1008) 63589* private_bundle: Darwin Options. (line 196) 63590* pthread: Preprocessor Options. 63591 (line 70) 63592* pthread <1>: Link Options. (line 192) 63593* pthreads: Solaris 2 Options. (line 30) 63594* Q: Developer Options. (line 850) 63595* Qn: System V Options. (line 18) 63596* Qy: System V Options. (line 14) 63597* r: Link Options. (line 199) 63598* rdynamic: Link Options. (line 203) 63599* read_only_relocs: Darwin Options. (line 196) 63600* remap: Preprocessor Options. 63601 (line 398) 63602* S: Overall Options. (line 180) 63603* S <1>: Link Options. (line 20) 63604* s: Link Options. (line 210) 63605* save-temps: Developer Options. (line 725) 63606* save-temps=obj: Developer Options. (line 751) 63607* sectalign: Darwin Options. (line 196) 63608* sectcreate: Darwin Options. (line 196) 63609* sectobjectsymbols: Darwin Options. (line 196) 63610* sectobjectsymbols <1>: Darwin Options. (line 196) 63611* sectorder: Darwin Options. (line 196) 63612* seg1addr: Darwin Options. (line 196) 63613* segaddr: Darwin Options. (line 196) 63614* seglinkedit: Darwin Options. (line 196) 63615* segprot: Darwin Options. (line 196) 63616* segs_read_only_addr: Darwin Options. (line 196) 63617* segs_read_only_addr <1>: Darwin Options. (line 196) 63618* segs_read_write_addr: Darwin Options. (line 196) 63619* segs_read_write_addr <1>: Darwin Options. (line 196) 63620* seg_addr_table: Darwin Options. (line 196) 63621* seg_addr_table_filename: Darwin Options. (line 196) 63622* shared: Link Options. (line 219) 63623* shared-libgcc: Link Options. (line 227) 63624* short-calls: Adapteva Epiphany Options. 63625 (line 61) 63626* sim: CRIS Options. (line 95) 63627* sim2: CRIS Options. (line 101) 63628* single_module: Darwin Options. (line 196) 63629* specs: Overall Options. (line 353) 63630* static: Link Options. (line 214) 63631* static <1>: Darwin Options. (line 196) 63632* static <2>: HPPA Options. (line 192) 63633* static-libasan: Link Options. (line 261) 63634* static-libgcc: Link Options. (line 227) 63635* static-liblsan: Link Options. (line 277) 63636* static-libstdc++: Link Options. (line 294) 63637* static-libtsan: Link Options. (line 269) 63638* static-libubsan: Link Options. (line 285) 63639* static-pie: Link Options. (line 184) 63640* std: Standards. (line 13) 63641* std <1>: C Dialect Options. (line 46) 63642* std <2>: Other Builtins. (line 31) 63643* std <3>: Non-bugs. (line 107) 63644* sub_library: Darwin Options. (line 196) 63645* sub_umbrella: Darwin Options. (line 196) 63646* symbolic: Link Options. (line 305) 63647* sysroot: Directory Options. (line 168) 63648* T: Link Options. (line 311) 63649* target-help: Overall Options. (line 229) 63650* threads: HPPA Options. (line 205) 63651* time: Developer Options. (line 766) 63652* tno-android-cc: GNU/Linux Options. (line 36) 63653* tno-android-ld: GNU/Linux Options. (line 40) 63654* traditional: Preprocessor Options. 63655 (line 374) 63656* traditional <1>: Incompatibilities. (line 6) 63657* traditional-cpp: Preprocessor Options. 63658 (line 374) 63659* trigraphs: Preprocessor Options. 63660 (line 384) 63661* twolevel_namespace: Darwin Options. (line 196) 63662* U: Preprocessor Options. 63663 (line 42) 63664* u: Link Options. (line 343) 63665* umbrella: Darwin Options. (line 196) 63666* undef: Preprocessor Options. 63667 (line 66) 63668* undefined: Darwin Options. (line 196) 63669* unexported_symbols_list: Darwin Options. (line 196) 63670* v: Overall Options. (line 208) 63671* version: Overall Options. (line 336) 63672* w: Warning Options. (line 25) 63673* W: Warning Options. (line 212) 63674* W <1>: Warning Options. (line 2618) 63675* W <2>: Warning Options. (line 2726) 63676* W <3>: Incompatibilities. (line 64) 63677* Wa: Assembler Options. (line 9) 63678* Wabi: Warning Options. (line 256) 63679* Wabi-tag: C++ Dialect Options. 63680 (line 494) 63681* Wabi-tag <1>: C++ Dialect Options. 63682 (line 494) 63683* Wabsolute-value: Warning Options. (line 2147) 63684* Waddr-space-convert: AVR Options. (line 308) 63685* Waddress: Warning Options. (line 2488) 63686* Waddress-of-packed-member: Warning Options. (line 2501) 63687* Waggregate-return: Warning Options. (line 2529) 63688* Waggressive-loop-optimizations: Warning Options. (line 2534) 63689* Waligned-new: C++ Dialect Options. 63690 (line 934) 63691* Wall: Warning Options. (line 138) 63692* Wall <1>: Standard Libraries. (line 6) 63693* Walloc-size-larger-than=: Warning Options. (line 1616) 63694* Walloc-zero: Warning Options. (line 1606) 63695* Walloca: Warning Options. (line 1631) 63696* Walloca-larger-than=: Warning Options. (line 1634) 63697* Wanalyzer-double-fclose: Static Analyzer Options. 63698 (line 47) 63699* Wanalyzer-double-free: Static Analyzer Options. 63700 (line 54) 63701* Wanalyzer-exposure-through-output-file: Static Analyzer Options. 63702 (line 61) 63703* Wanalyzer-file-leak: Static Analyzer Options. 63704 (line 69) 63705* Wanalyzer-free-of-non-heap: Static Analyzer Options. 63706 (line 76) 63707* Wanalyzer-malloc-leak: Static Analyzer Options. 63708 (line 84) 63709* Wanalyzer-null-argument: Static Analyzer Options. 63710 (line 106) 63711* Wanalyzer-null-dereference: Static Analyzer Options. 63712 (line 114) 63713* Wanalyzer-possible-null-argument: Static Analyzer Options. 63714 (line 91) 63715* Wanalyzer-possible-null-dereference: Static Analyzer Options. 63716 (line 99) 63717* Wanalyzer-stale-setjmp-buffer: Static Analyzer Options. 63718 (line 121) 63719* Wanalyzer-tainted-array-index: Static Analyzer Options. 63720 (line 135) 63721* Wanalyzer-too-complex: Static Analyzer Options. 63722 (line 37) 63723* Wanalyzer-unsafe-call-within-signal-handler: Static Analyzer Options. 63724 (line 144) 63725* Wanalyzer-use-after-free: Static Analyzer Options. 63726 (line 152) 63727* Wanalyzer-use-of-pointer-in-stale-stack-frame: Static Analyzer Options. 63728 (line 159) 63729* Warith-conversion: Warning Options. (line 1699) 63730* Warray-bounds: Warning Options. (line 1712) 63731* Wassign-intercept: Objective-C and Objective-C++ Dialect Options. 63732 (line 170) 63733* Wattribute-alias: Warning Options. (line 1729) 63734* Wattribute-warning: Warning Options. (line 2697) 63735* Wattributes: Warning Options. (line 2539) 63736* Wbad-function-cast: Warning Options. (line 2216) 63737* Wbool-compare: Warning Options. (line 1761) 63738* Wbool-operation: Warning Options. (line 1770) 63739* Wbuiltin-declaration-mismatch: Warning Options. (line 2545) 63740* Wbuiltin-macro-redefined: Warning Options. (line 2566) 63741* Wc++-compat: Warning Options. (line 2244) 63742* Wc++11-compat: Warning Options. (line 2249) 63743* Wc++14-compat: Warning Options. (line 2255) 63744* Wc++17-compat: Warning Options. (line 2259) 63745* Wc++20-compat: Warning Options. (line 2263) 63746* Wc11-c2x-compat: Warning Options. (line 2236) 63747* Wc90-c99-compat: Warning Options. (line 2221) 63748* Wc99-c11-compat: Warning Options. (line 2228) 63749* Wcast-align: Warning Options. (line 2283) 63750* Wcast-align=strict: Warning Options. (line 2289) 63751* Wcast-function-type: Warning Options. (line 2294) 63752* Wcast-qual: Warning Options. (line 2267) 63753* Wcatch-value: C++ Dialect Options. 63754 (line 981) 63755* Wchar-subscripts: Warning Options. (line 366) 63756* Wclass-conversion: C++ Dialect Options. 63757 (line 912) 63758* Wclass-memaccess: C++ Dialect Options. 63759 (line 630) 63760* Wclobbered: Warning Options. (line 2320) 63761* Wcomma-subscript: C++ Dialect Options. 63762 (line 499) 63763* Wcomment: Warning Options. (line 2158) 63764* Wcomments: Warning Options. (line 2158) 63765* Wconditionally-supported: C++ Dialect Options. 63766 (line 989) 63767* Wconversion: Warning Options. (line 2324) 63768* Wconversion-null: C++ Dialect Options. 63769 (line 1066) 63770* Wcoverage-mismatch: Warning Options. (line 371) 63771* Wcpp: Warning Options. (line 385) 63772* Wctor-dtor-privacy: C++ Dialect Options. 63773 (line 511) 63774* Wdangling-else: Warning Options. (line 2345) 63775* Wdate-time: Warning Options. (line 2379) 63776* Wdeclaration-after-statement: Warning Options. (line 1990) 63777* Wdelete-incomplete: C++ Dialect Options. 63778 (line 992) 63779* Wdelete-non-virtual-dtor: C++ Dialect Options. 63780 (line 518) 63781* Wdeprecated: Warning Options. (line 2704) 63782* Wdeprecated-copy: C++ Dialect Options. 63783 (line 525) 63784* Wdeprecated-declarations: Warning Options. (line 2708) 63785* Wdesignated-init: Warning Options. (line 2959) 63786* Wdisabled-optimization: Warning Options. (line 2910) 63787* Wdiscarded-array-qualifiers: Warning Options. (line 1809) 63788* Wdiscarded-qualifiers: Warning Options. (line 1803) 63789* Wdiv-by-zero: Warning Options. (line 1850) 63790* Wdouble-promotion: Warning Options. (line 389) 63791* Wduplicate-decl-specifier: Warning Options. (line 407) 63792* Wduplicated-branches: Warning Options. (line 1780) 63793* Wduplicated-cond: Warning Options. (line 1791) 63794* weak_reference_mismatches: Darwin Options. (line 196) 63795* Weffc++: C++ Dialect Options. 63796 (line 767) 63797* Wempty-body: Warning Options. (line 2384) 63798* Wendif-labels: Warning Options. (line 2202) 63799* Wendif-labels <1>: Warning Options. (line 2388) 63800* Wenum-compare: Warning Options. (line 2391) 63801* Wenum-conversion: Warning Options. (line 2397) 63802* Werror: Warning Options. (line 28) 63803* Werror=: Warning Options. (line 31) 63804* Wexpansion-to-defined: Warning Options. (line 2177) 63805* Wextra: Warning Options. (line 212) 63806* Wextra <1>: Warning Options. (line 2618) 63807* Wextra <2>: Warning Options. (line 2726) 63808* Wextra-semi: C++ Dialect Options. 63809 (line 997) 63810* Wfatal-errors: Warning Options. (line 48) 63811* Wfloat-conversion: Warning Options. (line 2424) 63812* Wfloat-equal: Warning Options. (line 1890) 63813* Wformat: Warning Options. (line 412) 63814* Wformat <1>: Warning Options. (line 437) 63815* Wformat <2>: Warning Options. (line 1577) 63816* Wformat <3>: Common Function Attributes. 63817 (line 381) 63818* Wformat-contains-nul: Warning Options. (line 451) 63819* Wformat-extra-args: Warning Options. (line 455) 63820* Wformat-nonliteral: Warning Options. (line 548) 63821* Wformat-nonliteral <1>: Common Function Attributes. 63822 (line 446) 63823* Wformat-overflow: Warning Options. (line 469) 63824* Wformat-overflow <1>: Warning Options. (line 480) 63825* Wformat-security: Warning Options. (line 553) 63826* Wformat-signedness: Warning Options. (line 564) 63827* Wformat-truncation: Warning Options. (line 569) 63828* Wformat-truncation <1>: Warning Options. (line 581) 63829* Wformat-y2k: Warning Options. (line 592) 63830* Wformat-zero-length: Warning Options. (line 544) 63831* Wformat=: Warning Options. (line 412) 63832* Wformat=1: Warning Options. (line 437) 63833* Wformat=2: Warning Options. (line 446) 63834* Wframe-address: Warning Options. (line 1797) 63835* Wframe-larger-than=: Warning Options. (line 2059) 63836* Wfree-nonheap-object: Warning Options. (line 2076) 63837* whatsloaded: Darwin Options. (line 196) 63838* Whsa: Warning Options. (line 2964) 63839* whyload: Darwin Options. (line 196) 63840* Wif-not-aligned: Warning Options. (line 765) 63841* Wignored-attributes: Warning Options. (line 780) 63842* Wignored-qualifiers: Warning Options. (line 769) 63843* Wimplicit: Warning Options. (line 643) 63844* Wimplicit-fallthrough: Warning Options. (line 647) 63845* Wimplicit-fallthrough=: Warning Options. (line 652) 63846* Wimplicit-function-declaration: Warning Options. (line 637) 63847* Wimplicit-int: Warning Options. (line 632) 63848* Winaccessible-base: C++ Dialect Options. 63849 (line 1001) 63850* Wincompatible-pointer-types: Warning Options. (line 1815) 63851* Winherited-variadic-ctor: C++ Dialect Options. 63852 (line 1012) 63853* Winit-list-lifetime: C++ Dialect Options. 63854 (line 533) 63855* Winit-self: Warning Options. (line 617) 63856* Winline: Warning Options. (line 2819) 63857* Winline <1>: Inline. (line 60) 63858* Wint-conversion: Warning Options. (line 1821) 63859* Wint-in-bool-context: Warning Options. (line 2832) 63860* Wint-to-pointer-cast: Warning Options. (line 2840) 63861* Winvalid-memory-model: Warning Options. (line 1237) 63862* Winvalid-offsetof: C++ Dialect Options. 63863 (line 1017) 63864* Winvalid-pch: Warning Options. (line 2849) 63865* Wjump-misses-init: Warning Options. (line 2401) 63866* Wl: Link Options. (line 335) 63867* Wlarger-than-BYTE-SIZE: Warning Options. (line 2049) 63868* Wlarger-than=: Warning Options. (line 2049) 63869* Wliteral-suffix: C++ Dialect Options. 63870 (line 568) 63871* Wlogical-not-parentheses: Warning Options. (line 2514) 63872* Wlogical-op: Warning Options. (line 2506) 63873* Wlong-long: Warning Options. (line 2853) 63874* Wlto-type-mismatch: Warning Options. (line 2953) 63875* Wmain: Warning Options. (line 787) 63876* Wmaybe-uninitialized: Warning Options. (line 1254) 63877* Wmemset-elt-size: Warning Options. (line 2469) 63878* Wmemset-transposed-args: Warning Options. (line 2477) 63879* Wmisleading-indentation: Warning Options. (line 794) 63880* Wmismatched-tags: C++ Dialect Options. 63881 (line 853) 63882* Wmissing-attributes: Warning Options. (line 828) 63883* Wmissing-braces: Warning Options. (line 872) 63884* Wmissing-declarations: Warning Options. (line 2608) 63885* Wmissing-field-initializers: Warning Options. (line 2618) 63886* Wmissing-format-attribute: Warning Options. (line 1577) 63887* Wmissing-include-dirs: Warning Options. (line 882) 63888* Wmissing-noreturn: Warning Options. (line 1563) 63889* Wmissing-parameter-type: Warning Options. (line 2590) 63890* Wmissing-profile: Warning Options. (line 885) 63891* Wmissing-prototypes: Warning Options. (line 2598) 63892* Wmisspelled-isr: AVR Options. (line 313) 63893* Wmultichar: Warning Options. (line 2648) 63894* Wmultiple-inheritance: C++ Dialect Options. 63895 (line 878) 63896* Wmultistatement-macros: Warning Options. (line 901) 63897* Wnamespaces: C++ Dialect Options. 63898 (line 901) 63899* Wnarrowing: C++ Dialect Options. 63900 (line 594) 63901* Wnested-externs: Warning Options. (line 2816) 63902* Wno-abi: Warning Options. (line 256) 63903* Wno-absolute-value: Warning Options. (line 2147) 63904* Wno-addr-space-convert: AVR Options. (line 308) 63905* Wno-address: Warning Options. (line 2488) 63906* Wno-address-of-packed-member: Warning Options. (line 2501) 63907* Wno-aggregate-return: Warning Options. (line 2529) 63908* Wno-aggressive-loop-optimizations: Warning Options. (line 2534) 63909* Wno-aligned-new: C++ Dialect Options. 63910 (line 934) 63911* Wno-all: Warning Options. (line 138) 63912* Wno-alloc-size-larger-than: Warning Options. (line 1616) 63913* Wno-alloc-size-larger-than <1>: Warning Options. (line 1627) 63914* Wno-alloc-zero: Warning Options. (line 1606) 63915* Wno-alloca: Warning Options. (line 1631) 63916* Wno-alloca-larger-than: Warning Options. (line 1634) 63917* Wno-alloca-larger-than <1>: Warning Options. (line 1695) 63918* Wno-analyzer-double-fclose: Static Analyzer Options. 63919 (line 47) 63920* Wno-analyzer-double-free: Static Analyzer Options. 63921 (line 54) 63922* Wno-analyzer-exposure-through-output-file: Static Analyzer Options. 63923 (line 61) 63924* Wno-analyzer-file-leak: Static Analyzer Options. 63925 (line 69) 63926* Wno-analyzer-free-of-non-heap: Static Analyzer Options. 63927 (line 76) 63928* Wno-analyzer-malloc-leak: Static Analyzer Options. 63929 (line 84) 63930* Wno-analyzer-null-argument: Static Analyzer Options. 63931 (line 106) 63932* Wno-analyzer-null-dereference: Static Analyzer Options. 63933 (line 114) 63934* Wno-analyzer-possible-null-argument: Static Analyzer Options. 63935 (line 91) 63936* Wno-analyzer-possible-null-dereference: Static Analyzer Options. 63937 (line 99) 63938* Wno-analyzer-stale-setjmp-buffer: Static Analyzer Options. 63939 (line 121) 63940* Wno-analyzer-tainted-array-index: Static Analyzer Options. 63941 (line 135) 63942* Wno-analyzer-too-complex: Static Analyzer Options. 63943 (line 37) 63944* Wno-analyzer-unsafe-call-within-signal-handler: Static Analyzer Options. 63945 (line 144) 63946* Wno-analyzer-use-after-free: Static Analyzer Options. 63947 (line 152) 63948* Wno-analyzer-use-of-pointer-in-stale-stack-frame: Static Analyzer Options. 63949 (line 159) 63950* Wno-arith-conversion: Warning Options. (line 1699) 63951* Wno-array-bounds: Warning Options. (line 1712) 63952* Wno-assign-intercept: Objective-C and Objective-C++ Dialect Options. 63953 (line 170) 63954* Wno-attribute-alias: Warning Options. (line 1729) 63955* Wno-attribute-warning: Warning Options. (line 2697) 63956* Wno-attributes: Warning Options. (line 2539) 63957* Wno-bad-function-cast: Warning Options. (line 2216) 63958* Wno-bool-compare: Warning Options. (line 1761) 63959* Wno-bool-operation: Warning Options. (line 1770) 63960* Wno-builtin-declaration-mismatch: Warning Options. (line 2545) 63961* Wno-builtin-macro-redefined: Warning Options. (line 2566) 63962* Wno-c++-compat: Warning Options. (line 2244) 63963* Wno-c++11-compat: Warning Options. (line 2249) 63964* Wno-c++14-compat: Warning Options. (line 2255) 63965* Wno-c++17-compat: Warning Options. (line 2259) 63966* Wno-c++20-compat: Warning Options. (line 2263) 63967* Wno-c11-c2x-compat: Warning Options. (line 2236) 63968* Wno-c90-c99-compat: Warning Options. (line 2221) 63969* Wno-c99-c11-compat: Warning Options. (line 2228) 63970* Wno-cast-align: Warning Options. (line 2283) 63971* Wno-cast-function-type: Warning Options. (line 2294) 63972* Wno-cast-qual: Warning Options. (line 2267) 63973* Wno-catch-value: C++ Dialect Options. 63974 (line 981) 63975* Wno-char-subscripts: Warning Options. (line 366) 63976* Wno-class-conversion: C++ Dialect Options. 63977 (line 912) 63978* Wno-class-memaccess: C++ Dialect Options. 63979 (line 630) 63980* Wno-clobbered: Warning Options. (line 2320) 63981* Wno-comma-subscript: C++ Dialect Options. 63982 (line 499) 63983* Wno-conditionally-supported: C++ Dialect Options. 63984 (line 989) 63985* Wno-conversion: Warning Options. (line 2324) 63986* Wno-conversion-null: C++ Dialect Options. 63987 (line 1066) 63988* Wno-coverage-mismatch: Warning Options. (line 371) 63989* Wno-cpp: Warning Options. (line 385) 63990* Wno-ctor-dtor-privacy: C++ Dialect Options. 63991 (line 511) 63992* Wno-dangling-else: Warning Options. (line 2345) 63993* Wno-date-time: Warning Options. (line 2379) 63994* Wno-declaration-after-statement: Warning Options. (line 1990) 63995* Wno-delete-incomplete: C++ Dialect Options. 63996 (line 992) 63997* Wno-delete-non-virtual-dtor: C++ Dialect Options. 63998 (line 518) 63999* Wno-deprecated: Warning Options. (line 2704) 64000* Wno-deprecated-copy: C++ Dialect Options. 64001 (line 525) 64002* Wno-deprecated-declarations: Warning Options. (line 2708) 64003* Wno-designated-init: Warning Options. (line 2959) 64004* Wno-disabled-optimization: Warning Options. (line 2910) 64005* Wno-discarded-array-qualifiers: Warning Options. (line 1809) 64006* Wno-discarded-qualifiers: Warning Options. (line 1803) 64007* Wno-div-by-zero: Warning Options. (line 1850) 64008* Wno-double-promotion: Warning Options. (line 389) 64009* Wno-duplicate-decl-specifier: Warning Options. (line 407) 64010* Wno-duplicated-branches: Warning Options. (line 1780) 64011* Wno-duplicated-cond: Warning Options. (line 1791) 64012* Wno-effc++: C++ Dialect Options. 64013 (line 767) 64014* Wno-empty-body: Warning Options. (line 2384) 64015* Wno-endif-labels: Warning Options. (line 2202) 64016* Wno-endif-labels <1>: Warning Options. (line 2388) 64017* Wno-enum-compare: Warning Options. (line 2391) 64018* Wno-enum-conversion: Warning Options. (line 2397) 64019* Wno-error: Warning Options. (line 28) 64020* Wno-error=: Warning Options. (line 31) 64021* Wno-extra: Warning Options. (line 212) 64022* Wno-extra <1>: Warning Options. (line 2618) 64023* Wno-extra <2>: Warning Options. (line 2726) 64024* Wno-extra-semi: C++ Dialect Options. 64025 (line 997) 64026* Wno-fatal-errors: Warning Options. (line 48) 64027* Wno-float-conversion: Warning Options. (line 2424) 64028* Wno-float-equal: Warning Options. (line 1890) 64029* Wno-format: Warning Options. (line 412) 64030* Wno-format <1>: Warning Options. (line 1577) 64031* Wno-format-contains-nul: Warning Options. (line 451) 64032* Wno-format-extra-args: Warning Options. (line 455) 64033* Wno-format-nonliteral: Warning Options. (line 548) 64034* Wno-format-overflow: Warning Options. (line 469) 64035* Wno-format-overflow <1>: Warning Options. (line 480) 64036* Wno-format-security: Warning Options. (line 553) 64037* Wno-format-signedness: Warning Options. (line 564) 64038* Wno-format-truncation: Warning Options. (line 569) 64039* Wno-format-truncation <1>: Warning Options. (line 581) 64040* Wno-format-y2k: Warning Options. (line 592) 64041* Wno-format-zero-length: Warning Options. (line 544) 64042* Wno-frame-address: Warning Options. (line 1797) 64043* Wno-frame-larger-than: Warning Options. (line 2059) 64044* Wno-frame-larger-than <1>: Warning Options. (line 2072) 64045* Wno-free-nonheap-object: Warning Options. (line 2076) 64046* Wno-hsa: Warning Options. (line 2964) 64047* Wno-if-not-aligned: Warning Options. (line 765) 64048* Wno-ignored-attributes: Warning Options. (line 780) 64049* Wno-ignored-qualifiers: Warning Options. (line 769) 64050* Wno-implicit: Warning Options. (line 643) 64051* Wno-implicit-fallthrough: Warning Options. (line 647) 64052* Wno-implicit-function-declaration: Warning Options. (line 637) 64053* Wno-implicit-int: Warning Options. (line 632) 64054* Wno-inaccessible-base: C++ Dialect Options. 64055 (line 1001) 64056* Wno-incompatible-pointer-types: Warning Options. (line 1815) 64057* Wno-inherited-variadic-ctor: C++ Dialect Options. 64058 (line 1012) 64059* Wno-init-list-lifetime: C++ Dialect Options. 64060 (line 533) 64061* Wno-init-self: Warning Options. (line 617) 64062* Wno-inline: Warning Options. (line 2819) 64063* Wno-int-conversion: Warning Options. (line 1821) 64064* Wno-int-in-bool-context: Warning Options. (line 2832) 64065* Wno-int-to-pointer-cast: Warning Options. (line 2840) 64066* Wno-invalid-memory-model: Warning Options. (line 1237) 64067* Wno-invalid-offsetof: C++ Dialect Options. 64068 (line 1017) 64069* Wno-invalid-pch: Warning Options. (line 2849) 64070* Wno-jump-misses-init: Warning Options. (line 2401) 64071* Wno-larger-than: Warning Options. (line 2055) 64072* Wno-literal-suffix: C++ Dialect Options. 64073 (line 568) 64074* Wno-logical-not-parentheses: Warning Options. (line 2514) 64075* Wno-logical-op: Warning Options. (line 2506) 64076* Wno-long-long: Warning Options. (line 2853) 64077* Wno-lto-type-mismatch: Warning Options. (line 2953) 64078* Wno-main: Warning Options. (line 787) 64079* Wno-maybe-uninitialized: Warning Options. (line 1254) 64080* Wno-memset-elt-size: Warning Options. (line 2469) 64081* Wno-memset-transposed-args: Warning Options. (line 2477) 64082* Wno-misleading-indentation: Warning Options. (line 794) 64083* Wno-mismatched-tags: C++ Dialect Options. 64084 (line 853) 64085* Wno-missing-attributes: Warning Options. (line 828) 64086* Wno-missing-braces: Warning Options. (line 872) 64087* Wno-missing-declarations: Warning Options. (line 2608) 64088* Wno-missing-field-initializers: Warning Options. (line 2618) 64089* Wno-missing-format-attribute: Warning Options. (line 1577) 64090* Wno-missing-include-dirs: Warning Options. (line 882) 64091* Wno-missing-noreturn: Warning Options. (line 1563) 64092* Wno-missing-parameter-type: Warning Options. (line 2590) 64093* Wno-missing-profile: Warning Options. (line 885) 64094* Wno-missing-prototypes: Warning Options. (line 2598) 64095* Wno-misspelled-isr: AVR Options. (line 313) 64096* Wno-multichar: Warning Options. (line 2648) 64097* Wno-multiple-inheritance: C++ Dialect Options. 64098 (line 878) 64099* Wno-multistatement-macros: Warning Options. (line 901) 64100* Wno-namespaces: C++ Dialect Options. 64101 (line 901) 64102* Wno-narrowing: C++ Dialect Options. 64103 (line 594) 64104* Wno-nested-externs: Warning Options. (line 2816) 64105* Wno-noexcept: C++ Dialect Options. 64106 (line 610) 64107* Wno-noexcept-type: C++ Dialect Options. 64108 (line 616) 64109* Wno-non-template-friend: C++ Dialect Options. 64110 (line 802) 64111* Wno-non-virtual-dtor: C++ Dialect Options. 64112 (line 650) 64113* Wno-nonnull: Warning Options. (line 596) 64114* Wno-nonnull-compare: Warning Options. (line 603) 64115* Wno-normalized: Warning Options. (line 2654) 64116* Wno-null-dereference: Warning Options. (line 610) 64117* Wno-odr: Warning Options. (line 2717) 64118* Wno-old-style-cast: C++ Dialect Options. 64119 (line 811) 64120* Wno-old-style-declaration: Warning Options. (line 2577) 64121* Wno-old-style-definition: Warning Options. (line 2583) 64122* Wno-openmp-simd: Warning Options. (line 2721) 64123* Wno-overflow: Warning Options. (line 2714) 64124* Wno-overlength-strings: Warning Options. (line 2930) 64125* Wno-overloaded-virtual: C++ Dialect Options. 64126 (line 817) 64127* Wno-override-init: Warning Options. (line 2726) 64128* Wno-override-init-side-effects: Warning Options. (line 2734) 64129* Wno-packed: Warning Options. (line 2739) 64130* Wno-packed-bitfield-compat: Warning Options. (line 2756) 64131* Wno-packed-not-aligned: Warning Options. (line 2773) 64132* Wno-padded: Warning Options. (line 2786) 64133* Wno-parentheses: Warning Options. (line 921) 64134* Wno-pedantic: Warning Options. (line 86) 64135* Wno-pedantic-ms-format: Warning Options. (line 2114) 64136* Wno-pessimizing-move: C++ Dialect Options. 64137 (line 679) 64138* Wno-placement-new: C++ Dialect Options. 64139 (line 945) 64140* Wno-pmf-conversions: C++ Dialect Options. 64141 (line 836) 64142* Wno-pmf-conversions <1>: Bound member functions. 64143 (line 35) 64144* Wno-pointer-arith: Warning Options. (line 2120) 64145* Wno-pointer-compare: Warning Options. (line 2127) 64146* Wno-pointer-sign: Warning Options. (line 2919) 64147* Wno-pointer-to-int-cast: Warning Options. (line 2845) 64148* Wno-pragmas: Warning Options. (line 1306) 64149* Wno-prio-ctor-dtor: Warning Options. (line 1311) 64150* Wno-property-assign-default: Objective-C and Objective-C++ Dialect Options. 64151 (line 174) 64152* Wno-protocol: Objective-C and Objective-C++ Dialect Options. 64153 (line 178) 64154* Wno-redundant-decls: Warning Options. (line 2793) 64155* Wno-redundant-move: C++ Dialect Options. 64156 (line 701) 64157* Wno-redundant-tags: C++ Dialect Options. 64158 (line 740) 64159* Wno-register: C++ Dialect Options. 64160 (line 658) 64161* Wno-reorder: C++ Dialect Options. 64162 (line 665) 64163* Wno-restrict: Warning Options. (line 2797) 64164* Wno-return-local-addr: Warning Options. (line 1001) 64165* Wno-return-type: Warning Options. (line 1005) 64166* Wno-scalar-storage-order: Warning Options. (line 2430) 64167* Wno-selector: Objective-C and Objective-C++ Dialect Options. 64168 (line 188) 64169* Wno-sequence-point: Warning Options. (line 948) 64170* Wno-shadow: Warning Options. (line 1996) 64171* Wno-shadow-ivar: Warning Options. (line 2007) 64172* Wno-shift-count-negative: Warning Options. (line 1026) 64173* Wno-shift-count-overflow: Warning Options. (line 1030) 64174* Wno-shift-negative-value: Warning Options. (line 1034) 64175* Wno-shift-overflow: Warning Options. (line 1039) 64176* Wno-sign-compare: Warning Options. (line 2412) 64177* Wno-sign-conversion: Warning Options. (line 2418) 64178* Wno-sign-promo: C++ Dialect Options. 64179 (line 840) 64180* Wno-sized-deallocation: C++ Dialect Options. 64181 (line 1029) 64182* Wno-sizeof-array-argument: Warning Options. (line 2464) 64183* Wno-sizeof-pointer-div: Warning Options. (line 2434) 64184* Wno-sizeof-pointer-memaccess: Warning Options. (line 2442) 64185* Wno-stack-protector: Warning Options. (line 2925) 64186* Wno-stack-usage: Warning Options. (line 2080) 64187* Wno-stack-usage <1>: Warning Options. (line 2104) 64188* Wno-strict-aliasing: Warning Options. (line 1319) 64189* Wno-strict-null-sentinel: C++ Dialect Options. 64190 (line 795) 64191* Wno-strict-overflow: Warning Options. (line 1358) 64192* Wno-strict-prototypes: Warning Options. (line 2571) 64193* Wno-strict-selector-match: Objective-C and Objective-C++ Dialect Options. 64194 (line 200) 64195* Wno-string-compare: Warning Options. (line 1406) 64196* Wno-stringop-overflow: Warning Options. (line 1427) 64197* Wno-stringop-overflow <1>: Warning Options. (line 1466) 64198* Wno-stringop-truncation: Warning Options. (line 1504) 64199* Wno-subobject-linkage: C++ Dialect Options. 64200 (line 754) 64201* Wno-suggest-attribute=: Warning Options. (line 1555) 64202* Wno-suggest-attribute=cold: Warning Options. (line 1598) 64203* Wno-suggest-attribute=const: Warning Options. (line 1563) 64204* Wno-suggest-attribute=format: Warning Options. (line 1577) 64205* Wno-suggest-attribute=malloc: Warning Options. (line 1563) 64206* Wno-suggest-attribute=noreturn: Warning Options. (line 1563) 64207* Wno-suggest-attribute=pure: Warning Options. (line 1563) 64208* Wno-suggest-final-methods: C++ Dialect Options. 64209 (line 1049) 64210* Wno-suggest-final-types: C++ Dialect Options. 64211 (line 1040) 64212* Wno-suggest-override: C++ Dialect Options. 64213 (line 1059) 64214* Wno-switch: Warning Options. (line 1055) 64215* Wno-switch-bool: Warning Options. (line 1075) 64216* Wno-switch-default: Warning Options. (line 1063) 64217* Wno-switch-enum: Warning Options. (line 1066) 64218* Wno-switch-outside-range: Warning Options. (line 1086) 64219* Wno-switch-unreachable: Warning Options. (line 1091) 64220* Wno-sync-nand: Warning Options. (line 1115) 64221* Wno-system-headers: Warning Options. (line 1855) 64222* Wno-tautological-compare: Warning Options. (line 1866) 64223* Wno-templates: C++ Dialect Options. 64224 (line 846) 64225* Wno-terminate: C++ Dialect Options. 64226 (line 908) 64227* Wno-traditional: Warning Options. (line 1905) 64228* Wno-traditional-conversion: Warning Options. (line 1982) 64229* Wno-trampolines: Warning Options. (line 1880) 64230* Wno-type-limits: Warning Options. (line 2140) 64231* Wno-undeclared-selector: Objective-C and Objective-C++ Dialect Options. 64232 (line 208) 64233* Wno-undef: Warning Options. (line 2173) 64234* Wno-uninitialized: Warning Options. (line 1216) 64235* Wno-unknown-pragmas: Warning Options. (line 1299) 64236* Wno-unsafe-loop-optimizations: Warning Options. (line 2108) 64237* Wno-unsuffixed-float-constants: Warning Options. (line 2945) 64238* Wno-unused: Warning Options. (line 1209) 64239* Wno-unused-but-set-parameter: Warning Options. (line 1120) 64240* Wno-unused-but-set-variable: Warning Options. (line 1129) 64241* Wno-unused-const-variable: Warning Options. (line 1176) 64242* Wno-unused-function: Warning Options. (line 1139) 64243* Wno-unused-label: Warning Options. (line 1144) 64244* Wno-unused-local-typedefs: Warning Options. (line 1151) 64245* Wno-unused-parameter: Warning Options. (line 1155) 64246* Wno-unused-result: Warning Options. (line 1162) 64247* Wno-unused-value: Warning Options. (line 1199) 64248* Wno-unused-variable: Warning Options. (line 1167) 64249* Wno-useless-cast: C++ Dialect Options. 64250 (line 1063) 64251* Wno-varargs: Warning Options. (line 2864) 64252* Wno-variadic-macros: Warning Options. (line 2858) 64253* Wno-vector-operation-performance: Warning Options. (line 2869) 64254* Wno-virtual-inheritance: C++ Dialect Options. 64255 (line 885) 64256* Wno-virtual-move-assign: C++ Dialect Options. 64257 (line 892) 64258* Wno-vla: Warning Options. (line 2879) 64259* Wno-vla-larger-than: Warning Options. (line 2883) 64260* Wno-vla-larger-than <1>: Warning Options. (line 2900) 64261* Wno-volatile: C++ Dialect Options. 64262 (line 917) 64263* Wno-volatile-register-var: Warning Options. (line 2904) 64264* Wno-write-strings: Warning Options. (line 2307) 64265* Wno-zero-as-null-pointer-constant: C++ Dialect Options. 64266 (line 930) 64267* Wnoexcept: C++ Dialect Options. 64268 (line 610) 64269* Wnoexcept-type: C++ Dialect Options. 64270 (line 616) 64271* Wnon-template-friend: C++ Dialect Options. 64272 (line 802) 64273* Wnon-virtual-dtor: C++ Dialect Options. 64274 (line 650) 64275* Wnonnull: Warning Options. (line 596) 64276* Wnonnull-compare: Warning Options. (line 603) 64277* Wnormalized: Warning Options. (line 2654) 64278* Wnormalized=: Warning Options. (line 2654) 64279* Wnull-dereference: Warning Options. (line 610) 64280* Wodr: Warning Options. (line 2717) 64281* Wold-style-cast: C++ Dialect Options. 64282 (line 811) 64283* Wold-style-declaration: Warning Options. (line 2577) 64284* Wold-style-definition: Warning Options. (line 2583) 64285* Wopenmp-simd: Warning Options. (line 2721) 64286* Woverflow: Warning Options. (line 2714) 64287* Woverlength-strings: Warning Options. (line 2930) 64288* Woverloaded-virtual: C++ Dialect Options. 64289 (line 817) 64290* Woverride-init: Warning Options. (line 2726) 64291* Woverride-init-side-effects: Warning Options. (line 2734) 64292* Wp: Preprocessor Options. 64293 (line 462) 64294* Wpacked: Warning Options. (line 2739) 64295* Wpacked-bitfield-compat: Warning Options. (line 2756) 64296* Wpacked-not-aligned: Warning Options. (line 2773) 64297* Wpadded: Warning Options. (line 2786) 64298* Wparentheses: Warning Options. (line 921) 64299* Wpedantic: Warning Options. (line 86) 64300* Wpedantic-ms-format: Warning Options. (line 2114) 64301* Wpessimizing-move: C++ Dialect Options. 64302 (line 679) 64303* Wplacement-new: C++ Dialect Options. 64304 (line 945) 64305* Wpmf-conversions: C++ Dialect Options. 64306 (line 836) 64307* Wpointer-arith: Warning Options. (line 2120) 64308* Wpointer-arith <1>: Pointer Arith. (line 13) 64309* Wpointer-compare: Warning Options. (line 2127) 64310* Wpointer-sign: Warning Options. (line 2919) 64311* Wpointer-to-int-cast: Warning Options. (line 2845) 64312* Wpragmas: Warning Options. (line 1306) 64313* Wprio-ctor-dtor: Warning Options. (line 1311) 64314* Wproperty-assign-default: Objective-C and Objective-C++ Dialect Options. 64315 (line 174) 64316* Wprotocol: Objective-C and Objective-C++ Dialect Options. 64317 (line 178) 64318* wrapper: Overall Options. (line 362) 64319* Wredundant-decls: Warning Options. (line 2793) 64320* Wredundant-move: C++ Dialect Options. 64321 (line 701) 64322* Wredundant-tags: C++ Dialect Options. 64323 (line 740) 64324* Wregister: C++ Dialect Options. 64325 (line 658) 64326* Wreorder: C++ Dialect Options. 64327 (line 665) 64328* Wrestrict: Warning Options. (line 2797) 64329* Wreturn-local-addr: Warning Options. (line 1001) 64330* Wreturn-type: Warning Options. (line 1005) 64331* Wscalar-storage-order: Warning Options. (line 2430) 64332* Wselector: Objective-C and Objective-C++ Dialect Options. 64333 (line 188) 64334* Wsequence-point: Warning Options. (line 948) 64335* Wshadow: Warning Options. (line 1996) 64336* Wshadow-ivar: Warning Options. (line 2007) 64337* Wshadow=compatible-local: Warning Options. (line 2018) 64338* Wshadow=global: Warning Options. (line 2011) 64339* Wshadow=local: Warning Options. (line 2014) 64340* Wshift-count-negative: Warning Options. (line 1026) 64341* Wshift-count-overflow: Warning Options. (line 1030) 64342* Wshift-negative-value: Warning Options. (line 1034) 64343* Wshift-overflow: Warning Options. (line 1039) 64344* Wsign-compare: Warning Options. (line 2412) 64345* Wsign-conversion: Warning Options. (line 2418) 64346* Wsign-promo: C++ Dialect Options. 64347 (line 840) 64348* Wsized-deallocation: C++ Dialect Options. 64349 (line 1029) 64350* Wsizeof-array-argument: Warning Options. (line 2464) 64351* Wsizeof-pointer-div: Warning Options. (line 2434) 64352* Wsizeof-pointer-memaccess: Warning Options. (line 2442) 64353* Wstack-protector: Warning Options. (line 2925) 64354* Wstack-usage: Warning Options. (line 2080) 64355* Wstrict-aliasing: Warning Options. (line 1319) 64356* Wstrict-aliasing=n: Warning Options. (line 1326) 64357* Wstrict-null-sentinel: C++ Dialect Options. 64358 (line 795) 64359* Wstrict-overflow: Warning Options. (line 1358) 64360* Wstrict-prototypes: Warning Options. (line 2571) 64361* Wstrict-selector-match: Objective-C and Objective-C++ Dialect Options. 64362 (line 200) 64363* Wstring-compare: Warning Options. (line 1406) 64364* Wstringop-overflow: Warning Options. (line 1427) 64365* Wstringop-overflow <1>: Warning Options. (line 1466) 64366* Wstringop-truncation: Warning Options. (line 1504) 64367* Wsubobject-linkage: C++ Dialect Options. 64368 (line 754) 64369* Wsuggest-attribute=: Warning Options. (line 1555) 64370* Wsuggest-attribute=cold: Warning Options. (line 1598) 64371* Wsuggest-attribute=const: Warning Options. (line 1563) 64372* Wsuggest-attribute=format: Warning Options. (line 1577) 64373* Wsuggest-attribute=malloc: Warning Options. (line 1563) 64374* Wsuggest-attribute=noreturn: Warning Options. (line 1563) 64375* Wsuggest-attribute=pure: Warning Options. (line 1563) 64376* Wsuggest-final-methods: C++ Dialect Options. 64377 (line 1049) 64378* Wsuggest-final-types: C++ Dialect Options. 64379 (line 1040) 64380* Wsuggest-override: C++ Dialect Options. 64381 (line 1059) 64382* Wswitch: Warning Options. (line 1055) 64383* Wswitch-bool: Warning Options. (line 1075) 64384* Wswitch-default: Warning Options. (line 1063) 64385* Wswitch-enum: Warning Options. (line 1066) 64386* Wswitch-outside-range: Warning Options. (line 1086) 64387* Wswitch-unreachable: Warning Options. (line 1091) 64388* Wsync-nand: Warning Options. (line 1115) 64389* Wsystem-headers: Warning Options. (line 1855) 64390* Wtautological-compare: Warning Options. (line 1866) 64391* Wtemplates: C++ Dialect Options. 64392 (line 846) 64393* Wterminate: C++ Dialect Options. 64394 (line 908) 64395* Wtraditional: Warning Options. (line 1905) 64396* Wtraditional-conversion: Warning Options. (line 1982) 64397* Wtrampolines: Warning Options. (line 1880) 64398* Wtrigraphs: Warning Options. (line 2163) 64399* Wtype-limits: Warning Options. (line 2140) 64400* Wundeclared-selector: Objective-C and Objective-C++ Dialect Options. 64401 (line 208) 64402* Wundef: Warning Options. (line 2173) 64403* Wuninitialized: Warning Options. (line 1216) 64404* Wunknown-pragmas: Warning Options. (line 1299) 64405* Wunsafe-loop-optimizations: Warning Options. (line 2108) 64406* Wunsuffixed-float-constants: Warning Options. (line 2945) 64407* Wunused: Warning Options. (line 1209) 64408* Wunused-but-set-parameter: Warning Options. (line 1120) 64409* Wunused-but-set-variable: Warning Options. (line 1129) 64410* Wunused-const-variable: Warning Options. (line 1176) 64411* Wunused-function: Warning Options. (line 1139) 64412* Wunused-label: Warning Options. (line 1144) 64413* Wunused-local-typedefs: Warning Options. (line 1151) 64414* Wunused-macros: Warning Options. (line 2183) 64415* Wunused-parameter: Warning Options. (line 1155) 64416* Wunused-result: Warning Options. (line 1162) 64417* Wunused-value: Warning Options. (line 1199) 64418* Wunused-variable: Warning Options. (line 1167) 64419* Wuseless-cast: C++ Dialect Options. 64420 (line 1063) 64421* Wvarargs: Warning Options. (line 2864) 64422* Wvariadic-macros: Warning Options. (line 2858) 64423* Wvector-operation-performance: Warning Options. (line 2869) 64424* Wvirtual-inheritance: C++ Dialect Options. 64425 (line 885) 64426* Wvirtual-move-assign: C++ Dialect Options. 64427 (line 892) 64428* Wvla: Warning Options. (line 2879) 64429* Wvla-larger-than=: Warning Options. (line 2883) 64430* Wvolatile: C++ Dialect Options. 64431 (line 917) 64432* Wvolatile-register-var: Warning Options. (line 2904) 64433* Wwrite-strings: Warning Options. (line 2307) 64434* Wzero-as-null-pointer-constant: C++ Dialect Options. 64435 (line 930) 64436* Wzero-length-bounds: Warning Options. (line 1827) 64437* Wzero-length-bounds <1>: Warning Options. (line 1827) 64438* x: Overall Options. (line 138) 64439* Xassembler: Assembler Options. (line 13) 64440* Xbind-lazy: VxWorks Options. (line 26) 64441* Xbind-now: VxWorks Options. (line 30) 64442* Xlinker: Link Options. (line 317) 64443* Xpreprocessor: Preprocessor Options. 64444 (line 473) 64445* Ym: System V Options. (line 26) 64446* YP: System V Options. (line 22) 64447* z: Link Options. (line 348) 64448 64449 64450File: gcc.info, Node: Keyword Index, Prev: Option Index, Up: Top 64451 64452Keyword Index 64453************* 64454 64455[index] 64456* Menu: 64457 64458* #pragma: Pragmas. (line 6) 64459* #pragma implementation: C++ Interface. (line 36) 64460* #pragma implementation, implied: C++ Interface. (line 43) 64461* #pragma interface: C++ Interface. (line 17) 64462* $: Dollar Signs. (line 6) 64463* % in constraint: Modifiers. (line 52) 64464* %include: Spec Files. (line 26) 64465* %include_noerr: Spec Files. (line 30) 64466* %rename: Spec Files. (line 34) 64467* & in constraint: Modifiers. (line 25) 64468* ': Incompatibilities. (line 116) 64469* *__builtin_alloca: Other Builtins. (line 129) 64470* *__builtin_alloca_with_align: Other Builtins. (line 166) 64471* *__builtin_alloca_with_align_and_max: Other Builtins. (line 211) 64472* + in constraint: Modifiers. (line 12) 64473* -lgcc, use with -nodefaultlibs: Link Options. (line 154) 64474* -lgcc, use with -nostdlib: Link Options. (line 154) 64475* -march feature modifiers: AArch64 Options. (line 339) 64476* -mcpu feature modifiers: AArch64 Options. (line 339) 64477* -nodefaultlibs and unresolved references: Link Options. (line 154) 64478* -nostdlib and unresolved references: Link Options. (line 154) 64479* .sdata/.sdata2 references (PowerPC): RS/6000 and PowerPC Options. 64480 (line 713) 64481* //: C++ Comments. (line 6) 64482* 0 in constraint: Simple Constraints. (line 125) 64483* < in constraint: Simple Constraints. (line 47) 64484* = in constraint: Modifiers. (line 8) 64485* > in constraint: Simple Constraints. (line 59) 64486* ?: extensions: Conditionals. (line 6) 64487* ?: side effect: Conditionals. (line 20) 64488* _ in variables in macros: Typeof. (line 46) 64489* _Accum data type: Fixed-Point. (line 6) 64490* _Complex keyword: Complex. (line 6) 64491* _Decimal128 data type: Decimal Float. (line 6) 64492* _Decimal32 data type: Decimal Float. (line 6) 64493* _Decimal64 data type: Decimal Float. (line 6) 64494* _Exit: Other Builtins. (line 6) 64495* _exit: Other Builtins. (line 6) 64496* _FloatN data types: Floating Types. (line 6) 64497* _FloatNx data types: Floating Types. (line 6) 64498* _Fract data type: Fixed-Point. (line 6) 64499* _get_ssp: x86 control-flow protection intrinsics. 64500 (line 6) 64501* _HTM_FIRST_USER_ABORT_CODE: S/390 System z Built-in Functions. 64502 (line 44) 64503* _inc_ssp: x86 control-flow protection intrinsics. 64504 (line 12) 64505* _Sat data type: Fixed-Point. (line 6) 64506* _xabort: x86 transactional memory intrinsics. 64507 (line 57) 64508* _xbegin: x86 transactional memory intrinsics. 64509 (line 19) 64510* _xend: x86 transactional memory intrinsics. 64511 (line 48) 64512* _xtest: x86 transactional memory intrinsics. 64513 (line 53) 64514* __atomic_add_fetch: __atomic Builtins. (line 179) 64515* __atomic_always_lock_free: __atomic Builtins. (line 267) 64516* __atomic_and_fetch: __atomic Builtins. (line 183) 64517* __atomic_clear: __atomic Builtins. (line 241) 64518* __atomic_compare_exchange: __atomic Builtins. (line 171) 64519* __atomic_compare_exchange_n: __atomic Builtins. (line 147) 64520* __atomic_exchange: __atomic Builtins. (line 141) 64521* __atomic_exchange_n: __atomic Builtins. (line 131) 64522* __atomic_fetch_add: __atomic Builtins. (line 204) 64523* __atomic_fetch_and: __atomic Builtins. (line 208) 64524* __atomic_fetch_nand: __atomic Builtins. (line 214) 64525* __atomic_fetch_or: __atomic Builtins. (line 212) 64526* __atomic_fetch_sub: __atomic Builtins. (line 206) 64527* __atomic_fetch_xor: __atomic Builtins. (line 210) 64528* __atomic_is_lock_free: __atomic Builtins. (line 281) 64529* __atomic_load: __atomic Builtins. (line 113) 64530* __atomic_load_n: __atomic Builtins. (line 106) 64531* __atomic_nand_fetch: __atomic Builtins. (line 189) 64532* __atomic_or_fetch: __atomic Builtins. (line 187) 64533* __atomic_signal_fence: __atomic Builtins. (line 260) 64534* __atomic_store: __atomic Builtins. (line 126) 64535* __atomic_store_n: __atomic Builtins. (line 118) 64536* __atomic_sub_fetch: __atomic Builtins. (line 181) 64537* __atomic_test_and_set: __atomic Builtins. (line 229) 64538* __atomic_thread_fence: __atomic Builtins. (line 253) 64539* __atomic_xor_fetch: __atomic Builtins. (line 185) 64540* __builtin_addf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64541 (line 17) 64542* __builtin_add_overflow: Integer Overflow Builtins. 64543 (line 9) 64544* __builtin_add_overflow_p: Integer Overflow Builtins. 64545 (line 86) 64546* __builtin_alloca: Other Builtins. (line 6) 64547* __builtin_alloca_with_align: Other Builtins. (line 6) 64548* __builtin_alloca_with_align_and_max: Other Builtins. (line 6) 64549* __builtin_apply: Constructing Calls. (line 29) 64550* __builtin_apply_args: Constructing Calls. (line 19) 64551* __builtin_arc_aligned: ARC Built-in Functions. 64552 (line 18) 64553* __builtin_arc_brk: ARC Built-in Functions. 64554 (line 28) 64555* __builtin_arc_core_read: ARC Built-in Functions. 64556 (line 32) 64557* __builtin_arc_core_write: ARC Built-in Functions. 64558 (line 39) 64559* __builtin_arc_divaw: ARC Built-in Functions. 64560 (line 46) 64561* __builtin_arc_flag: ARC Built-in Functions. 64562 (line 53) 64563* __builtin_arc_lr: ARC Built-in Functions. 64564 (line 57) 64565* __builtin_arc_mul64: ARC Built-in Functions. 64566 (line 64) 64567* __builtin_arc_mulu64: ARC Built-in Functions. 64568 (line 68) 64569* __builtin_arc_nop: ARC Built-in Functions. 64570 (line 73) 64571* __builtin_arc_norm: ARC Built-in Functions. 64572 (line 77) 64573* __builtin_arc_normw: ARC Built-in Functions. 64574 (line 84) 64575* __builtin_arc_rtie: ARC Built-in Functions. 64576 (line 91) 64577* __builtin_arc_sleep: ARC Built-in Functions. 64578 (line 95) 64579* __builtin_arc_sr: ARC Built-in Functions. 64580 (line 99) 64581* __builtin_arc_swap: ARC Built-in Functions. 64582 (line 106) 64583* __builtin_arc_swi: ARC Built-in Functions. 64584 (line 112) 64585* __builtin_arc_sync: ARC Built-in Functions. 64586 (line 116) 64587* __builtin_arc_trap_s: ARC Built-in Functions. 64588 (line 120) 64589* __builtin_arc_unimp_s: ARC Built-in Functions. 64590 (line 124) 64591* __builtin_assume_aligned: Other Builtins. (line 662) 64592* __builtin_bswap16: Other Builtins. (line 983) 64593* __builtin_bswap32: Other Builtins. (line 987) 64594* __builtin_bswap64: Other Builtins. (line 991) 64595* __builtin_call_with_static_chain: Other Builtins. (line 6) 64596* __builtin_call_with_static_chain <1>: Other Builtins. (line 385) 64597* __builtin_choose_expr: Other Builtins. (line 396) 64598* __builtin_clrsb: Other Builtins. (line 913) 64599* __builtin_clrsbl: Other Builtins. (line 935) 64600* __builtin_clrsbll: Other Builtins. (line 958) 64601* __builtin_clz: Other Builtins. (line 905) 64602* __builtin_clzl: Other Builtins. (line 927) 64603* __builtin_clzll: Other Builtins. (line 950) 64604* __builtin_complex: Other Builtins. (line 490) 64605* __builtin_constant_p: Other Builtins. (line 499) 64606* __builtin_convertvector: Vector Extensions. (line 165) 64607* __builtin_cpu_init: Basic PowerPC Built-in Functions Available on all Configurations. 64608 (line 6) 64609* __builtin_cpu_init <1>: x86 Built-in Functions. 64610 (line 68) 64611* __builtin_cpu_is: Basic PowerPC Built-in Functions Available on all Configurations. 64612 (line 10) 64613* __builtin_cpu_is <1>: x86 Built-in Functions. 64614 (line 96) 64615* __builtin_cpu_supports: Basic PowerPC Built-in Functions Available on all Configurations. 64616 (line 70) 64617* __builtin_cpu_supports <1>: x86 Built-in Functions. 64618 (line 240) 64619* __builtin_ctz: Other Builtins. (line 909) 64620* __builtin_ctzl: Other Builtins. (line 931) 64621* __builtin_ctzll: Other Builtins. (line 954) 64622* __builtin_divf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64623 (line 29) 64624* __builtin_expect: Other Builtins. (line 562) 64625* __builtin_expect_with_probability: Other Builtins. (line 597) 64626* __builtin_extend_pointer: Other Builtins. (line 6) 64627* __builtin_extend_pointer <1>: Other Builtins. (line 995) 64628* __builtin_extract_return_addr: Return Address. (line 50) 64629* __builtin_ffs: Other Builtins. (line 901) 64630* __builtin_ffsl: Other Builtins. (line 924) 64631* __builtin_ffsll: Other Builtins. (line 946) 64632* __builtin_FILE: Other Builtins. (line 695) 64633* __builtin_fmaf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64634 (line 37) 64635* __builtin_fpclassify: Other Builtins. (line 6) 64636* __builtin_fpclassify <1>: Other Builtins. (line 797) 64637* __builtin_frame_address: Return Address. (line 62) 64638* __builtin_frob_return_addr: Return Address. (line 59) 64639* __builtin_FUNCTION: Other Builtins. (line 687) 64640* __builtin_goacc_parlevel_id: Other Builtins. (line 1002) 64641* __builtin_goacc_parlevel_size: Other Builtins. (line 1006) 64642* __builtin_has_attribute: Other Builtins. (line 6) 64643* __builtin_has_attribute <1>: Other Builtins. (line 220) 64644* __builtin_huge_val: Other Builtins. (line 777) 64645* __builtin_huge_valf: Other Builtins. (line 782) 64646* __builtin_huge_valfN: Other Builtins. (line 789) 64647* __builtin_huge_valfNx: Other Builtins. (line 793) 64648* __builtin_huge_vall: Other Builtins. (line 785) 64649* __builtin_huge_valq: x86 Built-in Functions. 64650 (line 50) 64651* __builtin_inf: Other Builtins. (line 808) 64652* __builtin_infd128: Other Builtins. (line 818) 64653* __builtin_infd32: Other Builtins. (line 812) 64654* __builtin_infd64: Other Builtins. (line 815) 64655* __builtin_inff: Other Builtins. (line 822) 64656* __builtin_inffN: Other Builtins. (line 831) 64657* __builtin_inffNx: Other Builtins. (line 834) 64658* __builtin_infl: Other Builtins. (line 827) 64659* __builtin_infq: x86 Built-in Functions. 64660 (line 47) 64661* __builtin_isfinite: Other Builtins. (line 6) 64662* __builtin_isgreater: Other Builtins. (line 6) 64663* __builtin_isgreaterequal: Other Builtins. (line 6) 64664* __builtin_isinf_sign: Other Builtins. (line 6) 64665* __builtin_isinf_sign <1>: Other Builtins. (line 837) 64666* __builtin_isless: Other Builtins. (line 6) 64667* __builtin_islessequal: Other Builtins. (line 6) 64668* __builtin_islessgreater: Other Builtins. (line 6) 64669* __builtin_isnormal: Other Builtins. (line 6) 64670* __builtin_isunordered: Other Builtins. (line 6) 64671* __builtin_is_constant_evaluated: Other Builtins. (line 544) 64672* __builtin_LINE: Other Builtins. (line 680) 64673* __builtin_longjmp: Nonlocal Gotos. (line 37) 64674* __builtin_mulf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64675 (line 25) 64676* __builtin_mul_overflow: Integer Overflow Builtins. 64677 (line 63) 64678* __builtin_mul_overflow_p: Integer Overflow Builtins. 64679 (line 90) 64680* __builtin_nan: Other Builtins. (line 845) 64681* __builtin_nand128: Other Builtins. (line 867) 64682* __builtin_nand32: Other Builtins. (line 861) 64683* __builtin_nand64: Other Builtins. (line 864) 64684* __builtin_nanf: Other Builtins. (line 871) 64685* __builtin_nanfN: Other Builtins. (line 878) 64686* __builtin_nanfNx: Other Builtins. (line 881) 64687* __builtin_nanl: Other Builtins. (line 874) 64688* __builtin_nanq: x86 Built-in Functions. 64689 (line 54) 64690* __builtin_nans: Other Builtins. (line 884) 64691* __builtin_nansf: Other Builtins. (line 888) 64692* __builtin_nansfN: Other Builtins. (line 895) 64693* __builtin_nansfNx: Other Builtins. (line 898) 64694* __builtin_nansl: Other Builtins. (line 891) 64695* __builtin_nansq: x86 Built-in Functions. 64696 (line 57) 64697* __builtin_nds32_isb: NDS32 Built-in Functions. 64698 (line 12) 64699* __builtin_nds32_isync: NDS32 Built-in Functions. 64700 (line 8) 64701* __builtin_nds32_mfsr: NDS32 Built-in Functions. 64702 (line 15) 64703* __builtin_nds32_mfusr: NDS32 Built-in Functions. 64704 (line 18) 64705* __builtin_nds32_mtsr: NDS32 Built-in Functions. 64706 (line 21) 64707* __builtin_nds32_mtusr: NDS32 Built-in Functions. 64708 (line 24) 64709* __builtin_nds32_setgie_dis: NDS32 Built-in Functions. 64710 (line 30) 64711* __builtin_nds32_setgie_en: NDS32 Built-in Functions. 64712 (line 27) 64713* __builtin_non_tx_store: S/390 System z Built-in Functions. 64714 (line 98) 64715* __builtin_object_size: Object Size Checking. 64716 (line 6) 64717* __builtin_object_size <1>: Object Size Checking. 64718 (line 16) 64719* __builtin_object_size <2>: Other Builtins. (line 6) 64720* __builtin_object_size <3>: Other Builtins. (line 772) 64721* __builtin_offsetof: Offsetof. (line 6) 64722* __builtin_parity: Other Builtins. (line 921) 64723* __builtin_parityl: Other Builtins. (line 942) 64724* __builtin_parityll: Other Builtins. (line 966) 64725* __builtin_popcount: Other Builtins. (line 918) 64726* __builtin_popcountl: Other Builtins. (line 938) 64727* __builtin_popcountll: Other Builtins. (line 962) 64728* __builtin_powi: Other Builtins. (line 6) 64729* __builtin_powi <1>: Other Builtins. (line 970) 64730* __builtin_powif: Other Builtins. (line 6) 64731* __builtin_powif <1>: Other Builtins. (line 975) 64732* __builtin_powil: Other Builtins. (line 6) 64733* __builtin_powil <1>: Other Builtins. (line 979) 64734* __builtin_prefetch: Other Builtins. (line 733) 64735* __builtin_return: Constructing Calls. (line 47) 64736* __builtin_return_address: Return Address. (line 9) 64737* __builtin_rx_brk: RX Built-in Functions. 64738 (line 10) 64739* __builtin_rx_clrpsw: RX Built-in Functions. 64740 (line 13) 64741* __builtin_rx_int: RX Built-in Functions. 64742 (line 17) 64743* __builtin_rx_machi: RX Built-in Functions. 64744 (line 21) 64745* __builtin_rx_maclo: RX Built-in Functions. 64746 (line 26) 64747* __builtin_rx_mulhi: RX Built-in Functions. 64748 (line 31) 64749* __builtin_rx_mullo: RX Built-in Functions. 64750 (line 36) 64751* __builtin_rx_mvfachi: RX Built-in Functions. 64752 (line 41) 64753* __builtin_rx_mvfacmi: RX Built-in Functions. 64754 (line 45) 64755* __builtin_rx_mvfc: RX Built-in Functions. 64756 (line 49) 64757* __builtin_rx_mvtachi: RX Built-in Functions. 64758 (line 53) 64759* __builtin_rx_mvtaclo: RX Built-in Functions. 64760 (line 57) 64761* __builtin_rx_mvtc: RX Built-in Functions. 64762 (line 61) 64763* __builtin_rx_mvtipl: RX Built-in Functions. 64764 (line 65) 64765* __builtin_rx_racw: RX Built-in Functions. 64766 (line 69) 64767* __builtin_rx_revw: RX Built-in Functions. 64768 (line 73) 64769* __builtin_rx_rmpa: RX Built-in Functions. 64770 (line 78) 64771* __builtin_rx_round: RX Built-in Functions. 64772 (line 82) 64773* __builtin_rx_sat: RX Built-in Functions. 64774 (line 87) 64775* __builtin_rx_setpsw: RX Built-in Functions. 64776 (line 91) 64777* __builtin_rx_wait: RX Built-in Functions. 64778 (line 95) 64779* __builtin_saddll_overflow: Integer Overflow Builtins. 64780 (line 15) 64781* __builtin_saddl_overflow: Integer Overflow Builtins. 64782 (line 13) 64783* __builtin_sadd_overflow: Integer Overflow Builtins. 64784 (line 11) 64785* __builtin_setjmp: Nonlocal Gotos. (line 32) 64786* __builtin_set_thread_pointer: SH Built-in Functions. 64787 (line 9) 64788* __builtin_shuffle: Vector Extensions. (line 127) 64789* __builtin_sh_get_fpscr: SH Built-in Functions. 64790 (line 35) 64791* __builtin_sh_set_fpscr: SH Built-in Functions. 64792 (line 38) 64793* __builtin_smulll_overflow: Integer Overflow Builtins. 64794 (line 69) 64795* __builtin_smull_overflow: Integer Overflow Builtins. 64796 (line 67) 64797* __builtin_smul_overflow: Integer Overflow Builtins. 64798 (line 65) 64799* __builtin_speculation_safe_value: Other Builtins. (line 6) 64800* __builtin_speculation_safe_value <1>: Other Builtins. (line 261) 64801* __builtin_sqrtf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64802 (line 33) 64803* __builtin_ssubll_overflow: Integer Overflow Builtins. 64804 (line 49) 64805* __builtin_ssubl_overflow: Integer Overflow Builtins. 64806 (line 47) 64807* __builtin_ssub_overflow: Integer Overflow Builtins. 64808 (line 45) 64809* __builtin_subf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64810 (line 21) 64811* __builtin_sub_overflow: Integer Overflow Builtins. 64812 (line 43) 64813* __builtin_sub_overflow_p: Integer Overflow Builtins. 64814 (line 88) 64815* __builtin_tabort: S/390 System z Built-in Functions. 64816 (line 82) 64817* __builtin_tbegin: S/390 System z Built-in Functions. 64818 (line 6) 64819* __builtin_tbeginc: S/390 System z Built-in Functions. 64820 (line 73) 64821* __builtin_tbegin_nofloat: S/390 System z Built-in Functions. 64822 (line 54) 64823* __builtin_tbegin_retry: S/390 System z Built-in Functions. 64824 (line 60) 64825* __builtin_tbegin_retry_nofloat: S/390 System z Built-in Functions. 64826 (line 67) 64827* __builtin_tend: S/390 System z Built-in Functions. 64828 (line 77) 64829* __builtin_tgmath: Other Builtins. (line 436) 64830* __builtin_thread_pointer: RISC-V Built-in Functions. 64831 (line 9) 64832* __builtin_thread_pointer <1>: SH Built-in Functions. 64833 (line 18) 64834* __builtin_trap: Other Builtins. (line 606) 64835* __builtin_truncf128_round_to_odd: Basic PowerPC Built-in Functions Available on ISA 3.0. 64836 (line 41) 64837* __builtin_tx_assist: S/390 System z Built-in Functions. 64838 (line 87) 64839* __builtin_tx_nesting_depth: S/390 System z Built-in Functions. 64840 (line 93) 64841* __builtin_types_compatible_p: Other Builtins. (line 340) 64842* __builtin_uaddll_overflow: Integer Overflow Builtins. 64843 (line 21) 64844* __builtin_uaddl_overflow: Integer Overflow Builtins. 64845 (line 19) 64846* __builtin_uadd_overflow: Integer Overflow Builtins. 64847 (line 17) 64848* __builtin_umulll_overflow: Integer Overflow Builtins. 64849 (line 75) 64850* __builtin_umull_overflow: Integer Overflow Builtins. 64851 (line 73) 64852* __builtin_umul_overflow: Integer Overflow Builtins. 64853 (line 71) 64854* __builtin_unreachable: Other Builtins. (line 613) 64855* __builtin_usubll_overflow: Integer Overflow Builtins. 64856 (line 55) 64857* __builtin_usubl_overflow: Integer Overflow Builtins. 64858 (line 53) 64859* __builtin_usub_overflow: Integer Overflow Builtins. 64860 (line 51) 64861* __builtin_va_arg_pack: Constructing Calls. (line 52) 64862* __builtin_va_arg_pack_len: Constructing Calls. (line 75) 64863* __builtin___clear_cache: Other Builtins. (line 720) 64864* __builtin___fprintf_chk: Object Size Checking. 64865 (line 6) 64866* __builtin___memcpy_chk: Object Size Checking. 64867 (line 6) 64868* __builtin___memmove_chk: Object Size Checking. 64869 (line 6) 64870* __builtin___mempcpy_chk: Object Size Checking. 64871 (line 6) 64872* __builtin___memset_chk: Object Size Checking. 64873 (line 6) 64874* __builtin___printf_chk: Object Size Checking. 64875 (line 6) 64876* __builtin___snprintf_chk: Object Size Checking. 64877 (line 6) 64878* __builtin___sprintf_chk: Object Size Checking. 64879 (line 6) 64880* __builtin___stpcpy_chk: Object Size Checking. 64881 (line 6) 64882* __builtin___strcat_chk: Object Size Checking. 64883 (line 6) 64884* __builtin___strcpy_chk: Object Size Checking. 64885 (line 6) 64886* __builtin___strncat_chk: Object Size Checking. 64887 (line 6) 64888* __builtin___strncpy_chk: Object Size Checking. 64889 (line 6) 64890* __builtin___vfprintf_chk: Object Size Checking. 64891 (line 6) 64892* __builtin___vprintf_chk: Object Size Checking. 64893 (line 6) 64894* __builtin___vsnprintf_chk: Object Size Checking. 64895 (line 6) 64896* __builtin___vsprintf_chk: Object Size Checking. 64897 (line 6) 64898* __complex__ keyword: Complex. (line 6) 64899* __declspec(dllexport): Microsoft Windows Function Attributes. 64900 (line 10) 64901* __declspec(dllimport): Microsoft Windows Function Attributes. 64902 (line 42) 64903* __extension__: Alternate Keywords. (line 30) 64904* __far M32C Named Address Spaces: Named Address Spaces. 64905 (line 153) 64906* __far RL78 Named Address Spaces: Named Address Spaces. 64907 (line 162) 64908* __flash AVR Named Address Spaces: Named Address Spaces. 64909 (line 44) 64910* __flash1 AVR Named Address Spaces: Named Address Spaces. 64911 (line 53) 64912* __flash2 AVR Named Address Spaces: Named Address Spaces. 64913 (line 53) 64914* __flash3 AVR Named Address Spaces: Named Address Spaces. 64915 (line 53) 64916* __flash4 AVR Named Address Spaces: Named Address Spaces. 64917 (line 53) 64918* __flash5 AVR Named Address Spaces: Named Address Spaces. 64919 (line 53) 64920* __float128 data type: Floating Types. (line 6) 64921* __float80 data type: Floating Types. (line 6) 64922* __fp16 data type: Half-Precision. (line 6) 64923* __FUNCTION__ identifier: Function Names. (line 6) 64924* __func__ identifier: Function Names. (line 6) 64925* __ibm128 data type: Floating Types. (line 6) 64926* __imag__ keyword: Complex. (line 31) 64927* __int128 data types: __int128. (line 6) 64928* __memx AVR Named Address Spaces: Named Address Spaces. 64929 (line 59) 64930* __PRETTY_FUNCTION__ identifier: Function Names. (line 6) 64931* __real__ keyword: Complex. (line 31) 64932* __seg_fs x86 named address space: Named Address Spaces. 64933 (line 175) 64934* __seg_gs x86 named address space: Named Address Spaces. 64935 (line 175) 64936* __STDC_HOSTED__: Standards. (line 13) 64937* __sync_add_and_fetch: __sync Builtins. (line 72) 64938* __sync_and_and_fetch: __sync Builtins. (line 72) 64939* __sync_bool_compare_and_swap: __sync Builtins. (line 88) 64940* __sync_fetch_and_add: __sync Builtins. (line 50) 64941* __sync_fetch_and_and: __sync Builtins. (line 50) 64942* __sync_fetch_and_nand: __sync Builtins. (line 50) 64943* __sync_fetch_and_or: __sync Builtins. (line 50) 64944* __sync_fetch_and_sub: __sync Builtins. (line 50) 64945* __sync_fetch_and_xor: __sync Builtins. (line 50) 64946* __sync_lock_release: __sync Builtins. (line 118) 64947* __sync_lock_test_and_set: __sync Builtins. (line 100) 64948* __sync_nand_and_fetch: __sync Builtins. (line 72) 64949* __sync_or_and_fetch: __sync Builtins. (line 72) 64950* __sync_sub_and_fetch: __sync Builtins. (line 72) 64951* __sync_synchronize: __sync Builtins. (line 97) 64952* __sync_val_compare_and_swap: __sync Builtins. (line 88) 64953* __sync_xor_and_fetch: __sync Builtins. (line 72) 64954* __thread: Thread-Local. (line 6) 64955* AArch64 Options: AArch64 Options. (line 6) 64956* ABI: Compatibility. (line 6) 64957* abi_tag function attribute: C++ Attributes. (line 9) 64958* abi_tag type attribute: C++ Attributes. (line 9) 64959* abi_tag variable attribute: C++ Attributes. (line 9) 64960* abort: Other Builtins. (line 6) 64961* abs: Other Builtins. (line 6) 64962* absdata variable attribute, AVR: AVR Variable Attributes. 64963 (line 104) 64964* accessing volatiles: Volatiles. (line 6) 64965* accessing volatiles <1>: C++ Volatiles. (line 6) 64966* acos: Other Builtins. (line 6) 64967* acosf: Other Builtins. (line 6) 64968* acosh: Other Builtins. (line 6) 64969* acoshf: Other Builtins. (line 6) 64970* acoshl: Other Builtins. (line 6) 64971* acosl: Other Builtins. (line 6) 64972* Ada: G++ and GCC. (line 6) 64973* Ada <1>: G++ and GCC. (line 29) 64974* additional floating types: Floating Types. (line 6) 64975* address constraints: Simple Constraints. (line 152) 64976* address of a label: Labels as Values. (line 6) 64977* address variable attribute, AVR: AVR Variable Attributes. 64978 (line 97) 64979* address_operand: Simple Constraints. (line 156) 64980* alias function attribute: Common Function Attributes. 64981 (line 75) 64982* alias variable attribute: Common Variable Attributes. 64983 (line 9) 64984* aligned function attribute: Common Function Attributes. 64985 (line 94) 64986* aligned type attribute: Common Type Attributes. 64987 (line 8) 64988* aligned variable attribute: Common Variable Attributes. 64989 (line 31) 64990* alignment: Alignment. (line 6) 64991* alloca: Other Builtins. (line 6) 64992* alloca vs variable-length arrays: Variable Length. (line 35) 64993* alloc_align function attribute: Common Function Attributes. 64994 (line 122) 64995* alloc_size function attribute: Common Function Attributes. 64996 (line 142) 64997* alloc_size type attribute: Common Type Attributes. 64998 (line 136) 64999* alloc_size variable attribute: Common Variable Attributes. 65000 (line 137) 65001* Allow nesting in an interrupt handler on the Blackfin processor: Blackfin Function Attributes. 65002 (line 45) 65003* Altera Nios II options: Nios II Options. (line 6) 65004* alternate keywords: Alternate Keywords. (line 6) 65005* altivec type attribute, PowerPC: PowerPC Type Attributes. 65006 (line 12) 65007* altivec variable attribute, PowerPC: PowerPC Variable Attributes. 65008 (line 12) 65009* always_inline function attribute: Common Function Attributes. 65010 (line 168) 65011* AMD GCN Options: AMD GCN Options. (line 6) 65012* AMD1: Standards. (line 13) 65013* amdgpu_hsa_kernel function attribute, AMD GCN: AMD GCN Function Attributes. 65014 (line 9) 65015* ANSI C: Standards. (line 13) 65016* ANSI C standard: Standards. (line 13) 65017* ANSI C89: Standards. (line 13) 65018* ANSI support: C Dialect Options. (line 10) 65019* ANSI X3.159-1989: Standards. (line 13) 65020* apostrophes: Incompatibilities. (line 116) 65021* application binary interface: Compatibility. (line 6) 65022* ARC options: ARC Options. (line 6) 65023* arch= function attribute, AArch64: AArch64 Function Attributes. 65024 (line 53) 65025* arch= function attribute, ARM: ARM Function Attributes. 65026 (line 98) 65027* ARM options: ARM Options. (line 6) 65028* ARM [Annotated C++ Reference Manual]: Backwards Compatibility. 65029 (line 6) 65030* arrays of length zero: Zero Length. (line 6) 65031* arrays of variable length: Variable Length. (line 6) 65032* arrays, non-lvalue: Subscripting. (line 6) 65033* artificial function attribute: Common Function Attributes. 65034 (line 178) 65035* asin: Other Builtins. (line 6) 65036* asinf: Other Builtins. (line 6) 65037* asinh: Other Builtins. (line 6) 65038* asinhf: Other Builtins. (line 6) 65039* asinhl: Other Builtins. (line 6) 65040* asinl: Other Builtins. (line 6) 65041* asm assembler template: Extended Asm. (line 226) 65042* asm clobbers: Extended Asm. (line 693) 65043* asm constraints: Constraints. (line 6) 65044* asm expressions: Extended Asm. (line 598) 65045* asm flag output operands: Extended Asm. (line 488) 65046* asm goto labels: Extended Asm. (line 880) 65047* asm inline: Size of an asm. (line 25) 65048* asm input operands: Extended Asm. (line 598) 65049* asm keyword: Using Assembly Language with C. 65050 (line 6) 65051* asm output operands: Extended Asm. (line 329) 65052* asm scratch registers: Extended Asm. (line 693) 65053* asm volatile: Extended Asm. (line 116) 65054* assembler names for identifiers: Asm Labels. (line 6) 65055* assembly code, invalid: Bug Criteria. (line 12) 65056* assembly language in C: Using Assembly Language with C. 65057 (line 6) 65058* assembly language in C, basic: Basic Asm. (line 6) 65059* assembly language in C, extended: Extended Asm. (line 6) 65060* assume_aligned function attribute: Common Function Attributes. 65061 (line 186) 65062* atan: Other Builtins. (line 6) 65063* atan2: Other Builtins. (line 6) 65064* atan2f: Other Builtins. (line 6) 65065* atan2l: Other Builtins. (line 6) 65066* atanf: Other Builtins. (line 6) 65067* atanh: Other Builtins. (line 6) 65068* atanhf: Other Builtins. (line 6) 65069* atanhl: Other Builtins. (line 6) 65070* atanl: Other Builtins. (line 6) 65071* attribute of types: Type Attributes. (line 6) 65072* attribute of variables: Variable Attributes. 65073 (line 6) 65074* attribute syntax: Attribute Syntax. (line 6) 65075* autoincrement/decrement addressing: Simple Constraints. (line 30) 65076* automatic inline for C++ member fns: Inline. (line 68) 65077* aux variable attribute, ARC: ARC Variable Attributes. 65078 (line 7) 65079* AVR Options: AVR Options. (line 6) 65080* Backwards Compatibility: Backwards Compatibility. 65081 (line 6) 65082* bank_switch function attribute, M32C: M32C Function Attributes. 65083 (line 9) 65084* base class members: Name lookup. (line 6) 65085* based type attribute, MeP: MeP Type Attributes. 65086 (line 6) 65087* based variable attribute, MeP: MeP Variable Attributes. 65088 (line 16) 65089* basic asm: Basic Asm. (line 6) 65090* bcmp: Other Builtins. (line 6) 65091* below100 variable attribute, Xstormy16: Xstormy16 Variable Attributes. 65092 (line 10) 65093* binary compatibility: Compatibility. (line 6) 65094* Binary constants using the 0b prefix: Binary constants. (line 6) 65095* Blackfin Options: Blackfin Options. (line 6) 65096* bound pointer to member function: Bound member functions. 65097 (line 6) 65098* branch-protection function attribute, AArch64: AArch64 Function Attributes. 65099 (line 76) 65100* break handler functions: MicroBlaze Function Attributes. 65101 (line 17) 65102* break_handler function attribute, MicroBlaze: MicroBlaze Function Attributes. 65103 (line 17) 65104* brk_interrupt function attribute, RL78: RL78 Function Attributes. 65105 (line 10) 65106* bug criteria: Bug Criteria. (line 6) 65107* bugs: Bugs. (line 6) 65108* bugs, known: Trouble. (line 6) 65109* built-in functions: C Dialect Options. (line 266) 65110* built-in functions <1>: Other Builtins. (line 6) 65111* bzero: Other Builtins. (line 6) 65112* C compilation options: Invoking GCC. (line 18) 65113* C intermediate output, nonexistent: G++ and GCC. (line 34) 65114* C language extensions: C Extensions. (line 6) 65115* C language, traditional: Preprocessor Options. 65116 (line 372) 65117* C standard: Standards. (line 13) 65118* C standards: Standards. (line 13) 65119* c++: Invoking G++. (line 14) 65120* C++: G++ and GCC. (line 29) 65121* C++ comments: C++ Comments. (line 6) 65122* C++ interface and implementation headers: C++ Interface. (line 6) 65123* C++ language extensions: C++ Extensions. (line 6) 65124* C++ member fns, automatically inline: Inline. (line 68) 65125* C++ misunderstandings: C++ Misunderstandings. 65126 (line 6) 65127* C++ options, command-line: C++ Dialect Options. 65128 (line 6) 65129* C++ pragmas, effect on inlining: C++ Interface. (line 57) 65130* C++ source file suffixes: Invoking G++. (line 6) 65131* C++ static data, declaring and defining: Static Definitions. 65132 (line 6) 65133* C-SKY Options: C-SKY Options. (line 6) 65134* C11: Standards. (line 13) 65135* C17: Standards. (line 13) 65136* C1X: Standards. (line 13) 65137* C2X: Standards. (line 13) 65138* C6X Options: C6X Options. (line 6) 65139* C89: Standards. (line 13) 65140* C90: Standards. (line 13) 65141* C94: Standards. (line 13) 65142* C95: Standards. (line 13) 65143* C99: Standards. (line 13) 65144* C9X: Standards. (line 13) 65145* cabs: Other Builtins. (line 6) 65146* cabsf: Other Builtins. (line 6) 65147* cabsl: Other Builtins. (line 6) 65148* cacos: Other Builtins. (line 6) 65149* cacosf: Other Builtins. (line 6) 65150* cacosh: Other Builtins. (line 6) 65151* cacoshf: Other Builtins. (line 6) 65152* cacoshl: Other Builtins. (line 6) 65153* cacosl: Other Builtins. (line 6) 65154* callee_pop_aggregate_return function attribute, x86: x86 Function Attributes. 65155 (line 47) 65156* calling functions through the function vector on SH2A: SH Function Attributes. 65157 (line 9) 65158* calloc: Other Builtins. (line 6) 65159* carg: Other Builtins. (line 6) 65160* cargf: Other Builtins. (line 6) 65161* cargl: Other Builtins. (line 6) 65162* case labels in initializers: Designated Inits. (line 6) 65163* case ranges: Case Ranges. (line 6) 65164* casin: Other Builtins. (line 6) 65165* casinf: Other Builtins. (line 6) 65166* casinh: Other Builtins. (line 6) 65167* casinhf: Other Builtins. (line 6) 65168* casinhl: Other Builtins. (line 6) 65169* casinl: Other Builtins. (line 6) 65170* cast to a union: Cast to Union. (line 6) 65171* catan: Other Builtins. (line 6) 65172* catanf: Other Builtins. (line 6) 65173* catanh: Other Builtins. (line 6) 65174* catanhf: Other Builtins. (line 6) 65175* catanhl: Other Builtins. (line 6) 65176* catanl: Other Builtins. (line 6) 65177* cb variable attribute, MeP: MeP Variable Attributes. 65178 (line 46) 65179* cbrt: Other Builtins. (line 6) 65180* cbrtf: Other Builtins. (line 6) 65181* cbrtl: Other Builtins. (line 6) 65182* ccos: Other Builtins. (line 6) 65183* ccosf: Other Builtins. (line 6) 65184* ccosh: Other Builtins. (line 6) 65185* ccoshf: Other Builtins. (line 6) 65186* ccoshl: Other Builtins. (line 6) 65187* ccosl: Other Builtins. (line 6) 65188* cdecl function attribute, x86-32: x86 Function Attributes. 65189 (line 9) 65190* ceil: Other Builtins. (line 6) 65191* ceilf: Other Builtins. (line 6) 65192* ceill: Other Builtins. (line 6) 65193* cexp: Other Builtins. (line 6) 65194* cexpf: Other Builtins. (line 6) 65195* cexpl: Other Builtins. (line 6) 65196* cf_check function attribute, x86: x86 Function Attributes. 65197 (line 625) 65198* character set, execution: Preprocessor Options. 65199 (line 270) 65200* character set, input: Preprocessor Options. 65201 (line 285) 65202* character set, input normalization: Warning Options. (line 2654) 65203* character set, wide execution: Preprocessor Options. 65204 (line 275) 65205* cimag: Other Builtins. (line 6) 65206* cimagf: Other Builtins. (line 6) 65207* cimagl: Other Builtins. (line 6) 65208* cleanup variable attribute: Common Variable Attributes. 65209 (line 161) 65210* clog: Other Builtins. (line 6) 65211* clog10: Other Builtins. (line 6) 65212* clog10f: Other Builtins. (line 6) 65213* clog10l: Other Builtins. (line 6) 65214* clogf: Other Builtins. (line 6) 65215* clogl: Other Builtins. (line 6) 65216* cmodel= function attribute, AArch64: AArch64 Function Attributes. 65217 (line 27) 65218* COBOL: G++ and GCC. (line 23) 65219* code generation conventions: Code Gen Options. (line 6) 65220* code, mixed with declarations: Mixed Declarations. (line 6) 65221* cold function attribute: Common Function Attributes. 65222 (line 202) 65223* cold label attribute: Label Attributes. (line 45) 65224* command options: Invoking GCC. (line 6) 65225* comments, C++ style: C++ Comments. (line 6) 65226* common variable attribute: Common Variable Attributes. 65227 (line 176) 65228* comparison of signed and unsigned values, warning: Warning Options. 65229 (line 2412) 65230* compilation statistics: Developer Options. (line 6) 65231* compiler bugs, reporting: Bug Reporting. (line 6) 65232* compiler compared to C++ preprocessor: G++ and GCC. (line 34) 65233* compiler options, C++: C++ Dialect Options. 65234 (line 6) 65235* compiler options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options. 65236 (line 6) 65237* compiler version, specifying: Invoking GCC. (line 24) 65238* COMPILER_PATH: Environment Variables. 65239 (line 91) 65240* complex conjugation: Complex. (line 38) 65241* complex numbers: Complex. (line 6) 65242* compound literals: Compound Literals. (line 6) 65243* computed gotos: Labels as Values. (line 6) 65244* conditional expressions, extensions: Conditionals. (line 6) 65245* conflicting types: Disappointments. (line 21) 65246* conj: Other Builtins. (line 6) 65247* conjf: Other Builtins. (line 6) 65248* conjl: Other Builtins. (line 6) 65249* const applied to function: Function Attributes. 65250 (line 6) 65251* const function attribute: Common Function Attributes. 65252 (line 218) 65253* const qualifier: Pointers to Arrays. (line 6) 65254* constants in constraints: Simple Constraints. (line 68) 65255* constraint modifier characters: Modifiers. (line 6) 65256* constraint, matching: Simple Constraints. (line 137) 65257* constraints, asm: Constraints. (line 6) 65258* constraints, machine specific: Machine Constraints. 65259 (line 6) 65260* constructing calls: Constructing Calls. (line 6) 65261* constructor expressions: Compound Literals. (line 6) 65262* constructor function attribute: Common Function Attributes. 65263 (line 259) 65264* contributors: Contributors. (line 6) 65265* copy function attribute: Common Function Attributes. 65266 (line 287) 65267* copy type attribute: Common Type Attributes. 65268 (line 161) 65269* copy variable attribute: Common Variable Attributes. 65270 (line 185) 65271* copysign: Other Builtins. (line 6) 65272* copysignf: Other Builtins. (line 6) 65273* copysignl: Other Builtins. (line 6) 65274* core dump: Bug Criteria. (line 9) 65275* cos: Other Builtins. (line 6) 65276* cosf: Other Builtins. (line 6) 65277* cosh: Other Builtins. (line 6) 65278* coshf: Other Builtins. (line 6) 65279* coshl: Other Builtins. (line 6) 65280* cosl: Other Builtins. (line 6) 65281* CPATH: Environment Variables. 65282 (line 127) 65283* CPLUS_INCLUDE_PATH: Environment Variables. 65284 (line 129) 65285* cpow: Other Builtins. (line 6) 65286* cpowf: Other Builtins. (line 6) 65287* cpowl: Other Builtins. (line 6) 65288* cproj: Other Builtins. (line 6) 65289* cprojf: Other Builtins. (line 6) 65290* cprojl: Other Builtins. (line 6) 65291* cpu= function attribute, AArch64: AArch64 Function Attributes. 65292 (line 63) 65293* CR16 Options: CR16 Options. (line 6) 65294* creal: Other Builtins. (line 6) 65295* crealf: Other Builtins. (line 6) 65296* creall: Other Builtins. (line 6) 65297* CRIS Options: CRIS Options. (line 6) 65298* critical function attribute, MSP430: MSP430 Function Attributes. 65299 (line 9) 65300* cross compiling: Invoking GCC. (line 24) 65301* csin: Other Builtins. (line 6) 65302* csinf: Other Builtins. (line 6) 65303* csinh: Other Builtins. (line 6) 65304* csinhf: Other Builtins. (line 6) 65305* csinhl: Other Builtins. (line 6) 65306* csinl: Other Builtins. (line 6) 65307* csqrt: Other Builtins. (line 6) 65308* csqrtf: Other Builtins. (line 6) 65309* csqrtl: Other Builtins. (line 6) 65310* ctan: Other Builtins. (line 6) 65311* ctanf: Other Builtins. (line 6) 65312* ctanh: Other Builtins. (line 6) 65313* ctanhf: Other Builtins. (line 6) 65314* ctanhl: Other Builtins. (line 6) 65315* ctanl: Other Builtins. (line 6) 65316* C_INCLUDE_PATH: Environment Variables. 65317 (line 128) 65318* D: G++ and GCC. (line 6) 65319* Darwin options: Darwin Options. (line 6) 65320* dcgettext: Other Builtins. (line 6) 65321* dd integer suffix: Decimal Float. (line 6) 65322* DD integer suffix: Decimal Float. (line 6) 65323* deallocating variable length arrays: Variable Length. (line 22) 65324* debug dump options: Developer Options. (line 6) 65325* debugging GCC: Developer Options. (line 6) 65326* debugging information options: Debugging Options. (line 6) 65327* decimal floating types: Decimal Float. (line 6) 65328* declaration scope: Incompatibilities. (line 80) 65329* declarations inside expressions: Statement Exprs. (line 6) 65330* declarations, mixed with code: Mixed Declarations. (line 6) 65331* declaring attributes of functions: Function Attributes. 65332 (line 6) 65333* declaring static data in C++: Static Definitions. (line 6) 65334* defining static data in C++: Static Definitions. (line 6) 65335* dependencies for make as output: Environment Variables. 65336 (line 156) 65337* dependencies for make as output <1>: Environment Variables. 65338 (line 172) 65339* dependencies, make: Preprocessor Options. 65340 (line 77) 65341* DEPENDENCIES_OUTPUT: Environment Variables. 65342 (line 155) 65343* dependent name lookup: Name lookup. (line 6) 65344* deprecated enumerator attribute: Enumerator Attributes. 65345 (line 28) 65346* deprecated function attribute: Common Function Attributes. 65347 (line 319) 65348* deprecated type attribute: Common Type Attributes. 65349 (line 189) 65350* deprecated variable attribute: Common Variable Attributes. 65351 (line 201) 65352* designated initializers: Designated Inits. (line 6) 65353* designated_init type attribute: Common Type Attributes. 65354 (line 223) 65355* designator lists: Designated Inits. (line 96) 65356* designators: Designated Inits. (line 64) 65357* destructor function attribute: Common Function Attributes. 65358 (line 259) 65359* developer options: Developer Options. (line 6) 65360* df integer suffix: Decimal Float. (line 6) 65361* DF integer suffix: Decimal Float. (line 6) 65362* dgettext: Other Builtins. (line 6) 65363* diagnostic messages: Diagnostic Message Formatting Options. 65364 (line 6) 65365* dialect options: C Dialect Options. (line 6) 65366* diff-delete GCC_COLORS capability: Diagnostic Message Formatting Options. 65367 (line 119) 65368* diff-filename GCC_COLORS capability: Diagnostic Message Formatting Options. 65369 (line 112) 65370* diff-hunk GCC_COLORS capability: Diagnostic Message Formatting Options. 65371 (line 115) 65372* diff-insert GCC_COLORS capability: Diagnostic Message Formatting Options. 65373 (line 122) 65374* digits in constraint: Simple Constraints. (line 125) 65375* directory options: Directory Options. (line 6) 65376* disinterrupt function attribute, Epiphany: Epiphany Function Attributes. 65377 (line 9) 65378* disinterrupt function attribute, MeP: MeP Function Attributes. 65379 (line 9) 65380* dl integer suffix: Decimal Float. (line 6) 65381* DL integer suffix: Decimal Float. (line 6) 65382* dllexport function attribute: Microsoft Windows Function Attributes. 65383 (line 10) 65384* dllexport variable attribute: Microsoft Windows Variable Attributes. 65385 (line 12) 65386* dllimport function attribute: Microsoft Windows Function Attributes. 65387 (line 42) 65388* dllimport variable attribute: Microsoft Windows Variable Attributes. 65389 (line 12) 65390* dollar signs in identifier names: Dollar Signs. (line 6) 65391* double-word arithmetic: Long Long. (line 6) 65392* downward funargs: Nested Functions. (line 6) 65393* drem: Other Builtins. (line 6) 65394* dremf: Other Builtins. (line 6) 65395* dreml: Other Builtins. (line 6) 65396* dump options: Developer Options. (line 6) 65397* E in constraint: Simple Constraints. (line 87) 65398* earlyclobber operand: Modifiers. (line 25) 65399* eBPF Options: eBPF Options. (line 6) 65400* eight-bit data on the H8/300, H8/300H, and H8S: H8/300 Variable Attributes. 65401 (line 9) 65402* eightbit_data variable attribute, H8/300: H8/300 Variable Attributes. 65403 (line 9) 65404* EIND: AVR Options. (line 319) 65405* either function attribute, MSP430: MSP430 Function Attributes. 65406 (line 57) 65407* either variable attribute, MSP430: MSP430 Variable Attributes. 65408 (line 23) 65409* empty structures: Empty Structures. (line 6) 65410* Enumerator Attributes: Enumerator Attributes. 65411 (line 6) 65412* environment variables: Environment Variables. 65413 (line 6) 65414* erf: Other Builtins. (line 6) 65415* erfc: Other Builtins. (line 6) 65416* erfcf: Other Builtins. (line 6) 65417* erfcl: Other Builtins. (line 6) 65418* erff: Other Builtins. (line 6) 65419* erfl: Other Builtins. (line 6) 65420* error function attribute: Common Function Attributes. 65421 (line 343) 65422* error GCC_COLORS capability: Diagnostic Message Formatting Options. 65423 (line 77) 65424* error messages: Warnings and Errors. 65425 (line 6) 65426* escaped newlines: Escaped Newlines. (line 6) 65427* exception function attribute: NDS32 Function Attributes. 65428 (line 9) 65429* exception handler functions, Blackfin: Blackfin Function Attributes. 65430 (line 9) 65431* exception handler functions, NDS32: NDS32 Function Attributes. 65432 (line 9) 65433* exception_handler function attribute: Blackfin Function Attributes. 65434 (line 9) 65435* exit: Other Builtins. (line 6) 65436* exp: Other Builtins. (line 6) 65437* exp10: Other Builtins. (line 6) 65438* exp10f: Other Builtins. (line 6) 65439* exp10l: Other Builtins. (line 6) 65440* exp2: Other Builtins. (line 6) 65441* exp2f: Other Builtins. (line 6) 65442* exp2l: Other Builtins. (line 6) 65443* expf: Other Builtins. (line 6) 65444* expl: Other Builtins. (line 6) 65445* explicit register variables: Explicit Register Variables. 65446 (line 6) 65447* expm1: Other Builtins. (line 6) 65448* expm1f: Other Builtins. (line 6) 65449* expm1l: Other Builtins. (line 6) 65450* expressions containing statements: Statement Exprs. (line 6) 65451* expressions, constructor: Compound Literals. (line 6) 65452* extended asm: Extended Asm. (line 6) 65453* extensible constraints: Simple Constraints. (line 161) 65454* extensions, ?:: Conditionals. (line 6) 65455* extensions, C language: C Extensions. (line 6) 65456* extensions, C++ language: C++ Extensions. (line 6) 65457* external declaration scope: Incompatibilities. (line 80) 65458* externally_visible function attribute: Common Function Attributes. 65459 (line 360) 65460* extra NOP instructions at the function entry point: Common Function Attributes. 65461 (line 910) 65462* F in constraint: Simple Constraints. (line 92) 65463* fabs: Other Builtins. (line 6) 65464* fabsf: Other Builtins. (line 6) 65465* fabsl: Other Builtins. (line 6) 65466* fallthrough statement attribute: Statement Attributes. 65467 (line 26) 65468* far function attribute, MeP: MeP Function Attributes. 65469 (line 25) 65470* far function attribute, MIPS: MIPS Function Attributes. 65471 (line 63) 65472* far type attribute, MeP: MeP Type Attributes. 65473 (line 6) 65474* far variable attribute, MeP: MeP Variable Attributes. 65475 (line 30) 65476* fastcall function attribute, x86-32: x86 Function Attributes. 65477 (line 15) 65478* fast_interrupt function attribute, M32C: M32C Function Attributes. 65479 (line 14) 65480* fast_interrupt function attribute, MicroBlaze: MicroBlaze Function Attributes. 65481 (line 27) 65482* fast_interrupt function attribute, RX: RX Function Attributes. 65483 (line 9) 65484* fatal signal: Bug Criteria. (line 9) 65485* fdim: Other Builtins. (line 6) 65486* fdimf: Other Builtins. (line 6) 65487* fdiml: Other Builtins. (line 6) 65488* FDL, GNU Free Documentation License: GNU Free Documentation License. 65489 (line 6) 65490* fentry_name function attribute, x86: x86 Function Attributes. 65491 (line 637) 65492* fentry_section function attribute, x86: x86 Function Attributes. 65493 (line 643) 65494* ffs: Other Builtins. (line 6) 65495* file name suffix: Overall Options. (line 14) 65496* file names: Link Options. (line 10) 65497* fix-cortex-a53-835769 function attribute, AArch64: AArch64 Function Attributes. 65498 (line 19) 65499* fixed-point types: Fixed-Point. (line 6) 65500* fixit-delete GCC_COLORS capability: Diagnostic Message Formatting Options. 65501 (line 109) 65502* fixit-insert GCC_COLORS capability: Diagnostic Message Formatting Options. 65503 (line 105) 65504* flatten function attribute: Common Function Attributes. 65505 (line 373) 65506* flexible array members: Zero Length. (line 6) 65507* float as function value type: Incompatibilities. (line 141) 65508* floating point precision: Disappointments. (line 68) 65509* floating-point precision: Optimize Options. (line 2199) 65510* floor: Other Builtins. (line 6) 65511* floorf: Other Builtins. (line 6) 65512* floorl: Other Builtins. (line 6) 65513* fma: Other Builtins. (line 6) 65514* fmaf: Other Builtins. (line 6) 65515* fmal: Other Builtins. (line 6) 65516* fmax: Other Builtins. (line 6) 65517* fmaxf: Other Builtins. (line 6) 65518* fmaxl: Other Builtins. (line 6) 65519* fmin: Other Builtins. (line 6) 65520* fminf: Other Builtins. (line 6) 65521* fminl: Other Builtins. (line 6) 65522* fmod: Other Builtins. (line 6) 65523* fmodf: Other Builtins. (line 6) 65524* fmodl: Other Builtins. (line 6) 65525* force_align_arg_pointer function attribute, x86: x86 Function Attributes. 65526 (line 100) 65527* format function attribute: Common Function Attributes. 65528 (line 381) 65529* format_arg function attribute: Common Function Attributes. 65530 (line 446) 65531* Fortran: G++ and GCC. (line 6) 65532* forwarder_section function attribute, Epiphany: Epiphany Function Attributes. 65533 (line 13) 65534* forwarding calls: Constructing Calls. (line 6) 65535* fprintf: Other Builtins. (line 6) 65536* fprintf_unlocked: Other Builtins. (line 6) 65537* fputs: Other Builtins. (line 6) 65538* fputs_unlocked: Other Builtins. (line 6) 65539* FR30 Options: FR30 Options. (line 6) 65540* free: Other Builtins. (line 6) 65541* freestanding environment: Standards. (line 13) 65542* freestanding implementation: Standards. (line 13) 65543* frexp: Other Builtins. (line 6) 65544* frexpf: Other Builtins. (line 6) 65545* frexpl: Other Builtins. (line 6) 65546* FRV Options: FRV Options. (line 6) 65547* fscanf: Other Builtins. (line 6) 65548* fscanf, and constant strings: Incompatibilities. (line 17) 65549* FT32 Options: FT32 Options. (line 6) 65550* function addressability on the M32R/D: M32R/D Function Attributes. 65551 (line 15) 65552* function attributes: Function Attributes. 65553 (line 6) 65554* function pointers, arithmetic: Pointer Arith. (line 6) 65555* function prototype declarations: Function Prototypes. 65556 (line 6) 65557* function versions: Function Multiversioning. 65558 (line 6) 65559* function, size of pointer to: Pointer Arith. (line 6) 65560* functions in arbitrary sections: Common Function Attributes. 65561 (line 994) 65562* functions that are dynamically resolved: Common Function Attributes. 65563 (line 547) 65564* functions that are passed arguments in registers on x86-32: x86 Function Attributes. 65565 (line 76) 65566* functions that behave like malloc: Common Function Attributes. 65567 (line 674) 65568* functions that have no side effects: Common Function Attributes. 65569 (line 218) 65570* functions that have no side effects <1>: Common Function Attributes. 65571 (line 928) 65572* functions that never return: Common Function Attributes. 65573 (line 842) 65574* functions that pop the argument stack on x86-32: x86 Function Attributes. 65575 (line 9) 65576* functions that pop the argument stack on x86-32 <1>: x86 Function Attributes. 65577 (line 15) 65578* functions that pop the argument stack on x86-32 <2>: x86 Function Attributes. 65579 (line 23) 65580* functions that pop the argument stack on x86-32 <3>: x86 Function Attributes. 65581 (line 108) 65582* functions that return more than once: Common Function Attributes. 65583 (line 985) 65584* functions with non-null pointer arguments: Common Function Attributes. 65585 (line 788) 65586* functions with printf, scanf, strftime or strfmon style arguments: Common Function Attributes. 65587 (line 381) 65588* function_return function attribute, x86: x86 Function Attributes. 65589 (line 562) 65590* function_vector function attribute, H8/300: H8/300 Function Attributes. 65591 (line 9) 65592* function_vector function attribute, M16C/M32C: M32C Function Attributes. 65593 (line 20) 65594* function_vector function attribute, SH: SH Function Attributes. 65595 (line 9) 65596* G in constraint: Simple Constraints. (line 96) 65597* g in constraint: Simple Constraints. (line 118) 65598* g++: Invoking G++. (line 14) 65599* G++: G++ and GCC. (line 29) 65600* gamma: Other Builtins. (line 6) 65601* gammaf: Other Builtins. (line 6) 65602* gammaf_r: Other Builtins. (line 6) 65603* gammal: Other Builtins. (line 6) 65604* gammal_r: Other Builtins. (line 6) 65605* gamma_r: Other Builtins. (line 6) 65606* GCC: G++ and GCC. (line 6) 65607* GCC command options: Invoking GCC. (line 6) 65608* GCC_COLORS environment variable: Diagnostic Message Formatting Options. 65609 (line 40) 65610* GCC_COMPARE_DEBUG: Environment Variables. 65611 (line 52) 65612* GCC_EXEC_PREFIX: Environment Variables. 65613 (line 57) 65614* gcc_struct type attribute, PowerPC: PowerPC Type Attributes. 65615 (line 9) 65616* gcc_struct type attribute, x86: x86 Type Attributes. 65617 (line 11) 65618* gcc_struct variable attribute, PowerPC: PowerPC Variable Attributes. 65619 (line 9) 65620* gcc_struct variable attribute, x86: x86 Variable Attributes. 65621 (line 11) 65622* GCC_URLS environment variable: Diagnostic Message Formatting Options. 65623 (line 129) 65624* gcov: Instrumentation Options. 65625 (line 49) 65626* general-regs-only function attribute, AArch64: AArch64 Function Attributes. 65627 (line 12) 65628* general-regs-only function attribute, ARM: ARM Function Attributes. 65629 (line 9) 65630* gettext: Other Builtins. (line 6) 65631* global offset table: Code Gen Options. (line 353) 65632* global register after longjmp: Global Register Variables. 65633 (line 92) 65634* global register variables: Global Register Variables. 65635 (line 6) 65636* GNAT: G++ and GCC. (line 29) 65637* GNU C Compiler: G++ and GCC. (line 6) 65638* GNU Compiler Collection: G++ and GCC. (line 6) 65639* gnu_inline function attribute: Common Function Attributes. 65640 (line 501) 65641* Go: G++ and GCC. (line 6) 65642* goto with computed label: Labels as Values. (line 6) 65643* gprof: Instrumentation Options. 65644 (line 18) 65645* grouping options: Invoking GCC. (line 31) 65646* H in constraint: Simple Constraints. (line 96) 65647* half-precision floating point: Half-Precision. (line 6) 65648* hardware models and configurations, specifying: Submodel Options. 65649 (line 6) 65650* hex floats: Hex Floats. (line 6) 65651* highlight, color: Diagnostic Message Formatting Options. 65652 (line 40) 65653* hk fixed-suffix: Fixed-Point. (line 6) 65654* HK fixed-suffix: Fixed-Point. (line 6) 65655* hosted environment: Standards. (line 13) 65656* hosted environment <1>: C Dialect Options. (line 306) 65657* hosted environment <2>: C Dialect Options. (line 314) 65658* hosted implementation: Standards. (line 13) 65659* hot function attribute: Common Function Attributes. 65660 (line 537) 65661* hot label attribute: Label Attributes. (line 38) 65662* hotpatch function attribute, S/390: S/390 Function Attributes. 65663 (line 9) 65664* HPPA Options: HPPA Options. (line 6) 65665* hr fixed-suffix: Fixed-Point. (line 6) 65666* HR fixed-suffix: Fixed-Point. (line 6) 65667* hypot: Other Builtins. (line 6) 65668* hypotf: Other Builtins. (line 6) 65669* hypotl: Other Builtins. (line 6) 65670* i in constraint: Simple Constraints. (line 68) 65671* I in constraint: Simple Constraints. (line 79) 65672* IA-64 Options: IA-64 Options. (line 6) 65673* IBM RS/6000 and PowerPC Options: RS/6000 and PowerPC Options. 65674 (line 6) 65675* identifier names, dollar signs in: Dollar Signs. (line 6) 65676* identifiers, names in assembler code: Asm Labels. (line 6) 65677* ifunc function attribute: Common Function Attributes. 65678 (line 547) 65679* ilogb: Other Builtins. (line 6) 65680* ilogbf: Other Builtins. (line 6) 65681* ilogbl: Other Builtins. (line 6) 65682* imaxabs: Other Builtins. (line 6) 65683* implementation-defined behavior, C language: C Implementation. 65684 (line 6) 65685* implementation-defined behavior, C++ language: C++ Implementation. 65686 (line 6) 65687* implied #pragma implementation: C++ Interface. (line 43) 65688* incompatibilities of GCC: Incompatibilities. (line 6) 65689* increment operators: Bug Criteria. (line 17) 65690* index: Other Builtins. (line 6) 65691* indirect calls, ARC: ARC Function Attributes. 65692 (line 27) 65693* indirect calls, ARM: ARM Function Attributes. 65694 (line 38) 65695* indirect calls, Blackfin: Blackfin Function Attributes. 65696 (line 38) 65697* indirect calls, Epiphany: Epiphany Function Attributes. 65698 (line 57) 65699* indirect calls, MIPS: MIPS Function Attributes. 65700 (line 63) 65701* indirect calls, PowerPC: PowerPC Function Attributes. 65702 (line 10) 65703* indirect functions: Common Function Attributes. 65704 (line 547) 65705* indirect_branch function attribute, x86: x86 Function Attributes. 65706 (line 552) 65707* indirect_return function attribute, x86: x86 Function Attributes. 65708 (line 631) 65709* initializations in expressions: Compound Literals. (line 6) 65710* initializers with labeled elements: Designated Inits. (line 6) 65711* initializers, non-constant: Initializers. (line 6) 65712* init_priority variable attribute: C++ Attributes. (line 50) 65713* inline assembly language: Using Assembly Language with C. 65714 (line 6) 65715* inline automatic for C++ member fns: Inline. (line 68) 65716* inline functions: Inline. (line 6) 65717* inline functions, omission of: Inline. (line 51) 65718* inlining and C++ pragmas: C++ Interface. (line 57) 65719* installation trouble: Trouble. (line 6) 65720* instrumentation options: Instrumentation Options. 65721 (line 6) 65722* integrating function code: Inline. (line 6) 65723* interface and implementation headers, C++: C++ Interface. (line 6) 65724* intermediate C version, nonexistent: G++ and GCC. (line 34) 65725* interrupt function attribute, ARC: ARC Function Attributes. 65726 (line 9) 65727* interrupt function attribute, ARM: ARM Function Attributes. 65728 (line 16) 65729* interrupt function attribute, AVR: AVR Function Attributes. 65730 (line 9) 65731* interrupt function attribute, C-SKY: C-SKY Function Attributes. 65732 (line 10) 65733* interrupt function attribute, CR16: CR16 Function Attributes. 65734 (line 9) 65735* interrupt function attribute, Epiphany: Epiphany Function Attributes. 65736 (line 20) 65737* interrupt function attribute, M32C: M32C Function Attributes. 65738 (line 53) 65739* interrupt function attribute, M32R/D: M32R/D Function Attributes. 65740 (line 9) 65741* interrupt function attribute, m68k: m68k Function Attributes. 65742 (line 10) 65743* interrupt function attribute, MeP: MeP Function Attributes. 65744 (line 14) 65745* interrupt function attribute, MIPS: MIPS Function Attributes. 65746 (line 9) 65747* interrupt function attribute, MSP430: MSP430 Function Attributes. 65748 (line 19) 65749* interrupt function attribute, NDS32: NDS32 Function Attributes. 65750 (line 14) 65751* interrupt function attribute, RISC-V: RISC-V Function Attributes. 65752 (line 19) 65753* interrupt function attribute, RL78: RL78 Function Attributes. 65754 (line 10) 65755* interrupt function attribute, RX: RX Function Attributes. 65756 (line 15) 65757* interrupt function attribute, V850: V850 Function Attributes. 65758 (line 10) 65759* interrupt function attribute, Visium: Visium Function Attributes. 65760 (line 9) 65761* interrupt function attribute, x86: x86 Function Attributes. 65762 (line 124) 65763* interrupt function attribute, Xstormy16: Xstormy16 Function Attributes. 65764 (line 9) 65765* interrupt_handler function attribute, Blackfin: Blackfin Function Attributes. 65766 (line 15) 65767* interrupt_handler function attribute, H8/300: H8/300 Function Attributes. 65768 (line 17) 65769* interrupt_handler function attribute, m68k: m68k Function Attributes. 65770 (line 10) 65771* interrupt_handler function attribute, MicroBlaze: MicroBlaze Function Attributes. 65772 (line 27) 65773* interrupt_handler function attribute, SH: SH Function Attributes. 65774 (line 28) 65775* interrupt_handler function attribute, V850: V850 Function Attributes. 65776 (line 10) 65777* interrupt_thread function attribute, fido: m68k Function Attributes. 65778 (line 16) 65779* introduction: Top. (line 6) 65780* invalid assembly code: Bug Criteria. (line 12) 65781* invalid input: Bug Criteria. (line 42) 65782* invoking g++: Invoking G++. (line 22) 65783* io variable attribute, AVR: AVR Variable Attributes. 65784 (line 73) 65785* io variable attribute, MeP: MeP Variable Attributes. 65786 (line 36) 65787* io_low variable attribute, AVR: AVR Variable Attributes. 65788 (line 91) 65789* isalnum: Other Builtins. (line 6) 65790* isalpha: Other Builtins. (line 6) 65791* isascii: Other Builtins. (line 6) 65792* isblank: Other Builtins. (line 6) 65793* iscntrl: Other Builtins. (line 6) 65794* isdigit: Other Builtins. (line 6) 65795* isgraph: Other Builtins. (line 6) 65796* islower: Other Builtins. (line 6) 65797* ISO 9899: Standards. (line 13) 65798* ISO C: Standards. (line 13) 65799* ISO C standard: Standards. (line 13) 65800* ISO C11: Standards. (line 13) 65801* ISO C17: Standards. (line 13) 65802* ISO C1X: Standards. (line 13) 65803* ISO C2X: Standards. (line 13) 65804* ISO C90: Standards. (line 13) 65805* ISO C94: Standards. (line 13) 65806* ISO C95: Standards. (line 13) 65807* ISO C99: Standards. (line 13) 65808* ISO C9X: Standards. (line 13) 65809* ISO support: C Dialect Options. (line 10) 65810* ISO/IEC 9899: Standards. (line 13) 65811* isprint: Other Builtins. (line 6) 65812* ispunct: Other Builtins. (line 6) 65813* isr function attribute, ARM: ARM Function Attributes. 65814 (line 33) 65815* isr function attribute, C-SKY: C-SKY Function Attributes. 65816 (line 10) 65817* isspace: Other Builtins. (line 6) 65818* isupper: Other Builtins. (line 6) 65819* iswalnum: Other Builtins. (line 6) 65820* iswalpha: Other Builtins. (line 6) 65821* iswblank: Other Builtins. (line 6) 65822* iswcntrl: Other Builtins. (line 6) 65823* iswdigit: Other Builtins. (line 6) 65824* iswgraph: Other Builtins. (line 6) 65825* iswlower: Other Builtins. (line 6) 65826* iswprint: Other Builtins. (line 6) 65827* iswpunct: Other Builtins. (line 6) 65828* iswspace: Other Builtins. (line 6) 65829* iswupper: Other Builtins. (line 6) 65830* iswxdigit: Other Builtins. (line 6) 65831* isxdigit: Other Builtins. (line 6) 65832* j0: Other Builtins. (line 6) 65833* j0f: Other Builtins. (line 6) 65834* j0l: Other Builtins. (line 6) 65835* j1: Other Builtins. (line 6) 65836* j1f: Other Builtins. (line 6) 65837* j1l: Other Builtins. (line 6) 65838* jli_always function attribute, ARC: ARC Function Attributes. 65839 (line 44) 65840* jli_fixed function attribute, ARC: ARC Function Attributes. 65841 (line 50) 65842* jn: Other Builtins. (line 6) 65843* jnf: Other Builtins. (line 6) 65844* jnl: Other Builtins. (line 6) 65845* k fixed-suffix: Fixed-Point. (line 6) 65846* K fixed-suffix: Fixed-Point. (line 6) 65847* keep_interrupts_masked function attribute, MIPS: MIPS Function Attributes. 65848 (line 34) 65849* kernel attribute, Nvidia PTX: Nvidia PTX Function Attributes. 65850 (line 9) 65851* kernel helper, function attribute, BPF: BPF Function Attributes. 65852 (line 9) 65853* keywords, alternate: Alternate Keywords. (line 6) 65854* known causes of trouble: Trouble. (line 6) 65855* kspisusp function attribute, Blackfin: Blackfin Function Attributes. 65856 (line 21) 65857* l1_data variable attribute, Blackfin: Blackfin Variable Attributes. 65858 (line 11) 65859* l1_data_A variable attribute, Blackfin: Blackfin Variable Attributes. 65860 (line 11) 65861* l1_data_B variable attribute, Blackfin: Blackfin Variable Attributes. 65862 (line 11) 65863* l1_text function attribute, Blackfin: Blackfin Function Attributes. 65864 (line 26) 65865* l2 function attribute, Blackfin: Blackfin Function Attributes. 65866 (line 32) 65867* l2 variable attribute, Blackfin: Blackfin Variable Attributes. 65868 (line 19) 65869* Label Attributes: Label Attributes. (line 6) 65870* labeled elements in initializers: Designated Inits. (line 6) 65871* labels as values: Labels as Values. (line 6) 65872* labs: Other Builtins. (line 6) 65873* LANG: Environment Variables. 65874 (line 21) 65875* LANG <1>: Environment Variables. 65876 (line 106) 65877* language dialect options: C Dialect Options. (line 6) 65878* LC_ALL: Environment Variables. 65879 (line 21) 65880* LC_CTYPE: Environment Variables. 65881 (line 21) 65882* LC_MESSAGES: Environment Variables. 65883 (line 21) 65884* ldexp: Other Builtins. (line 6) 65885* ldexpf: Other Builtins. (line 6) 65886* ldexpl: Other Builtins. (line 6) 65887* leaf function attribute: Common Function Attributes. 65888 (line 637) 65889* length-zero arrays: Zero Length. (line 6) 65890* lgamma: Other Builtins. (line 6) 65891* lgammaf: Other Builtins. (line 6) 65892* lgammaf_r: Other Builtins. (line 6) 65893* lgammal: Other Builtins. (line 6) 65894* lgammal_r: Other Builtins. (line 6) 65895* lgamma_r: Other Builtins. (line 6) 65896* Libraries: Link Options. (line 82) 65897* LIBRARY_PATH: Environment Variables. 65898 (line 97) 65899* link options: Link Options. (line 6) 65900* linker script: Link Options. (line 311) 65901* lk fixed-suffix: Fixed-Point. (line 6) 65902* LK fixed-suffix: Fixed-Point. (line 6) 65903* LL integer suffix: Long Long. (line 6) 65904* llabs: Other Builtins. (line 6) 65905* llk fixed-suffix: Fixed-Point. (line 6) 65906* LLK fixed-suffix: Fixed-Point. (line 6) 65907* llr fixed-suffix: Fixed-Point. (line 6) 65908* LLR fixed-suffix: Fixed-Point. (line 6) 65909* llrint: Other Builtins. (line 6) 65910* llrintf: Other Builtins. (line 6) 65911* llrintl: Other Builtins. (line 6) 65912* llround: Other Builtins. (line 6) 65913* llroundf: Other Builtins. (line 6) 65914* llroundl: Other Builtins. (line 6) 65915* LM32 options: LM32 Options. (line 6) 65916* load address instruction: Simple Constraints. (line 152) 65917* local labels: Local Labels. (line 6) 65918* local variables in macros: Typeof. (line 46) 65919* local variables, specifying registers: Local Register Variables. 65920 (line 6) 65921* locale: Environment Variables. 65922 (line 21) 65923* locale definition: Environment Variables. 65924 (line 106) 65925* locus GCC_COLORS capability: Diagnostic Message Formatting Options. 65926 (line 98) 65927* log: Other Builtins. (line 6) 65928* log10: Other Builtins. (line 6) 65929* log10f: Other Builtins. (line 6) 65930* log10l: Other Builtins. (line 6) 65931* log1p: Other Builtins. (line 6) 65932* log1pf: Other Builtins. (line 6) 65933* log1pl: Other Builtins. (line 6) 65934* log2: Other Builtins. (line 6) 65935* log2f: Other Builtins. (line 6) 65936* log2l: Other Builtins. (line 6) 65937* logb: Other Builtins. (line 6) 65938* logbf: Other Builtins. (line 6) 65939* logbl: Other Builtins. (line 6) 65940* logf: Other Builtins. (line 6) 65941* logl: Other Builtins. (line 6) 65942* long: BPF Built-in Functions. 65943 (line 8) 65944* long <1>: BPF Built-in Functions. 65945 (line 13) 65946* long <2>: BPF Built-in Functions. 65947 (line 18) 65948* long long data types: Long Long. (line 6) 65949* longcall function attribute, Blackfin: Blackfin Function Attributes. 65950 (line 38) 65951* longcall function attribute, PowerPC: PowerPC Function Attributes. 65952 (line 10) 65953* longjmp: Global Register Variables. 65954 (line 92) 65955* longjmp incompatibilities: Incompatibilities. (line 39) 65956* longjmp warnings: Warning Options. (line 1284) 65957* long_call function attribute, ARC: ARC Function Attributes. 65958 (line 27) 65959* long_call function attribute, ARM: ARM Function Attributes. 65960 (line 38) 65961* long_call function attribute, Epiphany: Epiphany Function Attributes. 65962 (line 57) 65963* long_call function attribute, MIPS: MIPS Function Attributes. 65964 (line 63) 65965* lower function attribute, MSP430: MSP430 Function Attributes. 65966 (line 57) 65967* lower variable attribute, MSP430: MSP430 Variable Attributes. 65968 (line 27) 65969* lr fixed-suffix: Fixed-Point. (line 6) 65970* LR fixed-suffix: Fixed-Point. (line 6) 65971* lrint: Other Builtins. (line 6) 65972* lrintf: Other Builtins. (line 6) 65973* lrintl: Other Builtins. (line 6) 65974* lround: Other Builtins. (line 6) 65975* lroundf: Other Builtins. (line 6) 65976* lroundl: Other Builtins. (line 6) 65977* m in constraint: Simple Constraints. (line 17) 65978* M32C options: M32C Options. (line 6) 65979* M32R/D options: M32R/D Options. (line 6) 65980* M680x0 options: M680x0 Options. (line 6) 65981* machine specific constraints: Machine Constraints. 65982 (line 6) 65983* machine-dependent options: Submodel Options. (line 6) 65984* macro with variable arguments: Variadic Macros. (line 6) 65985* macros, inline alternative: Inline. (line 6) 65986* macros, local labels: Local Labels. (line 6) 65987* macros, local variables in: Typeof. (line 46) 65988* macros, statements in expressions: Statement Exprs. (line 6) 65989* macros, types of arguments: Typeof. (line 6) 65990* make: Preprocessor Options. 65991 (line 77) 65992* malloc: Other Builtins. (line 6) 65993* malloc function attribute: Common Function Attributes. 65994 (line 674) 65995* matching constraint: Simple Constraints. (line 137) 65996* may_alias type attribute: Common Type Attributes. 65997 (line 234) 65998* MCore options: MCore Options. (line 6) 65999* medium_call function attribute, ARC: ARC Function Attributes. 66000 (line 27) 66001* member fns, automatically inline: Inline. (line 68) 66002* memchr: Other Builtins. (line 6) 66003* memcmp: Other Builtins. (line 6) 66004* memcpy: Other Builtins. (line 6) 66005* memory references in constraints: Simple Constraints. (line 17) 66006* mempcpy: Other Builtins. (line 6) 66007* memset: Other Builtins. (line 6) 66008* MeP options: MeP Options. (line 6) 66009* Mercury: G++ and GCC. (line 23) 66010* message formatting: Diagnostic Message Formatting Options. 66011 (line 6) 66012* messages, warning: Warning Options. (line 6) 66013* messages, warning and error: Warnings and Errors. 66014 (line 6) 66015* MicroBlaze Options: MicroBlaze Options. (line 6) 66016* micromips function attribute: MIPS Function Attributes. 66017 (line 91) 66018* middle-operands, omitted: Conditionals. (line 6) 66019* MIPS options: MIPS Options. (line 6) 66020* mips16 function attribute, MIPS: MIPS Function Attributes. 66021 (line 75) 66022* misunderstandings in C++: C++ Misunderstandings. 66023 (line 6) 66024* mixed declarations and code: Mixed Declarations. (line 6) 66025* mixing assembly language and C: Using Assembly Language with C. 66026 (line 6) 66027* mktemp, and constant strings: Incompatibilities. (line 13) 66028* MMIX Options: MMIX Options. (line 6) 66029* MN10300 options: MN10300 Options. (line 6) 66030* mode type attribute: Common Type Attributes. 66031 (line 270) 66032* mode variable attribute: Common Variable Attributes. 66033 (line 225) 66034* model function attribute, M32R/D: M32R/D Function Attributes. 66035 (line 15) 66036* model variable attribute, IA-64: IA-64 Variable Attributes. 66037 (line 9) 66038* model-name variable attribute, M32R/D: M32R/D Variable Attributes. 66039 (line 9) 66040* modf: Other Builtins. (line 6) 66041* modff: Other Builtins. (line 6) 66042* modfl: Other Builtins. (line 6) 66043* modifiers in constraints: Modifiers. (line 6) 66044* Moxie Options: Moxie Options. (line 6) 66045* MSP430 Options: MSP430 Options. (line 6) 66046* ms_abi function attribute, x86: x86 Function Attributes. 66047 (line 34) 66048* ms_hook_prologue function attribute, x86: x86 Function Attributes. 66049 (line 59) 66050* ms_struct type attribute, PowerPC: PowerPC Type Attributes. 66051 (line 9) 66052* ms_struct type attribute, x86: x86 Type Attributes. 66053 (line 11) 66054* ms_struct variable attribute, PowerPC: PowerPC Variable Attributes. 66055 (line 9) 66056* ms_struct variable attribute, x86: x86 Variable Attributes. 66057 (line 11) 66058* multiple alternative constraints: Multi-Alternative. (line 6) 66059* multiprecision arithmetic: Long Long. (line 6) 66060* n in constraint: Simple Constraints. (line 73) 66061* naked function attribute, ARC: ARC Function Attributes. 66062 (line 59) 66063* naked function attribute, ARM: ARM Function Attributes. 66064 (line 48) 66065* naked function attribute, AVR: AVR Function Attributes. 66066 (line 23) 66067* naked function attribute, C-SKY: C-SKY Function Attributes. 66068 (line 20) 66069* naked function attribute, MCORE: MCORE Function Attributes. 66070 (line 9) 66071* naked function attribute, MSP430: MSP430 Function Attributes. 66072 (line 34) 66073* naked function attribute, NDS32: NDS32 Function Attributes. 66074 (line 35) 66075* naked function attribute, RISC-V: RISC-V Function Attributes. 66076 (line 9) 66077* naked function attribute, RL78: RL78 Function Attributes. 66078 (line 20) 66079* naked function attribute, RX: RX Function Attributes. 66080 (line 39) 66081* naked function attribute, x86: x86 Function Attributes. 66082 (line 66) 66083* Named Address Spaces: Named Address Spaces. 66084 (line 6) 66085* names used in assembler code: Asm Labels. (line 6) 66086* naming convention, implementation headers: C++ Interface. (line 43) 66087* NDS32 Options: NDS32 Options. (line 6) 66088* near function attribute, MeP: MeP Function Attributes. 66089 (line 20) 66090* near function attribute, MIPS: MIPS Function Attributes. 66091 (line 63) 66092* near type attribute, MeP: MeP Type Attributes. 66093 (line 6) 66094* near variable attribute, MeP: MeP Variable Attributes. 66095 (line 24) 66096* nearbyint: Other Builtins. (line 6) 66097* nearbyintf: Other Builtins. (line 6) 66098* nearbyintl: Other Builtins. (line 6) 66099* nested function attribute, NDS32: NDS32 Function Attributes. 66100 (line 19) 66101* nested functions: Nested Functions. (line 6) 66102* nested_ready function attribute, NDS32: NDS32 Function Attributes. 66103 (line 23) 66104* nesting function attribute, Blackfin: Blackfin Function Attributes. 66105 (line 45) 66106* newlines (escaped): Escaped Newlines. (line 6) 66107* nextafter: Other Builtins. (line 6) 66108* nextafterf: Other Builtins. (line 6) 66109* nextafterl: Other Builtins. (line 6) 66110* nexttoward: Other Builtins. (line 6) 66111* nexttowardf: Other Builtins. (line 6) 66112* nexttowardl: Other Builtins. (line 6) 66113* NFC: Warning Options. (line 2654) 66114* NFKC: Warning Options. (line 2654) 66115* Nios II options: Nios II Options. (line 6) 66116* nmi function attribute, NDS32: NDS32 Function Attributes. 66117 (line 50) 66118* NMI handler functions on the Blackfin processor: Blackfin Function Attributes. 66119 (line 50) 66120* nmi_handler function attribute, Blackfin: Blackfin Function Attributes. 66121 (line 50) 66122* nocf_check function attribute: x86 Function Attributes. 66123 (line 571) 66124* noclone function attribute: Common Function Attributes. 66125 (line 757) 66126* nocommon variable attribute: Common Variable Attributes. 66127 (line 176) 66128* nocompression function attribute, MIPS: MIPS Function Attributes. 66129 (line 108) 66130* noinit variable attribute: Common Variable Attributes. 66131 (line 397) 66132* noinit variable attribute, MSP430: MSP430 Variable Attributes. 66133 (line 7) 66134* noinline function attribute: Common Function Attributes. 66135 (line 763) 66136* noipa function attribute: Common Function Attributes. 66137 (line 774) 66138* nomicromips function attribute: MIPS Function Attributes. 66139 (line 91) 66140* nomips16 function attribute, MIPS: MIPS Function Attributes. 66141 (line 75) 66142* non-constant initializers: Initializers. (line 6) 66143* non-static inline function: Inline. (line 82) 66144* nonlocal gotos: Nonlocal Gotos. (line 6) 66145* nonnull function attribute: Common Function Attributes. 66146 (line 788) 66147* nonstring variable attribute: Common Variable Attributes. 66148 (line 237) 66149* noplt function attribute: Common Function Attributes. 66150 (line 818) 66151* noreturn function attribute: Common Function Attributes. 66152 (line 842) 66153* nosave_low_regs function attribute, SH: SH Function Attributes. 66154 (line 34) 66155* note GCC_COLORS capability: Diagnostic Message Formatting Options. 66156 (line 83) 66157* nothrow function attribute: Common Function Attributes. 66158 (line 876) 66159* notshared type attribute, ARM: ARM Type Attributes. 66160 (line 6) 66161* not_nested function attribute, NDS32: NDS32 Function Attributes. 66162 (line 21) 66163* no_caller_saved_registers function attribute, x86: x86 Function Attributes. 66164 (line 113) 66165* no_gccisr function attribute, AVR: AVR Function Attributes. 66166 (line 33) 66167* no_icf function attribute: Common Function Attributes. 66168 (line 692) 66169* no_instrument_function function attribute: Common Function Attributes. 66170 (line 696) 66171* no_profile_instrument_function function attribute: Common Function Attributes. 66172 (line 702) 66173* no_reorder function attribute: Common Function Attributes. 66174 (line 707) 66175* no_sanitize function attribute: Common Function Attributes. 66176 (line 715) 66177* no_sanitize_address function attribute: Common Function Attributes. 66178 (line 727) 66179* no_sanitize_thread function attribute: Common Function Attributes. 66180 (line 735) 66181* no_sanitize_undefined function attribute: Common Function Attributes. 66182 (line 740) 66183* no_split_stack function attribute: Common Function Attributes. 66184 (line 746) 66185* no_stack_limit function attribute: Common Function Attributes. 66186 (line 752) 66187* Nvidia PTX options: Nvidia PTX Options. (line 6) 66188* nvptx options: Nvidia PTX Options. (line 6) 66189* o in constraint: Simple Constraints. (line 23) 66190* OBJC_INCLUDE_PATH: Environment Variables. 66191 (line 130) 66192* Objective-C: G++ and GCC. (line 6) 66193* Objective-C <1>: Standards. (line 189) 66194* Objective-C and Objective-C++ options, command-line: Objective-C and Objective-C++ Dialect Options. 66195 (line 6) 66196* Objective-C++: G++ and GCC. (line 6) 66197* Objective-C++ <1>: Standards. (line 189) 66198* offsettable address: Simple Constraints. (line 23) 66199* old-style function definitions: Function Prototypes. 66200 (line 6) 66201* omit-leaf-frame-pointer function attribute, AArch64: AArch64 Function Attributes. 66202 (line 41) 66203* omitted middle-operands: Conditionals. (line 6) 66204* open coding: Inline. (line 6) 66205* OpenACC accelerator programming: C Dialect Options. (line 325) 66206* OpenACC accelerator programming <1>: C Dialect Options. (line 334) 66207* OpenMP parallel: C Dialect Options. (line 340) 66208* OpenMP SIMD: C Dialect Options. (line 349) 66209* OpenRISC Options: OpenRISC Options. (line 6) 66210* operand constraints, asm: Constraints. (line 6) 66211* optimize function attribute: Common Function Attributes. 66212 (line 884) 66213* optimize options: Optimize Options. (line 6) 66214* options to control diagnostics formatting: Diagnostic Message Formatting Options. 66215 (line 6) 66216* options to control warnings: Warning Options. (line 6) 66217* options, C++: C++ Dialect Options. 66218 (line 6) 66219* options, code generation: Code Gen Options. (line 6) 66220* options, debugging: Debugging Options. (line 6) 66221* options, dialect: C Dialect Options. (line 6) 66222* options, directory search: Directory Options. (line 6) 66223* options, GCC command: Invoking GCC. (line 6) 66224* options, grouping: Invoking GCC. (line 31) 66225* options, linking: Link Options. (line 6) 66226* options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options. 66227 (line 6) 66228* options, optimization: Optimize Options. (line 6) 66229* options, order: Invoking GCC. (line 35) 66230* options, preprocessor: Preprocessor Options. 66231 (line 6) 66232* options, profiling: Instrumentation Options. 66233 (line 6) 66234* options, program instrumentation: Instrumentation Options. 66235 (line 6) 66236* options, run-time error checking: Instrumentation Options. 66237 (line 6) 66238* order of evaluation, side effects: Non-bugs. (line 196) 66239* order of options: Invoking GCC. (line 35) 66240* OS_main function attribute, AVR: AVR Function Attributes. 66241 (line 56) 66242* OS_task function attribute, AVR: AVR Function Attributes. 66243 (line 56) 66244* other register constraints: Simple Constraints. (line 161) 66245* outline-atomics function attribute, AArch64: AArch64 Function Attributes. 66246 (line 82) 66247* output file option: Overall Options. (line 196) 66248* overloaded virtual function, warning: C++ Dialect Options. 66249 (line 817) 66250* p in constraint: Simple Constraints. (line 152) 66251* packed type attribute: Common Type Attributes. 66252 (line 282) 66253* packed variable attribute: Common Variable Attributes. 66254 (line 270) 66255* parameter forward declaration: Variable Length. (line 66) 66256* partial_save function attribute, NDS32: NDS32 Function Attributes. 66257 (line 31) 66258* patchable_function_entry function attribute: Common Function Attributes. 66259 (line 910) 66260* path GCC_COLORS capability: Diagnostic Message Formatting Options. 66261 (line 86) 66262* pcs function attribute, ARM: ARM Function Attributes. 66263 (line 58) 66264* PDP-11 Options: PDP-11 Options. (line 6) 66265* persistent variable attribute, MSP430: MSP430 Variable Attributes. 66266 (line 12) 66267* PIC: Code Gen Options. (line 353) 66268* picoChip options: picoChip Options. (line 6) 66269* pmf: Bound member functions. 66270 (line 6) 66271* pointer arguments: Common Function Attributes. 66272 (line 247) 66273* pointer arguments in variadic functions: Variadic Pointer Args. 66274 (line 6) 66275* pointer to member function: Bound member functions. 66276 (line 6) 66277* pointers to arrays: Pointers to Arrays. (line 6) 66278* portions of temporary objects, pointers to: Temporaries. (line 6) 66279* pow: Other Builtins. (line 6) 66280* pow10: Other Builtins. (line 6) 66281* pow10f: Other Builtins. (line 6) 66282* pow10l: Other Builtins. (line 6) 66283* PowerPC options: PowerPC Options. (line 6) 66284* powf: Other Builtins. (line 6) 66285* powl: Other Builtins. (line 6) 66286* pragma GCC ivdep: Loop-Specific Pragmas. 66287 (line 7) 66288* pragma GCC optimize: Function Specific Option Pragmas. 66289 (line 21) 66290* pragma GCC pop_options: Function Specific Option Pragmas. 66291 (line 33) 66292* pragma GCC push_options: Function Specific Option Pragmas. 66293 (line 33) 66294* pragma GCC reset_options: Function Specific Option Pragmas. 66295 (line 41) 66296* pragma GCC target: Function Specific Option Pragmas. 66297 (line 7) 66298* pragma GCC unroll N: Loop-Specific Pragmas. 66299 (line 37) 66300* pragma, address: M32C Pragmas. (line 15) 66301* pragma, align: Solaris Pragmas. (line 11) 66302* pragma, call: MeP Pragmas. (line 48) 66303* pragma, coprocessor available: MeP Pragmas. (line 13) 66304* pragma, coprocessor call_saved: MeP Pragmas. (line 20) 66305* pragma, coprocessor subclass: MeP Pragmas. (line 28) 66306* pragma, ctable_entry: PRU Pragmas. (line 7) 66307* pragma, custom io_volatile: MeP Pragmas. (line 7) 66308* pragma, diagnostic: Diagnostic Pragmas. (line 14) 66309* pragma, diagnostic <1>: Diagnostic Pragmas. (line 57) 66310* pragma, diagnostic <2>: Diagnostic Pragmas. (line 77) 66311* pragma, diagnostic <3>: Diagnostic Pragmas. (line 99) 66312* pragma, disinterrupt: MeP Pragmas. (line 38) 66313* pragma, fini: Solaris Pragmas. (line 20) 66314* pragma, init: Solaris Pragmas. (line 26) 66315* pragma, longcall: RS/6000 and PowerPC Pragmas. 66316 (line 14) 66317* pragma, long_calls: ARM Pragmas. (line 11) 66318* pragma, long_calls_off: ARM Pragmas. (line 17) 66319* pragma, mark: Darwin Pragmas. (line 11) 66320* pragma, memregs: M32C Pragmas. (line 7) 66321* pragma, no_long_calls: ARM Pragmas. (line 14) 66322* pragma, options align: Darwin Pragmas. (line 14) 66323* pragma, pop_macro: Push/Pop Macro Pragmas. 66324 (line 15) 66325* pragma, push_macro: Push/Pop Macro Pragmas. 66326 (line 11) 66327* pragma, redefine_extname: Symbol-Renaming Pragmas. 66328 (line 13) 66329* pragma, segment: Darwin Pragmas. (line 21) 66330* pragma, unused: Darwin Pragmas. (line 24) 66331* pragma, visibility: Visibility Pragmas. (line 8) 66332* pragma, weak: Weak Pragmas. (line 10) 66333* pragmas: Pragmas. (line 6) 66334* pragmas in C++, effect on inlining: C++ Interface. (line 57) 66335* pragmas, interface and implementation: C++ Interface. (line 6) 66336* pragmas, warning of unknown: Warning Options. (line 1299) 66337* precompiled headers: Precompiled Headers. 66338 (line 6) 66339* preprocessing numbers: Incompatibilities. (line 173) 66340* preprocessing tokens: Incompatibilities. (line 173) 66341* preprocessor options: Preprocessor Options. 66342 (line 6) 66343* printf: Other Builtins. (line 6) 66344* printf_unlocked: Other Builtins. (line 6) 66345* prof: Instrumentation Options. 66346 (line 18) 66347* profiling options: Instrumentation Options. 66348 (line 6) 66349* progmem variable attribute, AVR: AVR Variable Attributes. 66350 (line 7) 66351* program instrumentation options: Instrumentation Options. 66352 (line 6) 66353* promotion of formal parameters: Function Prototypes. 66354 (line 6) 66355* PRU Options: PRU Options. (line 6) 66356* pure function attribute: Common Function Attributes. 66357 (line 928) 66358* push address instruction: Simple Constraints. (line 152) 66359* putchar: Other Builtins. (line 6) 66360* puts: Other Builtins. (line 6) 66361* q floating point suffix: Floating Types. (line 6) 66362* Q floating point suffix: Floating Types. (line 6) 66363* qsort, and global register variables: Global Register Variables. 66364 (line 75) 66365* quote GCC_COLORS capability: Diagnostic Message Formatting Options. 66366 (line 102) 66367* r fixed-suffix: Fixed-Point. (line 6) 66368* R fixed-suffix: Fixed-Point. (line 6) 66369* r in constraint: Simple Constraints. (line 64) 66370* RAMPD: AVR Options. (line 430) 66371* RAMPX: AVR Options. (line 430) 66372* RAMPY: AVR Options. (line 430) 66373* RAMPZ: AVR Options. (line 430) 66374* range1 GCC_COLORS capability: Diagnostic Message Formatting Options. 66375 (line 92) 66376* range2 GCC_COLORS capability: Diagnostic Message Formatting Options. 66377 (line 95) 66378* ranges in case statements: Case Ranges. (line 6) 66379* read-only strings: Incompatibilities. (line 9) 66380* realloc: Other Builtins. (line 6) 66381* reentrant function attribute, MSP430: MSP430 Function Attributes. 66382 (line 44) 66383* register variable after longjmp: Global Register Variables. 66384 (line 92) 66385* registers for local variables: Local Register Variables. 66386 (line 6) 66387* registers in constraints: Simple Constraints. (line 64) 66388* registers, global allocation: Global Register Variables. 66389 (line 6) 66390* registers, global variables in: Global Register Variables. 66391 (line 6) 66392* regparm function attribute, x86: x86 Function Attributes. 66393 (line 76) 66394* relocation truncated to fit (ColdFire): M680x0 Options. (line 322) 66395* relocation truncated to fit (MIPS): MIPS Options. (line 237) 66396* remainder: Other Builtins. (line 6) 66397* remainderf: Other Builtins. (line 6) 66398* remainderl: Other Builtins. (line 6) 66399* remquo: Other Builtins. (line 6) 66400* remquof: Other Builtins. (line 6) 66401* remquol: Other Builtins. (line 6) 66402* renesas function attribute, SH: SH Function Attributes. 66403 (line 40) 66404* reordering, warning: C++ Dialect Options. 66405 (line 665) 66406* reporting bugs: Bugs. (line 6) 66407* resbank function attribute, SH: SH Function Attributes. 66408 (line 44) 66409* reset function attribute, NDS32: NDS32 Function Attributes. 66410 (line 45) 66411* reset handler functions: NDS32 Function Attributes. 66412 (line 45) 66413* rest argument (in macro): Variadic Macros. (line 6) 66414* restricted pointers: Restricted Pointers. 66415 (line 6) 66416* restricted references: Restricted Pointers. 66417 (line 6) 66418* restricted this pointer: Restricted Pointers. 66419 (line 6) 66420* returns_nonnull function attribute: Common Function Attributes. 66421 (line 975) 66422* returns_twice function attribute: Common Function Attributes. 66423 (line 985) 66424* rindex: Other Builtins. (line 6) 66425* rint: Other Builtins. (line 6) 66426* rintf: Other Builtins. (line 6) 66427* rintl: Other Builtins. (line 6) 66428* RISC-V Options: RISC-V Options. (line 6) 66429* RL78 Options: RL78 Options. (line 6) 66430* round: Other Builtins. (line 6) 66431* roundf: Other Builtins. (line 6) 66432* roundl: Other Builtins. (line 6) 66433* RS/6000 and PowerPC Options: RS/6000 and PowerPC Options. 66434 (line 6) 66435* RTTI: Vague Linkage. (line 42) 66436* run-time error checking options: Instrumentation Options. 66437 (line 6) 66438* run-time options: Code Gen Options. (line 6) 66439* RX Options: RX Options. (line 6) 66440* s in constraint: Simple Constraints. (line 100) 66441* S/390 and zSeries Options: S/390 and zSeries Options. 66442 (line 6) 66443* saddr variable attribute, RL78: RL78 Variable Attributes. 66444 (line 6) 66445* save all registers on the Blackfin: Blackfin Function Attributes. 66446 (line 56) 66447* save all registers on the H8/300, H8/300H, and H8S: H8/300 Function Attributes. 66448 (line 23) 66449* saveall function attribute, Blackfin: Blackfin Function Attributes. 66450 (line 56) 66451* saveall function attribute, H8/300: H8/300 Function Attributes. 66452 (line 23) 66453* save_all function attribute, NDS32: NDS32 Function Attributes. 66454 (line 28) 66455* save_volatiles function attribute, MicroBlaze: MicroBlaze Function Attributes. 66456 (line 9) 66457* scalar_storage_order type attribute: Common Type Attributes. 66458 (line 317) 66459* scalb: Other Builtins. (line 6) 66460* scalbf: Other Builtins. (line 6) 66461* scalbl: Other Builtins. (line 6) 66462* scalbln: Other Builtins. (line 6) 66463* scalblnf: Other Builtins. (line 6) 66464* scalblnf <1>: Other Builtins. (line 6) 66465* scalbn: Other Builtins. (line 6) 66466* scalbnf: Other Builtins. (line 6) 66467* scanf, and constant strings: Incompatibilities. (line 17) 66468* scanfnl: Other Builtins. (line 6) 66469* scope of a variable length array: Variable Length. (line 22) 66470* scope of declaration: Disappointments. (line 21) 66471* scope of external declarations: Incompatibilities. (line 80) 66472* Score Options: Score Options. (line 6) 66473* sda variable attribute, V850: V850 Variable Attributes. 66474 (line 9) 66475* search path: Directory Options. (line 6) 66476* section function attribute: Common Function Attributes. 66477 (line 994) 66478* section variable attribute: Common Variable Attributes. 66479 (line 292) 66480* secure_call function attribute, ARC: ARC Function Attributes. 66481 (line 54) 66482* selectany variable attribute: Microsoft Windows Variable Attributes. 66483 (line 16) 66484* sentinel function attribute: Common Function Attributes. 66485 (line 1011) 66486* setjmp: Global Register Variables. 66487 (line 92) 66488* setjmp incompatibilities: Incompatibilities. (line 39) 66489* shared attribute, Nvidia PTX: Nvidia PTX Variable Attributes. 66490 (line 9) 66491* shared strings: Incompatibilities. (line 9) 66492* shared variable attribute: Microsoft Windows Variable Attributes. 66493 (line 37) 66494* shortcall function attribute, Blackfin: Blackfin Function Attributes. 66495 (line 38) 66496* shortcall function attribute, PowerPC: PowerPC Function Attributes. 66497 (line 10) 66498* short_call function attribute, ARC: ARC Function Attributes. 66499 (line 27) 66500* short_call function attribute, ARM: ARM Function Attributes. 66501 (line 38) 66502* short_call function attribute, Epiphany: Epiphany Function Attributes. 66503 (line 57) 66504* short_call function attribute, MIPS: MIPS Function Attributes. 66505 (line 63) 66506* side effect in ?:: Conditionals. (line 20) 66507* side effects, macro argument: Statement Exprs. (line 35) 66508* side effects, order of evaluation: Non-bugs. (line 196) 66509* sign-return-address function attribute, AArch64: AArch64 Function Attributes. 66510 (line 69) 66511* signal function attribute, AVR: AVR Function Attributes. 66512 (line 80) 66513* signbit: Other Builtins. (line 6) 66514* signbitd128: Other Builtins. (line 6) 66515* signbitd32: Other Builtins. (line 6) 66516* signbitd64: Other Builtins. (line 6) 66517* signbitf: Other Builtins. (line 6) 66518* signbitl: Other Builtins. (line 6) 66519* signed and unsigned values, comparison warning: Warning Options. 66520 (line 2412) 66521* significand: Other Builtins. (line 6) 66522* significandf: Other Builtins. (line 6) 66523* significandl: Other Builtins. (line 6) 66524* SIMD: C Dialect Options. (line 349) 66525* simd function attribute: Common Function Attributes. 66526 (line 1038) 66527* simple constraints: Simple Constraints. (line 6) 66528* sin: Other Builtins. (line 6) 66529* sincos: Other Builtins. (line 6) 66530* sincosf: Other Builtins. (line 6) 66531* sincosl: Other Builtins. (line 6) 66532* sinf: Other Builtins. (line 6) 66533* sinh: Other Builtins. (line 6) 66534* sinhf: Other Builtins. (line 6) 66535* sinhl: Other Builtins. (line 6) 66536* sinl: Other Builtins. (line 6) 66537* sizeof: Typeof. (line 6) 66538* smaller data references: M32R/D Options. (line 57) 66539* smaller data references <1>: Nios II Options. (line 9) 66540* smaller data references (PowerPC): RS/6000 and PowerPC Options. 66541 (line 713) 66542* snprintf: Other Builtins. (line 6) 66543* Solaris 2 options: Solaris 2 Options. (line 6) 66544* SOURCE_DATE_EPOCH: Environment Variables. 66545 (line 177) 66546* SPARC options: SPARC Options. (line 6) 66547* Spec Files: Spec Files. (line 6) 66548* specified registers: Explicit Register Variables. 66549 (line 6) 66550* specifying compiler version and target machine: Invoking GCC. 66551 (line 24) 66552* specifying hardware config: Submodel Options. (line 6) 66553* specifying machine version: Invoking GCC. (line 24) 66554* specifying registers for local variables: Local Register Variables. 66555 (line 6) 66556* speed of compilation: Precompiled Headers. 66557 (line 6) 66558* sprintf: Other Builtins. (line 6) 66559* sp_switch function attribute, SH: SH Function Attributes. 66560 (line 58) 66561* sqrt: Other Builtins. (line 6) 66562* sqrtf: Other Builtins. (line 6) 66563* sqrtl: Other Builtins. (line 6) 66564* sscanf: Other Builtins. (line 6) 66565* sscanf, and constant strings: Incompatibilities. (line 17) 66566* sseregparm function attribute, x86: x86 Function Attributes. 66567 (line 93) 66568* stack_protect function attribute: Common Function Attributes. 66569 (line 1057) 66570* Statement Attributes: Statement Attributes. 66571 (line 6) 66572* statements inside expressions: Statement Exprs. (line 6) 66573* static data in C++, declaring and defining: Static Definitions. 66574 (line 6) 66575* stdcall function attribute, x86-32: x86 Function Attributes. 66576 (line 108) 66577* stpcpy: Other Builtins. (line 6) 66578* stpncpy: Other Builtins. (line 6) 66579* strcasecmp: Other Builtins. (line 6) 66580* strcat: Other Builtins. (line 6) 66581* strchr: Other Builtins. (line 6) 66582* strcmp: Other Builtins. (line 6) 66583* strcpy: Other Builtins. (line 6) 66584* strcspn: Other Builtins. (line 6) 66585* strdup: Other Builtins. (line 6) 66586* strfmon: Other Builtins. (line 6) 66587* strftime: Other Builtins. (line 6) 66588* strict-align function attribute, AArch64: AArch64 Function Attributes. 66589 (line 33) 66590* string constants: Incompatibilities. (line 9) 66591* strlen: Other Builtins. (line 6) 66592* strncasecmp: Other Builtins. (line 6) 66593* strncat: Other Builtins. (line 6) 66594* strncmp: Other Builtins. (line 6) 66595* strncpy: Other Builtins. (line 6) 66596* strndup: Other Builtins. (line 6) 66597* strnlen: Other Builtins. (line 6) 66598* strpbrk: Other Builtins. (line 6) 66599* strrchr: Other Builtins. (line 6) 66600* strspn: Other Builtins. (line 6) 66601* strstr: Other Builtins. (line 6) 66602* struct: Unnamed Fields. (line 6) 66603* struct __htm_tdb: S/390 System z Built-in Functions. 66604 (line 49) 66605* structures: Incompatibilities. (line 146) 66606* structures, constructor expression: Compound Literals. (line 6) 66607* submodel options: Submodel Options. (line 6) 66608* subscripting: Subscripting. (line 6) 66609* subscripting and function values: Subscripting. (line 6) 66610* suffixes for C++ source: Invoking G++. (line 6) 66611* SUNPRO_DEPENDENCIES: Environment Variables. 66612 (line 171) 66613* suppressing warnings: Warning Options. (line 6) 66614* surprises in C++: C++ Misunderstandings. 66615 (line 6) 66616* symver function attribute: Common Function Attributes. 66617 (line 1101) 66618* syntax checking: Warning Options. (line 13) 66619* syscall_linkage function attribute, IA-64: IA-64 Function Attributes. 66620 (line 9) 66621* system headers, warnings from: Warning Options. (line 1855) 66622* sysv_abi function attribute, x86: x86 Function Attributes. 66623 (line 34) 66624* tan: Other Builtins. (line 6) 66625* tanf: Other Builtins. (line 6) 66626* tanh: Other Builtins. (line 6) 66627* tanhf: Other Builtins. (line 6) 66628* tanhl: Other Builtins. (line 6) 66629* tanl: Other Builtins. (line 6) 66630* target function attribute: Common Function Attributes. 66631 (line 1062) 66632* target function attribute <1>: ARM Function Attributes. 66633 (line 77) 66634* target function attribute <2>: Nios II Function Attributes. 66635 (line 9) 66636* target function attribute <3>: PowerPC Function Attributes. 66637 (line 21) 66638* target function attribute <4>: S/390 Function Attributes. 66639 (line 22) 66640* target function attribute <5>: x86 Function Attributes. 66641 (line 180) 66642* target machine, specifying: Invoking GCC. (line 24) 66643* target("3dnow") function attribute, x86: x86 Function Attributes. 66644 (line 186) 66645* target("3dnowa") function attribute, x86: x86 Function Attributes. 66646 (line 190) 66647* target("abm") function attribute, x86: x86 Function Attributes. 66648 (line 195) 66649* target("adx") function attribute, x86: x86 Function Attributes. 66650 (line 200) 66651* target("aes") function attribute, x86: x86 Function Attributes. 66652 (line 204) 66653* target("align-stringops") function attribute, x86: x86 Function Attributes. 66654 (line 529) 66655* target("altivec") function attribute, PowerPC: PowerPC Function Attributes. 66656 (line 28) 66657* target("arch=ARCH") function attribute, x86: x86 Function Attributes. 66658 (line 538) 66659* target("arm") function attribute, ARM: ARM Function Attributes. 66660 (line 87) 66661* target("avoid-indexed-addresses") function attribute, PowerPC: PowerPC Function Attributes. 66662 (line 142) 66663* target("avx") function attribute, x86: x86 Function Attributes. 66664 (line 208) 66665* target("avx2") function attribute, x86: x86 Function Attributes. 66666 (line 212) 66667* target("avx5124fmaps") function attribute, x86: x86 Function Attributes. 66668 (line 216) 66669* target("avx5124vnniw") function attribute, x86: x86 Function Attributes. 66670 (line 221) 66671* target("avx512bitalg") function attribute, x86: x86 Function Attributes. 66672 (line 226) 66673* target("avx512bw") function attribute, x86: x86 Function Attributes. 66674 (line 231) 66675* target("avx512cd") function attribute, x86: x86 Function Attributes. 66676 (line 235) 66677* target("avx512dq") function attribute, x86: x86 Function Attributes. 66678 (line 239) 66679* target("avx512er") function attribute, x86: x86 Function Attributes. 66680 (line 243) 66681* target("avx512f") function attribute, x86: x86 Function Attributes. 66682 (line 247) 66683* target("avx512ifma") function attribute, x86: x86 Function Attributes. 66684 (line 251) 66685* target("avx512pf") function attribute, x86: x86 Function Attributes. 66686 (line 255) 66687* target("avx512vbmi") function attribute, x86: x86 Function Attributes. 66688 (line 259) 66689* target("avx512vbmi2") function attribute, x86: x86 Function Attributes. 66690 (line 263) 66691* target("avx512vl") function attribute, x86: x86 Function Attributes. 66692 (line 267) 66693* target("avx512vnni") function attribute, x86: x86 Function Attributes. 66694 (line 271) 66695* target("avx512vpopcntdq") function attribute, x86: x86 Function Attributes. 66696 (line 275) 66697* target("bmi") function attribute, x86: x86 Function Attributes. 66698 (line 280) 66699* target("bmi2") function attribute, x86: x86 Function Attributes. 66700 (line 284) 66701* target("cld") function attribute, x86: x86 Function Attributes. 66702 (line 505) 66703* target("cldemote") function attribute, x86: x86 Function Attributes. 66704 (line 288) 66705* target("clflushopt") function attribute, x86: x86 Function Attributes. 66706 (line 292) 66707* target("clwb") function attribute, x86: x86 Function Attributes. 66708 (line 296) 66709* target("clzero") function attribute, x86: x86 Function Attributes. 66710 (line 300) 66711* target("cmpb") function attribute, PowerPC: PowerPC Function Attributes. 66712 (line 34) 66713* target("cpu=CPU") function attribute, PowerPC: PowerPC Function Attributes. 66714 (line 157) 66715* target("crc32") function attribute, x86: x86 Function Attributes. 66716 (line 304) 66717* target("custom-fpu-cfg=NAME") function attribute, Nios II: Nios II Function Attributes. 66718 (line 25) 66719* target("custom-INSN=N") function attribute, Nios II: Nios II Function Attributes. 66720 (line 16) 66721* target("cx16") function attribute, x86: x86 Function Attributes. 66722 (line 308) 66723* target("default") function attribute, x86: x86 Function Attributes. 66724 (line 311) 66725* target("dlmzb") function attribute, PowerPC: PowerPC Function Attributes. 66726 (line 40) 66727* target("f16c") function attribute, x86: x86 Function Attributes. 66728 (line 316) 66729* target("fancy-math-387") function attribute, x86: x86 Function Attributes. 66730 (line 509) 66731* target("fma") function attribute, x86: x86 Function Attributes. 66732 (line 320) 66733* target("fma4") function attribute, x86: x86 Function Attributes. 66734 (line 324) 66735* target("fpmath=FPMATH") function attribute, x86: x86 Function Attributes. 66736 (line 546) 66737* target("fprnd") function attribute, PowerPC: PowerPC Function Attributes. 66738 (line 47) 66739* target("fpu=") function attribute, ARM: ARM Function Attributes. 66740 (line 93) 66741* target("friz") function attribute, PowerPC: PowerPC Function Attributes. 66742 (line 133) 66743* target("fsgsbase") function attribute, x86: x86 Function Attributes. 66744 (line 328) 66745* target("fxsr") function attribute, x86: x86 Function Attributes. 66746 (line 332) 66747* target("gfni") function attribute, x86: x86 Function Attributes. 66748 (line 336) 66749* target("hard-dfp") function attribute, PowerPC: PowerPC Function Attributes. 66750 (line 53) 66751* target("hle") function attribute, x86: x86 Function Attributes. 66752 (line 340) 66753* target("ieee-fp") function attribute, x86: x86 Function Attributes. 66754 (line 514) 66755* target("inline-all-stringops") function attribute, x86: x86 Function Attributes. 66756 (line 519) 66757* target("inline-stringops-dynamically") function attribute, x86: x86 Function Attributes. 66758 (line 523) 66759* target("isel") function attribute, PowerPC: PowerPC Function Attributes. 66760 (line 59) 66761* target("longcall") function attribute, PowerPC: PowerPC Function Attributes. 66762 (line 152) 66763* target("lwp") function attribute, x86: x86 Function Attributes. 66764 (line 344) 66765* target("lzcnt") function attribute, x86: x86 Function Attributes. 66766 (line 348) 66767* target("mfcrf") function attribute, PowerPC: PowerPC Function Attributes. 66768 (line 63) 66769* target("mmx") function attribute, x86: x86 Function Attributes. 66770 (line 352) 66771* target("movbe") function attribute, x86: x86 Function Attributes. 66772 (line 356) 66773* target("movdir64b") function attribute, x86: x86 Function Attributes. 66774 (line 360) 66775* target("movdiri") function attribute, x86: x86 Function Attributes. 66776 (line 364) 66777* target("mulhw") function attribute, PowerPC: PowerPC Function Attributes. 66778 (line 70) 66779* target("multiple") function attribute, PowerPC: PowerPC Function Attributes. 66780 (line 77) 66781* target("mwaitx") function attribute, x86: x86 Function Attributes. 66782 (line 368) 66783* target("no-custom-INSN") function attribute, Nios II: Nios II Function Attributes. 66784 (line 16) 66785* target("paired") function attribute, PowerPC: PowerPC Function Attributes. 66786 (line 147) 66787* target("pclmul") function attribute, x86: x86 Function Attributes. 66788 (line 372) 66789* target("pconfig") function attribute, x86: x86 Function Attributes. 66790 (line 376) 66791* target("pku") function attribute, x86: x86 Function Attributes. 66792 (line 380) 66793* target("popcnt") function attribute, x86: x86 Function Attributes. 66794 (line 384) 66795* target("popcntb") function attribute, PowerPC: PowerPC Function Attributes. 66796 (line 88) 66797* target("popcntd") function attribute, PowerPC: PowerPC Function Attributes. 66798 (line 95) 66799* target("powerpc-gfxopt") function attribute, PowerPC: PowerPC Function Attributes. 66800 (line 101) 66801* target("powerpc-gpopt") function attribute, PowerPC: PowerPC Function Attributes. 66802 (line 107) 66803* target("prefetchwt1") function attribute, x86: x86 Function Attributes. 66804 (line 388) 66805* target("prfchw") function attribute, x86: x86 Function Attributes. 66806 (line 392) 66807* target("ptwrite") function attribute, x86: x86 Function Attributes. 66808 (line 396) 66809* target("rdpid") function attribute, x86: x86 Function Attributes. 66810 (line 400) 66811* target("rdrnd") function attribute, x86: x86 Function Attributes. 66812 (line 404) 66813* target("rdseed") function attribute, x86: x86 Function Attributes. 66814 (line 408) 66815* target("recip") function attribute, x86: x86 Function Attributes. 66816 (line 533) 66817* target("recip-precision") function attribute, PowerPC: PowerPC Function Attributes. 66818 (line 113) 66819* target("rtm") function attribute, x86: x86 Function Attributes. 66820 (line 412) 66821* target("sahf") function attribute, x86: x86 Function Attributes. 66822 (line 416) 66823* target("sgx") function attribute, x86: x86 Function Attributes. 66824 (line 420) 66825* target("sha") function attribute, x86: x86 Function Attributes. 66826 (line 424) 66827* target("shstk") function attribute, x86: x86 Function Attributes. 66828 (line 428) 66829* target("sse") function attribute, x86: x86 Function Attributes. 66830 (line 432) 66831* target("sse2") function attribute, x86: x86 Function Attributes. 66832 (line 436) 66833* target("sse3") function attribute, x86: x86 Function Attributes. 66834 (line 440) 66835* target("sse4") function attribute, x86: x86 Function Attributes. 66836 (line 444) 66837* target("sse4.1") function attribute, x86: x86 Function Attributes. 66838 (line 449) 66839* target("sse4.2") function attribute, x86: x86 Function Attributes. 66840 (line 453) 66841* target("sse4a") function attribute, x86: x86 Function Attributes. 66842 (line 457) 66843* target("ssse3") function attribute, x86: x86 Function Attributes. 66844 (line 461) 66845* target("string") function attribute, PowerPC: PowerPC Function Attributes. 66846 (line 119) 66847* target("tbm") function attribute, x86: x86 Function Attributes. 66848 (line 465) 66849* target("thumb") function attribute, ARM: ARM Function Attributes. 66850 (line 83) 66851* target("tune=TUNE") function attribute, PowerPC: PowerPC Function Attributes. 66852 (line 164) 66853* target("tune=TUNE") function attribute, x86: x86 Function Attributes. 66854 (line 542) 66855* target("update") function attribute, PowerPC: PowerPC Function Attributes. 66856 (line 82) 66857* target("vaes") function attribute, x86: x86 Function Attributes. 66858 (line 469) 66859* target("vpclmulqdq") function attribute, x86: x86 Function Attributes. 66860 (line 473) 66861* target("vsx") function attribute, PowerPC: PowerPC Function Attributes. 66862 (line 125) 66863* target("waitpkg") function attribute, x86: x86 Function Attributes. 66864 (line 477) 66865* target("wbnoinvd") function attribute, x86: x86 Function Attributes. 66866 (line 481) 66867* target("xop") function attribute, x86: x86 Function Attributes. 66868 (line 485) 66869* target("xsave") function attribute, x86: x86 Function Attributes. 66870 (line 489) 66871* target("xsavec") function attribute, x86: x86 Function Attributes. 66872 (line 493) 66873* target("xsaveopt") function attribute, x86: x86 Function Attributes. 66874 (line 497) 66875* target("xsaves") function attribute, x86: x86 Function Attributes. 66876 (line 501) 66877* target-dependent options: Submodel Options. (line 6) 66878* target_clones function attribute: Common Function Attributes. 66879 (line 1131) 66880* TC1: Standards. (line 13) 66881* TC2: Standards. (line 13) 66882* TC3: Standards. (line 13) 66883* tda variable attribute, V850: V850 Variable Attributes. 66884 (line 13) 66885* Technical Corrigenda: Standards. (line 13) 66886* Technical Corrigendum 1: Standards. (line 13) 66887* Technical Corrigendum 2: Standards. (line 13) 66888* Technical Corrigendum 3: Standards. (line 13) 66889* template instantiation: Template Instantiation. 66890 (line 6) 66891* temporaries, lifetime of: Temporaries. (line 6) 66892* tentative definitions: Code Gen Options. (line 231) 66893* TERM_URLS environment variable: Diagnostic Message Formatting Options. 66894 (line 129) 66895* tgamma: Other Builtins. (line 6) 66896* tgammaf: Other Builtins. (line 6) 66897* tgammal: Other Builtins. (line 6) 66898* thiscall function attribute, x86-32: x86 Function Attributes. 66899 (line 23) 66900* Thread-Local Storage: Thread-Local. (line 6) 66901* thunks: Nested Functions. (line 6) 66902* TILE-Gx options: TILE-Gx Options. (line 6) 66903* TILEPro options: TILEPro Options. (line 6) 66904* tiny data section on the H8/300H and H8S: H8/300 Variable Attributes. 66905 (line 19) 66906* tiny type attribute, MeP: MeP Type Attributes. 66907 (line 6) 66908* tiny variable attribute, MeP: MeP Variable Attributes. 66909 (line 20) 66910* tiny_data variable attribute, H8/300: H8/300 Variable Attributes. 66911 (line 19) 66912* TLS: Thread-Local. (line 6) 66913* tls-dialect= function attribute, AArch64: AArch64 Function Attributes. 66914 (line 48) 66915* tls_model variable attribute: Common Variable Attributes. 66916 (line 337) 66917* TMPDIR: Environment Variables. 66918 (line 45) 66919* toascii: Other Builtins. (line 6) 66920* tolower: Other Builtins. (line 6) 66921* toupper: Other Builtins. (line 6) 66922* towlower: Other Builtins. (line 6) 66923* towupper: Other Builtins. (line 6) 66924* traditional C language: Preprocessor Options. 66925 (line 372) 66926* transparent_union type attribute: Common Type Attributes. 66927 (line 357) 66928* trapa_handler function attribute, SH: SH Function Attributes. 66929 (line 73) 66930* trap_exit function attribute, SH: SH Function Attributes. 66931 (line 68) 66932* trunc: Other Builtins. (line 6) 66933* truncf: Other Builtins. (line 6) 66934* truncl: Other Builtins. (line 6) 66935* tune= function attribute, AArch64: AArch64 Function Attributes. 66936 (line 58) 66937* two-stage name lookup: Name lookup. (line 6) 66938* type alignment: Alignment. (line 6) 66939* type attributes: Type Attributes. (line 6) 66940* type-diff GCC_COLORS capability: Diagnostic Message Formatting Options. 66941 (line 125) 66942* typedef names as function parameters: Incompatibilities. (line 97) 66943* typeof: Typeof. (line 6) 66944* type_info: Vague Linkage. (line 42) 66945* uhk fixed-suffix: Fixed-Point. (line 6) 66946* UHK fixed-suffix: Fixed-Point. (line 6) 66947* uhr fixed-suffix: Fixed-Point. (line 6) 66948* UHR fixed-suffix: Fixed-Point. (line 6) 66949* uk fixed-suffix: Fixed-Point. (line 6) 66950* UK fixed-suffix: Fixed-Point. (line 6) 66951* ulk fixed-suffix: Fixed-Point. (line 6) 66952* ULK fixed-suffix: Fixed-Point. (line 6) 66953* ULL integer suffix: Long Long. (line 6) 66954* ullk fixed-suffix: Fixed-Point. (line 6) 66955* ULLK fixed-suffix: Fixed-Point. (line 6) 66956* ullr fixed-suffix: Fixed-Point. (line 6) 66957* ULLR fixed-suffix: Fixed-Point. (line 6) 66958* ulr fixed-suffix: Fixed-Point. (line 6) 66959* ULR fixed-suffix: Fixed-Point. (line 6) 66960* uncached type attribute, ARC: ARC Type Attributes. 66961 (line 6) 66962* undefined behavior: Bug Criteria. (line 17) 66963* undefined function value: Bug Criteria. (line 17) 66964* underscores in variables in macros: Typeof. (line 46) 66965* union: Unnamed Fields. (line 6) 66966* union, casting to a: Cast to Union. (line 6) 66967* unions: Incompatibilities. (line 146) 66968* unknown pragmas, warning: Warning Options. (line 1299) 66969* unresolved references and -nodefaultlibs: Link Options. (line 154) 66970* unresolved references and -nostdlib: Link Options. (line 154) 66971* unused function attribute: Common Function Attributes. 66972 (line 1158) 66973* unused label attribute: Label Attributes. (line 31) 66974* unused type attribute: Common Type Attributes. 66975 (line 410) 66976* unused variable attribute: Common Variable Attributes. 66977 (line 346) 66978* upper function attribute, MSP430: MSP430 Function Attributes. 66979 (line 57) 66980* upper variable attribute, MSP430: MSP430 Variable Attributes. 66981 (line 23) 66982* ur fixed-suffix: Fixed-Point. (line 6) 66983* UR fixed-suffix: Fixed-Point. (line 6) 66984* urls: Diagnostic Message Formatting Options. 66985 (line 129) 66986* used function attribute: Common Function Attributes. 66987 (line 1163) 66988* used variable attribute: Common Variable Attributes. 66989 (line 351) 66990* User stack pointer in interrupts on the Blackfin: Blackfin Function Attributes. 66991 (line 21) 66992* use_debug_exception_return function attribute, MIPS: MIPS Function Attributes. 66993 (line 39) 66994* use_shadow_register_set function attribute, MIPS: MIPS Function Attributes. 66995 (line 28) 66996* V in constraint: Simple Constraints. (line 43) 66997* V850 Options: V850 Options. (line 6) 66998* vague linkage: Vague Linkage. (line 6) 66999* value after longjmp: Global Register Variables. 67000 (line 92) 67001* variable addressability on the M32R/D: M32R/D Variable Attributes. 67002 (line 9) 67003* variable alignment: Alignment. (line 6) 67004* variable attributes: Variable Attributes. 67005 (line 6) 67006* variable number of arguments: Variadic Macros. (line 6) 67007* variable-length array in a structure: Variable Length. (line 26) 67008* variable-length array scope: Variable Length. (line 22) 67009* variable-length arrays: Variable Length. (line 6) 67010* variables in specified registers: Explicit Register Variables. 67011 (line 6) 67012* variables, local, in macros: Typeof. (line 46) 67013* variadic functions, pointer arguments: Variadic Pointer Args. 67014 (line 6) 67015* variadic macros: Variadic Macros. (line 6) 67016* VAX options: VAX Options. (line 6) 67017* vector function attribute, RX: RX Function Attributes. 67018 (line 49) 67019* vector types, using with x86 intrinsics: Vector Extensions. 67020 (line 188) 67021* vector_size type attribute: Common Type Attributes. 67022 (line 419) 67023* vector_size variable attribute: Common Variable Attributes. 67024 (line 360) 67025* version_id function attribute, IA-64: IA-64 Function Attributes. 67026 (line 16) 67027* vfprintf: Other Builtins. (line 6) 67028* vfscanf: Other Builtins. (line 6) 67029* visibility function attribute: Common Function Attributes. 67030 (line 1173) 67031* visibility type attribute: Common Type Attributes. 67032 (line 446) 67033* visibility variable attribute: Common Variable Attributes. 67034 (line 388) 67035* Visium options: Visium Options. (line 6) 67036* VLAs: Variable Length. (line 6) 67037* vliw function attribute, MeP: MeP Function Attributes. 67038 (line 30) 67039* void pointers, arithmetic: Pointer Arith. (line 6) 67040* void, size of pointer to: Pointer Arith. (line 6) 67041* volatile access: Volatiles. (line 6) 67042* volatile access <1>: C++ Volatiles. (line 6) 67043* volatile applied to function: Function Attributes. 67044 (line 6) 67045* volatile asm: Extended Asm. (line 116) 67046* volatile read: Volatiles. (line 6) 67047* volatile read <1>: C++ Volatiles. (line 6) 67048* volatile write: Volatiles. (line 6) 67049* volatile write <1>: C++ Volatiles. (line 6) 67050* vprintf: Other Builtins. (line 6) 67051* vscanf: Other Builtins. (line 6) 67052* vsnprintf: Other Builtins. (line 6) 67053* vsprintf: Other Builtins. (line 6) 67054* vsscanf: Other Builtins. (line 6) 67055* vtable: Vague Linkage. (line 27) 67056* VxWorks Options: VxWorks Options. (line 6) 67057* w floating point suffix: Floating Types. (line 6) 67058* W floating point suffix: Floating Types. (line 6) 67059* wakeup function attribute, MSP430: MSP430 Function Attributes. 67060 (line 49) 67061* warm function attribute, NDS32: NDS32 Function Attributes. 67062 (line 52) 67063* warning for comparison of signed and unsigned values: Warning Options. 67064 (line 2412) 67065* warning for overloaded virtual function: C++ Dialect Options. 67066 (line 817) 67067* warning for reordering of member initializers: C++ Dialect Options. 67068 (line 665) 67069* warning for unknown pragmas: Warning Options. (line 1299) 67070* warning function attribute: Common Function Attributes. 67071 (line 343) 67072* warning GCC_COLORS capability: Diagnostic Message Formatting Options. 67073 (line 80) 67074* warning messages: Warning Options. (line 6) 67075* warnings from system headers: Warning Options. (line 1855) 67076* warnings vs errors: Warnings and Errors. 67077 (line 6) 67078* warn_if_not_aligned type attribute: Common Type Attributes. 67079 (line 91) 67080* warn_if_not_aligned variable attribute: Common Variable Attributes. 67081 (line 106) 67082* warn_unused type attribute: C++ Attributes. (line 71) 67083* warn_unused_result function attribute: Common Function Attributes. 67084 (line 1273) 67085* weak function attribute: Common Function Attributes. 67086 (line 1290) 67087* weak variable attribute: Common Variable Attributes. 67088 (line 393) 67089* weakref function attribute: Common Function Attributes. 67090 (line 1302) 67091* whitespace: Incompatibilities. (line 112) 67092* Windows Options for x86: x86 Windows Options. 67093 (line 6) 67094* X in constraint: Simple Constraints. (line 122) 67095* X3.159-1989: Standards. (line 13) 67096* x86 named address spaces: Named Address Spaces. 67097 (line 170) 67098* x86 Options: x86 Options. (line 6) 67099* x86 Windows Options: x86 Windows Options. 67100 (line 6) 67101* Xstormy16 Options: Xstormy16 Options. (line 6) 67102* Xtensa Options: Xtensa Options. (line 6) 67103* y0: Other Builtins. (line 6) 67104* y0f: Other Builtins. (line 6) 67105* y0l: Other Builtins. (line 6) 67106* y1: Other Builtins. (line 6) 67107* y1f: Other Builtins. (line 6) 67108* y1l: Other Builtins. (line 6) 67109* yn: Other Builtins. (line 6) 67110* ynf: Other Builtins. (line 6) 67111* ynl: Other Builtins. (line 6) 67112* zda variable attribute, V850: V850 Variable Attributes. 67113 (line 17) 67114* zero-length arrays: Zero Length. (line 6) 67115* zero-size structures: Empty Structures. (line 6) 67116* zSeries options: zSeries Options. (line 6) 67117 67118 67119 67120Tag Table: 67121Node: Top2135 67122Node: G++ and GCC4093 67123Node: Standards6153 67124Node: Invoking GCC19770 67125Node: Option Summary24946 67126Node: Overall Options80098 67127Node: Invoking G++95426 67128Node: C Dialect Options96949 67129Node: C++ Dialect Options116181 67130Node: Objective-C and Objective-C++ Dialect Options165323 67131Node: Diagnostic Message Formatting Options176719 67132Node: Warning Options202440 67133Ref: Wtrigraphs298518 67134Node: Static Analyzer Options334203 67135Node: Debugging Options346912 67136Node: Optimize Options366161 67137Ref: Type-punning435730 67138Node: Instrumentation Options543211 67139Node: Preprocessor Options583627 67140Ref: dashMF588474 67141Ref: fdollars-in-identifiers593053 67142Node: Assembler Options605553 67143Node: Link Options606244 67144Ref: Link Options-Footnote-1623120 67145Node: Directory Options623456 67146Node: Code Gen Options631859 67147Node: Developer Options660224 67148Node: Submodel Options702138 67149Node: AArch64 Options703948 67150Ref: aarch64-feature-modifiers719788 67151Node: Adapteva Epiphany Options724900 67152Node: AMD GCN Options730852 67153Node: ARC Options731746 67154Node: ARM Options752972 67155Node: AVR Options794095 67156Node: Blackfin Options820617 67157Node: C6X Options828509 67158Node: CRIS Options830052 67159Node: CR16 Options833791 67160Node: C-SKY Options834703 67161Node: Darwin Options839563 67162Node: DEC Alpha Options847004 67163Node: eBPF Options858620 67164Node: FR30 Options859902 67165Node: FT32 Options860462 67166Node: FRV Options861408 67167Node: GNU/Linux Options868172 67168Node: H8/300 Options869553 67169Node: HPPA Options871005 67170Node: IA-64 Options880537 67171Node: LM32 Options888665 67172Node: M32C Options889188 67173Node: M32R/D Options890461 67174Node: M680x0 Options894006 67175Node: MCore Options908081 67176Node: MeP Options909583 67177Node: MicroBlaze Options913543 67178Node: MIPS Options916633 67179Node: MMIX Options953172 67180Node: MN10300 Options955649 67181Node: Moxie Options958192 67182Node: MSP430 Options958679 67183Node: NDS32 Options965775 67184Node: Nios II Options967945 67185Node: Nvidia PTX Options980107 67186Node: OpenRISC Options982576 67187Node: PDP-11 Options985096 67188Node: picoChip Options986345 67189Node: PowerPC Options988483 67190Node: PRU Options988703 67191Node: RISC-V Options990912 67192Node: RL78 Options996416 67193Node: RS/6000 and PowerPC Options1000191 67194Node: RX Options1041174 67195Node: S/390 and zSeries Options1049776 67196Node: Score Options1060555 67197Node: SH Options1061404 67198Node: Solaris 2 Options1076544 67199Node: SPARC Options1077782 67200Node: System V Options1093434 67201Node: TILE-Gx Options1094262 67202Node: TILEPro Options1095280 67203Node: V850 Options1095784 67204Node: VAX Options1102471 67205Node: Visium Options1103009 67206Node: VMS Options1105317 67207Node: VxWorks Options1106133 67208Node: x86 Options1107285 67209Node: x86 Windows Options1170034 67210Node: Xstormy16 Options1172839 67211Node: Xtensa Options1173133 67212Node: zSeries Options1178282 67213Node: Spec Files1178478 67214Node: Environment Variables1200637 67215Node: Precompiled Headers1209363 67216Node: C Implementation1215552 67217Node: Translation implementation1217242 67218Node: Environment implementation1217833 67219Node: Identifiers implementation1218387 67220Node: Characters implementation1219473 67221Node: Integers implementation1223123 67222Node: Floating point implementation1225172 67223Node: Arrays and pointers implementation1228235 67224Ref: Arrays and pointers implementation-Footnote-11229695 67225Node: Hints implementation1229821 67226Node: Structures unions enumerations and bit-fields implementation1231316 67227Node: Qualifiers implementation1233540 67228Node: Declarators implementation1235601 67229Node: Statements implementation1235942 67230Node: Preprocessing directives implementation1236268 67231Node: Library functions implementation1238589 67232Node: Architecture implementation1239238 67233Node: Locale-specific behavior implementation1240883 67234Node: C++ Implementation1241188 67235Node: Conditionally-supported behavior1242471 67236Node: Exception handling1243088 67237Node: C Extensions1243555 67238Node: Statement Exprs1248759 67239Node: Local Labels1254131 67240Node: Labels as Values1257104 67241Ref: Labels as Values-Footnote-11259631 67242Node: Nested Functions1259816 67243Node: Nonlocal Gotos1263770 67244Node: Constructing Calls1266036 67245Node: Typeof1270748 67246Node: Conditionals1274677 67247Node: __int1281275566 67248Node: Long Long1276091 67249Node: Complex1277582 67250Node: Floating Types1280350 67251Node: Half-Precision1283817 67252Node: Decimal Float1286228 67253Node: Hex Floats1288082 67254Node: Fixed-Point1289156 67255Node: Named Address Spaces1292414 67256Ref: AVR Named Address Spaces1293095 67257Node: Zero Length1299707 67258Node: Empty Structures1303888 67259Node: Variable Length1304294 67260Node: Variadic Macros1307012 67261Node: Escaped Newlines1309390 67262Node: Subscripting1310251 67263Node: Pointer Arith1310976 67264Node: Variadic Pointer Args1311553 67265Node: Pointers to Arrays1312278 67266Node: Initializers1313031 67267Node: Compound Literals1313532 67268Node: Designated Inits1317099 67269Node: Case Ranges1321023 67270Node: Cast to Union1321704 67271Node: Mixed Declarations1323427 67272Node: Function Attributes1323937 67273Node: Common Function Attributes1328393 67274Node: AArch64 Function Attributes1393497 67275Node: AMD GCN Function Attributes1399691 67276Node: ARC Function Attributes1402744 67277Node: ARM Function Attributes1405681 67278Node: AVR Function Attributes1410820 67279Node: Blackfin Function Attributes1415355 67280Node: BPF Function Attributes1417852 67281Node: CR16 Function Attributes1418440 67282Node: C-SKY Function Attributes1418959 67283Node: Epiphany Function Attributes1420258 67284Node: H8/300 Function Attributes1423013 67285Node: IA-64 Function Attributes1424211 67286Node: M32C Function Attributes1425253 67287Node: M32R/D Function Attributes1427591 67288Node: m68k Function Attributes1429065 67289Node: MCORE Function Attributes1430009 67290Node: MeP Function Attributes1430820 67291Node: MicroBlaze Function Attributes1432121 67292Node: Microsoft Windows Function Attributes1433628 67293Node: MIPS Function Attributes1438197 67294Node: MSP430 Function Attributes1443815 67295Node: NDS32 Function Attributes1447894 67296Node: Nios II Function Attributes1450318 67297Node: Nvidia PTX Function Attributes1451615 67298Node: PowerPC Function Attributes1452230 67299Node: RISC-V Function Attributes1459004 67300Node: RL78 Function Attributes1460420 67301Node: RX Function Attributes1461659 67302Node: S/390 Function Attributes1464206 67303Node: SH Function Attributes1466034 67304Node: Symbian OS Function Attributes1469462 67305Node: V850 Function Attributes1469798 67306Node: Visium Function Attributes1470343 67307Node: x86 Function Attributes1470871 67308Node: Xstormy16 Function Attributes1493360 67309Node: Variable Attributes1493867 67310Node: Common Variable Attributes1495400 67311Node: ARC Variable Attributes1513530 67312Node: AVR Variable Attributes1513912 67313Node: Blackfin Variable Attributes1519074 67314Node: H8/300 Variable Attributes1519932 67315Node: IA-64 Variable Attributes1521005 67316Node: M32R/D Variable Attributes1521756 67317Node: MeP Variable Attributes1522539 67318Node: Microsoft Windows Variable Attributes1524632 67319Node: MSP430 Variable Attributes1527085 67320Node: Nvidia PTX Variable Attributes1529039 67321Node: PowerPC Variable Attributes1529656 67322Node: RL78 Variable Attributes1530213 67323Node: V850 Variable Attributes1530632 67324Node: x86 Variable Attributes1531265 67325Node: Xstormy16 Variable Attributes1532321 67326Node: Type Attributes1532891 67327Node: Common Type Attributes1534555 67328Node: ARC Type Attributes1556048 67329Node: ARM Type Attributes1556520 67330Node: MeP Type Attributes1557302 67331Node: PowerPC Type Attributes1557704 67332Node: x86 Type Attributes1558693 67333Node: Label Attributes1559685 67334Node: Enumerator Attributes1561618 67335Node: Statement Attributes1562937 67336Node: Attribute Syntax1564420 67337Node: Function Prototypes1575678 67338Node: C++ Comments1577458 67339Node: Dollar Signs1577977 67340Node: Character Escapes1578442 67341Node: Alignment1578726 67342Node: Inline1580379 67343Node: Volatiles1585196 67344Node: Using Assembly Language with C1588095 67345Node: Basic Asm1589332 67346Node: Extended Asm1594782 67347Ref: Volatile1598881 67348Ref: AssemblerTemplate1603001 67349Ref: OutputOperands1607241 67350Ref: FlagOutputOperands1614204 67351Ref: InputOperands1617151 67352Ref: Clobbers and Scratch Registers1621419 67353Ref: GotoLabels1630060 67354Ref: x86Operandmodifiers1632195 67355Ref: x86floatingpointasmoperands1635352 67356Node: Constraints1638681 67357Node: Simple Constraints1639787 67358Node: Multi-Alternative1647101 67359Node: Modifiers1648776 67360Node: Machine Constraints1651574 67361Node: Asm Labels1707799 67362Node: Explicit Register Variables1709419 67363Ref: Explicit Reg Vars1709633 67364Node: Global Register Variables1710242 67365Ref: Global Reg Vars1710450 67366Node: Local Register Variables1715232 67367Ref: Local Reg Vars1715452 67368Node: Size of an asm1719080 67369Node: Alternate Keywords1720558 67370Node: Incomplete Enums1722063 67371Node: Function Names1722820 67372Node: Return Address1724724 67373Node: Vector Extensions1729311 67374Node: Offsetof1739076 67375Node: __sync Builtins1739909 67376Node: __atomic Builtins1746352 67377Node: Integer Overflow Builtins1759977 67378Node: x86 specific memory model extensions for transactional memory1766460 67379Node: Object Size Checking1767726 67380Node: Other Builtins1773982 67381Node: Target Builtins1823433 67382Node: AArch64 Built-in Functions1825181 67383Node: Alpha Built-in Functions1825636 67384Node: Altera Nios II Built-in Functions1828684 67385Node: ARC Built-in Functions1833053 67386Node: ARC SIMD Built-in Functions1838265 67387Node: ARM iWMMXt Built-in Functions1847161 67388Node: ARM C Language Extensions (ACLE)1854157 67389Node: ARM Floating Point Status and Control Intrinsics1855434 67390Node: ARM ARMv8-M Security Extensions1855919 67391Node: AVR Built-in Functions1857199 67392Node: Blackfin Built-in Functions1860960 67393Node: BPF Built-in Functions1861578 67394Node: FR-V Built-in Functions1862478 67395Node: Argument Types1863341 67396Node: Directly-mapped Integer Functions1865095 67397Node: Directly-mapped Media Functions1866179 67398Node: Raw read/write Functions1874385 67399Node: Other Built-in Functions1875293 67400Node: MIPS DSP Built-in Functions1876479 67401Node: MIPS Paired-Single Support1888976 67402Node: MIPS Loongson Built-in Functions1890475 67403Node: Paired-Single Arithmetic1896997 67404Node: Paired-Single Built-in Functions1897945 67405Node: MIPS-3D Built-in Functions1900612 67406Node: MIPS SIMD Architecture (MSA) Support1906006 67407Node: MIPS SIMD Architecture Built-in Functions1908846 67408Node: Other MIPS Built-in Functions1935700 67409Node: MSP430 Built-in Functions1936709 67410Node: NDS32 Built-in Functions1938110 67411Node: picoChip Built-in Functions1939403 67412Node: Basic PowerPC Built-in Functions1940752 67413Node: Basic PowerPC Built-in Functions Available on all Configurations1941552 67414Node: Basic PowerPC Built-in Functions Available on ISA 2.051950061 67415Node: Basic PowerPC Built-in Functions Available on ISA 2.061954896 67416Node: Basic PowerPC Built-in Functions Available on ISA 2.071956976 67417Node: Basic PowerPC Built-in Functions Available on ISA 3.01957830 67418Node: PowerPC AltiVec/VSX Built-in Functions1964746 67419Node: PowerPC AltiVec Built-in Functions on ISA 2.051968428 67420Node: PowerPC AltiVec Built-in Functions Available on ISA 2.062057666 67421Node: PowerPC AltiVec Built-in Functions Available on ISA 2.072082111 67422Node: PowerPC AltiVec Built-in Functions Available on ISA 3.02103315 67423Node: PowerPC Hardware Transactional Memory Built-in Functions2138451 67424Node: PowerPC Atomic Memory Operation Functions2146966 67425Node: PowerPC Matrix-Multiply Assist Built-in Functions2149529 67426Node: RISC-V Built-in Functions2155661 67427Node: RX Built-in Functions2156099 67428Node: S/390 System z Built-in Functions2160101 67429Node: SH Built-in Functions2165331 67430Node: SPARC VIS Built-in Functions2167059 67431Node: TI C6X Built-in Functions2175591 67432Node: TILE-Gx Built-in Functions2176622 67433Node: TILEPro Built-in Functions2177741 67434Node: x86 Built-in Functions2178841 67435Node: x86 transactional memory intrinsics2242750 67436Node: x86 control-flow protection intrinsics2246017 67437Node: Target Format Checks2247788 67438Node: Solaris Format Checks2248220 67439Node: Darwin Format Checks2248646 67440Node: Pragmas2249609 67441Node: AArch64 Pragmas2250550 67442Node: ARM Pragmas2251007 67443Node: M32C Pragmas2251634 67444Node: MeP Pragmas2252706 67445Node: PRU Pragmas2254758 67446Node: RS/6000 and PowerPC Pragmas2255336 67447Node: S/390 Pragmas2256076 67448Node: Darwin Pragmas2256642 67449Node: Solaris Pragmas2257695 67450Node: Symbol-Renaming Pragmas2258859 67451Node: Structure-Layout Pragmas2260496 67452Node: Weak Pragmas2262776 67453Node: Diagnostic Pragmas2263511 67454Node: Visibility Pragmas2267702 67455Node: Push/Pop Macro Pragmas2268387 67456Node: Function Specific Option Pragmas2269360 67457Node: Loop-Specific Pragmas2271326 67458Node: Unnamed Fields2272926 67459Node: Thread-Local2275123 67460Node: C99 Thread-Local Edits2277229 67461Node: C++98 Thread-Local Edits2279227 67462Node: Binary constants2282672 67463Node: C++ Extensions2283343 67464Node: C++ Volatiles2284973 67465Node: Restricted Pointers2287321 67466Node: Vague Linkage2288912 67467Node: C++ Interface2292535 67468Ref: C++ Interface-Footnote-12296332 67469Node: Template Instantiation2296470 67470Node: Bound member functions2302561 67471Node: C++ Attributes2304093 67472Node: Function Multiversioning2308165 67473Node: Type Traits2309972 67474Node: C++ Concepts2316922 67475Node: Deprecated Features2318428 67476Node: Backwards Compatibility2320253 67477Node: Objective-C2321325 67478Node: GNU Objective-C runtime API2321932 67479Node: Modern GNU Objective-C runtime API2322939 67480Node: Traditional GNU Objective-C runtime API2325375 67481Node: Executing code before main2326102 67482Node: What you can and what you cannot do in +load2328846 67483Node: Type encoding2331216 67484Node: Legacy type encoding2336357 67485Node: @encode2337447 67486Node: Method signatures2337992 67487Node: Garbage Collection2339984 67488Node: Constant string objects2342674 67489Node: compatibility_alias2345183 67490Node: Exceptions2345908 67491Node: Synchronization2348618 67492Node: Fast enumeration2349802 67493Node: Using fast enumeration2350114 67494Node: c99-like fast enumeration syntax2351325 67495Node: Fast enumeration details2352028 67496Node: Fast enumeration protocol2354368 67497Node: Messaging with the GNU Objective-C runtime2357520 67498Node: Dynamically registering methods2358892 67499Node: Forwarding hook2360583 67500Node: Compatibility2363623 67501Node: Gcov2370179 67502Node: Gcov Intro2370714 67503Node: Invoking Gcov2373432 67504Node: Gcov and Optimization2396267 67505Node: Gcov Data Files2400010 67506Node: Cross-profiling2401419 67507Node: Gcov-tool2403273 67508Node: Gcov-tool Intro2403698 67509Node: Invoking Gcov-tool2405668 67510Node: Gcov-dump2408246 67511Node: Gcov-dump Intro2408569 67512Node: Invoking Gcov-dump2408836 67513Node: lto-dump2409437 67514Node: lto-dump Intro2409736 67515Node: Invoking lto-dump2409986 67516Node: Trouble2411082 67517Node: Actual Bugs2412499 67518Node: Interoperation2412946 67519Node: Incompatibilities2419837 67520Node: Fixed Headers2427989 67521Node: Standard Libraries2429647 67522Node: Disappointments2431019 67523Node: C++ Misunderstandings2435378 67524Node: Static Definitions2436189 67525Node: Name lookup2437242 67526Ref: Name lookup-Footnote-12442023 67527Node: Temporaries2442212 67528Node: Copy Assignment2444188 67529Node: Non-bugs2446023 67530Node: Warnings and Errors2456529 67531Node: Bugs2458291 67532Node: Bug Criteria2458758 67533Node: Bug Reporting2460968 67534Node: Service2461186 67535Node: Contributing2462006 67536Node: Funding2462747 67537Node: GNU Project2465237 67538Node: Copying2465883 67539Node: GNU Free Documentation License2503391 67540Node: Contributors2528509 67541Node: Option Index2569482 67542Node: Keyword Index2848961 67543 67544End Tag Table 67545