tm.texi revision 132718
1@c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001, 2@c 2002, 2003, 2004 Free Software Foundation, Inc. 3@c This is part of the GCC manual. 4@c For copying conditions, see the file gcc.texi. 5 6@node Target Macros 7@chapter Target Description Macros and Functions 8@cindex machine description macros 9@cindex target description macros 10@cindex macros, target description 11@cindex @file{tm.h} macros 12 13In addition to the file @file{@var{machine}.md}, a machine description 14includes a C header file conventionally given the name 15@file{@var{machine}.h} and a C source file named @file{@var{machine}.c}. 16The header file defines numerous macros that convey the information 17about the target machine that does not fit into the scheme of the 18@file{.md} file. The file @file{tm.h} should be a link to 19@file{@var{machine}.h}. The header file @file{config.h} includes 20@file{tm.h} and most compiler source files include @file{config.h}. The 21source file defines a variable @code{targetm}, which is a structure 22containing pointers to functions and data relating to the target 23machine. @file{@var{machine}.c} should also contain their definitions, 24if they are not defined elsewhere in GCC, and other functions called 25through the macros defined in the @file{.h} file. 26 27@menu 28* Target Structure:: The @code{targetm} variable. 29* Driver:: Controlling how the driver runs the compilation passes. 30* Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}. 31* Per-Function Data:: Defining data structures for per-function information. 32* Storage Layout:: Defining sizes and alignments of data. 33* Type Layout:: Defining sizes and properties of basic user data types. 34* Escape Sequences:: Defining the value of target character escape sequences 35* Registers:: Naming and describing the hardware registers. 36* Register Classes:: Defining the classes of hardware registers. 37* Stack and Calling:: Defining which way the stack grows and by how much. 38* Varargs:: Defining the varargs macros. 39* Trampolines:: Code set up at run time to enter a nested function. 40* Library Calls:: Controlling how library routines are implicitly called. 41* Addressing Modes:: Defining addressing modes valid for memory operands. 42* Condition Code:: Defining how insns update the condition code. 43* Costs:: Defining relative costs of different operations. 44* Scheduling:: Adjusting the behavior of the instruction scheduler. 45* Sections:: Dividing storage into text, data, and other sections. 46* PIC:: Macros for position independent code. 47* Assembler Format:: Defining how to write insns and pseudo-ops to output. 48* Debugging Info:: Defining the format of debugging output. 49* Floating Point:: Handling floating point for cross-compilers. 50* Mode Switching:: Insertion of mode-switching instructions. 51* Target Attributes:: Defining target-specific uses of @code{__attribute__}. 52* MIPS Coprocessors:: MIPS coprocessor support and how to customize it. 53* PCH Target:: Validity checking for precompiled headers. 54* Misc:: Everything else. 55@end menu 56 57@node Target Structure 58@section The Global @code{targetm} Variable 59@cindex target hooks 60@cindex target functions 61 62@deftypevar {struct gcc_target} targetm 63The target @file{.c} file must define the global @code{targetm} variable 64which contains pointers to functions and data relating to the target 65machine. The variable is declared in @file{target.h}; 66@file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is 67used to initialize the variable, and macros for the default initializers 68for elements of the structure. The @file{.c} file should override those 69macros for which the default definition is inappropriate. For example: 70@smallexample 71#include "target.h" 72#include "target-def.h" 73 74/* @r{Initialize the GCC target structure.} */ 75 76#undef TARGET_COMP_TYPE_ATTRIBUTES 77#define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes 78 79struct gcc_target targetm = TARGET_INITIALIZER; 80@end smallexample 81@end deftypevar 82 83Where a macro should be defined in the @file{.c} file in this manner to 84form part of the @code{targetm} structure, it is documented below as a 85``Target Hook'' with a prototype. Many macros will change in future 86from being defined in the @file{.h} file to being part of the 87@code{targetm} structure. 88 89@node Driver 90@section Controlling the Compilation Driver, @file{gcc} 91@cindex driver 92@cindex controlling the compilation driver 93 94@c prevent bad page break with this line 95You can control the compilation driver. 96 97@defmac SWITCH_TAKES_ARG (@var{char}) 98A C expression which determines whether the option @option{-@var{char}} 99takes arguments. The value should be the number of arguments that 100option takes--zero, for many options. 101 102By default, this macro is defined as 103@code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options 104properly. You need not define @code{SWITCH_TAKES_ARG} unless you 105wish to add additional options which take arguments. Any redefinition 106should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for 107additional options. 108@end defmac 109 110@defmac WORD_SWITCH_TAKES_ARG (@var{name}) 111A C expression which determines whether the option @option{-@var{name}} 112takes arguments. The value should be the number of arguments that 113option takes--zero, for many options. This macro rather than 114@code{SWITCH_TAKES_ARG} is used for multi-character option names. 115 116By default, this macro is defined as 117@code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options 118properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you 119wish to add additional options which take arguments. Any redefinition 120should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for 121additional options. 122@end defmac 123 124@defmac SWITCH_CURTAILS_COMPILATION (@var{char}) 125A C expression which determines whether the option @option{-@var{char}} 126stops compilation before the generation of an executable. The value is 127boolean, nonzero if the option does stop an executable from being 128generated, zero otherwise. 129 130By default, this macro is defined as 131@code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard 132options properly. You need not define 133@code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional 134options which affect the generation of an executable. Any redefinition 135should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check 136for additional options. 137@end defmac 138 139@defmac SWITCHES_NEED_SPACES 140A string-valued C expression which enumerates the options for which 141the linker needs a space between the option and its argument. 142 143If this macro is not defined, the default value is @code{""}. 144@end defmac 145 146@defmac TARGET_OPTION_TRANSLATE_TABLE 147If defined, a list of pairs of strings, the first of which is a 148potential command line target to the @file{gcc} driver program, and the 149second of which is a space-separated (tabs and other whitespace are not 150supported) list of options with which to replace the first option. The 151target defining this list is responsible for assuring that the results 152are valid. Replacement options may not be the @code{--opt} style, they 153must be the @code{-opt} style. It is the intention of this macro to 154provide a mechanism for substitution that affects the multilibs chosen, 155such as one option that enables many options, some of which select 156multilibs. Example nonsensical definition, where @code{-malt-abi}, 157@code{-EB}, and @code{-mspoo} cause different multilibs to be chosen: 158 159@smallexample 160#define TARGET_OPTION_TRANSLATE_TABLE \ 161@{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \ 162@{ "-compat", "-EB -malign=4 -mspoo" @} 163@end smallexample 164@end defmac 165 166@defmac DRIVER_SELF_SPECS 167A list of specs for the driver itself. It should be a suitable 168initializer for an array of strings, with no surrounding braces. 169 170The driver applies these specs to its own command line between loading 171default @file{specs} files (but not command-line specified ones) and 172choosing the multilib directory or running any subcommands. It 173applies them in the order given, so each spec can depend on the 174options added by earlier ones. It is also possible to remove options 175using @samp{%<@var{option}} in the usual way. 176 177This macro can be useful when a port has several interdependent target 178options. It provides a way of standardizing the command line so 179that the other specs are easier to write. 180 181Do not define this macro if it does not need to do anything. 182@end defmac 183 184@defmac OPTION_DEFAULT_SPECS 185A list of specs used to support configure-time default options (i.e.@: 186@option{--with} options) in the driver. It should be a suitable initializer 187for an array of structures, each containing two strings, without the 188outermost pair of surrounding braces. 189 190The first item in the pair is the name of the default. This must match 191the code in @file{config.gcc} for the target. The second item is a spec 192to apply if a default with this name was specified. The string 193@samp{%(VALUE)} in the spec will be replaced by the value of the default 194everywhere it occurs. 195 196The driver will apply these specs to its own command line between loading 197default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using 198the same mechanism as @code{DRIVER_SELF_SPECS}. 199 200Do not define this macro if it does not need to do anything. 201@end defmac 202 203@defmac CPP_SPEC 204A C string constant that tells the GCC driver program options to 205pass to CPP@. It can also specify how to translate options you 206give to GCC into options for GCC to pass to the CPP@. 207 208Do not define this macro if it does not need to do anything. 209@end defmac 210 211@defmac CPLUSPLUS_CPP_SPEC 212This macro is just like @code{CPP_SPEC}, but is used for C++, rather 213than C@. If you do not define this macro, then the value of 214@code{CPP_SPEC} (if any) will be used instead. 215@end defmac 216 217@defmac CC1_SPEC 218A C string constant that tells the GCC driver program options to 219pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language 220front ends. 221It can also specify how to translate options you give to GCC into options 222for GCC to pass to front ends. 223 224Do not define this macro if it does not need to do anything. 225@end defmac 226 227@defmac CC1PLUS_SPEC 228A C string constant that tells the GCC driver program options to 229pass to @code{cc1plus}. It can also specify how to translate options you 230give to GCC into options for GCC to pass to the @code{cc1plus}. 231 232Do not define this macro if it does not need to do anything. 233Note that everything defined in CC1_SPEC is already passed to 234@code{cc1plus} so there is no need to duplicate the contents of 235CC1_SPEC in CC1PLUS_SPEC@. 236@end defmac 237 238@defmac ASM_SPEC 239A C string constant that tells the GCC driver program options to 240pass to the assembler. It can also specify how to translate options 241you give to GCC into options for GCC to pass to the assembler. 242See the file @file{sun3.h} for an example of this. 243 244Do not define this macro if it does not need to do anything. 245@end defmac 246 247@defmac ASM_FINAL_SPEC 248A C string constant that tells the GCC driver program how to 249run any programs which cleanup after the normal assembler. 250Normally, this is not needed. See the file @file{mips.h} for 251an example of this. 252 253Do not define this macro if it does not need to do anything. 254@end defmac 255 256@defmac AS_NEEDS_DASH_FOR_PIPED_INPUT 257Define this macro, with no value, if the driver should give the assembler 258an argument consisting of a single dash, @option{-}, to instruct it to 259read from its standard input (which will be a pipe connected to the 260output of the compiler proper). This argument is given after any 261@option{-o} option specifying the name of the output file. 262 263If you do not define this macro, the assembler is assumed to read its 264standard input if given no non-option arguments. If your assembler 265cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct; 266see @file{mips.h} for instance. 267@end defmac 268 269@defmac LINK_SPEC 270A C string constant that tells the GCC driver program options to 271pass to the linker. It can also specify how to translate options you 272give to GCC into options for GCC to pass to the linker. 273 274Do not define this macro if it does not need to do anything. 275@end defmac 276 277@defmac LIB_SPEC 278Another C string constant used much like @code{LINK_SPEC}. The difference 279between the two is that @code{LIB_SPEC} is used at the end of the 280command given to the linker. 281 282If this macro is not defined, a default is provided that 283loads the standard C library from the usual place. See @file{gcc.c}. 284@end defmac 285 286@defmac LIBGCC_SPEC 287Another C string constant that tells the GCC driver program 288how and when to place a reference to @file{libgcc.a} into the 289linker command line. This constant is placed both before and after 290the value of @code{LIB_SPEC}. 291 292If this macro is not defined, the GCC driver provides a default that 293passes the string @option{-lgcc} to the linker. 294@end defmac 295 296@defmac STARTFILE_SPEC 297Another C string constant used much like @code{LINK_SPEC}. The 298difference between the two is that @code{STARTFILE_SPEC} is used at 299the very beginning of the command given to the linker. 300 301If this macro is not defined, a default is provided that loads the 302standard C startup file from the usual place. See @file{gcc.c}. 303@end defmac 304 305@defmac ENDFILE_SPEC 306Another C string constant used much like @code{LINK_SPEC}. The 307difference between the two is that @code{ENDFILE_SPEC} is used at 308the very end of the command given to the linker. 309 310Do not define this macro if it does not need to do anything. 311@end defmac 312 313@defmac THREAD_MODEL_SPEC 314GCC @code{-v} will print the thread model GCC was configured to use. 315However, this doesn't work on platforms that are multilibbed on thread 316models, such as AIX 4.3. On such platforms, define 317@code{THREAD_MODEL_SPEC} such that it evaluates to a string without 318blanks that names one of the recognized thread models. @code{%*}, the 319default value of this macro, will expand to the value of 320@code{thread_file} set in @file{config.gcc}. 321@end defmac 322 323@defmac SYSROOT_SUFFIX_SPEC 324Define this macro to add a suffix to the target sysroot when GCC is 325configured with a sysroot. This will cause GCC to search for usr/lib, 326et al, within sysroot+suffix. 327@end defmac 328 329@defmac SYSROOT_HEADERS_SUFFIX_SPEC 330Define this macro to add a headers_suffix to the target sysroot when 331GCC is configured with a sysroot. This will cause GCC to pass the 332updated sysroot+headers_suffix to CPP, causing it to search for 333usr/include, et al, within sysroot+headers_suffix. 334@end defmac 335 336@defmac EXTRA_SPECS 337Define this macro to provide additional specifications to put in the 338@file{specs} file that can be used in various specifications like 339@code{CC1_SPEC}. 340 341The definition should be an initializer for an array of structures, 342containing a string constant, that defines the specification name, and a 343string constant that provides the specification. 344 345Do not define this macro if it does not need to do anything. 346 347@code{EXTRA_SPECS} is useful when an architecture contains several 348related targets, which have various @code{@dots{}_SPECS} which are similar 349to each other, and the maintainer would like one central place to keep 350these definitions. 351 352For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to 353define either @code{_CALL_SYSV} when the System V calling sequence is 354used or @code{_CALL_AIX} when the older AIX-based calling sequence is 355used. 356 357The @file{config/rs6000/rs6000.h} target file defines: 358 359@smallexample 360#define EXTRA_SPECS \ 361 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, 362 363#define CPP_SYS_DEFAULT "" 364@end smallexample 365 366The @file{config/rs6000/sysv.h} target file defines: 367@smallexample 368#undef CPP_SPEC 369#define CPP_SPEC \ 370"%@{posix: -D_POSIX_SOURCE @} \ 371%@{mcall-sysv: -D_CALL_SYSV @} \ 372%@{!mcall-sysv: %(cpp_sysv_default) @} \ 373%@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" 374 375#undef CPP_SYSV_DEFAULT 376#define CPP_SYSV_DEFAULT "-D_CALL_SYSV" 377@end smallexample 378 379while the @file{config/rs6000/eabiaix.h} target file defines 380@code{CPP_SYSV_DEFAULT} as: 381 382@smallexample 383#undef CPP_SYSV_DEFAULT 384#define CPP_SYSV_DEFAULT "-D_CALL_AIX" 385@end smallexample 386@end defmac 387 388@defmac LINK_LIBGCC_SPECIAL 389Define this macro if the driver program should find the library 390@file{libgcc.a} itself and should not pass @option{-L} options to the 391linker. If you do not define this macro, the driver program will pass 392the argument @option{-lgcc} to tell the linker to do the search and will 393pass @option{-L} options to it. 394@end defmac 395 396@defmac LINK_LIBGCC_SPECIAL_1 397Define this macro if the driver program should find the library 398@file{libgcc.a}. If you do not define this macro, the driver program will pass 399the argument @option{-lgcc} to tell the linker to do the search. 400This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does 401not affect @option{-L} options. 402@end defmac 403 404@defmac LINK_GCC_C_SEQUENCE_SPEC 405The sequence in which libgcc and libc are specified to the linker. 406By default this is @code{%G %L %G}. 407@end defmac 408 409@defmac LINK_COMMAND_SPEC 410A C string constant giving the complete command line need to execute the 411linker. When you do this, you will need to update your port each time a 412change is made to the link command line within @file{gcc.c}. Therefore, 413define this macro only if you need to completely redefine the command 414line for invoking the linker and there is no other way to accomplish 415the effect you need. Overriding this macro may be avoidable by overriding 416@code{LINK_GCC_C_SEQUENCE_SPEC} instead. 417@end defmac 418 419@defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES 420A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search 421directories from linking commands. Do not give it a nonzero value if 422removing duplicate search directories changes the linker's semantics. 423@end defmac 424 425@defmac MULTILIB_DEFAULTS 426Define this macro as a C expression for the initializer of an array of 427string to tell the driver program which options are defaults for this 428target and thus do not need to be handled specially when using 429@code{MULTILIB_OPTIONS}. 430 431Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in 432the target makefile fragment or if none of the options listed in 433@code{MULTILIB_OPTIONS} are set by default. 434@xref{Target Fragment}. 435@end defmac 436 437@defmac RELATIVE_PREFIX_NOT_LINKDIR 438Define this macro to tell @command{gcc} that it should only translate 439a @option{-B} prefix into a @option{-L} linker option if the prefix 440indicates an absolute file name. 441@end defmac 442 443@defmac MD_EXEC_PREFIX 444If defined, this macro is an additional prefix to try after 445@code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched 446when the @option{-b} option is used, or the compiler is built as a cross 447compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it 448to the list of directories used to find the assembler in @file{configure.in}. 449@end defmac 450 451@defmac STANDARD_STARTFILE_PREFIX 452Define this macro as a C string constant if you wish to override the 453standard choice of @code{libdir} as the default prefix to 454try when searching for startup files such as @file{crt0.o}. 455@code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler 456is built as a cross compiler. 457@end defmac 458 459@defmac MD_STARTFILE_PREFIX 460If defined, this macro supplies an additional prefix to try after the 461standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the 462@option{-b} option is used, or when the compiler is built as a cross 463compiler. 464@end defmac 465 466@defmac MD_STARTFILE_PREFIX_1 467If defined, this macro supplies yet another prefix to try after the 468standard prefixes. It is not searched when the @option{-b} option is 469used, or when the compiler is built as a cross compiler. 470@end defmac 471 472@defmac INIT_ENVIRONMENT 473Define this macro as a C string constant if you wish to set environment 474variables for programs called by the driver, such as the assembler and 475loader. The driver passes the value of this macro to @code{putenv} to 476initialize the necessary environment variables. 477@end defmac 478 479@defmac LOCAL_INCLUDE_DIR 480Define this macro as a C string constant if you wish to override the 481standard choice of @file{/usr/local/include} as the default prefix to 482try when searching for local header files. @code{LOCAL_INCLUDE_DIR} 483comes before @code{SYSTEM_INCLUDE_DIR} in the search order. 484 485Cross compilers do not search either @file{/usr/local/include} or its 486replacement. 487@end defmac 488 489@defmac MODIFY_TARGET_NAME 490Define this macro if you wish to define command-line switches that 491modify the default target name. 492 493For each switch, you can include a string to be appended to the first 494part of the configuration name or a string to be deleted from the 495configuration name, if present. The definition should be an initializer 496for an array of structures. Each array element should have three 497elements: the switch name (a string constant, including the initial 498dash), one of the enumeration codes @code{ADD} or @code{DELETE} to 499indicate whether the string should be inserted or deleted, and the string 500to be inserted or deleted (a string constant). 501 502For example, on a machine where @samp{64} at the end of the 503configuration name denotes a 64-bit target and you want the @option{-32} 504and @option{-64} switches to select between 32- and 64-bit targets, you would 505code 506 507@smallexample 508#define MODIFY_TARGET_NAME \ 509 @{ @{ "-32", DELETE, "64"@}, \ 510 @{"-64", ADD, "64"@}@} 511@end smallexample 512@end defmac 513 514@defmac SYSTEM_INCLUDE_DIR 515Define this macro as a C string constant if you wish to specify a 516system-specific directory to search for header files before the standard 517directory. @code{SYSTEM_INCLUDE_DIR} comes before 518@code{STANDARD_INCLUDE_DIR} in the search order. 519 520Cross compilers do not use this macro and do not search the directory 521specified. 522@end defmac 523 524@defmac STANDARD_INCLUDE_DIR 525Define this macro as a C string constant if you wish to override the 526standard choice of @file{/usr/include} as the default prefix to 527try when searching for header files. 528 529Cross compilers ignore this macro and do not search either 530@file{/usr/include} or its replacement. 531@end defmac 532 533@defmac STANDARD_INCLUDE_COMPONENT 534The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}. 535See @code{INCLUDE_DEFAULTS}, below, for the description of components. 536If you do not define this macro, no component is used. 537@end defmac 538 539@defmac INCLUDE_DEFAULTS 540Define this macro if you wish to override the entire default search path 541for include files. For a native compiler, the default search path 542usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, 543@code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and 544@code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} 545and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, 546and specify private search areas for GCC@. The directory 547@code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. 548 549The definition should be an initializer for an array of structures. 550Each array element should have four elements: the directory name (a 551string constant), the component name (also a string constant), a flag 552for C++-only directories, 553and a flag showing that the includes in the directory don't need to be 554wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of 555the array with a null element. 556 557The component name denotes what GNU package the include file is part of, 558if any, in all uppercase letters. For example, it might be @samp{GCC} 559or @samp{BINUTILS}. If the package is part of a vendor-supplied 560operating system, code the component name as @samp{0}. 561 562For example, here is the definition used for VAX/VMS: 563 564@smallexample 565#define INCLUDE_DEFAULTS \ 566@{ \ 567 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \ 568 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \ 569 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \ 570 @{ ".", 0, 0, 0@}, \ 571 @{ 0, 0, 0, 0@} \ 572@} 573@end smallexample 574@end defmac 575 576Here is the order of prefixes tried for exec files: 577 578@enumerate 579@item 580Any prefixes specified by the user with @option{-B}. 581 582@item 583The environment variable @code{GCC_EXEC_PREFIX}, if any. 584 585@item 586The directories specified by the environment variable @code{COMPILER_PATH}. 587 588@item 589The macro @code{STANDARD_EXEC_PREFIX}. 590 591@item 592@file{/usr/lib/gcc/}. 593 594@item 595The macro @code{MD_EXEC_PREFIX}, if any. 596@end enumerate 597 598Here is the order of prefixes tried for startfiles: 599 600@enumerate 601@item 602Any prefixes specified by the user with @option{-B}. 603 604@item 605The environment variable @code{GCC_EXEC_PREFIX}, if any. 606 607@item 608The directories specified by the environment variable @code{LIBRARY_PATH} 609(or port-specific name; native only, cross compilers do not use this). 610 611@item 612The macro @code{STANDARD_EXEC_PREFIX}. 613 614@item 615@file{/usr/lib/gcc/}. 616 617@item 618The macro @code{MD_EXEC_PREFIX}, if any. 619 620@item 621The macro @code{MD_STARTFILE_PREFIX}, if any. 622 623@item 624The macro @code{STANDARD_STARTFILE_PREFIX}. 625 626@item 627@file{/lib/}. 628 629@item 630@file{/usr/lib/}. 631@end enumerate 632 633@node Run-time Target 634@section Run-time Target Specification 635@cindex run-time target specification 636@cindex predefined macros 637@cindex target specifications 638 639@c prevent bad page break with this line 640Here are run-time target specifications. 641 642@defmac TARGET_CPU_CPP_BUILTINS () 643This function-like macro expands to a block of code that defines 644built-in preprocessor macros and assertions for the target cpu, using 645the functions @code{builtin_define}, @code{builtin_define_std} and 646@code{builtin_assert}. When the front end 647calls this macro it provides a trailing semicolon, and since it has 648finished command line option processing your code can use those 649results freely. 650 651@code{builtin_assert} takes a string in the form you pass to the 652command-line option @option{-A}, such as @code{cpu=mips}, and creates 653the assertion. @code{builtin_define} takes a string in the form 654accepted by option @option{-D} and unconditionally defines the macro. 655 656@code{builtin_define_std} takes a string representing the name of an 657object-like macro. If it doesn't lie in the user's namespace, 658@code{builtin_define_std} defines it unconditionally. Otherwise, it 659defines a version with two leading underscores, and another version 660with two leading and trailing underscores, and defines the original 661only if an ISO standard was not requested on the command line. For 662example, passing @code{unix} defines @code{__unix}, @code{__unix__} 663and possibly @code{unix}; passing @code{_mips} defines @code{__mips}, 664@code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64} 665defines only @code{_ABI64}. 666 667You can also test for the C dialect being compiled. The variable 668@code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus} 669or @code{clk_objective_c}. Note that if we are preprocessing 670assembler, this variable will be @code{clk_c} but the function-like 671macro @code{preprocessing_asm_p()} will return true, so you might want 672to check for that first. If you need to check for strict ANSI, the 673variable @code{flag_iso} can be used. The function-like macro 674@code{preprocessing_trad_p()} can be used to check for traditional 675preprocessing. 676@end defmac 677 678@defmac TARGET_OS_CPP_BUILTINS () 679Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 680and is used for the target operating system instead. 681@end defmac 682 683@defmac TARGET_OBJFMT_CPP_BUILTINS () 684Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 685and is used for the target object format. @file{elfos.h} uses this 686macro to define @code{__ELF__}, so you probably do not need to define 687it yourself. 688@end defmac 689 690@deftypevar {extern int} target_flags 691This declaration should be present. 692@end deftypevar 693 694@cindex optional hardware or system features 695@cindex features, optional, in system conventions 696 697@defmac TARGET_@var{featurename} 698This series of macros is to allow compiler command arguments to 699enable or disable the use of optional features of the target machine. 700For example, one machine description serves both the 68000 and 701the 68020; a command argument tells the compiler whether it should 702use 68020-only instructions or not. This command argument works 703by means of a macro @code{TARGET_68020} that tests a bit in 704@code{target_flags}. 705 706Define a macro @code{TARGET_@var{featurename}} for each such option. 707Its definition should test a bit in @code{target_flags}. It is 708recommended that a helper macro @code{MASK_@var{featurename}} 709is defined for each bit-value to test, and used in 710@code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For 711example: 712 713@smallexample 714#define TARGET_MASK_68020 1 715#define TARGET_68020 (target_flags & MASK_68020) 716@end smallexample 717 718One place where these macros are used is in the condition-expressions 719of instruction patterns. Note how @code{TARGET_68020} appears 720frequently in the 68000 machine description file, @file{m68k.md}. 721Another place they are used is in the definitions of the other 722macros in the @file{@var{machine}.h} file. 723@end defmac 724 725@defmac TARGET_SWITCHES 726This macro defines names of command options to set and clear 727bits in @code{target_flags}. Its definition is an initializer 728with a subgrouping for each command option. 729 730Each subgrouping contains a string constant, that defines the option 731name, a number, which contains the bits to set in 732@code{target_flags}, and a second string which is the description 733displayed by @option{--help}. If the number is negative then the bits specified 734by the number are cleared instead of being set. If the description 735string is present but empty, then no help information will be displayed 736for that option, but it will not count as an undocumented option. The 737actual option name is made by appending @samp{-m} to the specified name. 738Non-empty description strings should be marked with @code{N_(@dots{})} for 739@command{xgettext}. Please do not mark empty strings because the empty 740string is reserved by GNU gettext. @code{gettext("")} returns the header entry 741of the message catalog with meta information, not the empty string. 742 743In addition to the description for @option{--help}, 744more detailed documentation for each option should be added to 745@file{invoke.texi}. 746 747One of the subgroupings should have a null string. The number in 748this grouping is the default value for @code{target_flags}. Any 749target options act starting with that value. 750 751Here is an example which defines @option{-m68000} and @option{-m68020} 752with opposite meanings, and picks the latter as the default: 753 754@smallexample 755#define TARGET_SWITCHES \ 756 @{ @{ "68020", MASK_68020, "" @}, \ 757 @{ "68000", -MASK_68020, \ 758 N_("Compile for the 68000") @}, \ 759 @{ "", MASK_68020, "" @}, \ 760 @} 761@end smallexample 762@end defmac 763 764@defmac TARGET_OPTIONS 765This macro is similar to @code{TARGET_SWITCHES} but defines names of command 766options that have values. Its definition is an initializer with a 767subgrouping for each command option. 768 769Each subgrouping contains a string constant, that defines the option 770name, the address of a variable, a description string, and a value. 771Non-empty description strings should be marked with @code{N_(@dots{})} 772for @command{xgettext}. Please do not mark empty strings because the 773empty string is reserved by GNU gettext. @code{gettext("")} returns the 774header entry of the message catalog with meta information, not the empty 775string. 776 777If the value listed in the table is @code{NULL}, then the variable, type 778@code{char *}, is set to the variable part of the given option if the 779fixed part matches. In other words, if the first part of the option 780matches what's in the table, the variable will be set to point to the 781rest of the option. This allows the user to specify a value for that 782option. The actual option name is made by appending @samp{-m} to the 783specified name. Again, each option should also be documented in 784@file{invoke.texi}. 785 786If the value listed in the table is non-@code{NULL}, then the option 787must match the option in the table exactly (with @samp{-m}), and the 788variable is set to point to the value listed in the table. 789 790Here is an example which defines @option{-mshort-data-@var{number}}. If the 791given option is @option{-mshort-data-512}, the variable @code{m88k_short_data} 792will be set to the string @code{"512"}. 793 794@smallexample 795extern char *m88k_short_data; 796#define TARGET_OPTIONS \ 797 @{ @{ "short-data-", &m88k_short_data, \ 798 N_("Specify the size of the short data section"), 0 @} @} 799@end smallexample 800 801Here is a variant of the above that allows the user to also specify 802just @option{-mshort-data} where a default of @code{"64"} is used. 803 804@smallexample 805extern char *m88k_short_data; 806#define TARGET_OPTIONS \ 807 @{ @{ "short-data-", &m88k_short_data, \ 808 N_("Specify the size of the short data section"), 0 @} \ 809 @{ "short-data", &m88k_short_data, "", "64" @}, 810 @} 811@end smallexample 812 813Here is an example which defines @option{-mno-alu}, @option{-malu1}, and 814@option{-malu2} as a three-state switch, along with suitable macros for 815checking the state of the option (documentation is elided for brevity). 816 817@smallexample 818[chip.c] 819char *chip_alu = ""; /* Specify default here. */ 820 821[chip.h] 822extern char *chip_alu; 823#define TARGET_OPTIONS \ 824 @{ @{ "no-alu", &chip_alu, "", "" @}, \ 825 @{ "alu1", &chip_alu, "", "1" @}, \ 826 @{ "alu2", &chip_alu, "", "2" @}, @} 827#define TARGET_ALU (chip_alu[0] != '\0') 828#define TARGET_ALU1 (chip_alu[0] == '1') 829#define TARGET_ALU2 (chip_alu[0] == '2') 830@end smallexample 831@end defmac 832 833@defmac TARGET_VERSION 834This macro is a C statement to print on @code{stderr} a string 835describing the particular machine description choice. Every machine 836description should define @code{TARGET_VERSION}. For example: 837 838@smallexample 839#ifdef MOTOROLA 840#define TARGET_VERSION \ 841 fprintf (stderr, " (68k, Motorola syntax)"); 842#else 843#define TARGET_VERSION \ 844 fprintf (stderr, " (68k, MIT syntax)"); 845#endif 846@end smallexample 847@end defmac 848 849@defmac OVERRIDE_OPTIONS 850Sometimes certain combinations of command options do not make sense on 851a particular target machine. You can define a macro 852@code{OVERRIDE_OPTIONS} to take account of this. This macro, if 853defined, is executed once just after all the command options have been 854parsed. 855 856Don't use this macro to turn on various extra optimizations for 857@option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for. 858@end defmac 859 860@defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size}) 861Some machines may desire to change what optimizations are performed for 862various optimization levels. This macro, if defined, is executed once 863just after the optimization level is determined and before the remainder 864of the command options have been parsed. Values set in this macro are 865used as the default values for the other command line options. 866 867@var{level} is the optimization level specified; 2 if @option{-O2} is 868specified, 1 if @option{-O} is specified, and 0 if neither is specified. 869 870@var{size} is nonzero if @option{-Os} is specified and zero otherwise. 871 872You should not use this macro to change options that are not 873machine-specific. These should uniformly selected by the same 874optimization level on all supported machines. Use this macro to enable 875machine-specific optimizations. 876 877@strong{Do not examine @code{write_symbols} in 878this macro!} The debugging options are not supposed to alter the 879generated code. 880@end defmac 881 882@defmac CAN_DEBUG_WITHOUT_FP 883Define this macro if debugging can be performed even without a frame 884pointer. If this macro is defined, GCC will turn on the 885@option{-fomit-frame-pointer} option whenever @option{-O} is specified. 886@end defmac 887 888@node Per-Function Data 889@section Defining data structures for per-function information. 890@cindex per-function data 891@cindex data structures 892 893If the target needs to store information on a per-function basis, GCC 894provides a macro and a couple of variables to allow this. Note, just 895using statics to store the information is a bad idea, since GCC supports 896nested functions, so you can be halfway through encoding one function 897when another one comes along. 898 899GCC defines a data structure called @code{struct function} which 900contains all of the data specific to an individual function. This 901structure contains a field called @code{machine} whose type is 902@code{struct machine_function *}, which can be used by targets to point 903to their own specific data. 904 905If a target needs per-function specific data it should define the type 906@code{struct machine_function} and also the macro @code{INIT_EXPANDERS}. 907This macro should be used to initialize the function pointer 908@code{init_machine_status}. This pointer is explained below. 909 910One typical use of per-function, target specific data is to create an 911RTX to hold the register containing the function's return address. This 912RTX can then be used to implement the @code{__builtin_return_address} 913function, for level 0. 914 915Note---earlier implementations of GCC used a single data area to hold 916all of the per-function information. Thus when processing of a nested 917function began the old per-function data had to be pushed onto a 918stack, and when the processing was finished, it had to be popped off the 919stack. GCC used to provide function pointers called 920@code{save_machine_status} and @code{restore_machine_status} to handle 921the saving and restoring of the target specific information. Since the 922single data area approach is no longer used, these pointers are no 923longer supported. 924 925@defmac INIT_EXPANDERS 926Macro called to initialize any target specific information. This macro 927is called once per function, before generation of any RTL has begun. 928The intention of this macro is to allow the initialization of the 929function pointer @code{init_machine_status}. 930@end defmac 931 932@deftypevar {void (*)(struct function *)} init_machine_status 933If this function pointer is non-@code{NULL} it will be called once per 934function, before function compilation starts, in order to allow the 935target to perform any target specific initialization of the 936@code{struct function} structure. It is intended that this would be 937used to initialize the @code{machine} of that structure. 938 939@code{struct machine_function} structures are expected to be freed by GC. 940Generally, any memory that they reference must be allocated by using 941@code{ggc_alloc}, including the structure itself. 942@end deftypevar 943 944@node Storage Layout 945@section Storage Layout 946@cindex storage layout 947 948Note that the definitions of the macros in this table which are sizes or 949alignments measured in bits do not need to be constant. They can be C 950expressions that refer to static variables, such as the @code{target_flags}. 951@xref{Run-time Target}. 952 953@defmac BITS_BIG_ENDIAN 954Define this macro to have the value 1 if the most significant bit in a 955byte has the lowest number; otherwise define it to have the value zero. 956This means that bit-field instructions count from the most significant 957bit. If the machine has no bit-field instructions, then this must still 958be defined, but it doesn't matter which value it is defined to. This 959macro need not be a constant. 960 961This macro does not affect the way structure fields are packed into 962bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. 963@end defmac 964 965@defmac BYTES_BIG_ENDIAN 966Define this macro to have the value 1 if the most significant byte in a 967word has the lowest number. This macro need not be a constant. 968@end defmac 969 970@defmac WORDS_BIG_ENDIAN 971Define this macro to have the value 1 if, in a multiword object, the 972most significant word has the lowest number. This applies to both 973memory locations and registers; GCC fundamentally assumes that the 974order of words in memory is the same as the order in registers. This 975macro need not be a constant. 976@end defmac 977 978@defmac LIBGCC2_WORDS_BIG_ENDIAN 979Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a 980constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be 981used only when compiling @file{libgcc2.c}. Typically the value will be set 982based on preprocessor defines. 983@end defmac 984 985@defmac FLOAT_WORDS_BIG_ENDIAN 986Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or 987@code{TFmode} floating point numbers are stored in memory with the word 988containing the sign bit at the lowest address; otherwise define it to 989have the value 0. This macro need not be a constant. 990 991You need not define this macro if the ordering is the same as for 992multi-word integers. 993@end defmac 994 995@defmac BITS_PER_UNIT 996Define this macro to be the number of bits in an addressable storage 997unit (byte). If you do not define this macro the default is 8. 998@end defmac 999 1000@defmac BITS_PER_WORD 1001Number of bits in a word. If you do not define this macro, the default 1002is @code{BITS_PER_UNIT * UNITS_PER_WORD}. 1003@end defmac 1004 1005@defmac MAX_BITS_PER_WORD 1006Maximum number of bits in a word. If this is undefined, the default is 1007@code{BITS_PER_WORD}. Otherwise, it is the constant value that is the 1008largest value that @code{BITS_PER_WORD} can have at run-time. 1009@end defmac 1010 1011@defmac UNITS_PER_WORD 1012Number of storage units in a word; normally 4. 1013@end defmac 1014 1015@defmac MIN_UNITS_PER_WORD 1016Minimum number of units in a word. If this is undefined, the default is 1017@code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the 1018smallest value that @code{UNITS_PER_WORD} can have at run-time. 1019@end defmac 1020 1021@defmac POINTER_SIZE 1022Width of a pointer, in bits. You must specify a value no wider than the 1023width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, 1024you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify 1025a value the default is @code{BITS_PER_WORD}. 1026@end defmac 1027 1028@defmac POINTERS_EXTEND_UNSIGNED 1029A C expression whose value is greater than zero if pointers that need to be 1030extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to 1031be zero-extended and zero if they are to be sign-extended. If the value 1032is less then zero then there must be an "ptr_extend" instruction that 1033extends a pointer from @code{POINTER_SIZE} to @code{Pmode}. 1034 1035You need not define this macro if the @code{POINTER_SIZE} is equal 1036to the width of @code{Pmode}. 1037@end defmac 1038 1039@defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) 1040A macro to update @var{m} and @var{unsignedp} when an object whose type 1041is @var{type} and which has the specified mode and signedness is to be 1042stored in a register. This macro is only called when @var{type} is a 1043scalar type. 1044 1045On most RISC machines, which only have operations that operate on a full 1046register, define this macro to set @var{m} to @code{word_mode} if 1047@var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most 1048cases, only integer modes should be widened because wider-precision 1049floating-point operations are usually more expensive than their narrower 1050counterparts. 1051 1052For most machines, the macro definition does not change @var{unsignedp}. 1053However, some machines, have instructions that preferentially handle 1054either signed or unsigned quantities of certain modes. For example, on 1055the DEC Alpha, 32-bit loads from memory and 32-bit add instructions 1056sign-extend the result to 64 bits. On such machines, set 1057@var{unsignedp} according to which kind of extension is more efficient. 1058 1059Do not define this macro if it would never modify @var{m}. 1060@end defmac 1061 1062@deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype}) 1063This target hook should return @code{true} if the promotion described by 1064@code{PROMOTE_MODE} should also be done for outgoing function arguments. 1065@end deftypefn 1066 1067@deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype}) 1068This target hook should return @code{true} if the promotion described by 1069@code{PROMOTE_MODE} should also be done for the return value of 1070functions. 1071 1072If this target hook returns @code{true}, @code{FUNCTION_VALUE} must 1073perform the same promotions done by @code{PROMOTE_MODE}. 1074@end deftypefn 1075 1076@defmac PROMOTE_FOR_CALL_ONLY 1077Define this macro if the promotion described by @code{PROMOTE_MODE} 1078should @emph{only} be performed for outgoing function arguments or 1079function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS} 1080and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively. 1081@end defmac 1082 1083@defmac PARM_BOUNDARY 1084Normal alignment required for function parameters on the stack, in 1085bits. All stack parameters receive at least this much alignment 1086regardless of data type. On most machines, this is the same as the 1087size of an integer. 1088@end defmac 1089 1090@defmac STACK_BOUNDARY 1091Define this macro to the minimum alignment enforced by hardware for the 1092stack pointer on this machine. The definition is a C expression for the 1093desired alignment (measured in bits). This value is used as a default 1094if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines, 1095this should be the same as @code{PARM_BOUNDARY}. 1096@end defmac 1097 1098@defmac PREFERRED_STACK_BOUNDARY 1099Define this macro if you wish to preserve a certain alignment for the 1100stack pointer, greater than what the hardware enforces. The definition 1101is a C expression for the desired alignment (measured in bits). This 1102macro must evaluate to a value equal to or larger than 1103@code{STACK_BOUNDARY}. 1104@end defmac 1105 1106@defmac FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN 1107A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is 1108not guaranteed by the runtime and we should emit code to align the stack 1109at the beginning of @code{main}. 1110 1111@cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY} 1112If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned 1113to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies 1114a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may 1115be momentarily unaligned while pushing arguments. 1116@end defmac 1117 1118@defmac FUNCTION_BOUNDARY 1119Alignment required for a function entry point, in bits. 1120@end defmac 1121 1122@defmac BIGGEST_ALIGNMENT 1123Biggest alignment that any data type can require on this machine, in bits. 1124@end defmac 1125 1126@defmac MINIMUM_ATOMIC_ALIGNMENT 1127If defined, the smallest alignment, in bits, that can be given to an 1128object that can be referenced in one operation, without disturbing any 1129nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger 1130on machines that don't have byte or half-word store operations. 1131@end defmac 1132 1133@defmac BIGGEST_FIELD_ALIGNMENT 1134Biggest alignment that any structure or union field can require on this 1135machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for 1136structure and union fields only, unless the field alignment has been set 1137by the @code{__attribute__ ((aligned (@var{n})))} construct. 1138@end defmac 1139 1140@defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed}) 1141An expression for the alignment of a structure field @var{field} if the 1142alignment computed in the usual way (including applying of 1143@code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the 1144alignment) is @var{computed}. It overrides alignment only if the 1145field alignment has not been set by the 1146@code{__attribute__ ((aligned (@var{n})))} construct. 1147@end defmac 1148 1149@defmac MAX_OFILE_ALIGNMENT 1150Biggest alignment supported by the object file format of this machine. 1151Use this macro to limit the alignment which can be specified using the 1152@code{__attribute__ ((aligned (@var{n})))} construct. If not defined, 1153the default value is @code{BIGGEST_ALIGNMENT}. 1154@end defmac 1155 1156@defmac DATA_ALIGNMENT (@var{type}, @var{basic-align}) 1157If defined, a C expression to compute the alignment for a variable in 1158the static store. @var{type} is the data type, and @var{basic-align} is 1159the alignment that the object would ordinarily have. The value of this 1160macro is used instead of that alignment to align the object. 1161 1162If this macro is not defined, then @var{basic-align} is used. 1163 1164@findex strcpy 1165One use of this macro is to increase alignment of medium-size data to 1166make it all fit in fewer cache lines. Another is to cause character 1167arrays to be word-aligned so that @code{strcpy} calls that copy 1168constants to character arrays can be done inline. 1169@end defmac 1170 1171@defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align}) 1172If defined, a C expression to compute the alignment given to a constant 1173that is being placed in memory. @var{constant} is the constant and 1174@var{basic-align} is the alignment that the object would ordinarily 1175have. The value of this macro is used instead of that alignment to 1176align the object. 1177 1178If this macro is not defined, then @var{basic-align} is used. 1179 1180The typical use of this macro is to increase alignment for string 1181constants to be word aligned so that @code{strcpy} calls that copy 1182constants can be done inline. 1183@end defmac 1184 1185@defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align}) 1186If defined, a C expression to compute the alignment for a variable in 1187the local store. @var{type} is the data type, and @var{basic-align} is 1188the alignment that the object would ordinarily have. The value of this 1189macro is used instead of that alignment to align the object. 1190 1191If this macro is not defined, then @var{basic-align} is used. 1192 1193One use of this macro is to increase alignment of medium-size data to 1194make it all fit in fewer cache lines. 1195@end defmac 1196 1197@defmac EMPTY_FIELD_BOUNDARY 1198Alignment in bits to be given to a structure bit-field that follows an 1199empty field such as @code{int : 0;}. 1200 1201If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro. 1202@end defmac 1203 1204@defmac STRUCTURE_SIZE_BOUNDARY 1205Number of bits which any structure or union's size must be a multiple of. 1206Each structure or union's size is rounded up to a multiple of this. 1207 1208If you do not define this macro, the default is the same as 1209@code{BITS_PER_UNIT}. 1210@end defmac 1211 1212@defmac STRICT_ALIGNMENT 1213Define this macro to be the value 1 if instructions will fail to work 1214if given data not on the nominal alignment. If instructions will merely 1215go slower in that case, define this macro as 0. 1216@end defmac 1217 1218@defmac PCC_BITFIELD_TYPE_MATTERS 1219Define this if you wish to imitate the way many other C compilers handle 1220alignment of bit-fields and the structures that contain them. 1221 1222The behavior is that the type written for a named bit-field (@code{int}, 1223@code{short}, or other integer type) imposes an alignment for the entire 1224structure, as if the structure really did contain an ordinary field of 1225that type. In addition, the bit-field is placed within the structure so 1226that it would fit within such a field, not crossing a boundary for it. 1227 1228Thus, on most machines, a named bit-field whose type is written as 1229@code{int} would not cross a four-byte boundary, and would force 1230four-byte alignment for the whole structure. (The alignment used may 1231not be four bytes; it is controlled by the other alignment parameters.) 1232 1233An unnamed bit-field will not affect the alignment of the containing 1234structure. 1235 1236If the macro is defined, its definition should be a C expression; 1237a nonzero value for the expression enables this behavior. 1238 1239Note that if this macro is not defined, or its value is zero, some 1240bit-fields may cross more than one alignment boundary. The compiler can 1241support such references if there are @samp{insv}, @samp{extv}, and 1242@samp{extzv} insns that can directly reference memory. 1243 1244The other known way of making bit-fields work is to define 1245@code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. 1246Then every structure can be accessed with fullwords. 1247 1248Unless the machine has bit-field instructions or you define 1249@code{STRUCTURE_SIZE_BOUNDARY} that way, you must define 1250@code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. 1251 1252If your aim is to make GCC use the same conventions for laying out 1253bit-fields as are used by another compiler, here is how to investigate 1254what the other compiler does. Compile and run this program: 1255 1256@smallexample 1257struct foo1 1258@{ 1259 char x; 1260 char :0; 1261 char y; 1262@}; 1263 1264struct foo2 1265@{ 1266 char x; 1267 int :0; 1268 char y; 1269@}; 1270 1271main () 1272@{ 1273 printf ("Size of foo1 is %d\n", 1274 sizeof (struct foo1)); 1275 printf ("Size of foo2 is %d\n", 1276 sizeof (struct foo2)); 1277 exit (0); 1278@} 1279@end smallexample 1280 1281If this prints 2 and 5, then the compiler's behavior is what you would 1282get from @code{PCC_BITFIELD_TYPE_MATTERS}. 1283@end defmac 1284 1285@defmac BITFIELD_NBYTES_LIMITED 1286Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited 1287to aligning a bit-field within the structure. 1288@end defmac 1289 1290@defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode}) 1291Return 1 if a structure or array containing @var{field} should be accessed using 1292@code{BLKMODE}. 1293 1294If @var{field} is the only field in the structure, @var{mode} is its 1295mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the 1296case where structures of one field would require the structure's mode to 1297retain the field's mode. 1298 1299Normally, this is not needed. See the file @file{c4x.h} for an example 1300of how to use this macro to prevent a structure having a floating point 1301field from being accessed in an integer mode. 1302@end defmac 1303 1304@defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified}) 1305Define this macro as an expression for the alignment of a type (given 1306by @var{type} as a tree node) if the alignment computed in the usual 1307way is @var{computed} and the alignment explicitly specified was 1308@var{specified}. 1309 1310The default is to use @var{specified} if it is larger; otherwise, use 1311the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} 1312@end defmac 1313 1314@defmac MAX_FIXED_MODE_SIZE 1315An integer expression for the size in bits of the largest integer 1316machine mode that should actually be used. All integer machine modes of 1317this size or smaller can be used for structures and unions with the 1318appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE 1319(DImode)} is assumed. 1320@end defmac 1321 1322@defmac VECTOR_MODE_SUPPORTED_P (@var{mode}) 1323Define this macro to be nonzero if the port is prepared to handle insns 1324involving vector mode @var{mode}. At the very least, it must have move 1325patterns for this mode. 1326@end defmac 1327 1328@defmac STACK_SAVEAREA_MODE (@var{save_level}) 1329If defined, an expression of type @code{enum machine_mode} that 1330specifies the mode of the save area operand of a 1331@code{save_stack_@var{level}} named pattern (@pxref{Standard Names}). 1332@var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or 1333@code{SAVE_NONLOCAL} and selects which of the three named patterns is 1334having its mode specified. 1335 1336You need not define this macro if it always returns @code{Pmode}. You 1337would most commonly define this macro if the 1338@code{save_stack_@var{level}} patterns need to support both a 32- and a 133964-bit mode. 1340@end defmac 1341 1342@defmac STACK_SIZE_MODE 1343If defined, an expression of type @code{enum machine_mode} that 1344specifies the mode of the size increment operand of an 1345@code{allocate_stack} named pattern (@pxref{Standard Names}). 1346 1347You need not define this macro if it always returns @code{word_mode}. 1348You would most commonly define this macro if the @code{allocate_stack} 1349pattern needs to support both a 32- and a 64-bit mode. 1350@end defmac 1351 1352@defmac TARGET_FLOAT_FORMAT 1353A code distinguishing the floating point format of the target machine. 1354There are four defined values: 1355 1356@ftable @code 1357@item IEEE_FLOAT_FORMAT 1358This code indicates IEEE floating point. It is the default; there is no 1359need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@. 1360 1361@item VAX_FLOAT_FORMAT 1362This code indicates the ``F float'' (for @code{float}) and ``D float'' 1363or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@. 1364 1365@item IBM_FLOAT_FORMAT 1366This code indicates the format used on the IBM System/370. 1367 1368@item C4X_FLOAT_FORMAT 1369This code indicates the format used on the TMS320C3x/C4x. 1370@end ftable 1371 1372If your target uses a floating point format other than these, you must 1373define a new @var{name}_FLOAT_FORMAT code for it, and add support for 1374it to @file{real.c}. 1375 1376The ordering of the component words of floating point values stored in 1377memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}. 1378@end defmac 1379 1380@defmac MODE_HAS_NANS (@var{mode}) 1381When defined, this macro should be true if @var{mode} has a NaN 1382representation. The compiler assumes that NaNs are not equal to 1383anything (including themselves) and that addition, subtraction, 1384multiplication and division all return NaNs when one operand is 1385NaN@. 1386 1387By default, this macro is true if @var{mode} is a floating-point 1388mode and the target floating-point format is IEEE@. 1389@end defmac 1390 1391@defmac MODE_HAS_INFINITIES (@var{mode}) 1392This macro should be true if @var{mode} can represent infinity. At 1393present, the compiler uses this macro to decide whether @samp{x - x} 1394is always defined. By default, the macro is true when @var{mode} 1395is a floating-point mode and the target format is IEEE@. 1396@end defmac 1397 1398@defmac MODE_HAS_SIGNED_ZEROS (@var{mode}) 1399True if @var{mode} distinguishes between positive and negative zero. 1400The rules are expected to follow the IEEE standard: 1401 1402@itemize @bullet 1403@item 1404@samp{x + x} has the same sign as @samp{x}. 1405 1406@item 1407If the sum of two values with opposite sign is zero, the result is 1408positive for all rounding modes expect towards @minus{}infinity, for 1409which it is negative. 1410 1411@item 1412The sign of a product or quotient is negative when exactly one 1413of the operands is negative. 1414@end itemize 1415 1416The default definition is true if @var{mode} is a floating-point 1417mode and the target format is IEEE@. 1418@end defmac 1419 1420@defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode}) 1421If defined, this macro should be true for @var{mode} if it has at 1422least one rounding mode in which @samp{x} and @samp{-x} can be 1423rounded to numbers of different magnitude. Two such modes are 1424towards @minus{}infinity and towards +infinity. 1425 1426The default definition of this macro is true if @var{mode} is 1427a floating-point mode and the target format is IEEE@. 1428@end defmac 1429 1430@defmac ROUND_TOWARDS_ZERO 1431If defined, this macro should be true if the prevailing rounding 1432mode is towards zero. A true value has the following effects: 1433 1434@itemize @bullet 1435@item 1436@code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes. 1437 1438@item 1439@file{libgcc.a}'s floating-point emulator will round towards zero 1440rather than towards nearest. 1441 1442@item 1443The compiler's floating-point emulator will round towards zero after 1444doing arithmetic, and when converting from the internal float format to 1445the target format. 1446@end itemize 1447 1448The macro does not affect the parsing of string literals. When the 1449primary rounding mode is towards zero, library functions like 1450@code{strtod} might still round towards nearest, and the compiler's 1451parser should behave like the target's @code{strtod} where possible. 1452 1453Not defining this macro is equivalent to returning zero. 1454@end defmac 1455 1456@defmac LARGEST_EXPONENT_IS_NORMAL (@var{size}) 1457This macro should return true if floats with @var{size} 1458bits do not have a NaN or infinity representation, but use the largest 1459exponent for normal numbers instead. 1460 1461Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS} 1462and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes. 1463It also affects the way @file{libgcc.a} and @file{real.c} emulate 1464floating-point arithmetic. 1465 1466The default definition of this macro returns false for all sizes. 1467@end defmac 1468 1469@deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type}) 1470This target hook should return @code{true} a vector is opaque. That 1471is, if no cast is needed when copying a vector value of type 1472@var{type} into another vector lvalue of the same size. Vector opaque 1473types cannot be initialized. The default is that there are no such 1474types. 1475@end deftypefn 1476 1477@deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type}) 1478This target hook returns @code{true} if bit-fields in the given 1479@var{record_type} are to be laid out following the rules of Microsoft 1480Visual C/C++, namely: (i) a bit-field won't share the same storage 1481unit with the previous bit-field if their underlying types have 1482different sizes, and the bit-field will be aligned to the highest 1483alignment of the underlying types of itself and of the previous 1484bit-field; (ii) a zero-sized bit-field will affect the alignment of 1485the whole enclosing structure, even if it is unnamed; except that 1486(iii) a zero-sized bit-field will be disregarded unless it follows 1487another bit-field of nonzero size. If this hook returns @code{true}, 1488other macros that control bit-field layout are ignored. 1489 1490When a bit-field is inserted into a packed record, the whole size 1491of the underlying type is used by one or more same-size adjacent 1492bit-fields (that is, if its long:3, 32 bits is used in the record, 1493and any additional adjacent long bit-fields are packed into the same 1494chunk of 32 bits. However, if the size changes, a new field of that 1495size is allocated). In an unpacked record, this is the same as using 1496alignment, but not equivalent when packing. 1497 1498If both MS bit-fields and @samp{__attribute__((packed))} are used, 1499the latter will take precedence. If @samp{__attribute__((packed))} is 1500used on a single field when MS bit-fields are in use, it will take 1501precedence for that field, but the alignment of the rest of the structure 1502may affect its placement. 1503@end deftypefn 1504 1505@node Type Layout 1506@section Layout of Source Language Data Types 1507 1508These macros define the sizes and other characteristics of the standard 1509basic data types used in programs being compiled. Unlike the macros in 1510the previous section, these apply to specific features of C and related 1511languages, rather than to fundamental aspects of storage layout. 1512 1513@defmac INT_TYPE_SIZE 1514A C expression for the size in bits of the type @code{int} on the 1515target machine. If you don't define this, the default is one word. 1516@end defmac 1517 1518@defmac SHORT_TYPE_SIZE 1519A C expression for the size in bits of the type @code{short} on the 1520target machine. If you don't define this, the default is half a word. 1521(If this would be less than one storage unit, it is rounded up to one 1522unit.) 1523@end defmac 1524 1525@defmac LONG_TYPE_SIZE 1526A C expression for the size in bits of the type @code{long} on the 1527target machine. If you don't define this, the default is one word. 1528@end defmac 1529 1530@defmac ADA_LONG_TYPE_SIZE 1531On some machines, the size used for the Ada equivalent of the type 1532@code{long} by a native Ada compiler differs from that used by C. In 1533that situation, define this macro to be a C expression to be used for 1534the size of that type. If you don't define this, the default is the 1535value of @code{LONG_TYPE_SIZE}. 1536@end defmac 1537 1538@defmac MAX_LONG_TYPE_SIZE 1539Maximum number for the size in bits of the type @code{long} on the 1540target machine. If this is undefined, the default is 1541@code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the 1542largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is 1543used in @code{cpp}. 1544@end defmac 1545 1546@defmac LONG_LONG_TYPE_SIZE 1547A C expression for the size in bits of the type @code{long long} on the 1548target machine. If you don't define this, the default is two 1549words. If you want to support GNU Ada on your machine, the value of this 1550macro must be at least 64. 1551@end defmac 1552 1553@defmac CHAR_TYPE_SIZE 1554A C expression for the size in bits of the type @code{char} on the 1555target machine. If you don't define this, the default is 1556@code{BITS_PER_UNIT}. 1557@end defmac 1558 1559@defmac BOOL_TYPE_SIZE 1560A C expression for the size in bits of the C++ type @code{bool} and 1561C99 type @code{_Bool} on the target machine. If you don't define 1562this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}. 1563@end defmac 1564 1565@defmac FLOAT_TYPE_SIZE 1566A C expression for the size in bits of the type @code{float} on the 1567target machine. If you don't define this, the default is one word. 1568@end defmac 1569 1570@defmac DOUBLE_TYPE_SIZE 1571A C expression for the size in bits of the type @code{double} on the 1572target machine. If you don't define this, the default is two 1573words. 1574@end defmac 1575 1576@defmac LONG_DOUBLE_TYPE_SIZE 1577A C expression for the size in bits of the type @code{long double} on 1578the target machine. If you don't define this, the default is two 1579words. 1580@end defmac 1581 1582@defmac MAX_LONG_DOUBLE_TYPE_SIZE 1583Maximum number for the size in bits of the type @code{long double} on the 1584target machine. If this is undefined, the default is 1585@code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is 1586the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time. 1587This is used in @code{cpp}. 1588@end defmac 1589 1590@defmac TARGET_FLT_EVAL_METHOD 1591A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h}, 1592assuming, if applicable, that the floating-point control word is in its 1593default state. If you do not define this macro the value of 1594@code{FLT_EVAL_METHOD} will be zero. 1595@end defmac 1596 1597@defmac WIDEST_HARDWARE_FP_SIZE 1598A C expression for the size in bits of the widest floating-point format 1599supported by the hardware. If you define this macro, you must specify a 1600value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}. 1601If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE} 1602is the default. 1603@end defmac 1604 1605@defmac DEFAULT_SIGNED_CHAR 1606An expression whose value is 1 or 0, according to whether the type 1607@code{char} should be signed or unsigned by default. The user can 1608always override this default with the options @option{-fsigned-char} 1609and @option{-funsigned-char}. 1610@end defmac 1611 1612@defmac DEFAULT_SHORT_ENUMS 1613A C expression to determine whether to give an @code{enum} type 1614only as many bytes as it takes to represent the range of possible values 1615of that type. A nonzero value means to do that; a zero value means all 1616@code{enum} types should be allocated like @code{int}. 1617 1618If you don't define the macro, the default is 0. 1619@end defmac 1620 1621@defmac SIZE_TYPE 1622A C expression for a string describing the name of the data type to use 1623for size values. The typedef name @code{size_t} is defined using the 1624contents of the string. 1625 1626The string can contain more than one keyword. If so, separate them with 1627spaces, and write first any length keyword, then @code{unsigned} if 1628appropriate, and finally @code{int}. The string must exactly match one 1629of the data type names defined in the function 1630@code{init_decl_processing} in the file @file{c-decl.c}. You may not 1631omit @code{int} or change the order---that would cause the compiler to 1632crash on startup. 1633 1634If you don't define this macro, the default is @code{"long unsigned 1635int"}. 1636@end defmac 1637 1638@defmac PTRDIFF_TYPE 1639A C expression for a string describing the name of the data type to use 1640for the result of subtracting two pointers. The typedef name 1641@code{ptrdiff_t} is defined using the contents of the string. See 1642@code{SIZE_TYPE} above for more information. 1643 1644If you don't define this macro, the default is @code{"long int"}. 1645@end defmac 1646 1647@defmac WCHAR_TYPE 1648A C expression for a string describing the name of the data type to use 1649for wide characters. The typedef name @code{wchar_t} is defined using 1650the contents of the string. See @code{SIZE_TYPE} above for more 1651information. 1652 1653If you don't define this macro, the default is @code{"int"}. 1654@end defmac 1655 1656@defmac WCHAR_TYPE_SIZE 1657A C expression for the size in bits of the data type for wide 1658characters. This is used in @code{cpp}, which cannot make use of 1659@code{WCHAR_TYPE}. 1660@end defmac 1661 1662@defmac MAX_WCHAR_TYPE_SIZE 1663Maximum number for the size in bits of the data type for wide 1664characters. If this is undefined, the default is 1665@code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the 1666largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is 1667used in @code{cpp}. 1668@end defmac 1669 1670@defmac GCOV_TYPE_SIZE 1671A C expression for the size in bits of the type used for gcov counters on the 1672target machine. If you don't define this, the default is one 1673@code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and 1674@code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to 1675ensure atomicity for counters in multithreaded programs. 1676@end defmac 1677 1678@defmac WINT_TYPE 1679A C expression for a string describing the name of the data type to 1680use for wide characters passed to @code{printf} and returned from 1681@code{getwc}. The typedef name @code{wint_t} is defined using the 1682contents of the string. See @code{SIZE_TYPE} above for more 1683information. 1684 1685If you don't define this macro, the default is @code{"unsigned int"}. 1686@end defmac 1687 1688@defmac INTMAX_TYPE 1689A C expression for a string describing the name of the data type that 1690can represent any value of any standard or extended signed integer type. 1691The typedef name @code{intmax_t} is defined using the contents of the 1692string. See @code{SIZE_TYPE} above for more information. 1693 1694If you don't define this macro, the default is the first of 1695@code{"int"}, @code{"long int"}, or @code{"long long int"} that has as 1696much precision as @code{long long int}. 1697@end defmac 1698 1699@defmac UINTMAX_TYPE 1700A C expression for a string describing the name of the data type that 1701can represent any value of any standard or extended unsigned integer 1702type. The typedef name @code{uintmax_t} is defined using the contents 1703of the string. See @code{SIZE_TYPE} above for more information. 1704 1705If you don't define this macro, the default is the first of 1706@code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long 1707unsigned int"} that has as much precision as @code{long long unsigned 1708int}. 1709@end defmac 1710 1711@defmac TARGET_PTRMEMFUNC_VBIT_LOCATION 1712The C++ compiler represents a pointer-to-member-function with a struct 1713that looks like: 1714 1715@smallexample 1716 struct @{ 1717 union @{ 1718 void (*fn)(); 1719 ptrdiff_t vtable_index; 1720 @}; 1721 ptrdiff_t delta; 1722 @}; 1723@end smallexample 1724 1725@noindent 1726The C++ compiler must use one bit to indicate whether the function that 1727will be called through a pointer-to-member-function is virtual. 1728Normally, we assume that the low-order bit of a function pointer must 1729always be zero. Then, by ensuring that the vtable_index is odd, we can 1730distinguish which variant of the union is in use. But, on some 1731platforms function pointers can be odd, and so this doesn't work. In 1732that case, we use the low-order bit of the @code{delta} field, and shift 1733the remainder of the @code{delta} field to the left. 1734 1735GCC will automatically make the right selection about where to store 1736this bit using the @code{FUNCTION_BOUNDARY} setting for your platform. 1737However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY} 1738set such that functions always start at even addresses, but the lowest 1739bit of pointers to functions indicate whether the function at that 1740address is in ARM or Thumb mode. If this is the case of your 1741architecture, you should define this macro to 1742@code{ptrmemfunc_vbit_in_delta}. 1743 1744In general, you should not have to define this macro. On architectures 1745in which function addresses are always even, according to 1746@code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to 1747@code{ptrmemfunc_vbit_in_pfn}. 1748@end defmac 1749 1750@defmac TARGET_VTABLE_USES_DESCRIPTORS 1751Normally, the C++ compiler uses function pointers in vtables. This 1752macro allows the target to change to use ``function descriptors'' 1753instead. Function descriptors are found on targets for whom a 1754function pointer is actually a small data structure. Normally the 1755data structure consists of the actual code address plus a data 1756pointer to which the function's data is relative. 1757 1758If vtables are used, the value of this macro should be the number 1759of words that the function descriptor occupies. 1760@end defmac 1761 1762@defmac TARGET_VTABLE_ENTRY_ALIGN 1763By default, the vtable entries are void pointers, the so the alignment 1764is the same as pointer alignment. The value of this macro specifies 1765the alignment of the vtable entry in bits. It should be defined only 1766when special alignment is necessary. */ 1767@end defmac 1768 1769@defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE 1770There are a few non-descriptor entries in the vtable at offsets below 1771zero. If these entries must be padded (say, to preserve the alignment 1772specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number 1773of words in each data entry. 1774@end defmac 1775 1776@node Escape Sequences 1777@section Target Character Escape Sequences 1778@cindex escape sequences 1779 1780By default, GCC assumes that the C character escape sequences take on 1781their ASCII values for the target. If this is not correct, you must 1782explicitly define all of the macros below. All of them must evaluate 1783to constants; they are used in @code{case} statements. 1784 1785@findex TARGET_BELL 1786@findex TARGET_CR 1787@findex TARGET_ESC 1788@findex TARGET_FF 1789@findex TARGET_NEWLINE 1790@findex TARGET_TAB 1791@findex TARGET_VT 1792@multitable {@code{TARGET_NEWLINE}} {Escape} {ASCII character} 1793@item Macro @tab Escape @tab ASCII character 1794@item @code{TARGET_BELL} @tab @kbd{\a} @tab @code{07}, @code{BEL} 1795@item @code{TARGET_CR} @tab @kbd{\r} @tab @code{0D}, @code{CR} 1796@item @code{TARGET_ESC} @tab @kbd{\e}, @kbd{\E} @tab @code{1B}, @code{ESC} 1797@item @code{TARGET_FF} @tab @kbd{\f} @tab @code{0C}, @code{FF} 1798@item @code{TARGET_NEWLINE} @tab @kbd{\n} @tab @code{0A}, @code{LF} 1799@item @code{TARGET_TAB} @tab @kbd{\t} @tab @code{09}, @code{HT} 1800@item @code{TARGET_VT} @tab @kbd{\v} @tab @code{0B}, @code{VT} 1801@end multitable 1802 1803@noindent 1804Note that the @kbd{\e} and @kbd{\E} escapes are GNU extensions, not 1805part of the C standard. 1806 1807@node Registers 1808@section Register Usage 1809@cindex register usage 1810 1811This section explains how to describe what registers the target machine 1812has, and how (in general) they can be used. 1813 1814The description of which registers a specific instruction can use is 1815done with register classes; see @ref{Register Classes}. For information 1816on using registers to access a stack frame, see @ref{Frame Registers}. 1817For passing values in registers, see @ref{Register Arguments}. 1818For returning values in registers, see @ref{Scalar Return}. 1819 1820@menu 1821* Register Basics:: Number and kinds of registers. 1822* Allocation Order:: Order in which registers are allocated. 1823* Values in Registers:: What kinds of values each reg can hold. 1824* Leaf Functions:: Renumbering registers for leaf functions. 1825* Stack Registers:: Handling a register stack such as 80387. 1826@end menu 1827 1828@node Register Basics 1829@subsection Basic Characteristics of Registers 1830 1831@c prevent bad page break with this line 1832Registers have various characteristics. 1833 1834@defmac FIRST_PSEUDO_REGISTER 1835Number of hardware registers known to the compiler. They receive 1836numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first 1837pseudo register's number really is assigned the number 1838@code{FIRST_PSEUDO_REGISTER}. 1839@end defmac 1840 1841@defmac FIXED_REGISTERS 1842@cindex fixed register 1843An initializer that says which registers are used for fixed purposes 1844all throughout the compiled code and are therefore not available for 1845general allocation. These would include the stack pointer, the frame 1846pointer (except on machines where that can be used as a general 1847register when no frame pointer is needed), the program counter on 1848machines where that is considered one of the addressable registers, 1849and any other numbered register with a standard use. 1850 1851This information is expressed as a sequence of numbers, separated by 1852commas and surrounded by braces. The @var{n}th number is 1 if 1853register @var{n} is fixed, 0 otherwise. 1854 1855The table initialized from this macro, and the table initialized by 1856the following one, may be overridden at run time either automatically, 1857by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by 1858the user with the command options @option{-ffixed-@var{reg}}, 1859@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}. 1860@end defmac 1861 1862@defmac CALL_USED_REGISTERS 1863@cindex call-used register 1864@cindex call-clobbered register 1865@cindex call-saved register 1866Like @code{FIXED_REGISTERS} but has 1 for each register that is 1867clobbered (in general) by function calls as well as for fixed 1868registers. This macro therefore identifies the registers that are not 1869available for general allocation of values that must live across 1870function calls. 1871 1872If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler 1873automatically saves it on function entry and restores it on function 1874exit, if the register is used within the function. 1875@end defmac 1876 1877@defmac CALL_REALLY_USED_REGISTERS 1878@cindex call-used register 1879@cindex call-clobbered register 1880@cindex call-saved register 1881Like @code{CALL_USED_REGISTERS} except this macro doesn't require 1882that the entire set of @code{FIXED_REGISTERS} be included. 1883(@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}). 1884This macro is optional. If not specified, it defaults to the value 1885of @code{CALL_USED_REGISTERS}. 1886@end defmac 1887 1888@defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode}) 1889@cindex call-used register 1890@cindex call-clobbered register 1891@cindex call-saved register 1892A C expression that is nonzero if it is not permissible to store a 1893value of mode @var{mode} in hard register number @var{regno} across a 1894call without some part of it being clobbered. For most machines this 1895macro need not be defined. It is only required for machines that do not 1896preserve the entire contents of a register across a call. 1897@end defmac 1898 1899@findex fixed_regs 1900@findex call_used_regs 1901@findex global_regs 1902@findex reg_names 1903@findex reg_class_contents 1904@defmac CONDITIONAL_REGISTER_USAGE 1905Zero or more C statements that may conditionally modify five variables 1906@code{fixed_regs}, @code{call_used_regs}, @code{global_regs}, 1907@code{reg_names}, and @code{reg_class_contents}, to take into account 1908any dependence of these register sets on target flags. The first three 1909of these are of type @code{char []} (interpreted as Boolean vectors). 1910@code{global_regs} is a @code{const char *[]}, and 1911@code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is 1912called, @code{fixed_regs}, @code{call_used_regs}, 1913@code{reg_class_contents}, and @code{reg_names} have been initialized 1914from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS}, 1915@code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively. 1916@code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}}, 1917@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}} 1918command options have been applied. 1919 1920You need not define this macro if it has no work to do. 1921 1922@cindex disabling certain registers 1923@cindex controlling register usage 1924If the usage of an entire class of registers depends on the target 1925flags, you may indicate this to GCC by using this macro to modify 1926@code{fixed_regs} and @code{call_used_regs} to 1 for each of the 1927registers in the classes which should not be used by GCC@. Also define 1928the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT} 1929to return @code{NO_REGS} if it 1930is called with a letter for a class that shouldn't be used. 1931 1932(However, if this class is not included in @code{GENERAL_REGS} and all 1933of the insn patterns whose constraints permit this class are 1934controlled by target switches, then GCC will automatically avoid using 1935these registers when the target switches are opposed to them.) 1936@end defmac 1937 1938@defmac NON_SAVING_SETJMP 1939If this macro is defined and has a nonzero value, it means that 1940@code{setjmp} and related functions fail to save the registers, or that 1941@code{longjmp} fails to restore them. To compensate, the compiler 1942avoids putting variables in registers in functions that use 1943@code{setjmp}. 1944@end defmac 1945 1946@defmac INCOMING_REGNO (@var{out}) 1947Define this macro if the target machine has register windows. This C 1948expression returns the register number as seen by the called function 1949corresponding to the register number @var{out} as seen by the calling 1950function. Return @var{out} if register number @var{out} is not an 1951outbound register. 1952@end defmac 1953 1954@defmac OUTGOING_REGNO (@var{in}) 1955Define this macro if the target machine has register windows. This C 1956expression returns the register number as seen by the calling function 1957corresponding to the register number @var{in} as seen by the called 1958function. Return @var{in} if register number @var{in} is not an inbound 1959register. 1960@end defmac 1961 1962@defmac LOCAL_REGNO (@var{regno}) 1963Define this macro if the target machine has register windows. This C 1964expression returns true if the register is call-saved but is in the 1965register window. Unlike most call-saved registers, such registers 1966need not be explicitly restored on function exit or during non-local 1967gotos. 1968@end defmac 1969 1970@defmac PC_REGNUM 1971If the program counter has a register number, define this as that 1972register number. Otherwise, do not define it. 1973@end defmac 1974 1975@node Allocation Order 1976@subsection Order of Allocation of Registers 1977@cindex order of register allocation 1978@cindex register allocation order 1979 1980@c prevent bad page break with this line 1981Registers are allocated in order. 1982 1983@defmac REG_ALLOC_ORDER 1984If defined, an initializer for a vector of integers, containing the 1985numbers of hard registers in the order in which GCC should prefer 1986to use them (from most preferred to least). 1987 1988If this macro is not defined, registers are used lowest numbered first 1989(all else being equal). 1990 1991One use of this macro is on machines where the highest numbered 1992registers must always be saved and the save-multiple-registers 1993instruction supports only sequences of consecutive registers. On such 1994machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists 1995the highest numbered allocable register first. 1996@end defmac 1997 1998@defmac ORDER_REGS_FOR_LOCAL_ALLOC 1999A C statement (sans semicolon) to choose the order in which to allocate 2000hard registers for pseudo-registers local to a basic block. 2001 2002Store the desired register order in the array @code{reg_alloc_order}. 2003Element 0 should be the register to allocate first; element 1, the next 2004register; and so on. 2005 2006The macro body should not assume anything about the contents of 2007@code{reg_alloc_order} before execution of the macro. 2008 2009On most machines, it is not necessary to define this macro. 2010@end defmac 2011 2012@node Values in Registers 2013@subsection How Values Fit in Registers 2014 2015This section discusses the macros that describe which kinds of values 2016(specifically, which machine modes) each register can hold, and how many 2017consecutive registers are needed for a given mode. 2018 2019@defmac HARD_REGNO_NREGS (@var{regno}, @var{mode}) 2020A C expression for the number of consecutive hard registers, starting 2021at register number @var{regno}, required to hold a value of mode 2022@var{mode}. 2023 2024On a machine where all registers are exactly one word, a suitable 2025definition of this macro is 2026 2027@smallexample 2028#define HARD_REGNO_NREGS(REGNO, MODE) \ 2029 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ 2030 / UNITS_PER_WORD) 2031@end smallexample 2032@end defmac 2033 2034@defmac REGMODE_NATURAL_SIZE (@var{mode}) 2035Define this macro if the natural size of registers that hold values 2036of mode @var{mode} is not the word size. It is a C expression that 2037should give the natural size in bytes for the specified mode. It is 2038used by the register allocator to try to optimize its results. This 2039happens for example on SPARC 64-bit where the natural size of 2040floating-point registers is still 32-bit. 2041@end defmac 2042 2043@defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode}) 2044A C expression that is nonzero if it is permissible to store a value 2045of mode @var{mode} in hard register number @var{regno} (or in several 2046registers starting with that one). For a machine where all registers 2047are equivalent, a suitable definition is 2048 2049@smallexample 2050#define HARD_REGNO_MODE_OK(REGNO, MODE) 1 2051@end smallexample 2052 2053You need not include code to check for the numbers of fixed registers, 2054because the allocation mechanism considers them to be always occupied. 2055 2056@cindex register pairs 2057On some machines, double-precision values must be kept in even/odd 2058register pairs. You can implement that by defining this macro to reject 2059odd register numbers for such modes. 2060 2061The minimum requirement for a mode to be OK in a register is that the 2062@samp{mov@var{mode}} instruction pattern support moves between the 2063register and other hard register in the same class and that moving a 2064value into the register and back out not alter it. 2065 2066Since the same instruction used to move @code{word_mode} will work for 2067all narrower integer modes, it is not necessary on any machine for 2068@code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided 2069you define patterns @samp{movhi}, etc., to take advantage of this. This 2070is useful because of the interaction between @code{HARD_REGNO_MODE_OK} 2071and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes 2072to be tieable. 2073 2074Many machines have special registers for floating point arithmetic. 2075Often people assume that floating point machine modes are allowed only 2076in floating point registers. This is not true. Any registers that 2077can hold integers can safely @emph{hold} a floating point machine 2078mode, whether or not floating arithmetic can be done on it in those 2079registers. Integer move instructions can be used to move the values. 2080 2081On some machines, though, the converse is true: fixed-point machine 2082modes may not go in floating registers. This is true if the floating 2083registers normalize any value stored in them, because storing a 2084non-floating value there would garble it. In this case, 2085@code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in 2086floating registers. But if the floating registers do not automatically 2087normalize, if you can store any bit pattern in one and retrieve it 2088unchanged without a trap, then any machine mode may go in a floating 2089register, so you can define this macro to say so. 2090 2091The primary significance of special floating registers is rather that 2092they are the registers acceptable in floating point arithmetic 2093instructions. However, this is of no concern to 2094@code{HARD_REGNO_MODE_OK}. You handle it by writing the proper 2095constraints for those instructions. 2096 2097On some machines, the floating registers are especially slow to access, 2098so that it is better to store a value in a stack frame than in such a 2099register if floating point arithmetic is not being done. As long as the 2100floating registers are not in class @code{GENERAL_REGS}, they will not 2101be used unless some pattern's constraint asks for one. 2102@end defmac 2103 2104@defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2}) 2105A C expression that is nonzero if a value of mode 2106@var{mode1} is accessible in mode @var{mode2} without copying. 2107 2108If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and 2109@code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for 2110any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})} 2111should be nonzero. If they differ for any @var{r}, you should define 2112this macro to return zero unless some other mechanism ensures the 2113accessibility of the value in a narrower mode. 2114 2115You should define this macro to return nonzero in as many cases as 2116possible since doing so will allow GCC to perform better register 2117allocation. 2118@end defmac 2119 2120@defmac AVOID_CCMODE_COPIES 2121Define this macro if the compiler should avoid copies to/from @code{CCmode} 2122registers. You should only define this macro if support for copying to/from 2123@code{CCmode} is incomplete. 2124@end defmac 2125 2126@node Leaf Functions 2127@subsection Handling Leaf Functions 2128 2129@cindex leaf functions 2130@cindex functions, leaf 2131On some machines, a leaf function (i.e., one which makes no calls) can run 2132more efficiently if it does not make its own register window. Often this 2133means it is required to receive its arguments in the registers where they 2134are passed by the caller, instead of the registers where they would 2135normally arrive. 2136 2137The special treatment for leaf functions generally applies only when 2138other conditions are met; for example, often they may use only those 2139registers for its own variables and temporaries. We use the term ``leaf 2140function'' to mean a function that is suitable for this special 2141handling, so that functions with no calls are not necessarily ``leaf 2142functions''. 2143 2144GCC assigns register numbers before it knows whether the function is 2145suitable for leaf function treatment. So it needs to renumber the 2146registers in order to output a leaf function. The following macros 2147accomplish this. 2148 2149@defmac LEAF_REGISTERS 2150Name of a char vector, indexed by hard register number, which 2151contains 1 for a register that is allowable in a candidate for leaf 2152function treatment. 2153 2154If leaf function treatment involves renumbering the registers, then the 2155registers marked here should be the ones before renumbering---those that 2156GCC would ordinarily allocate. The registers which will actually be 2157used in the assembler code, after renumbering, should not be marked with 1 2158in this vector. 2159 2160Define this macro only if the target machine offers a way to optimize 2161the treatment of leaf functions. 2162@end defmac 2163 2164@defmac LEAF_REG_REMAP (@var{regno}) 2165A C expression whose value is the register number to which @var{regno} 2166should be renumbered, when a function is treated as a leaf function. 2167 2168If @var{regno} is a register number which should not appear in a leaf 2169function before renumbering, then the expression should yield @minus{}1, which 2170will cause the compiler to abort. 2171 2172Define this macro only if the target machine offers a way to optimize the 2173treatment of leaf functions, and registers need to be renumbered to do 2174this. 2175@end defmac 2176 2177@findex current_function_is_leaf 2178@findex current_function_uses_only_leaf_regs 2179@code{TARGET_ASM_FUNCTION_PROLOGUE} and 2180@code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions 2181specially. They can test the C variable @code{current_function_is_leaf} 2182which is nonzero for leaf functions. @code{current_function_is_leaf} is 2183set prior to local register allocation and is valid for the remaining 2184compiler passes. They can also test the C variable 2185@code{current_function_uses_only_leaf_regs} which is nonzero for leaf 2186functions which only use leaf registers. 2187@code{current_function_uses_only_leaf_regs} is valid after reload and is 2188only useful if @code{LEAF_REGISTERS} is defined. 2189@c changed this to fix overfull. ALSO: why the "it" at the beginning 2190@c of the next paragraph?! --mew 2feb93 2191 2192@node Stack Registers 2193@subsection Registers That Form a Stack 2194 2195There are special features to handle computers where some of the 2196``registers'' form a stack. Stack registers are normally written by 2197pushing onto the stack, and are numbered relative to the top of the 2198stack. 2199 2200Currently, GCC can only handle one group of stack-like registers, and 2201they must be consecutively numbered. Furthermore, the existing 2202support for stack-like registers is specific to the 80387 floating 2203point coprocessor. If you have a new architecture that uses 2204stack-like registers, you will need to do substantial work on 2205@file{reg-stack.c} and write your machine description to cooperate 2206with it, as well as defining these macros. 2207 2208@defmac STACK_REGS 2209Define this if the machine has any stack-like registers. 2210@end defmac 2211 2212@defmac FIRST_STACK_REG 2213The number of the first stack-like register. This one is the top 2214of the stack. 2215@end defmac 2216 2217@defmac LAST_STACK_REG 2218The number of the last stack-like register. This one is the bottom of 2219the stack. 2220@end defmac 2221 2222@node Register Classes 2223@section Register Classes 2224@cindex register class definitions 2225@cindex class definitions, register 2226 2227On many machines, the numbered registers are not all equivalent. 2228For example, certain registers may not be allowed for indexed addressing; 2229certain registers may not be allowed in some instructions. These machine 2230restrictions are described to the compiler using @dfn{register classes}. 2231 2232You define a number of register classes, giving each one a name and saying 2233which of the registers belong to it. Then you can specify register classes 2234that are allowed as operands to particular instruction patterns. 2235 2236@findex ALL_REGS 2237@findex NO_REGS 2238In general, each register will belong to several classes. In fact, one 2239class must be named @code{ALL_REGS} and contain all the registers. Another 2240class must be named @code{NO_REGS} and contain no registers. Often the 2241union of two classes will be another class; however, this is not required. 2242 2243@findex GENERAL_REGS 2244One of the classes must be named @code{GENERAL_REGS}. There is nothing 2245terribly special about the name, but the operand constraint letters 2246@samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is 2247the same as @code{ALL_REGS}, just define it as a macro which expands 2248to @code{ALL_REGS}. 2249 2250Order the classes so that if class @var{x} is contained in class @var{y} 2251then @var{x} has a lower class number than @var{y}. 2252 2253The way classes other than @code{GENERAL_REGS} are specified in operand 2254constraints is through machine-dependent operand constraint letters. 2255You can define such letters to correspond to various classes, then use 2256them in operand constraints. 2257 2258You should define a class for the union of two classes whenever some 2259instruction allows both classes. For example, if an instruction allows 2260either a floating point (coprocessor) register or a general register for a 2261certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} 2262which includes both of them. Otherwise you will get suboptimal code. 2263 2264You must also specify certain redundant information about the register 2265classes: for each class, which classes contain it and which ones are 2266contained in it; for each pair of classes, the largest class contained 2267in their union. 2268 2269When a value occupying several consecutive registers is expected in a 2270certain class, all the registers used must belong to that class. 2271Therefore, register classes cannot be used to enforce a requirement for 2272a register pair to start with an even-numbered register. The way to 2273specify this requirement is with @code{HARD_REGNO_MODE_OK}. 2274 2275Register classes used for input-operands of bitwise-and or shift 2276instructions have a special requirement: each such class must have, for 2277each fixed-point machine mode, a subclass whose registers can transfer that 2278mode to or from memory. For example, on some machines, the operations for 2279single-byte values (@code{QImode}) are limited to certain registers. When 2280this is so, each register class that is used in a bitwise-and or shift 2281instruction must have a subclass consisting of registers from which 2282single-byte values can be loaded or stored. This is so that 2283@code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. 2284 2285@deftp {Data type} {enum reg_class} 2286An enumerated type that must be defined with all the register class names 2287as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS} 2288must be the last register class, followed by one more enumerated value, 2289@code{LIM_REG_CLASSES}, which is not a register class but rather 2290tells how many classes there are. 2291 2292Each register class has a number, which is the value of casting 2293the class name to type @code{int}. The number serves as an index 2294in many of the tables described below. 2295@end deftp 2296 2297@defmac N_REG_CLASSES 2298The number of distinct register classes, defined as follows: 2299 2300@smallexample 2301#define N_REG_CLASSES (int) LIM_REG_CLASSES 2302@end smallexample 2303@end defmac 2304 2305@defmac REG_CLASS_NAMES 2306An initializer containing the names of the register classes as C string 2307constants. These names are used in writing some of the debugging dumps. 2308@end defmac 2309 2310@defmac REG_CLASS_CONTENTS 2311An initializer containing the contents of the register classes, as integers 2312which are bit masks. The @var{n}th integer specifies the contents of class 2313@var{n}. The way the integer @var{mask} is interpreted is that 2314register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. 2315 2316When the machine has more than 32 registers, an integer does not suffice. 2317Then the integers are replaced by sub-initializers, braced groupings containing 2318several integers. Each sub-initializer must be suitable as an initializer 2319for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. 2320In this situation, the first integer in each sub-initializer corresponds to 2321registers 0 through 31, the second integer to registers 32 through 63, and 2322so on. 2323@end defmac 2324 2325@defmac REGNO_REG_CLASS (@var{regno}) 2326A C expression whose value is a register class containing hard register 2327@var{regno}. In general there is more than one such class; choose a class 2328which is @dfn{minimal}, meaning that no smaller class also contains the 2329register. 2330@end defmac 2331 2332@defmac BASE_REG_CLASS 2333A macro whose definition is the name of the class to which a valid 2334base register must belong. A base register is one used in an address 2335which is the register value plus a displacement. 2336@end defmac 2337 2338@defmac MODE_BASE_REG_CLASS (@var{mode}) 2339This is a variation of the @code{BASE_REG_CLASS} macro which allows 2340the selection of a base register in a mode dependent manner. If 2341@var{mode} is VOIDmode then it should return the same value as 2342@code{BASE_REG_CLASS}. 2343@end defmac 2344 2345@defmac INDEX_REG_CLASS 2346A macro whose definition is the name of the class to which a valid 2347index register must belong. An index register is one used in an 2348address where its value is either multiplied by a scale factor or 2349added to another register (as well as added to a displacement). 2350@end defmac 2351 2352@defmac CONSTRAINT_LEN (@var{char}, @var{str}) 2353For the constraint at the start of @var{str}, which starts with the letter 2354@var{c}, return the length. This allows you to have register class / 2355constant / extra constraints that are longer than a single letter; 2356you don't need to define this macro if you can do with single-letter 2357constraints only. The definition of this macro should use 2358DEFAULT_CONSTRAINT_LEN for all the characters that you don't want 2359to handle specially. 2360There are some sanity checks in genoutput.c that check the constraint lengths 2361for the md file, so you can also use this macro to help you while you are 2362transitioning from a byzantine single-letter-constraint scheme: when you 2363return a negative length for a constraint you want to re-use, genoutput 2364will complain about every instance where it is used in the md file. 2365@end defmac 2366 2367@defmac REG_CLASS_FROM_LETTER (@var{char}) 2368A C expression which defines the machine-dependent operand constraint 2369letters for register classes. If @var{char} is such a letter, the 2370value should be the register class corresponding to it. Otherwise, 2371the value should be @code{NO_REGS}. The register letter @samp{r}, 2372corresponding to class @code{GENERAL_REGS}, will not be passed 2373to this macro; you do not need to handle it. 2374@end defmac 2375 2376@defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str}) 2377Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string 2378passed in @var{str}, so that you can use suffixes to distinguish between 2379different variants. 2380@end defmac 2381 2382@defmac REGNO_OK_FOR_BASE_P (@var{num}) 2383A C expression which is nonzero if register number @var{num} is 2384suitable for use as a base register in operand addresses. It may be 2385either a suitable hard register or a pseudo register that has been 2386allocated such a hard register. 2387@end defmac 2388 2389@defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode}) 2390A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that 2391that expression may examine the mode of the memory reference in 2392@var{mode}. You should define this macro if the mode of the memory 2393reference affects whether a register may be used as a base register. If 2394you define this macro, the compiler will use it instead of 2395@code{REGNO_OK_FOR_BASE_P}. 2396@end defmac 2397 2398@defmac REGNO_OK_FOR_INDEX_P (@var{num}) 2399A C expression which is nonzero if register number @var{num} is 2400suitable for use as an index register in operand addresses. It may be 2401either a suitable hard register or a pseudo register that has been 2402allocated such a hard register. 2403 2404The difference between an index register and a base register is that 2405the index register may be scaled. If an address involves the sum of 2406two registers, neither one of them scaled, then either one may be 2407labeled the ``base'' and the other the ``index''; but whichever 2408labeling is used must fit the machine's constraints of which registers 2409may serve in each capacity. The compiler will try both labelings, 2410looking for one that is valid, and will reload one or both registers 2411only if neither labeling works. 2412@end defmac 2413 2414@defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) 2415A C expression that places additional restrictions on the register class 2416to use when it is necessary to copy value @var{x} into a register in class 2417@var{class}. The value is a register class; perhaps @var{class}, or perhaps 2418another, smaller class. On many machines, the following definition is 2419safe: 2420 2421@smallexample 2422#define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS 2423@end smallexample 2424 2425Sometimes returning a more restrictive class makes better code. For 2426example, on the 68000, when @var{x} is an integer constant that is in range 2427for a @samp{moveq} instruction, the value of this macro is always 2428@code{DATA_REGS} as long as @var{class} includes the data registers. 2429Requiring a data register guarantees that a @samp{moveq} will be used. 2430 2431One case where @code{PREFERRED_RELOAD_CLASS} must not return 2432@var{class} is if @var{x} is a legitimate constant which cannot be 2433loaded into some register class. By returning @code{NO_REGS} you can 2434force @var{x} into a memory location. For example, rs6000 can load 2435immediate values into general-purpose registers, but does not have an 2436instruction for loading an immediate value into a floating-point 2437register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when 2438@var{x} is a floating-point constant. If the constant can't be loaded 2439into any kind of register, code generation will be better if 2440@code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead 2441of using @code{PREFERRED_RELOAD_CLASS}. 2442@end defmac 2443 2444@defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class}) 2445Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of 2446input reloads. If you don't define this macro, the default is to use 2447@var{class}, unchanged. 2448@end defmac 2449 2450@defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) 2451A C expression that places additional restrictions on the register class 2452to use when it is necessary to be able to hold a value of mode 2453@var{mode} in a reload register for which class @var{class} would 2454ordinarily be used. 2455 2456Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when 2457there are certain modes that simply can't go in certain reload classes. 2458 2459The value is a register class; perhaps @var{class}, or perhaps another, 2460smaller class. 2461 2462Don't define this macro unless the target machine has limitations which 2463require the macro to do something nontrivial. 2464@end defmac 2465 2466@defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2467@defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2468@defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2469Many machines have some registers that cannot be copied directly to or 2470from memory or even from other types of registers. An example is the 2471@samp{MQ} register, which on most machines, can only be copied to or 2472from general registers, but not memory. Some machines allow copying all 2473registers to and from memory, but require a scratch register for stores 2474to some memory locations (e.g., those with symbolic address on the RT, 2475and those with certain symbolic address on the SPARC when compiling 2476PIC)@. In some cases, both an intermediate and a scratch register are 2477required. 2478 2479You should define these macros to indicate to the reload phase that it may 2480need to allocate at least one register for a reload in addition to the 2481register to contain the data. Specifically, if copying @var{x} to a 2482register @var{class} in @var{mode} requires an intermediate register, 2483you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the 2484largest register class all of whose registers can be used as 2485intermediate registers or scratch registers. 2486 2487If copying a register @var{class} in @var{mode} to @var{x} requires an 2488intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} 2489should be defined to return the largest register class required. If the 2490requirements for input and output reloads are the same, the macro 2491@code{SECONDARY_RELOAD_CLASS} should be used instead of defining both 2492macros identically. 2493 2494The values returned by these macros are often @code{GENERAL_REGS}. 2495Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} 2496can be directly copied to or from a register of @var{class} in 2497@var{mode} without requiring a scratch register. Do not define this 2498macro if it would always return @code{NO_REGS}. 2499 2500If a scratch register is required (either with or without an 2501intermediate register), you should define patterns for 2502@samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required 2503(@pxref{Standard Names}. These patterns, which will normally be 2504implemented with a @code{define_expand}, should be similar to the 2505@samp{mov@var{m}} patterns, except that operand 2 is the scratch 2506register. 2507 2508Define constraints for the reload register and scratch register that 2509contain a single register class. If the original reload register (whose 2510class is @var{class}) can meet the constraint given in the pattern, the 2511value returned by these macros is used for the class of the scratch 2512register. Otherwise, two additional reload registers are required. 2513Their classes are obtained from the constraints in the insn pattern. 2514 2515@var{x} might be a pseudo-register or a @code{subreg} of a 2516pseudo-register, which could either be in a hard register or in memory. 2517Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is 2518in memory and the hard register number if it is in a register. 2519 2520These macros should not be used in the case where a particular class of 2521registers can only be copied to memory and not to another class of 2522registers. In that case, secondary reload registers are not needed and 2523would not be helpful. Instead, a stack location must be used to perform 2524the copy and the @code{mov@var{m}} pattern should use memory as an 2525intermediate storage. This case often occurs between floating-point and 2526general registers. 2527@end defmac 2528 2529@defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m}) 2530Certain machines have the property that some registers cannot be copied 2531to some other registers without using memory. Define this macro on 2532those machines to be a C expression that is nonzero if objects of mode 2533@var{m} in registers of @var{class1} can only be copied to registers of 2534class @var{class2} by storing a register of @var{class1} into memory 2535and loading that memory location into a register of @var{class2}. 2536 2537Do not define this macro if its value would always be zero. 2538@end defmac 2539 2540@defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) 2541Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler 2542allocates a stack slot for a memory location needed for register copies. 2543If this macro is defined, the compiler instead uses the memory location 2544defined by this macro. 2545 2546Do not define this macro if you do not define 2547@code{SECONDARY_MEMORY_NEEDED}. 2548@end defmac 2549 2550@defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode}) 2551When the compiler needs a secondary memory location to copy between two 2552registers of mode @var{mode}, it normally allocates sufficient memory to 2553hold a quantity of @code{BITS_PER_WORD} bits and performs the store and 2554load operations in a mode that many bits wide and whose class is the 2555same as that of @var{mode}. 2556 2557This is right thing to do on most machines because it ensures that all 2558bits of the register are copied and prevents accesses to the registers 2559in a narrower mode, which some machines prohibit for floating-point 2560registers. 2561 2562However, this default behavior is not correct on some machines, such as 2563the DEC Alpha, that store short integers in floating-point registers 2564differently than in integer registers. On those machines, the default 2565widening will not work correctly and you must define this macro to 2566suppress that widening in some cases. See the file @file{alpha.h} for 2567details. 2568 2569Do not define this macro if you do not define 2570@code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that 2571is @code{BITS_PER_WORD} bits wide is correct for your machine. 2572@end defmac 2573 2574@defmac SMALL_REGISTER_CLASSES 2575On some machines, it is risky to let hard registers live across arbitrary 2576insns. Typically, these machines have instructions that require values 2577to be in specific registers (like an accumulator), and reload will fail 2578if the required hard register is used for another purpose across such an 2579insn. 2580 2581Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero 2582value on these machines. When this macro has a nonzero value, the 2583compiler will try to minimize the lifetime of hard registers. 2584 2585It is always safe to define this macro with a nonzero value, but if you 2586unnecessarily define it, you will reduce the amount of optimizations 2587that can be performed in some cases. If you do not define this macro 2588with a nonzero value when it is required, the compiler will run out of 2589spill registers and print a fatal error message. For most machines, you 2590should not define this macro at all. 2591@end defmac 2592 2593@defmac CLASS_LIKELY_SPILLED_P (@var{class}) 2594A C expression whose value is nonzero if pseudos that have been assigned 2595to registers of class @var{class} would likely be spilled because 2596registers of @var{class} are needed for spill registers. 2597 2598The default value of this macro returns 1 if @var{class} has exactly one 2599register and zero otherwise. On most machines, this default should be 2600used. Only define this macro to some other expression if pseudos 2601allocated by @file{local-alloc.c} end up in memory because their hard 2602registers were needed for spill registers. If this macro returns nonzero 2603for those classes, those pseudos will only be allocated by 2604@file{global.c}, which knows how to reallocate the pseudo to another 2605register. If there would not be another register available for 2606reallocation, you should not change the definition of this macro since 2607the only effect of such a definition would be to slow down register 2608allocation. 2609@end defmac 2610 2611@defmac CLASS_MAX_NREGS (@var{class}, @var{mode}) 2612A C expression for the maximum number of consecutive registers 2613of class @var{class} needed to hold a value of mode @var{mode}. 2614 2615This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, 2616the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} 2617should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, 2618@var{mode})} for all @var{regno} values in the class @var{class}. 2619 2620This macro helps control the handling of multiple-word values 2621in the reload pass. 2622@end defmac 2623 2624@defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class}) 2625If defined, a C expression that returns nonzero for a @var{class} for which 2626a change from mode @var{from} to mode @var{to} is invalid. 2627 2628For the example, loading 32-bit integer or floating-point objects into 2629floating-point registers on the Alpha extends them to 64 bits. 2630Therefore loading a 64-bit object and then storing it as a 32-bit object 2631does not store the low-order 32 bits, as would be the case for a normal 2632register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS} 2633as below: 2634 2635@smallexample 2636#define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ 2637 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \ 2638 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0) 2639@end smallexample 2640@end defmac 2641 2642Three other special macros describe which operands fit which constraint 2643letters. 2644 2645@defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c}) 2646A C expression that defines the machine-dependent operand constraint 2647letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify 2648particular ranges of integer values. If @var{c} is one of those 2649letters, the expression should check that @var{value}, an integer, is in 2650the appropriate range and return 1 if so, 0 otherwise. If @var{c} is 2651not one of those letters, the value should be 0 regardless of 2652@var{value}. 2653@end defmac 2654 2655@defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str}) 2656Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint 2657string passed in @var{str}, so that you can use suffixes to distinguish 2658between different variants. 2659@end defmac 2660 2661@defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c}) 2662A C expression that defines the machine-dependent operand constraint 2663letters that specify particular ranges of @code{const_double} values 2664(@samp{G} or @samp{H}). 2665 2666If @var{c} is one of those letters, the expression should check that 2667@var{value}, an RTX of code @code{const_double}, is in the appropriate 2668range and return 1 if so, 0 otherwise. If @var{c} is not one of those 2669letters, the value should be 0 regardless of @var{value}. 2670 2671@code{const_double} is used for all floating-point constants and for 2672@code{DImode} fixed-point constants. A given letter can accept either 2673or both kinds of values. It can use @code{GET_MODE} to distinguish 2674between these kinds. 2675@end defmac 2676 2677@defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str}) 2678Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint 2679string passed in @var{str}, so that you can use suffixes to distinguish 2680between different variants. 2681@end defmac 2682 2683@defmac EXTRA_CONSTRAINT (@var{value}, @var{c}) 2684A C expression that defines the optional machine-dependent constraint 2685letters that can be used to segregate specific types of operands, usually 2686memory references, for the target machine. Any letter that is not 2687elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} / 2688@code{REG_CLASS_FROM_CONSTRAINT} 2689may be used. Normally this macro will not be defined. 2690 2691If it is required for a particular target machine, it should return 1 2692if @var{value} corresponds to the operand type represented by the 2693constraint letter @var{c}. If @var{c} is not defined as an extra 2694constraint, the value returned should be 0 regardless of @var{value}. 2695 2696For example, on the ROMP, load instructions cannot have their output 2697in r0 if the memory reference contains a symbolic address. Constraint 2698letter @samp{Q} is defined as representing a memory address that does 2699@emph{not} contain a symbolic address. An alternative is specified with 2700a @samp{Q} constraint on the input and @samp{r} on the output. The next 2701alternative specifies @samp{m} on the input and a register class that 2702does not include r0 on the output. 2703@end defmac 2704 2705@defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str}) 2706Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed 2707in @var{str}, so that you can use suffixes to distinguish between different 2708variants. 2709@end defmac 2710 2711@defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str}) 2712A C expression that defines the optional machine-dependent constraint 2713letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should 2714be treated like memory constraints by the reload pass. 2715 2716It should return 1 if the operand type represented by the constraint 2717at the start of @var{str}, the first letter of which is the letter @var{c}, 2718 comprises a subset of all memory references including 2719all those whose address is simply a base register. This allows the reload 2720pass to reload an operand, if it does not directly correspond to the operand 2721type of @var{c}, by copying its address into a base register. 2722 2723For example, on the S/390, some instructions do not accept arbitrary 2724memory references, but only those that do not make use of an index 2725register. The constraint letter @samp{Q} is defined via 2726@code{EXTRA_CONSTRAINT} as representing a memory address of this type. 2727If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT}, 2728a @samp{Q} constraint can handle any memory operand, because the 2729reload pass knows it can be reloaded by copying the memory address 2730into a base register if required. This is analogous to the way 2731a @samp{o} constraint can handle any memory operand. 2732@end defmac 2733 2734@defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str}) 2735A C expression that defines the optional machine-dependent constraint 2736letters, amongst those accepted by @code{EXTRA_CONSTRAINT} / 2737@code{EXTRA_CONSTRAINT_STR}, that should 2738be treated like address constraints by the reload pass. 2739 2740It should return 1 if the operand type represented by the constraint 2741at the start of @var{str}, which starts with the letter @var{c}, comprises 2742a subset of all memory addresses including 2743all those that consist of just a base register. This allows the reload 2744pass to reload an operand, if it does not directly correspond to the operand 2745type of @var{str}, by copying it into a base register. 2746 2747Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only 2748be used with the @code{address_operand} predicate. It is treated 2749analogously to the @samp{p} constraint. 2750@end defmac 2751 2752@node Stack and Calling 2753@section Stack Layout and Calling Conventions 2754@cindex calling conventions 2755 2756@c prevent bad page break with this line 2757This describes the stack layout and calling conventions. 2758 2759@menu 2760* Frame Layout:: 2761* Exception Handling:: 2762* Stack Checking:: 2763* Frame Registers:: 2764* Elimination:: 2765* Stack Arguments:: 2766* Register Arguments:: 2767* Scalar Return:: 2768* Aggregate Return:: 2769* Caller Saves:: 2770* Function Entry:: 2771* Profiling:: 2772* Tail Calls:: 2773@end menu 2774 2775@node Frame Layout 2776@subsection Basic Stack Layout 2777@cindex stack frame layout 2778@cindex frame layout 2779 2780@c prevent bad page break with this line 2781Here is the basic stack layout. 2782 2783@defmac STACK_GROWS_DOWNWARD 2784Define this macro if pushing a word onto the stack moves the stack 2785pointer to a smaller address. 2786 2787When we say, ``define this macro if @dots{},'' it means that the 2788compiler checks this macro only with @code{#ifdef} so the precise 2789definition used does not matter. 2790@end defmac 2791 2792@defmac STACK_PUSH_CODE 2793This macro defines the operation used when something is pushed 2794on the stack. In RTL, a push operation will be 2795@code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})} 2796 2797The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC}, 2798and @code{POST_INC}. Which of these is correct depends on 2799the stack direction and on whether the stack pointer points 2800to the last item on the stack or whether it points to the 2801space for the next item on the stack. 2802 2803The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is 2804defined, which is almost always right, and @code{PRE_INC} otherwise, 2805which is often wrong. 2806@end defmac 2807 2808@defmac FRAME_GROWS_DOWNWARD 2809Define this macro if the addresses of local variable slots are at negative 2810offsets from the frame pointer. 2811@end defmac 2812 2813@defmac ARGS_GROW_DOWNWARD 2814Define this macro if successive arguments to a function occupy decreasing 2815addresses on the stack. 2816@end defmac 2817 2818@defmac STARTING_FRAME_OFFSET 2819Offset from the frame pointer to the first local variable slot to be allocated. 2820 2821If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by 2822subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}. 2823Otherwise, it is found by adding the length of the first slot to the 2824value @code{STARTING_FRAME_OFFSET}. 2825@c i'm not sure if the above is still correct.. had to change it to get 2826@c rid of an overfull. --mew 2feb93 2827@end defmac 2828 2829@defmac STACK_ALIGNMENT_NEEDED 2830Define to zero to disable final alignment of the stack during reload. 2831The nonzero default for this macro is suitable for most ports. 2832 2833On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there 2834is a register save block following the local block that doesn't require 2835alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable 2836stack alignment and do it in the backend. 2837@end defmac 2838 2839@defmac STACK_POINTER_OFFSET 2840Offset from the stack pointer register to the first location at which 2841outgoing arguments are placed. If not specified, the default value of 2842zero is used. This is the proper value for most machines. 2843 2844If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2845the first location at which outgoing arguments are placed. 2846@end defmac 2847 2848@defmac FIRST_PARM_OFFSET (@var{fundecl}) 2849Offset from the argument pointer register to the first argument's 2850address. On some machines it may depend on the data type of the 2851function. 2852 2853If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2854the first argument's address. 2855@end defmac 2856 2857@defmac STACK_DYNAMIC_OFFSET (@var{fundecl}) 2858Offset from the stack pointer register to an item dynamically allocated 2859on the stack, e.g., by @code{alloca}. 2860 2861The default value for this macro is @code{STACK_POINTER_OFFSET} plus the 2862length of the outgoing arguments. The default is correct for most 2863machines. See @file{function.c} for details. 2864@end defmac 2865 2866@defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) 2867A C expression whose value is RTL representing the address in a stack 2868frame where the pointer to the caller's frame is stored. Assume that 2869@var{frameaddr} is an RTL expression for the address of the stack frame 2870itself. 2871 2872If you don't define this macro, the default is to return the value 2873of @var{frameaddr}---that is, the stack frame address is also the 2874address of the stack word that points to the previous frame. 2875@end defmac 2876 2877@defmac SETUP_FRAME_ADDRESSES 2878If defined, a C expression that produces the machine-specific code to 2879setup the stack so that arbitrary frames can be accessed. For example, 2880on the SPARC, we must flush all of the register windows to the stack 2881before we can access arbitrary stack frames. You will seldom need to 2882define this macro. 2883@end defmac 2884 2885@defmac BUILTIN_SETJMP_FRAME_VALUE 2886If defined, a C expression that contains an rtx that is used to store 2887the address of the current frame into the built in @code{setjmp} buffer. 2888The default value, @code{virtual_stack_vars_rtx}, is correct for most 2889machines. One reason you may need to define this macro is if 2890@code{hard_frame_pointer_rtx} is the appropriate value on your machine. 2891@end defmac 2892 2893@defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) 2894A C expression whose value is RTL representing the value of the return 2895address for the frame @var{count} steps up from the current frame, after 2896the prologue. @var{frameaddr} is the frame pointer of the @var{count} 2897frame, or the frame pointer of the @var{count} @minus{} 1 frame if 2898@code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined. 2899 2900The value of the expression must always be the correct address when 2901@var{count} is zero, but may be @code{NULL_RTX} if there is not way to 2902determine the return address of other frames. 2903@end defmac 2904 2905@defmac RETURN_ADDR_IN_PREVIOUS_FRAME 2906Define this if the return address of a particular stack frame is accessed 2907from the frame pointer of the previous stack frame. 2908@end defmac 2909 2910@defmac INCOMING_RETURN_ADDR_RTX 2911A C expression whose value is RTL representing the location of the 2912incoming return address at the beginning of any function, before the 2913prologue. This RTL is either a @code{REG}, indicating that the return 2914value is saved in @samp{REG}, or a @code{MEM} representing a location in 2915the stack. 2916 2917You only need to define this macro if you want to support call frame 2918debugging information like that provided by DWARF 2. 2919 2920If this RTL is a @code{REG}, you should also define 2921@code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}. 2922@end defmac 2923 2924@defmac DWARF_ALT_FRAME_RETURN_COLUMN 2925A C expression whose value is an integer giving a DWARF 2 column 2926number that may be used as an alternate return column. This should 2927be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a 2928general register, but an alternate column needs to be used for 2929signal frames. 2930@end defmac 2931 2932@defmac INCOMING_FRAME_SP_OFFSET 2933A C expression whose value is an integer giving the offset, in bytes, 2934from the value of the stack pointer register to the top of the stack 2935frame at the beginning of any function, before the prologue. The top of 2936the frame is defined to be the value of the stack pointer in the 2937previous frame, just before the call instruction. 2938 2939You only need to define this macro if you want to support call frame 2940debugging information like that provided by DWARF 2. 2941@end defmac 2942 2943@defmac ARG_POINTER_CFA_OFFSET (@var{fundecl}) 2944A C expression whose value is an integer giving the offset, in bytes, 2945from the argument pointer to the canonical frame address (cfa). The 2946final value should coincide with that calculated by 2947@code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable 2948during virtual register instantiation. 2949 2950The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)}, 2951which is correct for most machines; in general, the arguments are found 2952immediately before the stack frame. Note that this is not the case on 2953some targets that save registers into the caller's frame, such as SPARC 2954and rs6000, and so such targets need to define this macro. 2955 2956You only need to define this macro if the default is incorrect, and you 2957want to support call frame debugging information like that provided by 2958DWARF 2. 2959@end defmac 2960 2961@node Exception Handling 2962@subsection Exception Handling Support 2963@cindex exception handling 2964 2965@defmac EH_RETURN_DATA_REGNO (@var{N}) 2966A C expression whose value is the @var{N}th register number used for 2967data by exception handlers, or @code{INVALID_REGNUM} if fewer than 2968@var{N} registers are usable. 2969 2970The exception handling library routines communicate with the exception 2971handlers via a set of agreed upon registers. Ideally these registers 2972should be call-clobbered; it is possible to use call-saved registers, 2973but may negatively impact code size. The target must support at least 29742 data registers, but should define 4 if there are enough free registers. 2975 2976You must define this macro if you want to support call frame exception 2977handling like that provided by DWARF 2. 2978@end defmac 2979 2980@defmac EH_RETURN_STACKADJ_RTX 2981A C expression whose value is RTL representing a location in which 2982to store a stack adjustment to be applied before function return. 2983This is used to unwind the stack to an exception handler's call frame. 2984It will be assigned zero on code paths that return normally. 2985 2986Typically this is a call-clobbered hard register that is otherwise 2987untouched by the epilogue, but could also be a stack slot. 2988 2989Do not define this macro if the stack pointer is saved and restored 2990by the regular prolog and epilog code in the call frame itself; in 2991this case, the exception handling library routines will update the 2992stack location to be restored in place. Otherwise, you must define 2993this macro if you want to support call frame exception handling like 2994that provided by DWARF 2. 2995@end defmac 2996 2997@defmac EH_RETURN_HANDLER_RTX 2998A C expression whose value is RTL representing a location in which 2999to store the address of an exception handler to which we should 3000return. It will not be assigned on code paths that return normally. 3001 3002Typically this is the location in the call frame at which the normal 3003return address is stored. For targets that return by popping an 3004address off the stack, this might be a memory address just below 3005the @emph{target} call frame rather than inside the current call 3006frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already 3007been assigned, so it may be used to calculate the location of the 3008target call frame. 3009 3010Some targets have more complex requirements than storing to an 3011address calculable during initial code generation. In that case 3012the @code{eh_return} instruction pattern should be used instead. 3013 3014If you want to support call frame exception handling, you must 3015define either this macro or the @code{eh_return} instruction pattern. 3016@end defmac 3017 3018@defmac RETURN_ADDR_OFFSET 3019If defined, an integer-valued C expression for which rtl will be generated 3020to add it to the exception handler address before it is searched in the 3021exception handling tables, and to subtract it again from the address before 3022using it to return to the exception handler. 3023@end defmac 3024 3025@defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global}) 3026This macro chooses the encoding of pointers embedded in the exception 3027handling sections. If at all possible, this should be defined such 3028that the exception handling section will not require dynamic relocations, 3029and so may be read-only. 3030 3031@var{code} is 0 for data, 1 for code labels, 2 for function pointers. 3032@var{global} is true if the symbol may be affected by dynamic relocations. 3033The macro should return a combination of the @code{DW_EH_PE_*} defines 3034as found in @file{dwarf2.h}. 3035 3036If this macro is not defined, pointers will not be encoded but 3037represented directly. 3038@end defmac 3039 3040@defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done}) 3041This macro allows the target to emit whatever special magic is required 3042to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}. 3043Generic code takes care of pc-relative and indirect encodings; this must 3044be defined if the target uses text-relative or data-relative encodings. 3045 3046This is a C statement that branches to @var{done} if the format was 3047handled. @var{encoding} is the format chosen, @var{size} is the number 3048of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF} 3049to be emitted. 3050@end defmac 3051 3052@defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}, @var{success}) 3053This macro allows the target to add cpu and operating system specific 3054code to the call-frame unwinder for use when there is no unwind data 3055available. The most common reason to implement this macro is to unwind 3056through signal frames. 3057 3058This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c} 3059and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; 3060@var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra} 3061for the address of the code being executed and @code{context->cfa} for 3062the stack pointer value. If the frame can be decoded, the register save 3063addresses should be updated in @var{fs} and the macro should branch to 3064@var{success}. If the frame cannot be decoded, the macro should do 3065nothing. 3066 3067For proper signal handling in Java this macro is accompanied by 3068@code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers. 3069@end defmac 3070 3071@defmac MD_HANDLE_UNWABI (@var{context}, @var{fs}) 3072This macro allows the target to add operating system specific code to the 3073call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive, 3074usually used for signal or interrupt frames. 3075 3076This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}. 3077@var{context} is an @code{_Unwind_Context}; 3078@var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi} 3079for the abi and context in the @code{.unwabi} directive. If the 3080@code{.unwabi} directive can be handled, the register save addresses should 3081be updated in @var{fs}. 3082@end defmac 3083 3084@node Stack Checking 3085@subsection Specifying How Stack Checking is Done 3086 3087GCC will check that stack references are within the boundaries of 3088the stack, if the @option{-fstack-check} is specified, in one of three ways: 3089 3090@enumerate 3091@item 3092If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC 3093will assume that you have arranged for stack checking to be done at 3094appropriate places in the configuration files, e.g., in 3095@code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special 3096processing. 3097 3098@item 3099If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern 3100called @code{check_stack} in your @file{md} file, GCC will call that 3101pattern with one argument which is the address to compare the stack 3102value against. You must arrange for this pattern to report an error if 3103the stack pointer is out of range. 3104 3105@item 3106If neither of the above are true, GCC will generate code to periodically 3107``probe'' the stack pointer using the values of the macros defined below. 3108@end enumerate 3109 3110Normally, you will use the default values of these macros, so GCC 3111will use the third approach. 3112 3113@defmac STACK_CHECK_BUILTIN 3114A nonzero value if stack checking is done by the configuration files in a 3115machine-dependent manner. You should define this macro if stack checking 3116is require by the ABI of your machine or if you would like to have to stack 3117checking in some more efficient way than GCC's portable approach. 3118The default value of this macro is zero. 3119@end defmac 3120 3121@defmac STACK_CHECK_PROBE_INTERVAL 3122An integer representing the interval at which GCC must generate stack 3123probe instructions. You will normally define this macro to be no larger 3124than the size of the ``guard pages'' at the end of a stack area. The 3125default value of 4096 is suitable for most systems. 3126@end defmac 3127 3128@defmac STACK_CHECK_PROBE_LOAD 3129A integer which is nonzero if GCC should perform the stack probe 3130as a load instruction and zero if GCC should use a store instruction. 3131The default is zero, which is the most efficient choice on most systems. 3132@end defmac 3133 3134@defmac STACK_CHECK_PROTECT 3135The number of bytes of stack needed to recover from a stack overflow, 3136for languages where such a recovery is supported. The default value of 313775 words should be adequate for most machines. 3138@end defmac 3139 3140@defmac STACK_CHECK_MAX_FRAME_SIZE 3141The maximum size of a stack frame, in bytes. GCC will generate probe 3142instructions in non-leaf functions to ensure at least this many bytes of 3143stack are available. If a stack frame is larger than this size, stack 3144checking will not be reliable and GCC will issue a warning. The 3145default is chosen so that GCC only generates one instruction on most 3146systems. You should normally not change the default value of this macro. 3147@end defmac 3148 3149@defmac STACK_CHECK_FIXED_FRAME_SIZE 3150GCC uses this value to generate the above warning message. It 3151represents the amount of fixed frame used by a function, not including 3152space for any callee-saved registers, temporaries and user variables. 3153You need only specify an upper bound for this amount and will normally 3154use the default of four words. 3155@end defmac 3156 3157@defmac STACK_CHECK_MAX_VAR_SIZE 3158The maximum size, in bytes, of an object that GCC will place in the 3159fixed area of the stack frame when the user specifies 3160@option{-fstack-check}. 3161GCC computed the default from the values of the above macros and you will 3162normally not need to override that default. 3163@end defmac 3164 3165@need 2000 3166@node Frame Registers 3167@subsection Registers That Address the Stack Frame 3168 3169@c prevent bad page break with this line 3170This discusses registers that address the stack frame. 3171 3172@defmac STACK_POINTER_REGNUM 3173The register number of the stack pointer register, which must also be a 3174fixed register according to @code{FIXED_REGISTERS}. On most machines, 3175the hardware determines which register this is. 3176@end defmac 3177 3178@defmac FRAME_POINTER_REGNUM 3179The register number of the frame pointer register, which is used to 3180access automatic variables in the stack frame. On some machines, the 3181hardware determines which register this is. On other machines, you can 3182choose any register you wish for this purpose. 3183@end defmac 3184 3185@defmac HARD_FRAME_POINTER_REGNUM 3186On some machines the offset between the frame pointer and starting 3187offset of the automatic variables is not known until after register 3188allocation has been done (for example, because the saved registers are 3189between these two locations). On those machines, define 3190@code{FRAME_POINTER_REGNUM} the number of a special, fixed register to 3191be used internally until the offset is known, and define 3192@code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number 3193used for the frame pointer. 3194 3195You should define this macro only in the very rare circumstances when it 3196is not possible to calculate the offset between the frame pointer and 3197the automatic variables until after register allocation has been 3198completed. When this macro is defined, you must also indicate in your 3199definition of @code{ELIMINABLE_REGS} how to eliminate 3200@code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} 3201or @code{STACK_POINTER_REGNUM}. 3202 3203Do not define this macro if it would be the same as 3204@code{FRAME_POINTER_REGNUM}. 3205@end defmac 3206 3207@defmac ARG_POINTER_REGNUM 3208The register number of the arg pointer register, which is used to access 3209the function's argument list. On some machines, this is the same as the 3210frame pointer register. On some machines, the hardware determines which 3211register this is. On other machines, you can choose any register you 3212wish for this purpose. If this is not the same register as the frame 3213pointer register, then you must mark it as a fixed register according to 3214@code{FIXED_REGISTERS}, or arrange to be able to eliminate it 3215(@pxref{Elimination}). 3216@end defmac 3217 3218@defmac RETURN_ADDRESS_POINTER_REGNUM 3219The register number of the return address pointer register, which is used to 3220access the current function's return address from the stack. On some 3221machines, the return address is not at a fixed offset from the frame 3222pointer or stack pointer or argument pointer. This register can be defined 3223to point to the return address on the stack, and then be converted by 3224@code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. 3225 3226Do not define this macro unless there is no other way to get the return 3227address from the stack. 3228@end defmac 3229 3230@defmac STATIC_CHAIN_REGNUM 3231@defmacx STATIC_CHAIN_INCOMING_REGNUM 3232Register numbers used for passing a function's static chain pointer. If 3233register windows are used, the register number as seen by the called 3234function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register 3235number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If 3236these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need 3237not be defined. 3238 3239The static chain register need not be a fixed register. 3240 3241If the static chain is passed in memory, these macros should not be 3242defined; instead, the next two macros should be defined. 3243@end defmac 3244 3245@defmac STATIC_CHAIN 3246@defmacx STATIC_CHAIN_INCOMING 3247If the static chain is passed in memory, these macros provide rtx giving 3248@code{mem} expressions that denote where they are stored. 3249@code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations 3250as seen by the calling and called functions, respectively. Often the former 3251will be at an offset from the stack pointer and the latter at an offset from 3252the frame pointer. 3253 3254@findex stack_pointer_rtx 3255@findex frame_pointer_rtx 3256@findex arg_pointer_rtx 3257The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and 3258@code{arg_pointer_rtx} will have been initialized prior to the use of these 3259macros and should be used to refer to those items. 3260 3261If the static chain is passed in a register, the two previous macros should 3262be defined instead. 3263@end defmac 3264 3265@defmac DWARF_FRAME_REGISTERS 3266This macro specifies the maximum number of hard registers that can be 3267saved in a call frame. This is used to size data structures used in 3268DWARF2 exception handling. 3269 3270Prior to GCC 3.0, this macro was needed in order to establish a stable 3271exception handling ABI in the face of adding new hard registers for ISA 3272extensions. In GCC 3.0 and later, the EH ABI is insulated from changes 3273in the number of hard registers. Nevertheless, this macro can still be 3274used to reduce the runtime memory requirements of the exception handling 3275routines, which can be substantial if the ISA contains a lot of 3276registers that are not call-saved. 3277 3278If this macro is not defined, it defaults to 3279@code{FIRST_PSEUDO_REGISTER}. 3280@end defmac 3281 3282@defmac PRE_GCC3_DWARF_FRAME_REGISTERS 3283 3284This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided 3285for backward compatibility in pre GCC 3.0 compiled code. 3286 3287If this macro is not defined, it defaults to 3288@code{DWARF_FRAME_REGISTERS}. 3289@end defmac 3290 3291@defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno}) 3292 3293Define this macro if the target's representation for dwarf registers 3294is different than the internal representation for unwind column. 3295Given a dwarf register, this macro should return the internal unwind 3296column number to use instead. 3297 3298See the PowerPC's SPE target for an example. 3299@end defmac 3300 3301@defmac DWARF_FRAME_REGNUM (@var{regno}) 3302 3303Define this macro if the target's representation for dwarf registers 3304used in .eh_frame or .debug_frame is different from that used in other 3305debug info sections. Given a GCC hard register number, this macro 3306should return the .eh_frame register number. The default is 3307@code{DBX_REGISTER_NUMBER (@var{regno})}. 3308 3309@end defmac 3310 3311@defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh}) 3312 3313Define this macro to map register numbers held in the call frame info 3314that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that 3315should be output in .debug_frame (@code{@var{for_eh}} is zero) and 3316.eh_frame (@code{@var{for_eh}} is nonzero). The default is to 3317return @code{@var{regno}}. 3318 3319@end defmac 3320 3321@node Elimination 3322@subsection Eliminating Frame Pointer and Arg Pointer 3323 3324@c prevent bad page break with this line 3325This is about eliminating the frame pointer and arg pointer. 3326 3327@defmac FRAME_POINTER_REQUIRED 3328A C expression which is nonzero if a function must have and use a frame 3329pointer. This expression is evaluated in the reload pass. If its value is 3330nonzero the function will have a frame pointer. 3331 3332The expression can in principle examine the current function and decide 3333according to the facts, but on most machines the constant 0 or the 3334constant 1 suffices. Use 0 when the machine allows code to be generated 3335with no frame pointer, and doing so saves some time or space. Use 1 3336when there is no possible advantage to avoiding a frame pointer. 3337 3338In certain cases, the compiler does not know how to produce valid code 3339without a frame pointer. The compiler recognizes those cases and 3340automatically gives the function a frame pointer regardless of what 3341@code{FRAME_POINTER_REQUIRED} says. You don't need to worry about 3342them. 3343 3344In a function that does not require a frame pointer, the frame pointer 3345register can be allocated for ordinary usage, unless you mark it as a 3346fixed register. See @code{FIXED_REGISTERS} for more information. 3347@end defmac 3348 3349@findex get_frame_size 3350@defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var}) 3351A C statement to store in the variable @var{depth-var} the difference 3352between the frame pointer and the stack pointer values immediately after 3353the function prologue. The value would be computed from information 3354such as the result of @code{get_frame_size ()} and the tables of 3355registers @code{regs_ever_live} and @code{call_used_regs}. 3356 3357If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and 3358need not be defined. Otherwise, it must be defined even if 3359@code{FRAME_POINTER_REQUIRED} is defined to always be true; in that 3360case, you may set @var{depth-var} to anything. 3361@end defmac 3362 3363@defmac ELIMINABLE_REGS 3364If defined, this macro specifies a table of register pairs used to 3365eliminate unneeded registers that point into the stack frame. If it is not 3366defined, the only elimination attempted by the compiler is to replace 3367references to the frame pointer with references to the stack pointer. 3368 3369The definition of this macro is a list of structure initializations, each 3370of which specifies an original and replacement register. 3371 3372On some machines, the position of the argument pointer is not known until 3373the compilation is completed. In such a case, a separate hard register 3374must be used for the argument pointer. This register can be eliminated by 3375replacing it with either the frame pointer or the argument pointer, 3376depending on whether or not the frame pointer has been eliminated. 3377 3378In this case, you might specify: 3379@smallexample 3380#define ELIMINABLE_REGS \ 3381@{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ 3382 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ 3383 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} 3384@end smallexample 3385 3386Note that the elimination of the argument pointer with the stack pointer is 3387specified first since that is the preferred elimination. 3388@end defmac 3389 3390@defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg}) 3391A C expression that returns nonzero if the compiler is allowed to try 3392to replace register number @var{from-reg} with register number 3393@var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS} 3394is defined, and will usually be the constant 1, since most of the cases 3395preventing register elimination are things that the compiler already 3396knows about. 3397@end defmac 3398 3399@defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) 3400This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It 3401specifies the initial difference between the specified pair of 3402registers. This macro must be defined if @code{ELIMINABLE_REGS} is 3403defined. 3404@end defmac 3405 3406@node Stack Arguments 3407@subsection Passing Function Arguments on the Stack 3408@cindex arguments on stack 3409@cindex stack arguments 3410 3411The macros in this section control how arguments are passed 3412on the stack. See the following section for other macros that 3413control passing certain arguments in registers. 3414 3415@deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype}) 3416This target hook returns @code{true} if an argument declared in a 3417prototype as an integral type smaller than @code{int} should actually be 3418passed as an @code{int}. In addition to avoiding errors in certain 3419cases of mismatch, it also makes for better code on certain machines. 3420The default is to not promote prototypes. 3421@end deftypefn 3422 3423@defmac PUSH_ARGS 3424A C expression. If nonzero, push insns will be used to pass 3425outgoing arguments. 3426If the target machine does not have a push instruction, set it to zero. 3427That directs GCC to use an alternate strategy: to 3428allocate the entire argument block and then store the arguments into 3429it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too. 3430@end defmac 3431 3432@defmac PUSH_ARGS_REVERSED 3433A C expression. If nonzero, function arguments will be evaluated from 3434last to first, rather than from first to last. If this macro is not 3435defined, it defaults to @code{PUSH_ARGS} on targets where the stack 3436and args grow in opposite directions, and 0 otherwise. 3437@end defmac 3438 3439@defmac PUSH_ROUNDING (@var{npushed}) 3440A C expression that is the number of bytes actually pushed onto the 3441stack when an instruction attempts to push @var{npushed} bytes. 3442 3443On some machines, the definition 3444 3445@smallexample 3446#define PUSH_ROUNDING(BYTES) (BYTES) 3447@end smallexample 3448 3449@noindent 3450will suffice. But on other machines, instructions that appear 3451to push one byte actually push two bytes in an attempt to maintain 3452alignment. Then the definition should be 3453 3454@smallexample 3455#define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) 3456@end smallexample 3457@end defmac 3458 3459@findex current_function_outgoing_args_size 3460@defmac ACCUMULATE_OUTGOING_ARGS 3461A C expression. If nonzero, the maximum amount of space required for outgoing arguments 3462will be computed and placed into the variable 3463@code{current_function_outgoing_args_size}. No space will be pushed 3464onto the stack for each call; instead, the function prologue should 3465increase the stack frame size by this amount. 3466 3467Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS} 3468is not proper. 3469@end defmac 3470 3471@defmac REG_PARM_STACK_SPACE (@var{fndecl}) 3472Define this macro if functions should assume that stack space has been 3473allocated for arguments even when their values are passed in 3474registers. 3475 3476The value of this macro is the size, in bytes, of the area reserved for 3477arguments passed in registers for the function represented by @var{fndecl}, 3478which can be zero if GCC is calling a library function. 3479 3480This space can be allocated by the caller, or be a part of the 3481machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says 3482which. 3483@end defmac 3484@c above is overfull. not sure what to do. --mew 5feb93 did 3485@c something, not sure if it looks good. --mew 10feb93 3486 3487@defmac MAYBE_REG_PARM_STACK_SPACE 3488@defmacx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size}) 3489Define these macros in addition to the one above if functions might 3490allocate stack space for arguments even when their values are passed 3491in registers. These should be used when the stack space allocated 3492for arguments in registers is not a simple constant independent of the 3493function declaration. 3494 3495The value of the first macro is the size, in bytes, of the area that 3496we should initially assume would be reserved for arguments passed in registers. 3497 3498The value of the second macro is the actual size, in bytes, of the area 3499that will be reserved for arguments passed in registers. This takes two 3500arguments: an integer representing the number of bytes of fixed sized 3501arguments on the stack, and a tree representing the number of bytes of 3502variable sized arguments on the stack. 3503 3504When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be 3505called for libcall functions, the current function, or for a function 3506being called when it is known that such stack space must be allocated. 3507In each case this value can be easily computed. 3508 3509When deciding whether a called function needs such stack space, and how 3510much space to reserve, GCC uses these two macros instead of 3511@code{REG_PARM_STACK_SPACE}. 3512@end defmac 3513 3514@defmac OUTGOING_REG_PARM_STACK_SPACE 3515Define this if it is the responsibility of the caller to allocate the area 3516reserved for arguments passed in registers. 3517 3518If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls 3519whether the space for these arguments counts in the value of 3520@code{current_function_outgoing_args_size}. 3521@end defmac 3522 3523@defmac STACK_PARMS_IN_REG_PARM_AREA 3524Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the 3525stack parameters don't skip the area specified by it. 3526@c i changed this, makes more sens and it should have taken care of the 3527@c overfull.. not as specific, tho. --mew 5feb93 3528 3529Normally, when a parameter is not passed in registers, it is placed on the 3530stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro 3531suppresses this behavior and causes the parameter to be passed on the 3532stack in its natural location. 3533@end defmac 3534 3535@defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size}) 3536A C expression that should indicate the number of bytes of its own 3537arguments that a function pops on returning, or 0 if the 3538function pops no arguments and the caller must therefore pop them all 3539after the function returns. 3540 3541@var{fundecl} is a C variable whose value is a tree node that describes 3542the function in question. Normally it is a node of type 3543@code{FUNCTION_DECL} that describes the declaration of the function. 3544From this you can obtain the @code{DECL_ATTRIBUTES} of the function. 3545 3546@var{funtype} is a C variable whose value is a tree node that 3547describes the function in question. Normally it is a node of type 3548@code{FUNCTION_TYPE} that describes the data type of the function. 3549From this it is possible to obtain the data types of the value and 3550arguments (if known). 3551 3552When a call to a library function is being considered, @var{fundecl} 3553will contain an identifier node for the library function. Thus, if 3554you need to distinguish among various library functions, you can do so 3555by their names. Note that ``library function'' in this context means 3556a function used to perform arithmetic, whose name is known specially 3557in the compiler and was not mentioned in the C code being compiled. 3558 3559@var{stack-size} is the number of bytes of arguments passed on the 3560stack. If a variable number of bytes is passed, it is zero, and 3561argument popping will always be the responsibility of the calling function. 3562 3563On the VAX, all functions always pop their arguments, so the definition 3564of this macro is @var{stack-size}. On the 68000, using the standard 3565calling convention, no functions pop their arguments, so the value of 3566the macro is always 0 in this case. But an alternative calling 3567convention is available in which functions that take a fixed number of 3568arguments pop them but other functions (such as @code{printf}) pop 3569nothing (the caller pops all). When this convention is in use, 3570@var{funtype} is examined to determine whether a function takes a fixed 3571number of arguments. 3572@end defmac 3573 3574@defmac CALL_POPS_ARGS (@var{cum}) 3575A C expression that should indicate the number of bytes a call sequence 3576pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS} 3577when compiling a function call. 3578 3579@var{cum} is the variable in which all arguments to the called function 3580have been accumulated. 3581 3582On certain architectures, such as the SH5, a call trampoline is used 3583that pops certain registers off the stack, depending on the arguments 3584that have been passed to the function. Since this is a property of the 3585call site, not of the called function, @code{RETURN_POPS_ARGS} is not 3586appropriate. 3587@end defmac 3588 3589@node Register Arguments 3590@subsection Passing Arguments in Registers 3591@cindex arguments in registers 3592@cindex registers arguments 3593 3594This section describes the macros which let you control how various 3595types of arguments are passed in registers or how they are arranged in 3596the stack. 3597 3598@defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) 3599A C expression that controls whether a function argument is passed 3600in a register, and which register. 3601 3602The arguments are @var{cum}, which summarizes all the previous 3603arguments; @var{mode}, the machine mode of the argument; @var{type}, 3604the data type of the argument as a tree node or 0 if that is not known 3605(which happens for C support library functions); and @var{named}, 3606which is 1 for an ordinary argument and 0 for nameless arguments that 3607correspond to @samp{@dots{}} in the called function's prototype. 3608@var{type} can be an incomplete type if a syntax error has previously 3609occurred. 3610 3611The value of the expression is usually either a @code{reg} RTX for the 3612hard register in which to pass the argument, or zero to pass the 3613argument on the stack. 3614 3615For machines like the VAX and 68000, where normally all arguments are 3616pushed, zero suffices as a definition. 3617 3618The value of the expression can also be a @code{parallel} RTX@. This is 3619used when an argument is passed in multiple locations. The mode of the 3620@code{parallel} should be the mode of the entire argument. The 3621@code{parallel} holds any number of @code{expr_list} pairs; each one 3622describes where part of the argument is passed. In each 3623@code{expr_list} the first operand must be a @code{reg} RTX for the hard 3624register in which to pass this part of the argument, and the mode of the 3625register RTX indicates how large this part of the argument is. The 3626second operand of the @code{expr_list} is a @code{const_int} which gives 3627the offset in bytes into the entire argument of where this part starts. 3628As a special exception the first @code{expr_list} in the @code{parallel} 3629RTX may have a first operand of zero. This indicates that the entire 3630argument is also stored on the stack. 3631 3632The last time this macro is called, it is called with @code{MODE == 3633VOIDmode}, and its result is passed to the @code{call} or @code{call_value} 3634pattern as operands 2 and 3 respectively. 3635 3636@cindex @file{stdarg.h} and register arguments 3637The usual way to make the ISO library @file{stdarg.h} work on a machine 3638where some arguments are usually passed in registers, is to cause 3639nameless arguments to be passed on the stack instead. This is done 3640by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0. 3641 3642@cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG} 3643@cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG} 3644You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})} 3645in the definition of this macro to determine if this argument is of a 3646type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE} 3647is not defined and @code{FUNCTION_ARG} returns nonzero for such an 3648argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is 3649defined, the argument will be computed in the stack and then loaded into 3650a register. 3651@end defmac 3652 3653@defmac MUST_PASS_IN_STACK (@var{mode}, @var{type}) 3654Define as a C expression that evaluates to nonzero if we do not know how 3655to pass TYPE solely in registers. The file @file{expr.h} defines a 3656definition that is usually appropriate, refer to @file{expr.h} for additional 3657documentation. 3658@end defmac 3659 3660@defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) 3661Define this macro if the target machine has ``register windows'', so 3662that the register in which a function sees an arguments is not 3663necessarily the same as the one in which the caller passed the 3664argument. 3665 3666For such machines, @code{FUNCTION_ARG} computes the register in which 3667the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should 3668be defined in a similar fashion to tell the function being called 3669where the arguments will arrive. 3670 3671If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG} 3672serves both purposes. 3673@end defmac 3674 3675@defmac FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named}) 3676A C expression for the number of words, at the beginning of an 3677argument, that must be put in registers. The value must be zero for 3678arguments that are passed entirely in registers or that are entirely 3679pushed on the stack. 3680 3681On some machines, certain arguments must be passed partially in 3682registers and partially in memory. On these machines, typically the 3683first @var{n} words of arguments are passed in registers, and the rest 3684on the stack. If a multi-word argument (a @code{double} or a 3685structure) crosses that boundary, its first few words must be passed 3686in registers and the rest must be pushed. This macro tells the 3687compiler when this occurs, and how many of the words should go in 3688registers. 3689 3690@code{FUNCTION_ARG} for these arguments should return the first 3691register to be used by the caller for this argument; likewise 3692@code{FUNCTION_INCOMING_ARG}, for the called function. 3693@end defmac 3694 3695@defmac FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named}) 3696A C expression that indicates when an argument must be passed by reference. 3697If nonzero for an argument, a copy of that argument is made in memory and a 3698pointer to the argument is passed instead of the argument itself. 3699The pointer is passed in whatever way is appropriate for passing a pointer 3700to that type. 3701 3702On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable 3703definition of this macro might be 3704@smallexample 3705#define FUNCTION_ARG_PASS_BY_REFERENCE\ 3706(CUM, MODE, TYPE, NAMED) \ 3707 MUST_PASS_IN_STACK (MODE, TYPE) 3708@end smallexample 3709@c this is *still* too long. --mew 5feb93 3710@end defmac 3711 3712@defmac FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named}) 3713If defined, a C expression that indicates when it is the called function's 3714responsibility to make a copy of arguments passed by invisible reference. 3715Normally, the caller makes a copy and passes the address of the copy to the 3716routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is 3717nonzero, the caller does not make a copy. Instead, it passes a pointer to the 3718``live'' value. The called function must not modify this value. If it can be 3719determined that the value won't be modified, it need not make a copy; 3720otherwise a copy must be made. 3721@end defmac 3722 3723@defmac CUMULATIVE_ARGS 3724A C type for declaring a variable that is used as the first argument of 3725@code{FUNCTION_ARG} and other related values. For some target machines, 3726the type @code{int} suffices and can hold the number of bytes of 3727argument so far. 3728 3729There is no need to record in @code{CUMULATIVE_ARGS} anything about the 3730arguments that have been passed on the stack. The compiler has other 3731variables to keep track of that. For target machines on which all 3732arguments are passed on the stack, there is no need to store anything in 3733@code{CUMULATIVE_ARGS}; however, the data structure must exist and 3734should not be empty, so use @code{int}. 3735@end defmac 3736 3737@defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args}) 3738A C statement (sans semicolon) for initializing the variable 3739@var{cum} for the state at the beginning of the argument list. The 3740variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype} 3741is the tree node for the data type of the function which will receive 3742the args, or 0 if the args are to a compiler support library function. 3743For direct calls that are not libcalls, @var{fndecl} contain the 3744declaration node of the function. @var{fndecl} is also set when 3745@code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function 3746being compiled. @var{n_named_args} is set to the number of named 3747arguments, including a structure return address if it is passed as a 3748parameter, when making a call. When processing incoming arguments, 3749@var{n_named_args} is set to -1. 3750 3751When processing a call to a compiler support library function, 3752@var{libname} identifies which one. It is a @code{symbol_ref} rtx which 3753contains the name of the function, as a string. @var{libname} is 0 when 3754an ordinary C function call is being processed. Thus, each time this 3755macro is called, either @var{libname} or @var{fntype} is nonzero, but 3756never both of them at once. 3757@end defmac 3758 3759@defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname}) 3760Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls, 3761it gets a @code{MODE} argument instead of @var{fntype}, that would be 3762@code{NULL}. @var{indirect} would always be zero, too. If this macro 3763is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, 37640)} is used instead. 3765@end defmac 3766 3767@defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) 3768Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of 3769finding the arguments for the function being compiled. If this macro is 3770undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. 3771 3772The value passed for @var{libname} is always 0, since library routines 3773with special calling conventions are never compiled with GCC@. The 3774argument @var{libname} exists for symmetry with 3775@code{INIT_CUMULATIVE_ARGS}. 3776@c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. 3777@c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 3778@end defmac 3779 3780@defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named}) 3781A C statement (sans semicolon) to update the summarizer variable 3782@var{cum} to advance past an argument in the argument list. The 3783values @var{mode}, @var{type} and @var{named} describe that argument. 3784Once this is done, the variable @var{cum} is suitable for analyzing 3785the @emph{following} argument with @code{FUNCTION_ARG}, etc. 3786 3787This macro need not do anything if the argument in question was passed 3788on the stack. The compiler knows how to track the amount of stack space 3789used for arguments without any special help. 3790@end defmac 3791 3792@defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type}) 3793If defined, a C expression which determines whether, and in which direction, 3794to pad out an argument with extra space. The value should be of type 3795@code{enum direction}: either @code{upward} to pad above the argument, 3796@code{downward} to pad below, or @code{none} to inhibit padding. 3797 3798The @emph{amount} of padding is always just enough to reach the next 3799multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control 3800it. 3801 3802This macro has a default definition which is right for most systems. 3803For little-endian machines, the default is to pad upward. For 3804big-endian machines, the default is to pad downward for an argument of 3805constant size shorter than an @code{int}, and upward otherwise. 3806@end defmac 3807 3808@defmac PAD_VARARGS_DOWN 3809If defined, a C expression which determines whether the default 3810implementation of va_arg will attempt to pad down before reading the 3811next argument, if that argument is smaller than its aligned space as 3812controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such 3813arguments are padded down if @code{BYTES_BIG_ENDIAN} is true. 3814@end defmac 3815 3816@defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first}) 3817Specify padding for the last element of a block move between registers and 3818memory. @var{first} is nonzero if this is the only element. Defining this 3819macro allows better control of register function parameters on big-endian 3820machines, without using @code{PARALLEL} rtl. In particular, 3821@code{MUST_PASS_IN_STACK} need not test padding and mode of types in 3822registers, as there is no longer a "wrong" part of a register; For example, 3823a three byte aggregate may be passed in the high part of a register if so 3824required. 3825@end defmac 3826 3827@defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type}) 3828If defined, a C expression that gives the alignment boundary, in bits, 3829of an argument with the specified mode and type. If it is not defined, 3830@code{PARM_BOUNDARY} is used for all arguments. 3831@end defmac 3832 3833@defmac FUNCTION_ARG_REGNO_P (@var{regno}) 3834A C expression that is nonzero if @var{regno} is the number of a hard 3835register in which function arguments are sometimes passed. This does 3836@emph{not} include implicit arguments such as the static chain and 3837the structure-value address. On many machines, no registers can be 3838used for this purpose since all function arguments are pushed on the 3839stack. 3840@end defmac 3841 3842@deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type}) 3843This hook should return true if parameter of type @var{type} are passed 3844as two scalar parameters. By default, GCC will attempt to pack complex 3845arguments into the target's word size. Some ABIs require complex arguments 3846to be split and treated as their individual components. For example, on 3847AIX64, complex floats should be passed in a pair of floating point 3848registers, even though a complex float would fit in one 64-bit floating 3849point register. 3850 3851The default value of this hook is @code{NULL}, which is treated as always 3852false. 3853@end deftypefn 3854 3855@node Scalar Return 3856@subsection How Scalar Function Values Are Returned 3857@cindex return values in registers 3858@cindex values, returned by functions 3859@cindex scalars, returned as values 3860 3861This section discusses the macros that control returning scalars as 3862values---values that can fit in registers. 3863 3864@defmac FUNCTION_VALUE (@var{valtype}, @var{func}) 3865A C expression to create an RTX representing the place where a 3866function returns a value of data type @var{valtype}. @var{valtype} is 3867a tree node representing a data type. Write @code{TYPE_MODE 3868(@var{valtype})} to get the machine mode used to represent that type. 3869On many machines, only the mode is relevant. (Actually, on most 3870machines, scalar values are returned in the same place regardless of 3871mode). 3872 3873The value of the expression is usually a @code{reg} RTX for the hard 3874register where the return value is stored. The value can also be a 3875@code{parallel} RTX, if the return value is in multiple places. See 3876@code{FUNCTION_ARG} for an explanation of the @code{parallel} form. 3877 3878If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply the same 3879promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a 3880scalar type. 3881 3882If the precise function being called is known, @var{func} is a tree 3883node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null 3884pointer. This makes it possible to use a different value-returning 3885convention for specific functions when all their calls are 3886known. 3887 3888@code{FUNCTION_VALUE} is not used for return vales with aggregate data 3889types, because these are returned in another way. See 3890@code{TARGET_STRUCT_VALUE_RTX} and related macros, below. 3891@end defmac 3892 3893@defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func}) 3894Define this macro if the target machine has ``register windows'' 3895so that the register in which a function returns its value is not 3896the same as the one in which the caller sees the value. 3897 3898For such machines, @code{FUNCTION_VALUE} computes the register in which 3899the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be 3900defined in a similar fashion to tell the function where to put the 3901value. 3902 3903If @code{FUNCTION_OUTGOING_VALUE} is not defined, 3904@code{FUNCTION_VALUE} serves both purposes. 3905 3906@code{FUNCTION_OUTGOING_VALUE} is not used for return vales with 3907aggregate data types, because these are returned in another way. See 3908@code{TARGET_STRUCT_VALUE_RTX} and related macros, below. 3909@end defmac 3910 3911@defmac LIBCALL_VALUE (@var{mode}) 3912A C expression to create an RTX representing the place where a library 3913function returns a value of mode @var{mode}. If the precise function 3914being called is known, @var{func} is a tree node 3915(@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null 3916pointer. This makes it possible to use a different value-returning 3917convention for specific functions when all their calls are 3918known. 3919 3920Note that ``library function'' in this context means a compiler 3921support routine, used to perform arithmetic, whose name is known 3922specially by the compiler and was not mentioned in the C code being 3923compiled. 3924 3925The definition of @code{LIBRARY_VALUE} need not be concerned aggregate 3926data types, because none of the library functions returns such types. 3927@end defmac 3928 3929@defmac FUNCTION_VALUE_REGNO_P (@var{regno}) 3930A C expression that is nonzero if @var{regno} is the number of a hard 3931register in which the values of called function may come back. 3932 3933A register whose use for returning values is limited to serving as the 3934second of a pair (for a value of type @code{double}, say) need not be 3935recognized by this macro. So for most machines, this definition 3936suffices: 3937 3938@smallexample 3939#define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) 3940@end smallexample 3941 3942If the machine has register windows, so that the caller and the called 3943function use different registers for the return value, this macro 3944should recognize only the caller's register numbers. 3945@end defmac 3946 3947@defmac APPLY_RESULT_SIZE 3948Define this macro if @samp{untyped_call} and @samp{untyped_return} 3949need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for 3950saving and restoring an arbitrary return value. 3951@end defmac 3952 3953@deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type}) 3954This hook should return true if values of type @var{type} are returned 3955at the most significant end of a register (in other words, if they are 3956padded at the least significant end). You can assume that @var{type} 3957is returned in a register; the caller is required to check this. 3958 3959Note that the register provided by @code{FUNCTION_VALUE} must be able 3960to hold the complete return value. For example, if a 1-, 2- or 3-byte 3961structure is returned at the most significant end of a 4-byte register, 3962@code{FUNCTION_VALUE} should provide an @code{SImode} rtx. 3963@end deftypefn 3964 3965@node Aggregate Return 3966@subsection How Large Values Are Returned 3967@cindex aggregates as return values 3968@cindex large return values 3969@cindex returning aggregate values 3970@cindex structure value address 3971 3972When a function value's mode is @code{BLKmode} (and in some other 3973cases), the value is not returned according to @code{FUNCTION_VALUE} 3974(@pxref{Scalar Return}). Instead, the caller passes the address of a 3975block of memory in which the value should be stored. This address 3976is called the @dfn{structure value address}. 3977 3978This section describes how to control returning structure values in 3979memory. 3980 3981@deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype}) 3982This target hook should return a nonzero value to say to return the 3983function value in memory, just as large structures are always returned. 3984Here @var{type} will be the data type of the value, and @var{fntype} 3985will be the type of the function doing the returning, or @code{NULL} for 3986libcalls. 3987 3988Note that values of mode @code{BLKmode} must be explicitly handled 3989by this function. Also, the option @option{-fpcc-struct-return} 3990takes effect regardless of this macro. On most systems, it is 3991possible to leave the hook undefined; this causes a default 3992definition to be used, whose value is the constant 1 for @code{BLKmode} 3993values, and 0 otherwise. 3994 3995Do not use this hook to indicate that structures and unions should always 3996be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN} 3997to indicate this. 3998@end deftypefn 3999 4000@defmac DEFAULT_PCC_STRUCT_RETURN 4001Define this macro to be 1 if all structure and union return values must be 4002in memory. Since this results in slower code, this should be defined 4003only if needed for compatibility with other compilers or with an ABI@. 4004If you define this macro to be 0, then the conventions used for structure 4005and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY} 4006target hook. 4007 4008If not defined, this defaults to the value 1. 4009@end defmac 4010 4011@deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming}) 4012This target hook should return the location of the structure value 4013address (normally a @code{mem} or @code{reg}), or 0 if the address is 4014passed as an ``invisible'' first argument. Note that @var{fndecl} may 4015be @code{NULL}, for libcalls. 4016 4017On some architectures the place where the structure value address 4018is found by the called function is not the same place that the 4019caller put it. This can be due to register windows, or it could 4020be because the function prologue moves it to a different place. 4021@var{incoming} is @code{true} when the location is needed in 4022the context of the called function, and @code{false} in the context of 4023the caller. 4024 4025If @var{incoming} is @code{true} and the address is to be found on the 4026stack, return a @code{mem} which refers to the frame pointer. 4027@end deftypefn 4028 4029@defmac PCC_STATIC_STRUCT_RETURN 4030Define this macro if the usual system convention on the target machine 4031for returning structures and unions is for the called function to return 4032the address of a static variable containing the value. 4033 4034Do not define this if the usual system convention is for the caller to 4035pass an address to the subroutine. 4036 4037This macro has effect in @option{-fpcc-struct-return} mode, but it does 4038nothing when you use @option{-freg-struct-return} mode. 4039@end defmac 4040 4041@node Caller Saves 4042@subsection Caller-Saves Register Allocation 4043 4044If you enable it, GCC can save registers around function calls. This 4045makes it possible to use call-clobbered registers to hold variables that 4046must live across calls. 4047 4048@defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls}) 4049A C expression to determine whether it is worthwhile to consider placing 4050a pseudo-register in a call-clobbered hard register and saving and 4051restoring it around each function call. The expression should be 1 when 4052this is worth doing, and 0 otherwise. 4053 4054If you don't define this macro, a default is used which is good on most 4055machines: @code{4 * @var{calls} < @var{refs}}. 4056@end defmac 4057 4058@defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs}) 4059A C expression specifying which mode is required for saving @var{nregs} 4060of a pseudo-register in call-clobbered hard register @var{regno}. If 4061@var{regno} is unsuitable for caller save, @code{VOIDmode} should be 4062returned. For most machines this macro need not be defined since GCC 4063will select the smallest suitable mode. 4064@end defmac 4065 4066@node Function Entry 4067@subsection Function Entry and Exit 4068@cindex function entry and exit 4069@cindex prologue 4070@cindex epilogue 4071 4072This section describes the macros that output function entry 4073(@dfn{prologue}) and exit (@dfn{epilogue}) code. 4074 4075@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4076If defined, a function that outputs the assembler code for entry to a 4077function. The prologue is responsible for setting up the stack frame, 4078initializing the frame pointer register, saving registers that must be 4079saved, and allocating @var{size} additional bytes of storage for the 4080local variables. @var{size} is an integer. @var{file} is a stdio 4081stream to which the assembler code should be output. 4082 4083The label for the beginning of the function need not be output by this 4084macro. That has already been done when the macro is run. 4085 4086@findex regs_ever_live 4087To determine which registers to save, the macro can refer to the array 4088@code{regs_ever_live}: element @var{r} is nonzero if hard register 4089@var{r} is used anywhere within the function. This implies the function 4090prologue should save register @var{r}, provided it is not one of the 4091call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use 4092@code{regs_ever_live}.) 4093 4094On machines that have ``register windows'', the function entry code does 4095not save on the stack the registers that are in the windows, even if 4096they are supposed to be preserved by function calls; instead it takes 4097appropriate steps to ``push'' the register stack, if any non-call-used 4098registers are used in the function. 4099 4100@findex frame_pointer_needed 4101On machines where functions may or may not have frame-pointers, the 4102function entry code must vary accordingly; it must set up the frame 4103pointer if one is wanted, and not otherwise. To determine whether a 4104frame pointer is in wanted, the macro can refer to the variable 4105@code{frame_pointer_needed}. The variable's value will be 1 at run 4106time in a function that needs a frame pointer. @xref{Elimination}. 4107 4108The function entry code is responsible for allocating any stack space 4109required for the function. This stack space consists of the regions 4110listed below. In most cases, these regions are allocated in the 4111order listed, with the last listed region closest to the top of the 4112stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and 4113the highest address if it is not defined). You can use a different order 4114for a machine if doing so is more convenient or required for 4115compatibility reasons. Except in cases where required by standard 4116or by a debugger, there is no reason why the stack layout used by GCC 4117need agree with that used by other compilers for a machine. 4118@end deftypefn 4119 4120@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file}) 4121If defined, a function that outputs assembler code at the end of a 4122prologue. This should be used when the function prologue is being 4123emitted as RTL, and you have some extra assembler that needs to be 4124emitted. @xref{prologue instruction pattern}. 4125@end deftypefn 4126 4127@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file}) 4128If defined, a function that outputs assembler code at the start of an 4129epilogue. This should be used when the function epilogue is being 4130emitted as RTL, and you have some extra assembler that needs to be 4131emitted. @xref{epilogue instruction pattern}. 4132@end deftypefn 4133 4134@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4135If defined, a function that outputs the assembler code for exit from a 4136function. The epilogue is responsible for restoring the saved 4137registers and stack pointer to their values when the function was 4138called, and returning control to the caller. This macro takes the 4139same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the 4140registers to restore are determined from @code{regs_ever_live} and 4141@code{CALL_USED_REGISTERS} in the same way. 4142 4143On some machines, there is a single instruction that does all the work 4144of returning from the function. On these machines, give that 4145instruction the name @samp{return} and do not define the macro 4146@code{TARGET_ASM_FUNCTION_EPILOGUE} at all. 4147 4148Do not define a pattern named @samp{return} if you want the 4149@code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target 4150switches to control whether return instructions or epilogues are used, 4151define a @samp{return} pattern with a validity condition that tests the 4152target switches appropriately. If the @samp{return} pattern's validity 4153condition is false, epilogues will be used. 4154 4155On machines where functions may or may not have frame-pointers, the 4156function exit code must vary accordingly. Sometimes the code for these 4157two cases is completely different. To determine whether a frame pointer 4158is wanted, the macro can refer to the variable 4159@code{frame_pointer_needed}. The variable's value will be 1 when compiling 4160a function that needs a frame pointer. 4161 4162Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and 4163@code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially. 4164The C variable @code{current_function_is_leaf} is nonzero for such a 4165function. @xref{Leaf Functions}. 4166 4167On some machines, some functions pop their arguments on exit while 4168others leave that for the caller to do. For example, the 68020 when 4169given @option{-mrtd} pops arguments in functions that take a fixed 4170number of arguments. 4171 4172@findex current_function_pops_args 4173Your definition of the macro @code{RETURN_POPS_ARGS} decides which 4174functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE} 4175needs to know what was decided. The variable that is called 4176@code{current_function_pops_args} is the number of bytes of its 4177arguments that a function should pop. @xref{Scalar Return}. 4178@c what is the "its arguments" in the above sentence referring to, pray 4179@c tell? --mew 5feb93 4180@end deftypefn 4181 4182@itemize @bullet 4183@item 4184@findex current_function_pretend_args_size 4185A region of @code{current_function_pretend_args_size} bytes of 4186uninitialized space just underneath the first argument arriving on the 4187stack. (This may not be at the very start of the allocated stack region 4188if the calling sequence has pushed anything else since pushing the stack 4189arguments. But usually, on such machines, nothing else has been pushed 4190yet, because the function prologue itself does all the pushing.) This 4191region is used on machines where an argument may be passed partly in 4192registers and partly in memory, and, in some cases to support the 4193features in @code{<stdarg.h>}. 4194 4195@item 4196An area of memory used to save certain registers used by the function. 4197The size of this area, which may also include space for such things as 4198the return address and pointers to previous stack frames, is 4199machine-specific and usually depends on which registers have been used 4200in the function. Machines with register windows often do not require 4201a save area. 4202 4203@item 4204A region of at least @var{size} bytes, possibly rounded up to an allocation 4205boundary, to contain the local variables of the function. On some machines, 4206this region and the save area may occur in the opposite order, with the 4207save area closer to the top of the stack. 4208 4209@item 4210@cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames 4211Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of 4212@code{current_function_outgoing_args_size} bytes to be used for outgoing 4213argument lists of the function. @xref{Stack Arguments}. 4214@end itemize 4215 4216Normally, it is necessary for the macros 4217@code{TARGET_ASM_FUNCTION_PROLOGUE} and 4218@code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially. 4219The C variable @code{current_function_is_leaf} is nonzero for such a 4220function. 4221 4222@defmac EXIT_IGNORE_STACK 4223Define this macro as a C expression that is nonzero if the return 4224instruction or the function epilogue ignores the value of the stack 4225pointer; in other words, if it is safe to delete an instruction to 4226adjust the stack pointer before a return from the function. The 4227default is 0. 4228 4229Note that this macro's value is relevant only for functions for which 4230frame pointers are maintained. It is never safe to delete a final 4231stack adjustment in a function that has no frame pointer, and the 4232compiler knows this regardless of @code{EXIT_IGNORE_STACK}. 4233@end defmac 4234 4235@defmac EPILOGUE_USES (@var{regno}) 4236Define this macro as a C expression that is nonzero for registers that are 4237used by the epilogue or the @samp{return} pattern. The stack and frame 4238pointer registers are already be assumed to be used as needed. 4239@end defmac 4240 4241@defmac EH_USES (@var{regno}) 4242Define this macro as a C expression that is nonzero for registers that are 4243used by the exception handling mechanism, and so should be considered live 4244on entry to an exception edge. 4245@end defmac 4246 4247@defmac DELAY_SLOTS_FOR_EPILOGUE 4248Define this macro if the function epilogue contains delay slots to which 4249instructions from the rest of the function can be ``moved''. The 4250definition should be a C expression whose value is an integer 4251representing the number of delay slots there. 4252@end defmac 4253 4254@defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n}) 4255A C expression that returns 1 if @var{insn} can be placed in delay 4256slot number @var{n} of the epilogue. 4257 4258The argument @var{n} is an integer which identifies the delay slot now 4259being considered (since different slots may have different rules of 4260eligibility). It is never negative and is always less than the number 4261of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns). 4262If you reject a particular insn for a given delay slot, in principle, it 4263may be reconsidered for a subsequent delay slot. Also, other insns may 4264(at least in principle) be considered for the so far unfilled delay 4265slot. 4266 4267@findex current_function_epilogue_delay_list 4268@findex final_scan_insn 4269The insns accepted to fill the epilogue delay slots are put in an RTL 4270list made with @code{insn_list} objects, stored in the variable 4271@code{current_function_epilogue_delay_list}. The insn for the first 4272delay slot comes first in the list. Your definition of the macro 4273@code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by 4274outputting the insns in this list, usually by calling 4275@code{final_scan_insn}. 4276 4277You need not define this macro if you did not define 4278@code{DELAY_SLOTS_FOR_EPILOGUE}. 4279@end defmac 4280 4281@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, tree @var{function}) 4282A function that outputs the assembler code for a thunk 4283function, used to implement C++ virtual function calls with multiple 4284inheritance. The thunk acts as a wrapper around a virtual function, 4285adjusting the implicit object parameter before handing control off to 4286the real function. 4287 4288First, emit code to add the integer @var{delta} to the location that 4289contains the incoming first argument. Assume that this argument 4290contains a pointer, and is the one used to pass the @code{this} pointer 4291in C++. This is the incoming argument @emph{before} the function prologue, 4292e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of 4293all other incoming arguments. 4294 4295After the addition, emit code to jump to @var{function}, which is a 4296@code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does 4297not touch the return address. Hence returning from @var{FUNCTION} will 4298return to whoever called the current @samp{thunk}. 4299 4300The effect must be as if @var{function} had been called directly with 4301the adjusted first argument. This macro is responsible for emitting all 4302of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE} 4303and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked. 4304 4305The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function} 4306have already been extracted from it.) It might possibly be useful on 4307some targets, but probably not. 4308 4309If you do not define this macro, the target-independent code in the C++ 4310front end will generate a less efficient heavyweight thunk that calls 4311@var{function} instead of jumping to it. The generic approach does 4312not support varargs. 4313@end deftypefn 4314 4315@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_VCALL_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, int @var{vcall_offset}, tree @var{function}) 4316A function like @code{TARGET_ASM_OUTPUT_MI_THUNK}, except that if 4317@var{vcall_offset} is nonzero, an additional adjustment should be made 4318after adding @code{delta}. In particular, if @var{p} is the 4319adjusted pointer, the following adjustment should be made: 4320 4321@smallexample 4322p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)] 4323@end smallexample 4324 4325@noindent 4326If this function is defined, it will always be used in place of 4327@code{TARGET_ASM_OUTPUT_MI_THUNK}. 4328@end deftypefn 4329 4330@node Profiling 4331@subsection Generating Code for Profiling 4332@cindex profiling, code generation 4333 4334These macros will help you generate code for profiling. 4335 4336@defmac FUNCTION_PROFILER (@var{file}, @var{labelno}) 4337A C statement or compound statement to output to @var{file} some 4338assembler code to call the profiling subroutine @code{mcount}. 4339 4340@findex mcount 4341The details of how @code{mcount} expects to be called are determined by 4342your operating system environment, not by GCC@. To figure them out, 4343compile a small program for profiling using the system's installed C 4344compiler and look at the assembler code that results. 4345 4346Older implementations of @code{mcount} expect the address of a counter 4347variable to be loaded into some register. The name of this variable is 4348@samp{LP} followed by the number @var{labelno}, so you would generate 4349the name using @samp{LP%d} in a @code{fprintf}. 4350@end defmac 4351 4352@defmac PROFILE_HOOK 4353A C statement or compound statement to output to @var{file} some assembly 4354code to call the profiling subroutine @code{mcount} even the target does 4355not support profiling. 4356@end defmac 4357 4358@defmac NO_PROFILE_COUNTERS 4359Define this macro if the @code{mcount} subroutine on your system does 4360not need a counter variable allocated for each function. This is true 4361for almost all modern implementations. If you define this macro, you 4362must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}. 4363@end defmac 4364 4365@defmac PROFILE_BEFORE_PROLOGUE 4366Define this macro if the code for function profiling should come before 4367the function prologue. Normally, the profiling code comes after. 4368@end defmac 4369 4370@node Tail Calls 4371@subsection Permitting tail calls 4372@cindex tail calls 4373 4374@deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp}) 4375True if it is ok to do sibling call optimization for the specified 4376call expression @var{exp}. @var{decl} will be the called function, 4377or @code{NULL} if this is an indirect call. 4378 4379It is not uncommon for limitations of calling conventions to prevent 4380tail calls to functions outside the current unit of translation, or 4381during PIC compilation. The hook is used to enforce these restrictions, 4382as the @code{sibcall} md pattern can not fail, or fall over to a 4383``normal'' call. The criteria for successful sibling call optimization 4384may vary greatly between different architectures. 4385@end deftypefn 4386 4387@node Varargs 4388@section Implementing the Varargs Macros 4389@cindex varargs implementation 4390 4391GCC comes with an implementation of @code{<varargs.h>} and 4392@code{<stdarg.h>} that work without change on machines that pass arguments 4393on the stack. Other machines require their own implementations of 4394varargs, and the two machine independent header files must have 4395conditionals to include it. 4396 4397ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in 4398the calling convention for @code{va_start}. The traditional 4399implementation takes just one argument, which is the variable in which 4400to store the argument pointer. The ISO implementation of 4401@code{va_start} takes an additional second argument. The user is 4402supposed to write the last named argument of the function here. 4403 4404However, @code{va_start} should not use this argument. The way to find 4405the end of the named arguments is with the built-in functions described 4406below. 4407 4408@defmac __builtin_saveregs () 4409Use this built-in function to save the argument registers in memory so 4410that the varargs mechanism can access them. Both ISO and traditional 4411versions of @code{va_start} must use @code{__builtin_saveregs}, unless 4412you use @code{SETUP_INCOMING_VARARGS} (see below) instead. 4413 4414On some machines, @code{__builtin_saveregs} is open-coded under the 4415control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines, 4416it calls a routine written in assembler language, found in 4417@file{libgcc2.c}. 4418 4419Code generated for the call to @code{__builtin_saveregs} appears at the 4420beginning of the function, as opposed to where the call to 4421@code{__builtin_saveregs} is written, regardless of what the code is. 4422This is because the registers must be saved before the function starts 4423to use them for its own purposes. 4424@c i rewrote the first sentence above to fix an overfull hbox. --mew 4425@c 10feb93 4426@end defmac 4427 4428@defmac __builtin_args_info (@var{category}) 4429Use this built-in function to find the first anonymous arguments in 4430registers. 4431 4432In general, a machine may have several categories of registers used for 4433arguments, each for a particular category of data types. (For example, 4434on some machines, floating-point registers are used for floating-point 4435arguments while other arguments are passed in the general registers.) 4436To make non-varargs functions use the proper calling convention, you 4437have defined the @code{CUMULATIVE_ARGS} data type to record how many 4438registers in each category have been used so far 4439 4440@code{__builtin_args_info} accesses the same data structure of type 4441@code{CUMULATIVE_ARGS} after the ordinary argument layout is finished 4442with it, with @var{category} specifying which word to access. Thus, the 4443value indicates the first unused register in a given category. 4444 4445Normally, you would use @code{__builtin_args_info} in the implementation 4446of @code{va_start}, accessing each category just once and storing the 4447value in the @code{va_list} object. This is because @code{va_list} will 4448have to update the values, and there is no way to alter the 4449values accessed by @code{__builtin_args_info}. 4450@end defmac 4451 4452@defmac __builtin_next_arg (@var{lastarg}) 4453This is the equivalent of @code{__builtin_args_info}, for stack 4454arguments. It returns the address of the first anonymous stack 4455argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it 4456returns the address of the location above the first anonymous stack 4457argument. Use it in @code{va_start} to initialize the pointer for 4458fetching arguments from the stack. Also use it in @code{va_start} to 4459verify that the second parameter @var{lastarg} is the last named argument 4460of the current function. 4461@end defmac 4462 4463@defmac __builtin_classify_type (@var{object}) 4464Since each machine has its own conventions for which data types are 4465passed in which kind of register, your implementation of @code{va_arg} 4466has to embody these conventions. The easiest way to categorize the 4467specified data type is to use @code{__builtin_classify_type} together 4468with @code{sizeof} and @code{__alignof__}. 4469 4470@code{__builtin_classify_type} ignores the value of @var{object}, 4471considering only its data type. It returns an integer describing what 4472kind of type that is---integer, floating, pointer, structure, and so on. 4473 4474The file @file{typeclass.h} defines an enumeration that you can use to 4475interpret the values of @code{__builtin_classify_type}. 4476@end defmac 4477 4478These machine description macros help implement varargs: 4479 4480@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void) 4481If defined, this hook produces the machine-specific code for a call to 4482@code{__builtin_saveregs}. This code will be moved to the very 4483beginning of the function, before any parameter access are made. The 4484return value of this function should be an RTX that contains the value 4485to use as the return of @code{__builtin_saveregs}. 4486@end deftypefn 4487 4488@deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS *@var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time}) 4489This target hook offers an alternative to using 4490@code{__builtin_saveregs} and defining the hook 4491@code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous 4492register arguments into the stack so that all the arguments appear to 4493have been passed consecutively on the stack. Once this is done, you can 4494use the standard implementation of varargs that works for machines that 4495pass all their arguments on the stack. 4496 4497The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data 4498structure, containing the values that are obtained after processing the 4499named arguments. The arguments @var{mode} and @var{type} describe the 4500last named argument---its machine mode and its data type as a tree node. 4501 4502The target hook should do two things: first, push onto the stack all the 4503argument registers @emph{not} used for the named arguments, and second, 4504store the size of the data thus pushed into the @code{int}-valued 4505variable pointed to by @var{pretend_args_size}. The value that you 4506store here will serve as additional offset for setting up the stack 4507frame. 4508 4509Because you must generate code to push the anonymous arguments at 4510compile time without knowing their data types, 4511@code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that 4512have just a single category of argument register and use it uniformly 4513for all data types. 4514 4515If the argument @var{second_time} is nonzero, it means that the 4516arguments of the function are being analyzed for the second time. This 4517happens for an inline function, which is not actually compiled until the 4518end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should 4519not generate any instructions in this case. 4520@end deftypefn 4521 4522@deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca}) 4523Define this hook to return @code{true} if the location where a function 4524argument is passed depends on whether or not it is a named argument. 4525 4526This hook controls how the @var{named} argument to @code{FUNCTION_ARG} 4527is set for varargs and stdarg functions. If this hook returns 4528@code{true}, the @var{named} argument is always true for named 4529arguments, and false for unnamed arguments. If it returns @code{false}, 4530but @code{TARGET_PRETEND_OUTOGOING_VARARGS_NAMED} returns @code{true}, 4531then all arguments are treated as named. Otherwise, all named arguments 4532except the last are treated as named. 4533 4534You need not define this hook if it always returns zero. 4535@end deftypefn 4536 4537@deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED 4538If you need to conditionally change ABIs so that one works with 4539@code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither 4540@code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was 4541defined, then define this hook to return @code{true} if 4542@code{SETUP_INCOMING_VARARGS} is used, @code{false} otherwise. 4543Otherwise, you should not define this hook. 4544@end deftypefn 4545 4546@node Trampolines 4547@section Trampolines for Nested Functions 4548@cindex trampolines for nested functions 4549@cindex nested functions, trampolines for 4550 4551A @dfn{trampoline} is a small piece of code that is created at run time 4552when the address of a nested function is taken. It normally resides on 4553the stack, in the stack frame of the containing function. These macros 4554tell GCC how to generate code to allocate and initialize a 4555trampoline. 4556 4557The instructions in the trampoline must do two things: load a constant 4558address into the static chain register, and jump to the real address of 4559the nested function. On CISC machines such as the m68k, this requires 4560two instructions, a move immediate and a jump. Then the two addresses 4561exist in the trampoline as word-long immediate operands. On RISC 4562machines, it is often necessary to load each address into a register in 4563two parts. Then pieces of each address form separate immediate 4564operands. 4565 4566The code generated to initialize the trampoline must store the variable 4567parts---the static chain value and the function address---into the 4568immediate operands of the instructions. On a CISC machine, this is 4569simply a matter of copying each address to a memory reference at the 4570proper offset from the start of the trampoline. On a RISC machine, it 4571may be necessary to take out pieces of the address and store them 4572separately. 4573 4574@defmac TRAMPOLINE_TEMPLATE (@var{file}) 4575A C statement to output, on the stream @var{file}, assembler code for a 4576block of data that contains the constant parts of a trampoline. This 4577code should not include a label---the label is taken care of 4578automatically. 4579 4580If you do not define this macro, it means no template is needed 4581for the target. Do not define this macro on systems where the block move 4582code to copy the trampoline into place would be larger than the code 4583to generate it on the spot. 4584@end defmac 4585 4586@defmac TRAMPOLINE_SECTION 4587The name of a subroutine to switch to the section in which the 4588trampoline template is to be placed (@pxref{Sections}). The default is 4589a value of @samp{readonly_data_section}, which places the trampoline in 4590the section containing read-only data. 4591@end defmac 4592 4593@defmac TRAMPOLINE_SIZE 4594A C expression for the size in bytes of the trampoline, as an integer. 4595@end defmac 4596 4597@defmac TRAMPOLINE_ALIGNMENT 4598Alignment required for trampolines, in bits. 4599 4600If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT} 4601is used for aligning trampolines. 4602@end defmac 4603 4604@defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain}) 4605A C statement to initialize the variable parts of a trampoline. 4606@var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is 4607an RTX for the address of the nested function; @var{static_chain} is an 4608RTX for the static chain value that should be passed to the function 4609when it is called. 4610@end defmac 4611 4612@defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr}) 4613A C statement that should perform any machine-specific adjustment in 4614the address of the trampoline. Its argument contains the address that 4615was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be 4616used for a function call should be different from the address in which 4617the template was stored, the different address should be assigned to 4618@var{addr}. If this macro is not defined, @var{addr} will be used for 4619function calls. 4620 4621@cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines 4622@cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines 4623If this macro is not defined, by default the trampoline is allocated as 4624a stack slot. This default is right for most machines. The exceptions 4625are machines where it is impossible to execute instructions in the stack 4626area. On such machines, you may have to implement a separate stack, 4627using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE} 4628and @code{TARGET_ASM_FUNCTION_EPILOGUE}. 4629 4630@var{fp} points to a data structure, a @code{struct function}, which 4631describes the compilation status of the immediate containing function of 4632the function which the trampoline is for. The stack slot for the 4633trampoline is in the stack frame of this containing function. Other 4634allocation strategies probably must do something analogous with this 4635information. 4636@end defmac 4637 4638Implementing trampolines is difficult on many machines because they have 4639separate instruction and data caches. Writing into a stack location 4640fails to clear the memory in the instruction cache, so when the program 4641jumps to that location, it executes the old contents. 4642 4643Here are two possible solutions. One is to clear the relevant parts of 4644the instruction cache whenever a trampoline is set up. The other is to 4645make all trampolines identical, by having them jump to a standard 4646subroutine. The former technique makes trampoline execution faster; the 4647latter makes initialization faster. 4648 4649To clear the instruction cache when a trampoline is initialized, define 4650the following macro. 4651 4652@defmac CLEAR_INSN_CACHE (@var{beg}, @var{end}) 4653If defined, expands to a C expression clearing the @emph{instruction 4654cache} in the specified interval. The definition of this macro would 4655typically be a series of @code{asm} statements. Both @var{beg} and 4656@var{end} are both pointer expressions. 4657@end defmac 4658 4659The operating system may also require the stack to be made executable 4660before calling the trampoline. To implement this requirement, define 4661the following macro. 4662 4663@defmac ENABLE_EXECUTE_STACK 4664Define this macro if certain operations must be performed before executing 4665code located on the stack. The macro should expand to a series of C 4666file-scope constructs (e.g. functions) and provide a unique entry point 4667named @code{__enable_execute_stack}. The target is responsible for 4668emitting calls to the entry point in the code, for example from the 4669@code{INITIALIZE_TRAMPOLINE} macro. 4670@end defmac 4671 4672To use a standard subroutine, define the following macro. In addition, 4673you must make sure that the instructions in a trampoline fill an entire 4674cache line with identical instructions, or else ensure that the 4675beginning of the trampoline code is always aligned at the same point in 4676its cache line. Look in @file{m68k.h} as a guide. 4677 4678@defmac TRANSFER_FROM_TRAMPOLINE 4679Define this macro if trampolines need a special subroutine to do their 4680work. The macro should expand to a series of @code{asm} statements 4681which will be compiled with GCC@. They go in a library function named 4682@code{__transfer_from_trampoline}. 4683 4684If you need to avoid executing the ordinary prologue code of a compiled 4685C function when you jump to the subroutine, you can do so by placing a 4686special label of your own in the assembler code. Use one @code{asm} 4687statement to generate an assembler label, and another to make the label 4688global. Then trampolines can use that label to jump directly to your 4689special assembler code. 4690@end defmac 4691 4692@node Library Calls 4693@section Implicit Calls to Library Routines 4694@cindex library subroutine names 4695@cindex @file{libgcc.a} 4696 4697@c prevent bad page break with this line 4698Here is an explanation of implicit calls to library routines. 4699 4700@defmac DECLARE_LIBRARY_RENAMES 4701This macro, if defined, should expand to a piece of C code that will get 4702expanded when compiling functions for libgcc.a. It can be used to 4703provide alternate names for GCC's internal library functions if there 4704are ABI-mandated names that the compiler should provide. 4705@end defmac 4706 4707@findex init_one_libfunc 4708@findex set_optab_libfunc 4709@deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void) 4710This hook should declare additional library routines or rename 4711existing ones, using the functions @code{set_optab_libfunc} and 4712@code{init_one_libfunc} defined in @file{optabs.c}. 4713@code{init_optabs} calls this macro after initializing all the normal 4714library routines. 4715 4716The default is to do nothing. Most ports don't need to define this hook. 4717@end deftypefn 4718 4719@defmac TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison}) 4720This macro should return @code{true} if the library routine that 4721implements the floating point comparison operator @var{comparison} in 4722mode @var{mode} will return a boolean, and @var{false} if it will 4723return a tristate. 4724 4725GCC's own floating point libraries return tristates from the 4726comparison operators, so the default returns false always. Most ports 4727don't need to define this macro. 4728@end defmac 4729 4730@cindex US Software GOFAST, floating point emulation library 4731@cindex floating point emulation library, US Software GOFAST 4732@cindex GOFAST, floating point emulation library 4733@findex gofast_maybe_init_libfuncs 4734@defmac US_SOFTWARE_GOFAST 4735Define this macro if your system C library uses the US Software GOFAST 4736library to provide floating point emulation. 4737 4738In addition to defining this macro, your architecture must set 4739@code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or 4740else call that function from its version of that hook. It is defined 4741in @file{config/gofast.h}, which must be included by your 4742architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for 4743an example. 4744 4745If this macro is defined, the 4746@code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return 4747false for @code{SFmode} and @code{DFmode} comparisons. 4748@end defmac 4749 4750@cindex @code{EDOM}, implicit usage 4751@findex matherr 4752@defmac TARGET_EDOM 4753The value of @code{EDOM} on the target machine, as a C integer constant 4754expression. If you don't define this macro, GCC does not attempt to 4755deposit the value of @code{EDOM} into @code{errno} directly. Look in 4756@file{/usr/include/errno.h} to find the value of @code{EDOM} on your 4757system. 4758 4759If you do not define @code{TARGET_EDOM}, then compiled code reports 4760domain errors by calling the library function and letting it report the 4761error. If mathematical functions on your system use @code{matherr} when 4762there is an error, then you should leave @code{TARGET_EDOM} undefined so 4763that @code{matherr} is used normally. 4764@end defmac 4765 4766@cindex @code{errno}, implicit usage 4767@defmac GEN_ERRNO_RTX 4768Define this macro as a C expression to create an rtl expression that 4769refers to the global ``variable'' @code{errno}. (On certain systems, 4770@code{errno} may not actually be a variable.) If you don't define this 4771macro, a reasonable default is used. 4772@end defmac 4773 4774@cindex @code{bcopy}, implicit usage 4775@cindex @code{memcpy}, implicit usage 4776@cindex @code{memmove}, implicit usage 4777@cindex @code{bzero}, implicit usage 4778@cindex @code{memset}, implicit usage 4779@defmac TARGET_MEM_FUNCTIONS 4780Define this macro if GCC should generate calls to the ISO C 4781(and System V) library functions @code{memcpy}, @code{memmove} and 4782@code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}. 4783@end defmac 4784 4785@cindex C99 math functions, implicit usage 4786@defmac TARGET_C99_FUNCTIONS 4787When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into 4788@code{sinf} and similarly for other functions defined by C99 standard. The 4789default is nonzero that should be proper value for most modern systems, however 4790number of existing systems lacks support for these functions in the runtime so 4791they needs this macro to be redefined to 0. 4792@end defmac 4793 4794@defmac NEXT_OBJC_RUNTIME 4795Define this macro to generate code for Objective-C message sending using 4796the calling convention of the NeXT system. This calling convention 4797involves passing the object, the selector and the method arguments all 4798at once to the method-lookup library function. 4799 4800The default calling convention passes just the object and the selector 4801to the lookup function, which returns a pointer to the method. 4802@end defmac 4803 4804@node Addressing Modes 4805@section Addressing Modes 4806@cindex addressing modes 4807 4808@c prevent bad page break with this line 4809This is about addressing modes. 4810 4811@defmac HAVE_PRE_INCREMENT 4812@defmacx HAVE_PRE_DECREMENT 4813@defmacx HAVE_POST_INCREMENT 4814@defmacx HAVE_POST_DECREMENT 4815A C expression that is nonzero if the machine supports pre-increment, 4816pre-decrement, post-increment, or post-decrement addressing respectively. 4817@end defmac 4818 4819@defmac HAVE_PRE_MODIFY_DISP 4820@defmacx HAVE_POST_MODIFY_DISP 4821A C expression that is nonzero if the machine supports pre- or 4822post-address side-effect generation involving constants other than 4823the size of the memory operand. 4824@end defmac 4825 4826@defmac HAVE_PRE_MODIFY_REG 4827@defmacx HAVE_POST_MODIFY_REG 4828A C expression that is nonzero if the machine supports pre- or 4829post-address side-effect generation involving a register displacement. 4830@end defmac 4831 4832@defmac CONSTANT_ADDRESS_P (@var{x}) 4833A C expression that is 1 if the RTX @var{x} is a constant which 4834is a valid address. On most machines, this can be defined as 4835@code{CONSTANT_P (@var{x})}, but a few machines are more restrictive 4836in which constant addresses are supported. 4837@end defmac 4838 4839@defmac CONSTANT_P (@var{x}) 4840@code{CONSTANT_P}, which is defined by target-independent code, 4841accepts integer-values expressions whose values are not explicitly 4842known, such as @code{symbol_ref}, @code{label_ref}, and @code{high} 4843expressions and @code{const} arithmetic expressions, in addition to 4844@code{const_int} and @code{const_double} expressions. 4845@end defmac 4846 4847@defmac MAX_REGS_PER_ADDRESS 4848A number, the maximum number of registers that can appear in a valid 4849memory address. Note that it is up to you to specify a value equal to 4850the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever 4851accept. 4852@end defmac 4853 4854@defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label}) 4855A C compound statement with a conditional @code{goto @var{label};} 4856executed if @var{x} (an RTX) is a legitimate memory address on the 4857target machine for a memory operand of mode @var{mode}. 4858 4859It usually pays to define several simpler macros to serve as 4860subroutines for this one. Otherwise it may be too complicated to 4861understand. 4862 4863This macro must exist in two variants: a strict variant and a 4864non-strict one. The strict variant is used in the reload pass. It 4865must be defined so that any pseudo-register that has not been 4866allocated a hard register is considered a memory reference. In 4867contexts where some kind of register is required, a pseudo-register 4868with no hard register must be rejected. 4869 4870The non-strict variant is used in other passes. It must be defined to 4871accept all pseudo-registers in every context where some kind of 4872register is required. 4873 4874@findex REG_OK_STRICT 4875Compiler source files that want to use the strict variant of this 4876macro define the macro @code{REG_OK_STRICT}. You should use an 4877@code{#ifdef REG_OK_STRICT} conditional to define the strict variant 4878in that case and the non-strict variant otherwise. 4879 4880Subroutines to check for acceptable registers for various purposes (one 4881for base registers, one for index registers, and so on) are typically 4882among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}. 4883Then only these subroutine macros need have two variants; the higher 4884levels of macros may be the same whether strict or not. 4885 4886Normally, constant addresses which are the sum of a @code{symbol_ref} 4887and an integer are stored inside a @code{const} RTX to mark them as 4888constant. Therefore, there is no need to recognize such sums 4889specifically as legitimate addresses. Normally you would simply 4890recognize any @code{const} as legitimate. 4891 4892Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant 4893sums that are not marked with @code{const}. It assumes that a naked 4894@code{plus} indicates indexing. If so, then you @emph{must} reject such 4895naked constant sums as illegitimate addresses, so that none of them will 4896be given to @code{PRINT_OPERAND_ADDRESS}. 4897 4898@cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation 4899On some machines, whether a symbolic address is legitimate depends on 4900the section that the address refers to. On these machines, define the 4901target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information 4902into the @code{symbol_ref}, and then check for it here. When you see a 4903@code{const}, you will have to look inside it to find the 4904@code{symbol_ref} in order to determine the section. @xref{Assembler 4905Format}. 4906@end defmac 4907 4908@defmac REG_OK_FOR_BASE_P (@var{x}) 4909A C expression that is nonzero if @var{x} (assumed to be a @code{reg} 4910RTX) is valid for use as a base register. For hard registers, it 4911should always accept those which the hardware permits and reject the 4912others. Whether the macro accepts or rejects pseudo registers must be 4913controlled by @code{REG_OK_STRICT} as described above. This usually 4914requires two variant definitions, of which @code{REG_OK_STRICT} 4915controls the one actually used. 4916@end defmac 4917 4918@defmac REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode}) 4919A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that 4920that expression may examine the mode of the memory reference in 4921@var{mode}. You should define this macro if the mode of the memory 4922reference affects whether a register may be used as a base register. If 4923you define this macro, the compiler will use it instead of 4924@code{REG_OK_FOR_BASE_P}. 4925@end defmac 4926 4927@defmac REG_OK_FOR_INDEX_P (@var{x}) 4928A C expression that is nonzero if @var{x} (assumed to be a @code{reg} 4929RTX) is valid for use as an index register. 4930 4931The difference between an index register and a base register is that 4932the index register may be scaled. If an address involves the sum of 4933two registers, neither one of them scaled, then either one may be 4934labeled the ``base'' and the other the ``index''; but whichever 4935labeling is used must fit the machine's constraints of which registers 4936may serve in each capacity. The compiler will try both labelings, 4937looking for one that is valid, and will reload one or both registers 4938only if neither labeling works. 4939@end defmac 4940 4941@defmac FIND_BASE_TERM (@var{x}) 4942A C expression to determine the base term of address @var{x}. 4943This macro is used in only one place: `find_base_term' in alias.c. 4944 4945It is always safe for this macro to not be defined. It exists so 4946that alias analysis can understand machine-dependent addresses. 4947 4948The typical use of this macro is to handle addresses containing 4949a label_ref or symbol_ref within an UNSPEC@. 4950@end defmac 4951 4952@defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win}) 4953A C compound statement that attempts to replace @var{x} with a valid 4954memory address for an operand of mode @var{mode}. @var{win} will be a 4955C statement label elsewhere in the code; the macro definition may use 4956 4957@smallexample 4958GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win}); 4959@end smallexample 4960 4961@noindent 4962to avoid further processing if the address has become legitimate. 4963 4964@findex break_out_memory_refs 4965@var{x} will always be the result of a call to @code{break_out_memory_refs}, 4966and @var{oldx} will be the operand that was given to that function to produce 4967@var{x}. 4968 4969The code generated by this macro should not alter the substructure of 4970@var{x}. If it transforms @var{x} into a more legitimate form, it 4971should assign @var{x} (which will always be a C variable) a new value. 4972 4973It is not necessary for this macro to come up with a legitimate 4974address. The compiler has standard ways of doing so in all cases. In 4975fact, it is safe for this macro to do nothing. But often a 4976machine-dependent strategy can generate better code. 4977@end defmac 4978 4979@defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win}) 4980A C compound statement that attempts to replace @var{x}, which is an address 4981that needs reloading, with a valid memory address for an operand of mode 4982@var{mode}. @var{win} will be a C statement label elsewhere in the code. 4983It is not necessary to define this macro, but it might be useful for 4984performance reasons. 4985 4986For example, on the i386, it is sometimes possible to use a single 4987reload register instead of two by reloading a sum of two pseudo 4988registers into a register. On the other hand, for number of RISC 4989processors offsets are limited so that often an intermediate address 4990needs to be generated in order to address a stack slot. By defining 4991@code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses 4992generated for adjacent some stack slots can be made identical, and thus 4993be shared. 4994 4995@emph{Note}: This macro should be used with caution. It is necessary 4996to know something of how reload works in order to effectively use this, 4997and it is quite easy to produce macros that build in too much knowledge 4998of reload internals. 4999 5000@emph{Note}: This macro must be able to reload an address created by a 5001previous invocation of this macro. If it fails to handle such addresses 5002then the compiler may generate incorrect code or abort. 5003 5004@findex push_reload 5005The macro definition should use @code{push_reload} to indicate parts that 5006need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually 5007suitable to be passed unaltered to @code{push_reload}. 5008 5009The code generated by this macro must not alter the substructure of 5010@var{x}. If it transforms @var{x} into a more legitimate form, it 5011should assign @var{x} (which will always be a C variable) a new value. 5012This also applies to parts that you change indirectly by calling 5013@code{push_reload}. 5014 5015@findex strict_memory_address_p 5016The macro definition may use @code{strict_memory_address_p} to test if 5017the address has become legitimate. 5018 5019@findex copy_rtx 5020If you want to change only a part of @var{x}, one standard way of doing 5021this is to use @code{copy_rtx}. Note, however, that is unshares only a 5022single level of rtl. Thus, if the part to be changed is not at the 5023top level, you'll need to replace first the top level. 5024It is not necessary for this macro to come up with a legitimate 5025address; but often a machine-dependent strategy can generate better code. 5026@end defmac 5027 5028@defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label}) 5029A C statement or compound statement with a conditional @code{goto 5030@var{label};} executed if memory address @var{x} (an RTX) can have 5031different meanings depending on the machine mode of the memory 5032reference it is used for or if the address is valid for some modes 5033but not others. 5034 5035Autoincrement and autodecrement addresses typically have mode-dependent 5036effects because the amount of the increment or decrement is the size 5037of the operand being addressed. Some machines have other mode-dependent 5038addresses. Many RISC machines have no mode-dependent addresses. 5039 5040You may assume that @var{addr} is a valid address for the machine. 5041@end defmac 5042 5043@defmac LEGITIMATE_CONSTANT_P (@var{x}) 5044A C expression that is nonzero if @var{x} is a legitimate constant for 5045an immediate operand on the target machine. You can assume that 5046@var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact, 5047@samp{1} is a suitable definition for this macro on machines where 5048anything @code{CONSTANT_P} is valid. 5049@end defmac 5050 5051@node Condition Code 5052@section Condition Code Status 5053@cindex condition code status 5054 5055@c prevent bad page break with this line 5056This describes the condition code status. 5057 5058@findex cc_status 5059The file @file{conditions.h} defines a variable @code{cc_status} to 5060describe how the condition code was computed (in case the interpretation of 5061the condition code depends on the instruction that it was set by). This 5062variable contains the RTL expressions on which the condition code is 5063currently based, and several standard flags. 5064 5065Sometimes additional machine-specific flags must be defined in the machine 5066description header file. It can also add additional machine-specific 5067information by defining @code{CC_STATUS_MDEP}. 5068 5069@defmac CC_STATUS_MDEP 5070C code for a data type which is used for declaring the @code{mdep} 5071component of @code{cc_status}. It defaults to @code{int}. 5072 5073This macro is not used on machines that do not use @code{cc0}. 5074@end defmac 5075 5076@defmac CC_STATUS_MDEP_INIT 5077A C expression to initialize the @code{mdep} field to ``empty''. 5078The default definition does nothing, since most machines don't use 5079the field anyway. If you want to use the field, you should probably 5080define this macro to initialize it. 5081 5082This macro is not used on machines that do not use @code{cc0}. 5083@end defmac 5084 5085@defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn}) 5086A C compound statement to set the components of @code{cc_status} 5087appropriately for an insn @var{insn} whose body is @var{exp}. It is 5088this macro's responsibility to recognize insns that set the condition 5089code as a byproduct of other activity as well as those that explicitly 5090set @code{(cc0)}. 5091 5092This macro is not used on machines that do not use @code{cc0}. 5093 5094If there are insns that do not set the condition code but do alter 5095other machine registers, this macro must check to see whether they 5096invalidate the expressions that the condition code is recorded as 5097reflecting. For example, on the 68000, insns that store in address 5098registers do not set the condition code, which means that usually 5099@code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such 5100insns. But suppose that the previous insn set the condition code 5101based on location @samp{a4@@(102)} and the current insn stores a new 5102value in @samp{a4}. Although the condition code is not changed by 5103this, it will no longer be true that it reflects the contents of 5104@samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter 5105@code{cc_status} in this case to say that nothing is known about the 5106condition code value. 5107 5108The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal 5109with the results of peephole optimization: insns whose patterns are 5110@code{parallel} RTXs containing various @code{reg}, @code{mem} or 5111constants which are just the operands. The RTL structure of these 5112insns is not sufficient to indicate what the insns actually do. What 5113@code{NOTICE_UPDATE_CC} should do when it sees one is just to run 5114@code{CC_STATUS_INIT}. 5115 5116A possible definition of @code{NOTICE_UPDATE_CC} is to call a function 5117that looks at an attribute (@pxref{Insn Attributes}) named, for example, 5118@samp{cc}. This avoids having detailed information about patterns in 5119two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}. 5120@end defmac 5121 5122@defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) 5123Returns a mode from class @code{MODE_CC} to be used when comparison 5124operation code @var{op} is applied to rtx @var{x} and @var{y}. For 5125example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see 5126@pxref{Jump Patterns} for a description of the reason for this 5127definition) 5128 5129@smallexample 5130#define SELECT_CC_MODE(OP,X,Y) \ 5131 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ 5132 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ 5133 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ 5134 || GET_CODE (X) == NEG) \ 5135 ? CC_NOOVmode : CCmode)) 5136@end smallexample 5137 5138You should define this macro if and only if you define extra CC modes 5139in @file{@var{machine}-modes.def}. 5140@end defmac 5141 5142@defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1}) 5143On some machines not all possible comparisons are defined, but you can 5144convert an invalid comparison into a valid one. For example, the Alpha 5145does not have a @code{GT} comparison, but you can use an @code{LT} 5146comparison instead and swap the order of the operands. 5147 5148On such machines, define this macro to be a C statement to do any 5149required conversions. @var{code} is the initial comparison code 5150and @var{op0} and @var{op1} are the left and right operands of the 5151comparison, respectively. You should modify @var{code}, @var{op0}, and 5152@var{op1} as required. 5153 5154GCC will not assume that the comparison resulting from this macro is 5155valid but will see if the resulting insn matches a pattern in the 5156@file{md} file. 5157 5158You need not define this macro if it would never change the comparison 5159code or operands. 5160@end defmac 5161 5162@defmac REVERSIBLE_CC_MODE (@var{mode}) 5163A C expression whose value is one if it is always safe to reverse a 5164comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} 5165can ever return @var{mode} for a floating-point inequality comparison, 5166then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. 5167 5168You need not define this macro if it would always returns zero or if the 5169floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. 5170For example, here is the definition used on the SPARC, where floating-point 5171inequality comparisons are always given @code{CCFPEmode}: 5172 5173@smallexample 5174#define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) 5175@end smallexample 5176@end defmac 5177 5178@defmac REVERSE_CONDITION (@var{code}, @var{mode}) 5179A C expression whose value is reversed condition code of the @var{code} for 5180comparison done in CC_MODE @var{mode}. The macro is used only in case 5181@code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case 5182machine has some non-standard way how to reverse certain conditionals. For 5183instance in case all floating point conditions are non-trapping, compiler may 5184freely convert unordered compares to ordered one. Then definition may look 5185like: 5186 5187@smallexample 5188#define REVERSE_CONDITION(CODE, MODE) \ 5189 ((MODE) != CCFPmode ? reverse_condition (CODE) \ 5190 : reverse_condition_maybe_unordered (CODE)) 5191@end smallexample 5192@end defmac 5193 5194@defmac REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2}) 5195A C expression that returns true if the conditional execution predicate 5196@var{code1} is the inverse of @var{code2} and vice versa. Define this to 5197return 0 if the target has conditional execution predicates that cannot be 5198reversed safely. If no expansion is specified, this macro is defined as 5199follows: 5200 5201@smallexample 5202#define REVERSE_CONDEXEC_PREDICATES_P (x, y) \ 5203 ((x) == reverse_condition (y)) 5204@end smallexample 5205@end defmac 5206 5207@deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *) 5208On targets which do not use @code{(cc0)}, and which use a hard 5209register rather than a pseudo-register to hold condition codes, the 5210regular CSE passes are often not able to identify cases in which the 5211hard register is set to a common value. Use this hook to enable a 5212small pass which optimizes such cases. This hook should return true 5213to enable this pass, and it should set the integers to which its 5214arguments point to the hard register numbers used for condition codes. 5215When there is only one such register, as is true on most systems, the 5216integer pointed to by the second argument should be set to 5217@code{INVALID_REGNUM}. 5218 5219The default version of this hook returns false. 5220@end deftypefn 5221 5222@deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode) 5223On targets which use multiple condition code modes in class 5224@code{MODE_CC}, it is sometimes the case that a comparison can be 5225validly done in more than one mode. On such a system, define this 5226target hook to take two mode arguments and to return a mode in which 5227both comparisons may be validly done. If there is no such mode, 5228return @code{VOIDmode}. 5229 5230The default version of this hook checks whether the modes are the 5231same. If they are, it returns that mode. If they are different, it 5232returns @code{VOIDmode}. 5233@end deftypefn 5234 5235@node Costs 5236@section Describing Relative Costs of Operations 5237@cindex costs of instructions 5238@cindex relative costs 5239@cindex speed of instructions 5240 5241These macros let you describe the relative speed of various operations 5242on the target machine. 5243 5244@defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to}) 5245A C expression for the cost of moving data of mode @var{mode} from a 5246register in class @var{from} to one in class @var{to}. The classes are 5247expressed using the enumeration values such as @code{GENERAL_REGS}. A 5248value of 2 is the default; other values are interpreted relative to 5249that. 5250 5251It is not required that the cost always equal 2 when @var{from} is the 5252same as @var{to}; on some machines it is expensive to move between 5253registers if they are not general registers. 5254 5255If reload sees an insn consisting of a single @code{set} between two 5256hard registers, and if @code{REGISTER_MOVE_COST} applied to their 5257classes returns a value of 2, reload does not check to ensure that the 5258constraints of the insn are met. Setting a cost of other than 2 will 5259allow reload to verify that the constraints are met. You should do this 5260if the @samp{mov@var{m}} pattern's constraints do not allow such copying. 5261@end defmac 5262 5263@defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in}) 5264A C expression for the cost of moving data of mode @var{mode} between a 5265register of class @var{class} and memory; @var{in} is zero if the value 5266is to be written to memory, nonzero if it is to be read in. This cost 5267is relative to those in @code{REGISTER_MOVE_COST}. If moving between 5268registers and memory is more expensive than between two registers, you 5269should define this macro to express the relative cost. 5270 5271If you do not define this macro, GCC uses a default cost of 4 plus 5272the cost of copying via a secondary reload register, if one is 5273needed. If your machine requires a secondary reload register to copy 5274between memory and a register of @var{class} but the reload mechanism is 5275more complex than copying via an intermediate, define this macro to 5276reflect the actual cost of the move. 5277 5278GCC defines the function @code{memory_move_secondary_cost} if 5279secondary reloads are needed. It computes the costs due to copying via 5280a secondary register. If your machine copies from memory using a 5281secondary register in the conventional way but the default base value of 52824 is not correct for your machine, define this macro to add some other 5283value to the result of that function. The arguments to that function 5284are the same as to this macro. 5285@end defmac 5286 5287@defmac BRANCH_COST 5288A C expression for the cost of a branch instruction. A value of 1 is 5289the default; other values are interpreted relative to that. 5290@end defmac 5291 5292Here are additional macros which do not specify precise relative costs, 5293but only that certain actions are more expensive than GCC would 5294ordinarily expect. 5295 5296@defmac SLOW_BYTE_ACCESS 5297Define this macro as a C expression which is nonzero if accessing less 5298than a word of memory (i.e.@: a @code{char} or a @code{short}) is no 5299faster than accessing a word of memory, i.e., if such access 5300require more than one instruction or if there is no difference in cost 5301between byte and (aligned) word loads. 5302 5303When this macro is not defined, the compiler will access a field by 5304finding the smallest containing object; when it is defined, a fullword 5305load will be used if alignment permits. Unless bytes accesses are 5306faster than word accesses, using word accesses is preferable since it 5307may eliminate subsequent memory access if subsequent accesses occur to 5308other fields in the same word of the structure, but to different bytes. 5309@end defmac 5310 5311@defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment}) 5312Define this macro to be the value 1 if memory accesses described by the 5313@var{mode} and @var{alignment} parameters have a cost many times greater 5314than aligned accesses, for example if they are emulated in a trap 5315handler. 5316 5317When this macro is nonzero, the compiler will act as if 5318@code{STRICT_ALIGNMENT} were nonzero when generating code for block 5319moves. This can cause significantly more instructions to be produced. 5320Therefore, do not set this macro nonzero if unaligned accesses only add a 5321cycle or two to the time for a memory access. 5322 5323If the value of this macro is always zero, it need not be defined. If 5324this macro is defined, it should produce a nonzero value when 5325@code{STRICT_ALIGNMENT} is nonzero. 5326@end defmac 5327 5328@defmac MOVE_RATIO 5329The threshold of number of scalar memory-to-memory move insns, @emph{below} 5330which a sequence of insns should be generated instead of a 5331string move insn or a library call. Increasing the value will always 5332make code faster, but eventually incurs high cost in increased code size. 5333 5334Note that on machines where the corresponding move insn is a 5335@code{define_expand} that emits a sequence of insns, this macro counts 5336the number of such sequences. 5337 5338If you don't define this, a reasonable default is used. 5339@end defmac 5340 5341@defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment}) 5342A C expression used to determine whether @code{move_by_pieces} will be used to 5343copy a chunk of memory, or whether some other block move mechanism 5344will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less 5345than @code{MOVE_RATIO}. 5346@end defmac 5347 5348@defmac MOVE_MAX_PIECES 5349A C expression used by @code{move_by_pieces} to determine the largest unit 5350a load or store used to copy memory is. Defaults to @code{MOVE_MAX}. 5351@end defmac 5352 5353@defmac CLEAR_RATIO 5354The threshold of number of scalar move insns, @emph{below} which a sequence 5355of insns should be generated to clear memory instead of a string clear insn 5356or a library call. Increasing the value will always make code faster, but 5357eventually incurs high cost in increased code size. 5358 5359If you don't define this, a reasonable default is used. 5360@end defmac 5361 5362@defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment}) 5363A C expression used to determine whether @code{clear_by_pieces} will be used 5364to clear a chunk of memory, or whether some other block clear mechanism 5365will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less 5366than @code{CLEAR_RATIO}. 5367@end defmac 5368 5369@defmac STORE_BY_PIECES_P (@var{size}, @var{alignment}) 5370A C expression used to determine whether @code{store_by_pieces} will be 5371used to set a chunk of memory to a constant value, or whether some other 5372mechanism will be used. Used by @code{__builtin_memset} when storing 5373values other than constant zero and by @code{__builtin_strcpy} when 5374when called with a constant source string. 5375Defaults to @code{MOVE_BY_PIECES_P}. 5376@end defmac 5377 5378@defmac USE_LOAD_POST_INCREMENT (@var{mode}) 5379A C expression used to determine whether a load postincrement is a good 5380thing to use for a given mode. Defaults to the value of 5381@code{HAVE_POST_INCREMENT}. 5382@end defmac 5383 5384@defmac USE_LOAD_POST_DECREMENT (@var{mode}) 5385A C expression used to determine whether a load postdecrement is a good 5386thing to use for a given mode. Defaults to the value of 5387@code{HAVE_POST_DECREMENT}. 5388@end defmac 5389 5390@defmac USE_LOAD_PRE_INCREMENT (@var{mode}) 5391A C expression used to determine whether a load preincrement is a good 5392thing to use for a given mode. Defaults to the value of 5393@code{HAVE_PRE_INCREMENT}. 5394@end defmac 5395 5396@defmac USE_LOAD_PRE_DECREMENT (@var{mode}) 5397A C expression used to determine whether a load predecrement is a good 5398thing to use for a given mode. Defaults to the value of 5399@code{HAVE_PRE_DECREMENT}. 5400@end defmac 5401 5402@defmac USE_STORE_POST_INCREMENT (@var{mode}) 5403A C expression used to determine whether a store postincrement is a good 5404thing to use for a given mode. Defaults to the value of 5405@code{HAVE_POST_INCREMENT}. 5406@end defmac 5407 5408@defmac USE_STORE_POST_DECREMENT (@var{mode}) 5409A C expression used to determine whether a store postdecrement is a good 5410thing to use for a given mode. Defaults to the value of 5411@code{HAVE_POST_DECREMENT}. 5412@end defmac 5413 5414@defmac USE_STORE_PRE_INCREMENT (@var{mode}) 5415This macro is used to determine whether a store preincrement is a good 5416thing to use for a given mode. Defaults to the value of 5417@code{HAVE_PRE_INCREMENT}. 5418@end defmac 5419 5420@defmac USE_STORE_PRE_DECREMENT (@var{mode}) 5421This macro is used to determine whether a store predecrement is a good 5422thing to use for a given mode. Defaults to the value of 5423@code{HAVE_PRE_DECREMENT}. 5424@end defmac 5425 5426@defmac NO_FUNCTION_CSE 5427Define this macro if it is as good or better to call a constant 5428function address than to call an address kept in a register. 5429@end defmac 5430 5431@defmac NO_RECURSIVE_FUNCTION_CSE 5432Define this macro if it is as good or better for a function to call 5433itself with an explicit address than to call an address kept in a 5434register. 5435@end defmac 5436 5437@defmac RANGE_TEST_NON_SHORT_CIRCUIT 5438Define this macro if a non-short-circuit operation produced by 5439@samp{fold_range_test ()} is optimal. This macro defaults to true if 5440@code{BRANCH_COST} is greater than or equal to the value 2. 5441@end defmac 5442 5443@deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}) 5444This target hook describes the relative costs of RTL expressions. 5445 5446The cost may depend on the precise form of the expression, which is 5447available for examination in @var{x}, and the rtx code of the expression 5448in which it is contained, found in @var{outer_code}. @var{code} is the 5449expression code---redundant, since it can be obtained with 5450@code{GET_CODE (@var{x})}. 5451 5452In implementing this hook, you can use the construct 5453@code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast 5454instructions. 5455 5456On entry to the hook, @code{*@var{total}} contains a default estimate 5457for the cost of the expression. The hook should modify this value as 5458necessary. 5459 5460The hook returns true when all subexpressions of @var{x} have been 5461processed, and false when @code{rtx_cost} should recurse. 5462@end deftypefn 5463 5464@deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}) 5465This hook computes the cost of an addressing mode that contains 5466@var{address}. If not defined, the cost is computed from 5467the @var{address} expression and the @code{TARGET_RTX_COST} hook. 5468 5469For most CISC machines, the default cost is a good approximation of the 5470true cost of the addressing mode. However, on RISC machines, all 5471instructions normally have the same length and execution time. Hence 5472all addresses will have equal costs. 5473 5474In cases where more than one form of an address is known, the form with 5475the lowest cost will be used. If multiple forms have the same, lowest, 5476cost, the one that is the most complex will be used. 5477 5478For example, suppose an address that is equal to the sum of a register 5479and a constant is used twice in the same basic block. When this macro 5480is not defined, the address will be computed in a register and memory 5481references will be indirect through that register. On machines where 5482the cost of the addressing mode containing the sum is no higher than 5483that of a simple indirect reference, this will produce an additional 5484instruction and possibly require an additional register. Proper 5485specification of this macro eliminates this overhead for such machines. 5486 5487This hook is never called with an invalid address. 5488 5489On machines where an address involving more than one register is as 5490cheap as an address computation involving only one register, defining 5491@code{TARGET_ADDRESS_COST} to reflect this can cause two registers to 5492be live over a region of code where only one would have been if 5493@code{TARGET_ADDRESS_COST} were not defined in that manner. This effect 5494should be considered in the definition of this macro. Equivalent costs 5495should probably only be given to addresses with different numbers of 5496registers on machines with lots of registers. 5497@end deftypefn 5498 5499@node Scheduling 5500@section Adjusting the Instruction Scheduler 5501 5502The instruction scheduler may need a fair amount of machine-specific 5503adjustment in order to produce good code. GCC provides several target 5504hooks for this purpose. It is usually enough to define just a few of 5505them: try the first ones in this list first. 5506 5507@deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void) 5508This hook returns the maximum number of instructions that can ever 5509issue at the same time on the target machine. The default is one. 5510Although the insn scheduler can define itself the possibility of issue 5511an insn on the same cycle, the value can serve as an additional 5512constraint to issue insns on the same simulated processor cycle (see 5513hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}). 5514This value must be constant over the entire compilation. If you need 5515it to vary depending on what the instructions are, you must use 5516@samp{TARGET_SCHED_VARIABLE_ISSUE}. 5517 5518For the automaton based pipeline interface, you could define this hook 5519to return the value of the macro @code{MAX_DFA_ISSUE_RATE}. 5520@end deftypefn 5521 5522@deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more}) 5523This hook is executed by the scheduler after it has scheduled an insn 5524from the ready list. It should return the number of insns which can 5525still be issued in the current cycle. The default is 5526@samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and 5527@code{USE}, which normally are not counted against the issue rate. 5528You should define this hook if some insns take more machine resources 5529than others, so that fewer insns can follow them in the same cycle. 5530@var{file} is either a null pointer, or a stdio stream to write any 5531debug output to. @var{verbose} is the verbose level provided by 5532@option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that 5533was scheduled. 5534@end deftypefn 5535 5536@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost}) 5537This function corrects the value of @var{cost} based on the 5538relationship between @var{insn} and @var{dep_insn} through the 5539dependence @var{link}. It should return the new value. The default 5540is to make no adjustment to @var{cost}. This can be used for example 5541to specify to the scheduler using the traditional pipeline description 5542that an output- or anti-dependence does not incur the same cost as a 5543data-dependence. If the scheduler using the automaton based pipeline 5544description, the cost of anti-dependence is zero and the cost of 5545output-dependence is maximum of one and the difference of latency 5546times of the first and the second insns. If these values are not 5547acceptable, you could use the hook to modify them too. See also 5548@pxref{Automaton pipeline description}. 5549@end deftypefn 5550 5551@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority}) 5552This hook adjusts the integer scheduling priority @var{priority} of 5553@var{insn}. It should return the new priority. Reduce the priority to 5554execute @var{insn} earlier, increase the priority to execute @var{insn} 5555later. Do not define this hook if you do not need to adjust the 5556scheduling priorities of insns. 5557@end deftypefn 5558 5559@deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock}) 5560This hook is executed by the scheduler after it has scheduled the ready 5561list, to allow the machine description to reorder it (for example to 5562combine two small instructions together on @samp{VLIW} machines). 5563@var{file} is either a null pointer, or a stdio stream to write any 5564debug output to. @var{verbose} is the verbose level provided by 5565@option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready 5566list of instructions that are ready to be scheduled. @var{n_readyp} is 5567a pointer to the number of elements in the ready list. The scheduler 5568reads the ready list in reverse order, starting with 5569@var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock} 5570is the timer tick of the scheduler. You may modify the ready list and 5571the number of ready insns. The return value is the number of insns that 5572can issue this cycle; normally this is just @code{issue_rate}. See also 5573@samp{TARGET_SCHED_REORDER2}. 5574@end deftypefn 5575 5576@deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock}) 5577Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That 5578function is called whenever the scheduler starts a new cycle. This one 5579is called once per iteration over a cycle, immediately after 5580@samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and 5581return the number of insns to be scheduled in the same cycle. Defining 5582this hook can be useful if there are frequent situations where 5583scheduling one insn causes other insns to become ready in the same 5584cycle. These other insns can then be taken into account properly. 5585@end deftypefn 5586 5587@deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail}) 5588This hook is called after evaluation forward dependencies of insns in 5589chain given by two parameter values (@var{head} and @var{tail} 5590correspondingly) but before insns scheduling of the insn chain. For 5591example, it can be used for better insn classification if it requires 5592analysis of dependencies. This hook can use backward and forward 5593dependencies of the insn scheduler because they are already 5594calculated. 5595@end deftypefn 5596 5597@deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready}) 5598This hook is executed by the scheduler at the beginning of each block of 5599instructions that are to be scheduled. @var{file} is either a null 5600pointer, or a stdio stream to write any debug output to. @var{verbose} 5601is the verbose level provided by @option{-fsched-verbose-@var{n}}. 5602@var{max_ready} is the maximum number of insns in the current scheduling 5603region that can be live at the same time. This can be used to allocate 5604scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}. 5605@end deftypefn 5606 5607@deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose}) 5608This hook is executed by the scheduler at the end of each block of 5609instructions that are to be scheduled. It can be used to perform 5610cleanup of any actions done by the other scheduling hooks. @var{file} 5611is either a null pointer, or a stdio stream to write any debug output 5612to. @var{verbose} is the verbose level provided by 5613@option{-fsched-verbose-@var{n}}. 5614@end deftypefn 5615 5616@deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void) 5617This hook is called many times during insn scheduling. If the hook 5618returns nonzero, the automaton based pipeline description is used for 5619insn scheduling. Otherwise the traditional pipeline description is 5620used. The default is usage of the traditional pipeline description. 5621 5622You should also remember that to simplify the insn scheduler sources 5623an empty traditional pipeline description interface is generated even 5624if there is no a traditional pipeline description in the @file{.md} 5625file. The same is true for the automaton based pipeline description. 5626That means that you should be accurate in defining the hook. 5627@end deftypefn 5628 5629@deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) 5630The hook returns an RTL insn. The automaton state used in the 5631pipeline hazard recognizer is changed as if the insn were scheduled 5632when the new simulated processor cycle starts. Usage of the hook may 5633simplify the automaton pipeline description for some @acronym{VLIW} 5634processors. If the hook is defined, it is used only for the automaton 5635based pipeline description. The default is not to change the state 5636when the new simulated processor cycle starts. 5637@end deftypefn 5638 5639@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) 5640The hook can be used to initialize data used by the previous hook. 5641@end deftypefn 5642 5643@deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void) 5644The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used 5645to changed the state as if the insn were scheduled when the new 5646simulated processor cycle finishes. 5647@end deftypefn 5648 5649@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) 5650The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but 5651used to initialize data used by the previous hook. 5652@end deftypefn 5653 5654@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void) 5655This hook controls better choosing an insn from the ready insn queue 5656for the @acronym{DFA}-based insn scheduler. Usually the scheduler 5657chooses the first insn from the queue. If the hook returns a positive 5658value, an additional scheduler code tries all permutations of 5659@samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()} 5660subsequent ready insns to choose an insn whose issue will result in 5661maximal number of issued insns on the same cycle. For the 5662@acronym{VLIW} processor, the code could actually solve the problem of 5663packing simple insns into the @acronym{VLIW} insn. Of course, if the 5664rules of @acronym{VLIW} packing are described in the automaton. 5665 5666This code also could be used for superscalar @acronym{RISC} 5667processors. Let us consider a superscalar @acronym{RISC} processor 5668with 3 pipelines. Some insns can be executed in pipelines @var{A} or 5669@var{B}, some insns can be executed only in pipelines @var{B} or 5670@var{C}, and one insn can be executed in pipeline @var{B}. The 5671processor may issue the 1st insn into @var{A} and the 2nd one into 5672@var{B}. In this case, the 3rd insn will wait for freeing @var{B} 5673until the next cycle. If the scheduler issues the 3rd insn the first, 5674the processor could issue all 3 insns per cycle. 5675 5676Actually this code demonstrates advantages of the automaton based 5677pipeline hazard recognizer. We try quickly and easy many insn 5678schedules to choose the best one. 5679 5680The default is no multipass scheduling. 5681@end deftypefn 5682 5683@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx) 5684 5685This hook controls what insns from the ready insn queue will be 5686considered for the multipass insn scheduling. If the hook returns 5687zero for insn passed as the parameter, the insn will be not chosen to 5688be issued. 5689 5690The default is that any ready insns can be chosen to be issued. 5691@end deftypefn 5692 5693@deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *) 5694 5695This hook is called by the insn scheduler before issuing insn passed 5696as the third parameter on given cycle. If the hook returns nonzero, 5697the insn is not issued on given processors cycle. Instead of that, 5698the processor cycle is advanced. If the value passed through the last 5699parameter is zero, the insn ready queue is not sorted on the new cycle 5700start as usually. The first parameter passes file for debugging 5701output. The second one passes the scheduler verbose level of the 5702debugging output. The forth and the fifth parameter values are 5703correspondingly processor cycle on which the previous insn has been 5704issued and the current processor cycle. 5705@end deftypefn 5706 5707@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void) 5708The @acronym{DFA}-based scheduler could take the insertion of nop 5709operations for better insn scheduling into account. It can be done 5710only if the multi-pass insn scheduling works (see hook 5711@samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}). 5712 5713Let us consider a @acronym{VLIW} processor insn with 3 slots. Each 5714insn can be placed only in one of the three slots. We have 3 ready 5715insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be 5716placed only in the 1st slot, @var{B} can be placed only in the 3rd 5717slot. We described the automaton which does not permit empty slot 5718gaps between insns (usually such description is simpler). Without 5719this code the scheduler would place each insn in 3 separate 5720@acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd 5721slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is 5722the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook 5723@samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or 5724create the nop insns. 5725 5726You should remember that the scheduler does not insert the nop insns. 5727It is not wise because of the following optimizations. The scheduler 5728only considers such possibility to improve the result schedule. The 5729nop insns should be inserted lately, e.g. on the final phase. 5730@end deftypefn 5731 5732@deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index}) 5733This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert 5734nop operations for better insn scheduling when @acronym{DFA}-based 5735scheduler makes multipass insn scheduling (see also description of 5736hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook 5737returns a nop insn with given @var{index}. The indexes start with 5738zero. The hook should return @code{NULL} if there are no more nop 5739insns with indexes greater than given index. 5740@end deftypefn 5741 5742@deftypefn {Target Hook} bool IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance}) 5743This hook is used to define which dependences are considered costly by 5744the target, so costly that it is not advisable to schedule the insns that 5745are involved in the dependence too close to one another. The parameters 5746to this hook are as follows: The second parameter @var{insn2} is dependent 5747upon the first parameter @var{insn1}. The dependence between @var{insn1} 5748and @var{insn2} is represented by the third parameter @var{dep_link}. The 5749fourth parameter @var{cost} is the cost of the dependence, and the fifth 5750parameter @var{distance} is the distance in cycles between the two insns. 5751The hook returns @code{true} if considering the distance between the two 5752insns the dependence between them is considered costly by the target, 5753and @code{false} otherwise. 5754 5755Defining this hook can be useful in multiple-issue out-of-order machines, 5756where (a) it's practically hopeless to predict the actual data/resource 5757delays, however: (b) there's a better chance to predict the actual grouping 5758that will be formed, and (c) correctly emulating the grouping can be very 5759important. In such targets one may want to allow issuing dependent insns 5760closer to one another - i.e, closer than the dependence distance; however, 5761not in cases of "costly dependences", which this hooks allows to define. 5762@end deftypefn 5763 5764Macros in the following table are generated by the program 5765@file{genattr} and can be useful for writing the hooks. 5766 5767@defmac TRADITIONAL_PIPELINE_INTERFACE 5768The macro definition is generated if there is a traditional pipeline 5769description in @file{.md} file. You should also remember that to 5770simplify the insn scheduler sources an empty traditional pipeline 5771description interface is generated even if there is no a traditional 5772pipeline description in the @file{.md} file. The macro can be used to 5773distinguish the two types of the traditional interface. 5774@end defmac 5775 5776@defmac DFA_PIPELINE_INTERFACE 5777The macro definition is generated if there is an automaton pipeline 5778description in @file{.md} file. You should also remember that to 5779simplify the insn scheduler sources an empty automaton pipeline 5780description interface is generated even if there is no an automaton 5781pipeline description in the @file{.md} file. The macro can be used to 5782distinguish the two types of the automaton interface. 5783@end defmac 5784 5785@defmac MAX_DFA_ISSUE_RATE 5786The macro definition is generated in the automaton based pipeline 5787description interface. Its value is calculated from the automaton 5788based pipeline description and is equal to maximal number of all insns 5789described in constructions @samp{define_insn_reservation} which can be 5790issued on the same processor cycle. 5791@end defmac 5792 5793@node Sections 5794@section Dividing the Output into Sections (Texts, Data, @dots{}) 5795@c the above section title is WAY too long. maybe cut the part between 5796@c the (...)? --mew 10feb93 5797 5798An object file is divided into sections containing different types of 5799data. In the most common case, there are three sections: the @dfn{text 5800section}, which holds instructions and read-only data; the @dfn{data 5801section}, which holds initialized writable data; and the @dfn{bss 5802section}, which holds uninitialized data. Some systems have other kinds 5803of sections. 5804 5805The compiler must tell the assembler when to switch sections. These 5806macros control what commands to output to tell the assembler this. You 5807can also define additional sections. 5808 5809@defmac TEXT_SECTION_ASM_OP 5810A C expression whose value is a string, including spacing, containing the 5811assembler operation that should precede instructions and read-only data. 5812Normally @code{"\t.text"} is right. 5813@end defmac 5814 5815@defmac HOT_TEXT_SECTION_NAME 5816If defined, a C string constant for the name of the section containing most 5817frequently executed functions of the program. If not defined, GCC will provide 5818a default definition if the target supports named sections. 5819@end defmac 5820 5821@defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME 5822If defined, a C string constant for the name of the section containing unlikely 5823executed functions in the program. 5824@end defmac 5825 5826@defmac DATA_SECTION_ASM_OP 5827A C expression whose value is a string, including spacing, containing the 5828assembler operation to identify the following data as writable initialized 5829data. Normally @code{"\t.data"} is right. 5830@end defmac 5831 5832@defmac READONLY_DATA_SECTION_ASM_OP 5833A C expression whose value is a string, including spacing, containing the 5834assembler operation to identify the following data as read-only initialized 5835data. 5836@end defmac 5837 5838@defmac READONLY_DATA_SECTION 5839A macro naming a function to call to switch to the proper section for 5840read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP} 5841if defined, else fall back to @code{text_section}. 5842 5843The most common definition will be @code{data_section}, if the target 5844does not have a special read-only data section, and does not put data 5845in the text section. 5846@end defmac 5847 5848@defmac SHARED_SECTION_ASM_OP 5849If defined, a C expression whose value is a string, including spacing, 5850containing the assembler operation to identify the following data as 5851shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used. 5852@end defmac 5853 5854@defmac BSS_SECTION_ASM_OP 5855If defined, a C expression whose value is a string, including spacing, 5856containing the assembler operation to identify the following data as 5857uninitialized global data. If not defined, and neither 5858@code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined, 5859uninitialized global data will be output in the data section if 5860@option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be 5861used. 5862@end defmac 5863 5864@defmac INIT_SECTION_ASM_OP 5865If defined, a C expression whose value is a string, including spacing, 5866containing the assembler operation to identify the following data as 5867initialization code. If not defined, GCC will assume such a section does 5868not exist. 5869@end defmac 5870 5871@defmac FINI_SECTION_ASM_OP 5872If defined, a C expression whose value is a string, including spacing, 5873containing the assembler operation to identify the following data as 5874finalization code. If not defined, GCC will assume such a section does 5875not exist. 5876@end defmac 5877 5878@defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function}) 5879If defined, an ASM statement that switches to a different section 5880via @var{section_op}, calls @var{function}, and switches back to 5881the text section. This is used in @file{crtstuff.c} if 5882@code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls 5883to initialization and finalization functions from the init and fini 5884sections. By default, this macro uses a simple function call. Some 5885ports need hand-crafted assembly code to avoid dependencies on 5886registers initialized in the function prologue or to ensure that 5887constant pools don't end up too far way in the text section. 5888@end defmac 5889 5890@defmac FORCE_CODE_SECTION_ALIGN 5891If defined, an ASM statement that aligns a code section to some 5892arbitrary boundary. This is used to force all fragments of the 5893@code{.init} and @code{.fini} sections to have to same alignment 5894and thus prevent the linker from having to add any padding. 5895@end defmac 5896 5897@findex in_text 5898@findex in_data 5899@defmac EXTRA_SECTIONS 5900A list of names for sections other than the standard two, which are 5901@code{in_text} and @code{in_data}. You need not define this macro 5902on a system with no other sections (that GCC needs to use). 5903@end defmac 5904 5905@findex text_section 5906@findex data_section 5907@defmac EXTRA_SECTION_FUNCTIONS 5908One or more functions to be defined in @file{varasm.c}. These 5909functions should do jobs analogous to those of @code{text_section} and 5910@code{data_section}, for your additional sections. Do not define this 5911macro if you do not define @code{EXTRA_SECTIONS}. 5912@end defmac 5913 5914@defmac JUMP_TABLES_IN_TEXT_SECTION 5915Define this macro to be an expression with a nonzero value if jump 5916tables (for @code{tablejump} insns) should be output in the text 5917section, along with the assembler instructions. Otherwise, the 5918readonly data section is used. 5919 5920This macro is irrelevant if there is no separate readonly data section. 5921@end defmac 5922 5923@deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align}) 5924Switches to the appropriate section for output of @var{exp}. You can 5925assume that @var{exp} is either a @code{VAR_DECL} node or a constant of 5926some sort. @var{reloc} indicates whether the initial value of @var{exp} 5927requires link-time relocations. Bit 0 is set when variable contains 5928local relocations only, while bit 1 is set for global relocations. 5929Select the section by calling @code{data_section} or one of the 5930alternatives for other sections. @var{align} is the constant alignment 5931in bits. 5932 5933The default version of this function takes care of putting read-only 5934variables in @code{readonly_data_section}. 5935@end deftypefn 5936 5937@deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc}) 5938Build up a unique section name, expressed as a @code{STRING_CST} node, 5939and assign it to @samp{DECL_SECTION_NAME (@var{decl})}. 5940As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether 5941the initial value of @var{exp} requires link-time relocations. 5942 5943The default version of this function appends the symbol name to the 5944ELF section name that would normally be used for the symbol. For 5945example, the function @code{foo} would be placed in @code{.text.foo}. 5946Whatever the actual target object format, this is often good enough. 5947@end deftypefn 5948 5949@deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align}) 5950Switches to the appropriate section for output of constant pool entry 5951@var{x} in @var{mode}. You can assume that @var{x} is some kind of 5952constant in RTL@. The argument @var{mode} is redundant except in the 5953case of a @code{const_int} rtx. Select the section by calling 5954@code{readonly_data_section} or one of the alternatives for other 5955sections. @var{align} is the constant alignment in bits. 5956 5957The default version of this function takes care of putting symbolic 5958constants in @code{flag_pic} mode in @code{data_section} and everything 5959else in @code{readonly_data_section}. 5960@end deftypefn 5961 5962@deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p}) 5963Define this hook if references to a symbol or a constant must be 5964treated differently depending on something about the variable or 5965function named by the symbol (such as what section it is in). 5966 5967The hook is executed immediately after rtl has been created for 5968@var{decl}, which may be a variable or function declaration or 5969an entry in the constant pool. In either case, @var{rtl} is the 5970rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})} 5971in this hook; that field may not have been initialized yet. 5972 5973In the case of a constant, it is safe to assume that the rtl is 5974a @code{mem} whose address is a @code{symbol_ref}. Most decls 5975will also have this form, but that is not guaranteed. Global 5976register variables, for instance, will have a @code{reg} for their 5977rtl. (Normally the right thing to do with such unusual rtl is 5978leave it alone.) 5979 5980The @var{new_decl_p} argument will be true if this is the first time 5981that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will 5982be false for subsequent invocations, which will happen for duplicate 5983declarations. Whether or not anything must be done for the duplicate 5984declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}. 5985@var{new_decl_p} is always true when the hook is called for a constant. 5986 5987@cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO} 5988The usual thing for this hook to do is to record flags in the 5989@code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}. 5990Historically, the name string was modified if it was necessary to 5991encode more than one bit of information, but this practice is now 5992discouraged; use @code{SYMBOL_REF_FLAGS}. 5993 5994The default definition of this hook, @code{default_encode_section_info} 5995in @file{varasm.c}, sets a number of commonly-useful bits in 5996@code{SYMBOL_REF_FLAGS}. Check whether the default does what you need 5997before overriding it. 5998@end deftypefn 5999 6000@deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name) 6001Decode @var{name} and return the real name part, sans 6002the characters that @code{TARGET_ENCODE_SECTION_INFO} 6003may have added. 6004@end deftypefn 6005 6006@deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp}) 6007Returns true if @var{exp} should be placed into a ``small data'' section. 6008The default version of this hook always returns false. 6009@end deftypefn 6010 6011@deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION 6012Contains the value true if the target places read-only 6013``small data'' into a separate section. The default value is false. 6014@end deftypevar 6015 6016@deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp}) 6017Returns true if @var{exp} names an object for which name resolution 6018rules must resolve to the current ``module'' (dynamic shared library 6019or executable image). 6020 6021The default version of this hook implements the name resolution rules 6022for ELF, which has a looser model of global name binding than other 6023currently supported object file formats. 6024@end deftypefn 6025 6026@deftypevar {Target Hook} bool TARGET_HAVE_TLS 6027Contains the value true if the target supports thread-local storage. 6028The default value is false. 6029@end deftypevar 6030 6031 6032@node PIC 6033@section Position Independent Code 6034@cindex position independent code 6035@cindex PIC 6036 6037This section describes macros that help implement generation of position 6038independent code. Simply defining these macros is not enough to 6039generate valid PIC; you must also add support to the macros 6040@code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as 6041well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of 6042@samp{movsi} to do something appropriate when the source operand 6043contains a symbolic address. You may also need to alter the handling of 6044switch statements so that they use relative addresses. 6045@c i rearranged the order of the macros above to try to force one of 6046@c them to the next line, to eliminate an overfull hbox. --mew 10feb93 6047 6048@defmac PIC_OFFSET_TABLE_REGNUM 6049The register number of the register used to address a table of static 6050data addresses in memory. In some cases this register is defined by a 6051processor's ``application binary interface'' (ABI)@. When this macro 6052is defined, RTL is generated for this register once, as with the stack 6053pointer and frame pointer registers. If this macro is not defined, it 6054is up to the machine-dependent files to allocate such a register (if 6055necessary). Note that this register must be fixed when in use (e.g.@: 6056when @code{flag_pic} is true). 6057@end defmac 6058 6059@defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED 6060Define this macro if the register defined by 6061@code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define 6062this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined. 6063@end defmac 6064 6065@defmac FINALIZE_PIC 6066By generating position-independent code, when two different programs (A 6067and B) share a common library (libC.a), the text of the library can be 6068shared whether or not the library is linked at the same address for both 6069programs. In some of these environments, position-independent code 6070requires not only the use of different addressing modes, but also 6071special code to enable the use of these addressing modes. 6072 6073The @code{FINALIZE_PIC} macro serves as a hook to emit these special 6074codes once the function is being compiled into assembly code, but not 6075before. (It is not done before, because in the case of compiling an 6076inline function, it would lead to multiple PIC prologues being 6077included in functions which used inline functions and were compiled to 6078assembly language.) 6079@end defmac 6080 6081@defmac LEGITIMATE_PIC_OPERAND_P (@var{x}) 6082A C expression that is nonzero if @var{x} is a legitimate immediate 6083operand on the target machine when generating position independent code. 6084You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not 6085check this. You can also assume @var{flag_pic} is true, so you need not 6086check it either. You need not define this macro if all constants 6087(including @code{SYMBOL_REF}) can be immediate operands when generating 6088position independent code. 6089@end defmac 6090 6091@node Assembler Format 6092@section Defining the Output Assembler Language 6093 6094This section describes macros whose principal purpose is to describe how 6095to write instructions in assembler language---rather than what the 6096instructions do. 6097 6098@menu 6099* File Framework:: Structural information for the assembler file. 6100* Data Output:: Output of constants (numbers, strings, addresses). 6101* Uninitialized Data:: Output of uninitialized variables. 6102* Label Output:: Output and generation of labels. 6103* Initialization:: General principles of initialization 6104 and termination routines. 6105* Macros for Initialization:: 6106 Specific macros that control the handling of 6107 initialization and termination routines. 6108* Instruction Output:: Output of actual instructions. 6109* Dispatch Tables:: Output of jump tables. 6110* Exception Region Output:: Output of exception region code. 6111* Alignment Output:: Pseudo ops for alignment and skipping data. 6112@end menu 6113 6114@node File Framework 6115@subsection The Overall Framework of an Assembler File 6116@cindex assembler format 6117@cindex output of assembler code 6118 6119@c prevent bad page break with this line 6120This describes the overall framework of an assembly file. 6121 6122@deftypefn {Target Hook} void TARGET_ASM_FILE_START () 6123@findex default_file_start 6124Output to @code{asm_out_file} any text which the assembler expects to 6125find at the beginning of a file. The default behavior is controlled 6126by two flags, documented below. Unless your target's assembler is 6127quite unusual, if you override the default, you should call 6128@code{default_file_start} at some point in your target hook. This 6129lets other target files rely on these variables. 6130@end deftypefn 6131 6132@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF 6133If this flag is true, the text of the macro @code{ASM_APP_OFF} will be 6134printed as the very first line in the assembly file, unless 6135@option{-fverbose-asm} is in effect. (If that macro has been defined 6136to the empty string, this variable has no effect.) With the normal 6137definition of @code{ASM_APP_OFF}, the effect is to notify the GNU 6138assembler that it need not bother stripping comments or extra 6139whitespace from its input. This allows it to work a bit faster. 6140 6141The default is false. You should not set it to true unless you have 6142verified that your port does not generate any extra whitespace or 6143comments that will cause GAS to issue errors in NO_APP mode. 6144@end deftypevr 6145 6146@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE 6147If this flag is true, @code{output_file_directive} will be called 6148for the primary source file, immediately after printing 6149@code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect 6150this to be done. The default is false. 6151@end deftypevr 6152 6153@deftypefn {Target Hook} void TARGET_ASM_FILE_END () 6154Output to @code{asm_out_file} any text which the assembler expects 6155to find at the end of a file. The default is to output nothing. 6156@end deftypefn 6157 6158@deftypefun void file_end_indicate_exec_stack () 6159Some systems use a common convention, the @samp{.note.GNU-stack} 6160special section, to indicate whether or not an object file relies on 6161the stack being executable. If your system uses this convention, you 6162should define @code{TARGET_ASM_FILE_END} to this function. If you 6163need to do other things in that hook, have your hook function call 6164this function. 6165@end deftypefun 6166 6167@defmac ASM_COMMENT_START 6168A C string constant describing how to begin a comment in the target 6169assembler language. The compiler assumes that the comment will end at 6170the end of the line. 6171@end defmac 6172 6173@defmac ASM_APP_ON 6174A C string constant for text to be output before each @code{asm} 6175statement or group of consecutive ones. Normally this is 6176@code{"#APP"}, which is a comment that has no effect on most 6177assemblers but tells the GNU assembler that it must check the lines 6178that follow for all valid assembler constructs. 6179@end defmac 6180 6181@defmac ASM_APP_OFF 6182A C string constant for text to be output after each @code{asm} 6183statement or group of consecutive ones. Normally this is 6184@code{"#NO_APP"}, which tells the GNU assembler to resume making the 6185time-saving assumptions that are valid for ordinary compiler output. 6186@end defmac 6187 6188@defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) 6189A C statement to output COFF information or DWARF debugging information 6190which indicates that filename @var{name} is the current source file to 6191the stdio stream @var{stream}. 6192 6193This macro need not be defined if the standard form of output 6194for the file format in use is appropriate. 6195@end defmac 6196 6197@defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string}) 6198A C statement to output the string @var{string} to the stdio stream 6199@var{stream}. If you do not call the function @code{output_quoted_string} 6200in your config files, GCC will only call it to output filenames to 6201the assembler source. So you can use it to canonicalize the format 6202of the filename using this macro. 6203@end defmac 6204 6205@defmac ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter}) 6206A C statement to output DBX or SDB debugging information before code 6207for line number @var{line} of the current source file to the 6208stdio stream @var{stream}. @var{counter} is the number of time the 6209macro was invoked, including the current invocation; it is intended 6210to generate unique labels in the assembly output. 6211 6212This macro need not be defined if the standard form of debugging 6213information for the debugger in use is appropriate. 6214@end defmac 6215 6216@defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string}) 6217A C statement to output something to the assembler file to handle a 6218@samp{#ident} directive containing the text @var{string}. If this 6219macro is not defined, nothing is output for a @samp{#ident} directive. 6220@end defmac 6221 6222@deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align}) 6223Output assembly directives to switch to section @var{name}. The section 6224should have attributes as specified by @var{flags}, which is a bit mask 6225of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align} 6226is nonzero, it contains an alignment in bytes to be used for the section, 6227otherwise some target default should be used. Only targets that must 6228specify an alignment within the section directive need pay attention to 6229@var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}. 6230@end deftypefn 6231 6232@deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS 6233This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}. 6234@end deftypefn 6235 6236@deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc}) 6237Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION} 6238based on a variable or function decl, a section name, and whether or not the 6239declaration's initializer may contain runtime relocations. @var{decl} may be 6240 null, in which case read-write data should be assumed. 6241 6242The default version if this function handles choosing code vs data, 6243read-only vs read-write data, and @code{flag_pic}. You should only 6244need to override this if your target has special flags that might be 6245set via @code{__attribute__}. 6246@end deftypefn 6247 6248@need 2000 6249@node Data Output 6250@subsection Output of Data 6251 6252 6253@deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP 6254@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP 6255@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP 6256@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP 6257@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP 6258@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP 6259@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP 6260@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP 6261@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP 6262These hooks specify assembly directives for creating certain kinds 6263of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a 6264byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an 6265aligned two-byte object, and so on. Any of the hooks may be 6266@code{NULL}, indicating that no suitable directive is available. 6267 6268The compiler will print these strings at the start of a new line, 6269followed immediately by the object's initial value. In most cases, 6270the string should contain a tab, a pseudo-op, and then another tab. 6271@end deftypevr 6272 6273@deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p}) 6274The @code{assemble_integer} function uses this hook to output an 6275integer object. @var{x} is the object's value, @var{size} is its size 6276in bytes and @var{aligned_p} indicates whether it is aligned. The 6277function should return @code{true} if it was able to output the 6278object. If it returns false, @code{assemble_integer} will try to 6279split the object into smaller parts. 6280 6281The default implementation of this hook will use the 6282@code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false} 6283when the relevant string is @code{NULL}. 6284@end deftypefn 6285 6286@defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail}) 6287A C statement to recognize @var{rtx} patterns that 6288@code{output_addr_const} can't deal with, and output assembly code to 6289@var{stream} corresponding to the pattern @var{x}. This may be used to 6290allow machine-dependent @code{UNSPEC}s to appear within constants. 6291 6292If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must 6293@code{goto fail}, so that a standard error message is printed. If it 6294prints an error message itself, by calling, for example, 6295@code{output_operand_lossage}, it may just complete normally. 6296@end defmac 6297 6298@defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) 6299A C statement to output to the stdio stream @var{stream} an assembler 6300instruction to assemble a string constant containing the @var{len} 6301bytes at @var{ptr}. @var{ptr} will be a C expression of type 6302@code{char *} and @var{len} a C expression of type @code{int}. 6303 6304If the assembler has a @code{.ascii} pseudo-op as found in the 6305Berkeley Unix assembler, do not define the macro 6306@code{ASM_OUTPUT_ASCII}. 6307@end defmac 6308 6309@defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n}) 6310A C statement to output word @var{n} of a function descriptor for 6311@var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS} 6312is defined, and is otherwise unused. 6313@end defmac 6314 6315@defmac CONSTANT_POOL_BEFORE_FUNCTION 6316You may define this macro as a C expression. You should define the 6317expression to have a nonzero value if GCC should output the constant 6318pool for a function before the code for the function, or a zero value if 6319GCC should output the constant pool after the function. If you do 6320not define this macro, the usual case, GCC will output the constant 6321pool before the function. 6322@end defmac 6323 6324@defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size}) 6325A C statement to output assembler commands to define the start of the 6326constant pool for a function. @var{funname} is a string giving 6327the name of the function. Should the return type of the function 6328be required, it can be obtained via @var{fundecl}. @var{size} 6329is the size, in bytes, of the constant pool that will be written 6330immediately after this call. 6331 6332If no constant-pool prefix is required, the usual case, this macro need 6333not be defined. 6334@end defmac 6335 6336@defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) 6337A C statement (with or without semicolon) to output a constant in the 6338constant pool, if it needs special treatment. (This macro need not do 6339anything for RTL expressions that can be output normally.) 6340 6341The argument @var{file} is the standard I/O stream to output the 6342assembler code on. @var{x} is the RTL expression for the constant to 6343output, and @var{mode} is the machine mode (in case @var{x} is a 6344@samp{const_int}). @var{align} is the required alignment for the value 6345@var{x}; you should output an assembler directive to force this much 6346alignment. 6347 6348The argument @var{labelno} is a number to use in an internal label for 6349the address of this pool entry. The definition of this macro is 6350responsible for outputting the label definition at the proper place. 6351Here is how to do this: 6352 6353@smallexample 6354@code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno}); 6355@end smallexample 6356 6357When you output a pool entry specially, you should end with a 6358@code{goto} to the label @var{jumpto}. This will prevent the same pool 6359entry from being output a second time in the usual manner. 6360 6361You need not define this macro if it would do nothing. 6362@end defmac 6363 6364@defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) 6365A C statement to output assembler commands to at the end of the constant 6366pool for a function. @var{funname} is a string giving the name of the 6367function. Should the return type of the function be required, you can 6368obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the 6369constant pool that GCC wrote immediately before this call. 6370 6371If no constant-pool epilogue is required, the usual case, you need not 6372define this macro. 6373@end defmac 6374 6375@defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}) 6376Define this macro as a C expression which is nonzero if @var{C} is 6377used as a logical line separator by the assembler. 6378 6379If you do not define this macro, the default is that only 6380the character @samp{;} is treated as a logical line separator. 6381@end defmac 6382 6383@deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN 6384@deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN 6385These target hooks are C string constants, describing the syntax in the 6386assembler for grouping arithmetic expressions. If not overridden, they 6387default to normal parentheses, which is correct for most assemblers. 6388@end deftypevr 6389 6390 These macros are provided by @file{real.h} for writing the definitions 6391of @code{ASM_OUTPUT_DOUBLE} and the like: 6392 6393@defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) 6394@defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) 6395@defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) 6396These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's 6397floating point representation, and store its bit pattern in the variable 6398@var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should 6399be a simple @code{long int}. For the others, it should be an array of 6400@code{long int}. The number of elements in this array is determined by 6401the size of the desired target floating point data type: 32 bits of it 6402go in each @code{long int} array element. Each array element holds 32 6403bits of the result, even if @code{long int} is wider than 32 bits on the 6404host machine. 6405 6406The array element values are designed so that you can print them out 6407using @code{fprintf} in the order they should appear in the target 6408machine's memory. 6409@end defmac 6410 6411@node Uninitialized Data 6412@subsection Output of Uninitialized Variables 6413 6414Each of the macros in this section is used to do the whole job of 6415outputting a single uninitialized variable. 6416 6417@defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6418A C statement (sans semicolon) to output to the stdio stream 6419@var{stream} the assembler definition of a common-label named 6420@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6421is the size rounded up to whatever alignment the caller wants. 6422 6423Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6424output the name itself; before and after that, output the additional 6425assembler syntax for defining the name, and a newline. 6426 6427This macro controls how the assembler definitions of uninitialized 6428common global variables are output. 6429@end defmac 6430 6431@defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) 6432Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a 6433separate, explicit argument. If you define this macro, it is used in 6434place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in 6435handling the required alignment of the variable. The alignment is specified 6436as the number of bits. 6437@end defmac 6438 6439@defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6440Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the 6441variable to be output, if there is one, or @code{NULL_TREE} if there 6442is no corresponding variable. If you define this macro, GCC will use it 6443in place of both @code{ASM_OUTPUT_COMMON} and 6444@code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see 6445the variable's decl in order to chose what to output. 6446@end defmac 6447 6448@defmac ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6449If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it 6450is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON} 6451will be used. 6452@end defmac 6453 6454@defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) 6455A C statement (sans semicolon) to output to the stdio stream 6456@var{stream} the assembler definition of uninitialized global @var{decl} named 6457@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6458is the size rounded up to whatever alignment the caller wants. 6459 6460Try to use function @code{asm_output_bss} defined in @file{varasm.c} when 6461defining this macro. If unable, use the expression 6462@code{assemble_name (@var{stream}, @var{name})} to output the name itself; 6463before and after that, output the additional assembler syntax for defining 6464the name, and a newline. 6465 6466This macro controls how the assembler definitions of uninitialized global 6467variables are output. This macro exists to properly support languages like 6468C++ which do not have @code{common} data. However, this macro currently 6469is not defined for all targets. If this macro and 6470@code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON} 6471or @code{ASM_OUTPUT_ALIGNED_COMMON} or 6472@code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used. 6473@end defmac 6474 6475@defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6476Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a 6477separate, explicit argument. If you define this macro, it is used in 6478place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in 6479handling the required alignment of the variable. The alignment is specified 6480as the number of bits. 6481 6482Try to use function @code{asm_output_aligned_bss} defined in file 6483@file{varasm.c} when defining this macro. 6484@end defmac 6485 6486@defmac ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) 6487If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it 6488is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS} 6489will be used. 6490@end defmac 6491 6492@defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6493A C statement (sans semicolon) to output to the stdio stream 6494@var{stream} the assembler definition of a local-common-label named 6495@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6496is the size rounded up to whatever alignment the caller wants. 6497 6498Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6499output the name itself; before and after that, output the additional 6500assembler syntax for defining the name, and a newline. 6501 6502This macro controls how the assembler definitions of uninitialized 6503static variables are output. 6504@end defmac 6505 6506@defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) 6507Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a 6508separate, explicit argument. If you define this macro, it is used in 6509place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in 6510handling the required alignment of the variable. The alignment is specified 6511as the number of bits. 6512@end defmac 6513 6514@defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6515Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the 6516variable to be output, if there is one, or @code{NULL_TREE} if there 6517is no corresponding variable. If you define this macro, GCC will use it 6518in place of both @code{ASM_OUTPUT_DECL} and 6519@code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see 6520the variable's decl in order to chose what to output. 6521@end defmac 6522 6523@defmac ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6524If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it 6525is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL} 6526will be used. 6527@end defmac 6528 6529@node Label Output 6530@subsection Output and Generation of Labels 6531 6532@c prevent bad page break with this line 6533This is about outputting labels. 6534 6535@findex assemble_name 6536@defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name}) 6537A C statement (sans semicolon) to output to the stdio stream 6538@var{stream} the assembler definition of a label named @var{name}. 6539Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6540output the name itself; before and after that, output the additional 6541assembler syntax for defining the name, and a newline. A default 6542definition of this macro is provided which is correct for most systems. 6543@end defmac 6544 6545@defmac SIZE_ASM_OP 6546A C string containing the appropriate assembler directive to specify the 6547size of a symbol, without any arguments. On systems that use ELF, the 6548default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other 6549systems, the default is not to define this macro. 6550 6551Define this macro only if it is correct to use the default definitions 6552of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE} 6553for your system. If you need your own custom definitions of those 6554macros, or if you do not need explicit symbol sizes at all, do not 6555define this macro. 6556@end defmac 6557 6558@defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size}) 6559A C statement (sans semicolon) to output to the stdio stream 6560@var{stream} a directive telling the assembler that the size of the 6561symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}. 6562If you define @code{SIZE_ASM_OP}, a default definition of this macro is 6563provided. 6564@end defmac 6565 6566@defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name}) 6567A C statement (sans semicolon) to output to the stdio stream 6568@var{stream} a directive telling the assembler to calculate the size of 6569the symbol @var{name} by subtracting its address from the current 6570address. 6571 6572If you define @code{SIZE_ASM_OP}, a default definition of this macro is 6573provided. The default assumes that the assembler recognizes a special 6574@samp{.} symbol as referring to the current address, and can calculate 6575the difference between this and another symbol. If your assembler does 6576not recognize @samp{.} or cannot do calculations with it, you will need 6577to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique. 6578@end defmac 6579 6580@defmac TYPE_ASM_OP 6581A C string containing the appropriate assembler directive to specify the 6582type of a symbol, without any arguments. On systems that use ELF, the 6583default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other 6584systems, the default is not to define this macro. 6585 6586Define this macro only if it is correct to use the default definition of 6587@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 6588custom definition of this macro, or if you do not need explicit symbol 6589types at all, do not define this macro. 6590@end defmac 6591 6592@defmac TYPE_OPERAND_FMT 6593A C string which specifies (using @code{printf} syntax) the format of 6594the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the 6595default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems, 6596the default is not to define this macro. 6597 6598Define this macro only if it is correct to use the default definition of 6599@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 6600custom definition of this macro, or if you do not need explicit symbol 6601types at all, do not define this macro. 6602@end defmac 6603 6604@defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type}) 6605A C statement (sans semicolon) to output to the stdio stream 6606@var{stream} a directive telling the assembler that the type of the 6607symbol @var{name} is @var{type}. @var{type} is a C string; currently, 6608that string is always either @samp{"function"} or @samp{"object"}, but 6609you should not count on this. 6610 6611If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default 6612definition of this macro is provided. 6613@end defmac 6614 6615@defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) 6616A C statement (sans semicolon) to output to the stdio stream 6617@var{stream} any text necessary for declaring the name @var{name} of a 6618function which is being defined. This macro is responsible for 6619outputting the label definition (perhaps using 6620@code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the 6621@code{FUNCTION_DECL} tree node representing the function. 6622 6623If this macro is not defined, then the function name is defined in the 6624usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 6625 6626You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition 6627of this macro. 6628@end defmac 6629 6630@defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) 6631A C statement (sans semicolon) to output to the stdio stream 6632@var{stream} any text necessary for declaring the size of a function 6633which is being defined. The argument @var{name} is the name of the 6634function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node 6635representing the function. 6636 6637If this macro is not defined, then the function size is not defined. 6638 6639You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition 6640of this macro. 6641@end defmac 6642 6643@defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) 6644A C statement (sans semicolon) to output to the stdio stream 6645@var{stream} any text necessary for declaring the name @var{name} of an 6646initialized variable which is being defined. This macro must output the 6647label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument 6648@var{decl} is the @code{VAR_DECL} tree node representing the variable. 6649 6650If this macro is not defined, then the variable name is defined in the 6651usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 6652 6653You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or 6654@code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro. 6655@end defmac 6656 6657@defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size}) 6658A C statement (sans semicolon) to output to the stdio stream 6659@var{stream} any text necessary for declaring the name @var{name} of a 6660constant which is being defined. This macro is responsible for 6661outputting the label definition (perhaps using 6662@code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the 6663value of the constant, and @var{size} is the size of the constant 6664in bytes. @var{name} will be an internal label. 6665 6666If this macro is not defined, then the @var{name} is defined in the 6667usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 6668 6669You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition 6670of this macro. 6671@end defmac 6672 6673@defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name}) 6674A C statement (sans semicolon) to output to the stdio stream 6675@var{stream} any text necessary for claiming a register @var{regno} 6676for a global variable @var{decl} with name @var{name}. 6677 6678If you don't define this macro, that is equivalent to defining it to do 6679nothing. 6680@end defmac 6681 6682@defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) 6683A C statement (sans semicolon) to finish up declaring a variable name 6684once the compiler has processed its initializer fully and thus has had a 6685chance to determine the size of an array when controlled by an 6686initializer. This is used on systems where it's necessary to declare 6687something about the size of the object. 6688 6689If you don't define this macro, that is equivalent to defining it to do 6690nothing. 6691 6692You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or 6693@code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro. 6694@end defmac 6695 6696@deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name}) 6697This target hook is a function to output to the stdio stream 6698@var{stream} some commands that will make the label @var{name} global; 6699that is, available for reference from other files. 6700 6701The default implementation relies on a proper definition of 6702@code{GLOBAL_ASM_OP}. 6703@end deftypefn 6704 6705@defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name}) 6706A C statement (sans semicolon) to output to the stdio stream 6707@var{stream} some commands that will make the label @var{name} weak; 6708that is, available for reference from other files but only used if 6709no other definition is available. Use the expression 6710@code{assemble_name (@var{stream}, @var{name})} to output the name 6711itself; before and after that, output the additional assembler syntax 6712for making that name weak, and a newline. 6713 6714If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not 6715support weak symbols and you should not define the @code{SUPPORTS_WEAK} 6716macro. 6717@end defmac 6718 6719@defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value}) 6720Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and 6721@code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function 6722or variable decl. If @var{value} is not @code{NULL}, this C statement 6723should output to the stdio stream @var{stream} assembler code which 6724defines (equates) the weak symbol @var{name} to have the value 6725@var{value}. If @var{value} is @code{NULL}, it should output commands 6726to make @var{name} weak. 6727@end defmac 6728 6729@defmac SUPPORTS_WEAK 6730A C expression which evaluates to true if the target supports weak symbols. 6731 6732If you don't define this macro, @file{defaults.h} provides a default 6733definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL} 6734is defined, the default definition is @samp{1}; otherwise, it is 6735@samp{0}. Define this macro if you want to control weak symbol support 6736with a compiler flag such as @option{-melf}. 6737@end defmac 6738 6739@defmac MAKE_DECL_ONE_ONLY (@var{decl}) 6740A C statement (sans semicolon) to mark @var{decl} to be emitted as a 6741public symbol such that extra copies in multiple translation units will 6742be discarded by the linker. Define this macro if your object file 6743format provides support for this concept, such as the @samp{COMDAT} 6744section flags in the Microsoft Windows PE/COFF format, and this support 6745requires changes to @var{decl}, such as putting it in a separate section. 6746@end defmac 6747 6748@defmac SUPPORTS_ONE_ONLY 6749A C expression which evaluates to true if the target supports one-only 6750semantics. 6751 6752If you don't define this macro, @file{varasm.c} provides a default 6753definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default 6754definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if 6755you want to control one-only symbol support with a compiler flag, or if 6756setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to 6757be emitted as one-only. 6758@end defmac 6759 6760@deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility}) 6761This target hook is a function to output to @var{asm_out_file} some 6762commands that will make the symbol(s) associated with @var{decl} have 6763hidden, protected or internal visibility as specified by @var{visibility}. 6764@end deftypefn 6765 6766@defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) 6767A C statement (sans semicolon) to output to the stdio stream 6768@var{stream} any text necessary for declaring the name of an external 6769symbol named @var{name} which is referenced in this compilation but 6770not defined. The value of @var{decl} is the tree node for the 6771declaration. 6772 6773This macro need not be defined if it does not need to output anything. 6774The GNU assembler and most Unix assemblers don't require anything. 6775@end defmac 6776 6777@deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref}) 6778This target hook is a function to output to @var{asm_out_file} an assembler 6779pseudo-op to declare a library function name external. The name of the 6780library function is given by @var{symref}, which is a @code{symbol_ref}. 6781@end deftypefn 6782 6783@defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) 6784A C statement (sans semicolon) to output to the stdio stream 6785@var{stream} a reference in assembler syntax to a label named 6786@var{name}. This should add @samp{_} to the front of the name, if that 6787is customary on your operating system, as it is in most Berkeley Unix 6788systems. This macro is used in @code{assemble_name}. 6789@end defmac 6790 6791@defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym}) 6792A C statement (sans semicolon) to output a reference to 6793@code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name} 6794will be used to output the name of the symbol. This macro may be used 6795to modify the way a symbol is referenced depending on information 6796encoded by @code{TARGET_ENCODE_SECTION_INFO}. 6797@end defmac 6798 6799@defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf}) 6800A C statement (sans semicolon) to output a reference to @var{buf}, the 6801result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined, 6802@code{assemble_name} will be used to output the name of the symbol. 6803This macro is not used by @code{output_asm_label}, or the @code{%l} 6804specifier that calls it; the intention is that this macro should be set 6805when it is necessary to output a label differently when its address is 6806being taken. 6807@end defmac 6808 6809@deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno}) 6810A function to output to the stdio stream @var{stream} a label whose 6811name is made from the string @var{prefix} and the number @var{labelno}. 6812 6813It is absolutely essential that these labels be distinct from the labels 6814used for user-level functions and variables. Otherwise, certain programs 6815will have name conflicts with internal labels. 6816 6817It is desirable to exclude internal labels from the symbol table of the 6818object file. Most assemblers have a naming convention for labels that 6819should be excluded; on many systems, the letter @samp{L} at the 6820beginning of a label has this effect. You should find out what 6821convention your system uses, and follow it. 6822 6823The default version of this function utilizes ASM_GENERATE_INTERNAL_LABEL. 6824@end deftypefn 6825 6826@defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num}) 6827A C statement to output to the stdio stream @var{stream} a debug info 6828label whose name is made from the string @var{prefix} and the number 6829@var{num}. This is useful for VLIW targets, where debug info labels 6830may need to be treated differently than branch target labels. On some 6831systems, branch target labels must be at the beginning of instruction 6832bundles, but debug info labels can occur in the middle of instruction 6833bundles. 6834 6835If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be 6836used. 6837@end defmac 6838 6839@defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) 6840A C statement to store into the string @var{string} a label whose name 6841is made from the string @var{prefix} and the number @var{num}. 6842 6843This string, when output subsequently by @code{assemble_name}, should 6844produce the output that @code{(*targetm.asm_out.internal_label)} would produce 6845with the same @var{prefix} and @var{num}. 6846 6847If the string begins with @samp{*}, then @code{assemble_name} will 6848output the rest of the string unchanged. It is often convenient for 6849@code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the 6850string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets 6851to output the string, and may change it. (Of course, 6852@code{ASM_OUTPUT_LABELREF} is also part of your machine description, so 6853you should know what it does on your machine.) 6854@end defmac 6855 6856@defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) 6857A C expression to assign to @var{outvar} (which is a variable of type 6858@code{char *}) a newly allocated string made from the string 6859@var{name} and the number @var{number}, with some suitable punctuation 6860added. Use @code{alloca} to get space for the string. 6861 6862The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to 6863produce an assembler label for an internal static variable whose name is 6864@var{name}. Therefore, the string must be such as to result in valid 6865assembler code. The argument @var{number} is different each time this 6866macro is executed; it prevents conflicts between similarly-named 6867internal static variables in different scopes. 6868 6869Ideally this string should not be a valid C identifier, to prevent any 6870conflict with the user's own symbols. Most assemblers allow periods 6871or percent signs in assembler symbols; putting at least one of these 6872between the name and the number will suffice. 6873 6874If this macro is not defined, a default definition will be provided 6875which is correct for most systems. 6876@end defmac 6877 6878@defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) 6879A C statement to output to the stdio stream @var{stream} assembler code 6880which defines (equates) the symbol @var{name} to have the value @var{value}. 6881 6882@findex SET_ASM_OP 6883If @code{SET_ASM_OP} is defined, a default definition is provided which is 6884correct for most systems. 6885@end defmac 6886 6887@defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value}) 6888A C statement to output to the stdio stream @var{stream} assembler code 6889which defines (equates) the symbol whose tree node is @var{decl_of_name} 6890to have the value of the tree node @var{decl_of_value}. This macro will 6891be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if 6892the tree nodes are available. 6893 6894@findex SET_ASM_OP 6895If @code{SET_ASM_OP} is defined, a default definition is provided which is 6896correct for most systems. 6897@end defmac 6898 6899@defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value}) 6900A C statement to output to the stdio stream @var{stream} assembler code 6901which defines (equates) the weak symbol @var{name} to have the value 6902@var{value}. If @var{value} is @code{NULL}, it defines @var{name} as 6903an undefined weak symbol. 6904 6905Define this macro if the target only supports weak aliases; define 6906@code{ASM_OUTPUT_DEF} instead if possible. 6907@end defmac 6908 6909@defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name}) 6910Define this macro to override the default assembler names used for 6911Objective-C methods. 6912 6913The default name is a unique method number followed by the name of the 6914class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of 6915the category is also included in the assembler name (e.g.@: 6916@samp{_1_Foo_Bar}). 6917 6918These names are safe on most systems, but make debugging difficult since 6919the method's selector is not present in the name. Therefore, particular 6920systems define other ways of computing names. 6921 6922@var{buf} is an expression of type @code{char *} which gives you a 6923buffer in which to store the name; its length is as long as 6924@var{class_name}, @var{cat_name} and @var{sel_name} put together, plus 692550 characters extra. 6926 6927The argument @var{is_inst} specifies whether the method is an instance 6928method or a class method; @var{class_name} is the name of the class; 6929@var{cat_name} is the name of the category (or @code{NULL} if the method is not 6930in a category); and @var{sel_name} is the name of the selector. 6931 6932On systems where the assembler can handle quoted names, you can use this 6933macro to provide more human-readable names. 6934@end defmac 6935 6936@defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name}) 6937A C statement (sans semicolon) to output to the stdio stream 6938@var{stream} commands to declare that the label @var{name} is an 6939Objective-C class reference. This is only needed for targets whose 6940linkers have special support for NeXT-style runtimes. 6941@end defmac 6942 6943@defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name}) 6944A C statement (sans semicolon) to output to the stdio stream 6945@var{stream} commands to declare that the label @var{name} is an 6946unresolved Objective-C class reference. This is only needed for targets 6947whose linkers have special support for NeXT-style runtimes. 6948@end defmac 6949 6950@node Initialization 6951@subsection How Initialization Functions Are Handled 6952@cindex initialization routines 6953@cindex termination routines 6954@cindex constructors, output of 6955@cindex destructors, output of 6956 6957The compiled code for certain languages includes @dfn{constructors} 6958(also called @dfn{initialization routines})---functions to initialize 6959data in the program when the program is started. These functions need 6960to be called before the program is ``started''---that is to say, before 6961@code{main} is called. 6962 6963Compiling some languages generates @dfn{destructors} (also called 6964@dfn{termination routines}) that should be called when the program 6965terminates. 6966 6967To make the initialization and termination functions work, the compiler 6968must output something in the assembler code to cause those functions to 6969be called at the appropriate time. When you port the compiler to a new 6970system, you need to specify how to do this. 6971 6972There are two major ways that GCC currently supports the execution of 6973initialization and termination functions. Each way has two variants. 6974Much of the structure is common to all four variations. 6975 6976@findex __CTOR_LIST__ 6977@findex __DTOR_LIST__ 6978The linker must build two lists of these functions---a list of 6979initialization functions, called @code{__CTOR_LIST__}, and a list of 6980termination functions, called @code{__DTOR_LIST__}. 6981 6982Each list always begins with an ignored function pointer (which may hold 69830, @minus{}1, or a count of the function pointers after it, depending on 6984the environment). This is followed by a series of zero or more function 6985pointers to constructors (or destructors), followed by a function 6986pointer containing zero. 6987 6988Depending on the operating system and its executable file format, either 6989@file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup 6990time and exit time. Constructors are called in reverse order of the 6991list; destructors in forward order. 6992 6993The best way to handle static constructors works only for object file 6994formats which provide arbitrarily-named sections. A section is set 6995aside for a list of constructors, and another for a list of destructors. 6996Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each 6997object file that defines an initialization function also puts a word in 6998the constructor section to point to that function. The linker 6999accumulates all these words into one contiguous @samp{.ctors} section. 7000Termination functions are handled similarly. 7001 7002This method will be chosen as the default by @file{target-def.h} if 7003@code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not 7004support arbitrary sections, but does support special designated 7005constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP} 7006and @code{DTORS_SECTION_ASM_OP} to achieve the same effect. 7007 7008When arbitrary sections are available, there are two variants, depending 7009upon how the code in @file{crtstuff.c} is called. On systems that 7010support a @dfn{.init} section which is executed at program startup, 7011parts of @file{crtstuff.c} are compiled into that section. The 7012program is linked by the @command{gcc} driver like this: 7013 7014@smallexample 7015ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o 7016@end smallexample 7017 7018The prologue of a function (@code{__init}) appears in the @code{.init} 7019section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise 7020for the function @code{__fini} in the @dfn{.fini} section. Normally these 7021files are provided by the operating system or by the GNU C library, but 7022are provided by GCC for a few targets. 7023 7024The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets) 7025compiled from @file{crtstuff.c}. They contain, among other things, code 7026fragments within the @code{.init} and @code{.fini} sections that branch 7027to routines in the @code{.text} section. The linker will pull all parts 7028of a section together, which results in a complete @code{__init} function 7029that invokes the routines we need at startup. 7030 7031To use this variant, you must define the @code{INIT_SECTION_ASM_OP} 7032macro properly. 7033 7034If no init section is available, when GCC compiles any function called 7035@code{main} (or more accurately, any function designated as a program 7036entry point by the language front end calling @code{expand_main_function}), 7037it inserts a procedure call to @code{__main} as the first executable code 7038after the function prologue. The @code{__main} function is defined 7039in @file{libgcc2.c} and runs the global constructors. 7040 7041In file formats that don't support arbitrary sections, there are again 7042two variants. In the simplest variant, the GNU linker (GNU @code{ld}) 7043and an `a.out' format must be used. In this case, 7044@code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs} 7045entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, 7046and with the address of the void function containing the initialization 7047code as its value. The GNU linker recognizes this as a request to add 7048the value to a @dfn{set}; the values are accumulated, and are eventually 7049placed in the executable as a vector in the format described above, with 7050a leading (ignored) count and a trailing zero element. 7051@code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init 7052section is available, the absence of @code{INIT_SECTION_ASM_OP} causes 7053the compilation of @code{main} to call @code{__main} as above, starting 7054the initialization process. 7055 7056The last variant uses neither arbitrary sections nor the GNU linker. 7057This is preferable when you want to do dynamic linking and when using 7058file formats which the GNU linker does not support, such as `ECOFF'@. In 7059this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and 7060termination functions are recognized simply by their names. This requires 7061an extra program in the linkage step, called @command{collect2}. This program 7062pretends to be the linker, for use with GCC; it does its job by running 7063the ordinary linker, but also arranges to include the vectors of 7064initialization and termination functions. These functions are called 7065via @code{__main} as described above. In order to use this method, 7066@code{use_collect2} must be defined in the target in @file{config.gcc}. 7067 7068@ifinfo 7069The following section describes the specific macros that control and 7070customize the handling of initialization and termination functions. 7071@end ifinfo 7072 7073@node Macros for Initialization 7074@subsection Macros Controlling Initialization Routines 7075 7076Here are the macros that control how the compiler handles initialization 7077and termination functions: 7078 7079@defmac INIT_SECTION_ASM_OP 7080If defined, a C string constant, including spacing, for the assembler 7081operation to identify the following data as initialization code. If not 7082defined, GCC will assume such a section does not exist. When you are 7083using special sections for initialization and termination functions, this 7084macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to 7085run the initialization functions. 7086@end defmac 7087 7088@defmac HAS_INIT_SECTION 7089If defined, @code{main} will not call @code{__main} as described above. 7090This macro should be defined for systems that control start-up code 7091on a symbol-by-symbol basis, such as OSF/1, and should not 7092be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}. 7093@end defmac 7094 7095@defmac LD_INIT_SWITCH 7096If defined, a C string constant for a switch that tells the linker that 7097the following symbol is an initialization routine. 7098@end defmac 7099 7100@defmac LD_FINI_SWITCH 7101If defined, a C string constant for a switch that tells the linker that 7102the following symbol is a finalization routine. 7103@end defmac 7104 7105@defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func}) 7106If defined, a C statement that will write a function that can be 7107automatically called when a shared library is loaded. The function 7108should call @var{func}, which takes no arguments. If not defined, and 7109the object format requires an explicit initialization function, then a 7110function called @code{_GLOBAL__DI} will be generated. 7111 7112This function and the following one are used by collect2 when linking a 7113shared library that needs constructors or destructors, or has DWARF2 7114exception tables embedded in the code. 7115@end defmac 7116 7117@defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func}) 7118If defined, a C statement that will write a function that can be 7119automatically called when a shared library is unloaded. The function 7120should call @var{func}, which takes no arguments. If not defined, and 7121the object format requires an explicit finalization function, then a 7122function called @code{_GLOBAL__DD} will be generated. 7123@end defmac 7124 7125@defmac INVOKE__main 7126If defined, @code{main} will call @code{__main} despite the presence of 7127@code{INIT_SECTION_ASM_OP}. This macro should be defined for systems 7128where the init section is not actually run automatically, but is still 7129useful for collecting the lists of constructors and destructors. 7130@end defmac 7131 7132@defmac SUPPORTS_INIT_PRIORITY 7133If nonzero, the C++ @code{init_priority} attribute is supported and the 7134compiler should emit instructions to control the order of initialization 7135of objects. If zero, the compiler will issue an error message upon 7136encountering an @code{init_priority} attribute. 7137@end defmac 7138 7139@deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS 7140This value is true if the target supports some ``native'' method of 7141collecting constructors and destructors to be run at startup and exit. 7142It is false if we must use @command{collect2}. 7143@end deftypefn 7144 7145@deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority}) 7146If defined, a function that outputs assembler code to arrange to call 7147the function referenced by @var{symbol} at initialization time. 7148 7149Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking 7150no arguments and with no return value. If the target supports initialization 7151priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY}; 7152otherwise it must be @code{DEFAULT_INIT_PRIORITY}. 7153 7154If this macro is not defined by the target, a suitable default will 7155be chosen if (1) the target supports arbitrary section names, (2) the 7156target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2} 7157is not defined. 7158@end deftypefn 7159 7160@deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority}) 7161This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination 7162functions rather than initialization functions. 7163@end deftypefn 7164 7165If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine 7166generated for the generated object file will have static linkage. 7167 7168If your system uses @command{collect2} as the means of processing 7169constructors, then that program normally uses @command{nm} to scan 7170an object file for constructor functions to be called. 7171 7172On certain kinds of systems, you can define this macro to make 7173@command{collect2} work faster (and, in some cases, make it work at all): 7174 7175@defmac OBJECT_FORMAT_COFF 7176Define this macro if the system uses COFF (Common Object File Format) 7177object files, so that @command{collect2} can assume this format and scan 7178object files directly for dynamic constructor/destructor functions. 7179 7180This macro is effective only in a native compiler; @command{collect2} as 7181part of a cross compiler always uses @command{nm} for the target machine. 7182@end defmac 7183 7184@defmac COLLECT_PARSE_FLAG (@var{flag}) 7185Define this macro to be C code that examines @command{collect2} command 7186line option @var{flag} and performs special actions if 7187@command{collect2} needs to behave differently depending on @var{flag}. 7188@end defmac 7189 7190@defmac REAL_NM_FILE_NAME 7191Define this macro as a C string constant containing the file name to use 7192to execute @command{nm}. The default is to search the path normally for 7193@command{nm}. 7194 7195If your system supports shared libraries and has a program to list the 7196dynamic dependencies of a given library or executable, you can define 7197these macros to enable support for running initialization and 7198termination functions in shared libraries: 7199@end defmac 7200 7201@defmac LDD_SUFFIX 7202Define this macro to a C string constant containing the name of the program 7203which lists dynamic dependencies, like @command{"ldd"} under SunOS 4. 7204@end defmac 7205 7206@defmac PARSE_LDD_OUTPUT (@var{ptr}) 7207Define this macro to be C code that extracts filenames from the output 7208of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable 7209of type @code{char *} that points to the beginning of a line of output 7210from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the 7211code must advance @var{ptr} to the beginning of the filename on that 7212line. Otherwise, it must set @var{ptr} to @code{NULL}. 7213@end defmac 7214 7215@node Instruction Output 7216@subsection Output of Assembler Instructions 7217 7218@c prevent bad page break with this line 7219This describes assembler instruction output. 7220 7221@defmac REGISTER_NAMES 7222A C initializer containing the assembler's names for the machine 7223registers, each one as a C string constant. This is what translates 7224register numbers in the compiler into assembler language. 7225@end defmac 7226 7227@defmac ADDITIONAL_REGISTER_NAMES 7228If defined, a C initializer for an array of structures containing a name 7229and a register number. This macro defines additional names for hard 7230registers, thus allowing the @code{asm} option in declarations to refer 7231to registers using alternate names. 7232@end defmac 7233 7234@defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) 7235Define this macro if you are using an unusual assembler that 7236requires different names for the machine instructions. 7237 7238The definition is a C statement or statements which output an 7239assembler instruction opcode to the stdio stream @var{stream}. The 7240macro-operand @var{ptr} is a variable of type @code{char *} which 7241points to the opcode name in its ``internal'' form---the form that is 7242written in the machine description. The definition should output the 7243opcode name to @var{stream}, performing any translation you desire, and 7244increment the variable @var{ptr} to point at the end of the opcode 7245so that it will not be output twice. 7246 7247In fact, your macro definition may process less than the entire opcode 7248name, or more than the opcode name; but if you want to process text 7249that includes @samp{%}-sequences to substitute operands, you must take 7250care of the substitution yourself. Just be sure to increment 7251@var{ptr} over whatever text should not be output normally. 7252 7253@findex recog_data.operand 7254If you need to look at the operand values, they can be found as the 7255elements of @code{recog_data.operand}. 7256 7257If the macro definition does nothing, the instruction is output 7258in the usual way. 7259@end defmac 7260 7261@defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) 7262If defined, a C statement to be executed just prior to the output of 7263assembler code for @var{insn}, to modify the extracted operands so 7264they will be output differently. 7265 7266Here the argument @var{opvec} is the vector containing the operands 7267extracted from @var{insn}, and @var{noperands} is the number of 7268elements of the vector which contain meaningful data for this insn. 7269The contents of this vector are what will be used to convert the insn 7270template into assembler code, so you can change the assembler output 7271by changing the contents of the vector. 7272 7273This macro is useful when various assembler syntaxes share a single 7274file of instruction patterns; by defining this macro differently, you 7275can cause a large class of instructions to be output differently (such 7276as with rearranged operands). Naturally, variations in assembler 7277syntax affecting individual insn patterns ought to be handled by 7278writing conditional output routines in those patterns. 7279 7280If this macro is not defined, it is equivalent to a null statement. 7281@end defmac 7282 7283@defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) 7284A C compound statement to output to stdio stream @var{stream} the 7285assembler syntax for an instruction operand @var{x}. @var{x} is an 7286RTL expression. 7287 7288@var{code} is a value that can be used to specify one of several ways 7289of printing the operand. It is used when identical operands must be 7290printed differently depending on the context. @var{code} comes from 7291the @samp{%} specification that was used to request printing of the 7292operand. If the specification was just @samp{%@var{digit}} then 7293@var{code} is 0; if the specification was @samp{%@var{ltr} 7294@var{digit}} then @var{code} is the ASCII code for @var{ltr}. 7295 7296@findex reg_names 7297If @var{x} is a register, this macro should print the register's name. 7298The names can be found in an array @code{reg_names} whose type is 7299@code{char *[]}. @code{reg_names} is initialized from 7300@code{REGISTER_NAMES}. 7301 7302When the machine description has a specification @samp{%@var{punct}} 7303(a @samp{%} followed by a punctuation character), this macro is called 7304with a null pointer for @var{x} and the punctuation character for 7305@var{code}. 7306@end defmac 7307 7308@defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code}) 7309A C expression which evaluates to true if @var{code} is a valid 7310punctuation character for use in the @code{PRINT_OPERAND} macro. If 7311@code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no 7312punctuation characters (except for the standard one, @samp{%}) are used 7313in this way. 7314@end defmac 7315 7316@defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) 7317A C compound statement to output to stdio stream @var{stream} the 7318assembler syntax for an instruction operand that is a memory reference 7319whose address is @var{x}. @var{x} is an RTL expression. 7320 7321@cindex @code{TARGET_ENCODE_SECTION_INFO} usage 7322On some machines, the syntax for a symbolic address depends on the 7323section that the address refers to. On these machines, define the hook 7324@code{TARGET_ENCODE_SECTION_INFO} to store the information into the 7325@code{symbol_ref}, and then check for it here. @xref{Assembler 7326Format}. 7327@end defmac 7328 7329@findex dbr_sequence_length 7330@defmac DBR_OUTPUT_SEQEND (@var{file}) 7331A C statement, to be executed after all slot-filler instructions have 7332been output. If necessary, call @code{dbr_sequence_length} to 7333determine the number of slots filled in a sequence (zero if not 7334currently outputting a sequence), to decide how many no-ops to output, 7335or whatever. 7336 7337Don't define this macro if it has nothing to do, but it is helpful in 7338reading assembly output if the extent of the delay sequence is made 7339explicit (e.g.@: with white space). 7340@end defmac 7341 7342@findex final_sequence 7343Note that output routines for instructions with delay slots must be 7344prepared to deal with not being output as part of a sequence 7345(i.e.@: when the scheduling pass is not run, or when no slot fillers could be 7346found.) The variable @code{final_sequence} is null when not 7347processing a sequence, otherwise it contains the @code{sequence} rtx 7348being output. 7349 7350@findex asm_fprintf 7351@defmac REGISTER_PREFIX 7352@defmacx LOCAL_LABEL_PREFIX 7353@defmacx USER_LABEL_PREFIX 7354@defmacx IMMEDIATE_PREFIX 7355If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, 7356@samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see 7357@file{final.c}). These are useful when a single @file{md} file must 7358support multiple assembler formats. In that case, the various @file{tm.h} 7359files can define these macros differently. 7360@end defmac 7361 7362@defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format}) 7363If defined this macro should expand to a series of @code{case} 7364statements which will be parsed inside the @code{switch} statement of 7365the @code{asm_fprintf} function. This allows targets to define extra 7366printf formats which may useful when generating their assembler 7367statements. Note that uppercase letters are reserved for future 7368generic extensions to asm_fprintf, and so are not available to target 7369specific code. The output file is given by the parameter @var{file}. 7370The varargs input pointer is @var{argptr} and the rest of the format 7371string, starting the character after the one that is being switched 7372upon, is pointed to by @var{format}. 7373@end defmac 7374 7375@defmac ASSEMBLER_DIALECT 7376If your target supports multiple dialects of assembler language (such as 7377different opcodes), define this macro as a C expression that gives the 7378numeric index of the assembler language dialect to use, with zero as the 7379first variant. 7380 7381If this macro is defined, you may use constructs of the form 7382@smallexample 7383@samp{@{option0|option1|option2@dots{}@}} 7384@end smallexample 7385@noindent 7386in the output templates of patterns (@pxref{Output Template}) or in the 7387first argument of @code{asm_fprintf}. This construct outputs 7388@samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of 7389@code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters 7390within these strings retain their usual meaning. If there are fewer 7391alternatives within the braces than the value of 7392@code{ASSEMBLER_DIALECT}, the construct outputs nothing. 7393 7394If you do not define this macro, the characters @samp{@{}, @samp{|} and 7395@samp{@}} do not have any special meaning when used in templates or 7396operands to @code{asm_fprintf}. 7397 7398Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, 7399@code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express 7400the variations in assembler language syntax with that mechanism. Define 7401@code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax 7402if the syntax variant are larger and involve such things as different 7403opcodes or operand order. 7404@end defmac 7405 7406@defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) 7407A C expression to output to @var{stream} some assembler code 7408which will push hard register number @var{regno} onto the stack. 7409The code need not be optimal, since this macro is used only when 7410profiling. 7411@end defmac 7412 7413@defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) 7414A C expression to output to @var{stream} some assembler code 7415which will pop hard register number @var{regno} off of the stack. 7416The code need not be optimal, since this macro is used only when 7417profiling. 7418@end defmac 7419 7420@node Dispatch Tables 7421@subsection Output of Dispatch Tables 7422 7423@c prevent bad page break with this line 7424This concerns dispatch tables. 7425 7426@cindex dispatch table 7427@defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel}) 7428A C statement to output to the stdio stream @var{stream} an assembler 7429pseudo-instruction to generate a difference between two labels. 7430@var{value} and @var{rel} are the numbers of two internal labels. The 7431definitions of these labels are output using 7432@code{(*targetm.asm_out.internal_label)}, and they must be printed in the same 7433way here. For example, 7434 7435@smallexample 7436fprintf (@var{stream}, "\t.word L%d-L%d\n", 7437 @var{value}, @var{rel}) 7438@end smallexample 7439 7440You must provide this macro on machines where the addresses in a 7441dispatch table are relative to the table's own address. If defined, GCC 7442will also use this macro on all machines when producing PIC@. 7443@var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the 7444mode and flags can be read. 7445@end defmac 7446 7447@defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) 7448This macro should be provided on machines where the addresses 7449in a dispatch table are absolute. 7450 7451The definition should be a C statement to output to the stdio stream 7452@var{stream} an assembler pseudo-instruction to generate a reference to 7453a label. @var{value} is the number of an internal label whose 7454definition is output using @code{(*targetm.asm_out.internal_label)}. 7455For example, 7456 7457@smallexample 7458fprintf (@var{stream}, "\t.word L%d\n", @var{value}) 7459@end smallexample 7460@end defmac 7461 7462@defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) 7463Define this if the label before a jump-table needs to be output 7464specially. The first three arguments are the same as for 7465@code{(*targetm.asm_out.internal_label)}; the fourth argument is the 7466jump-table which follows (a @code{jump_insn} containing an 7467@code{addr_vec} or @code{addr_diff_vec}). 7468 7469This feature is used on system V to output a @code{swbeg} statement 7470for the table. 7471 7472If this macro is not defined, these labels are output with 7473@code{(*targetm.asm_out.internal_label)}. 7474@end defmac 7475 7476@defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) 7477Define this if something special must be output at the end of a 7478jump-table. The definition should be a C statement to be executed 7479after the assembler code for the table is written. It should write 7480the appropriate code to stdio stream @var{stream}. The argument 7481@var{table} is the jump-table insn, and @var{num} is the label-number 7482of the preceding label. 7483 7484If this macro is not defined, nothing special is output at the end of 7485the jump-table. 7486@end defmac 7487 7488@node Exception Region Output 7489@subsection Assembler Commands for Exception Regions 7490 7491@c prevent bad page break with this line 7492 7493This describes commands marking the start and the end of an exception 7494region. 7495 7496@defmac EH_FRAME_SECTION_NAME 7497If defined, a C string constant for the name of the section containing 7498exception handling frame unwind information. If not defined, GCC will 7499provide a default definition if the target supports named sections. 7500@file{crtstuff.c} uses this macro to switch to the appropriate section. 7501 7502You should define this symbol if your target supports DWARF 2 frame 7503unwind information and the default definition does not work. 7504@end defmac 7505 7506@defmac EH_FRAME_IN_DATA_SECTION 7507If defined, DWARF 2 frame unwind information will be placed in the 7508data section even though the target supports named sections. This 7509might be necessary, for instance, if the system linker does garbage 7510collection and sections cannot be marked as not to be collected. 7511 7512Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is 7513also defined. 7514@end defmac 7515 7516@defmac MASK_RETURN_ADDR 7517An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so 7518that it does not contain any extraneous set bits in it. 7519@end defmac 7520 7521@defmac DWARF2_UNWIND_INFO 7522Define this macro to 0 if your target supports DWARF 2 frame unwind 7523information, but it does not yet work with exception handling. 7524Otherwise, if your target supports this information (if it defines 7525@samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP} 7526or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 75271. 7528 7529If this macro is defined to 1, the DWARF 2 unwinder will be the default 7530exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by 7531default. 7532 7533If this macro is defined to anything, the DWARF 2 unwinder will be used 7534instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case. 7535@end defmac 7536 7537@defmac MUST_USE_SJLJ_EXCEPTIONS 7538This macro need only be defined if @code{DWARF2_UNWIND_INFO} is 7539runtime-variable. In that case, @file{except.h} cannot correctly 7540determine the corresponding definition of 7541@code{MUST_USE_SJLJ_EXCEPTIONS}, so the target must provide it directly. 7542@end defmac 7543 7544@defmac DWARF_CIE_DATA_ALIGNMENT 7545This macro need only be defined if the target might save registers in the 7546function prologue at an offset to the stack pointer that is not aligned to 7547@code{UNITS_PER_WORD}. The definition should be the negative minimum 7548alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive 7549minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if 7550the target supports DWARF 2 frame unwind information. 7551@end defmac 7552 7553@deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION () 7554If defined, a function that switches to the section in which the main 7555exception table is to be placed (@pxref{Sections}). The default is a 7556function that switches to a section named @code{.gcc_except_table} on 7557machines that support named sections via 7558@code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or 7559@option{-fPIC} is in effect, the @code{data_section}, otherwise the 7560@code{readonly_data_section}. 7561@end deftypefn 7562 7563@deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION () 7564If defined, a function that switches to the section in which the DWARF 2 7565frame unwind information to be placed (@pxref{Sections}). The default 7566is a function that outputs a standard GAS section directive, if 7567@code{EH_FRAME_SECTION_NAME} is defined, or else a data section 7568directive followed by a synthetic label. 7569@end deftypefn 7570 7571@deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO 7572Contains the value true if the target should add a zero word onto the 7573end of a Dwarf-2 frame info section when used for exception handling. 7574Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and 7575true otherwise. 7576@end deftypevar 7577 7578@deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg}) 7579Given a register, this hook should return a parallel of registers to 7580represent where to find the register pieces. Define this hook if the 7581register and its mode are represented in Dwarf in non-contiguous 7582locations, or if the register should be represented in more than one 7583register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}. 7584If not defined, the default is to return @code{NULL_RTX}. 7585@end deftypefn 7586 7587@node Alignment Output 7588@subsection Assembler Commands for Alignment 7589 7590@c prevent bad page break with this line 7591This describes commands for alignment. 7592 7593@defmac JUMP_ALIGN (@var{label}) 7594The alignment (log base 2) to put in front of @var{label}, which is 7595a common destination of jumps and has no fallthru incoming edge. 7596 7597This macro need not be defined if you don't want any special alignment 7598to be done at such a time. Most machine descriptions do not currently 7599define the macro. 7600 7601Unless it's necessary to inspect the @var{label} parameter, it is better 7602to set the variable @var{align_jumps} in the target's 7603@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 7604selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation. 7605@end defmac 7606 7607@defmac LABEL_ALIGN_AFTER_BARRIER (@var{label}) 7608The alignment (log base 2) to put in front of @var{label}, which follows 7609a @code{BARRIER}. 7610 7611This macro need not be defined if you don't want any special alignment 7612to be done at such a time. Most machine descriptions do not currently 7613define the macro. 7614@end defmac 7615 7616@defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP 7617The maximum number of bytes to skip when applying 7618@code{LABEL_ALIGN_AFTER_BARRIER}. This works only if 7619@code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 7620@end defmac 7621 7622@defmac LOOP_ALIGN (@var{label}) 7623The alignment (log base 2) to put in front of @var{label}, which follows 7624a @code{NOTE_INSN_LOOP_BEG} note. 7625 7626This macro need not be defined if you don't want any special alignment 7627to be done at such a time. Most machine descriptions do not currently 7628define the macro. 7629 7630Unless it's necessary to inspect the @var{label} parameter, it is better 7631to set the variable @code{align_loops} in the target's 7632@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 7633selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation. 7634@end defmac 7635 7636@defmac LOOP_ALIGN_MAX_SKIP 7637The maximum number of bytes to skip when applying @code{LOOP_ALIGN}. 7638This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 7639@end defmac 7640 7641@defmac LABEL_ALIGN (@var{label}) 7642The alignment (log base 2) to put in front of @var{label}. 7643If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment, 7644the maximum of the specified values is used. 7645 7646Unless it's necessary to inspect the @var{label} parameter, it is better 7647to set the variable @code{align_labels} in the target's 7648@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 7649selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation. 7650@end defmac 7651 7652@defmac LABEL_ALIGN_MAX_SKIP 7653The maximum number of bytes to skip when applying @code{LABEL_ALIGN}. 7654This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 7655@end defmac 7656 7657@defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) 7658A C statement to output to the stdio stream @var{stream} an assembler 7659instruction to advance the location counter by @var{nbytes} bytes. 7660Those bytes should be zero when loaded. @var{nbytes} will be a C 7661expression of type @code{int}. 7662@end defmac 7663 7664@defmac ASM_NO_SKIP_IN_TEXT 7665Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the 7666text section because it fails to put zeros in the bytes that are skipped. 7667This is true on many Unix systems, where the pseudo--op to skip bytes 7668produces no-op instructions rather than zeros when used in the text 7669section. 7670@end defmac 7671 7672@defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) 7673A C statement to output to the stdio stream @var{stream} an assembler 7674command to advance the location counter to a multiple of 2 to the 7675@var{power} bytes. @var{power} will be a C expression of type @code{int}. 7676@end defmac 7677 7678@defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power}) 7679Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used 7680for padding, if necessary. 7681@end defmac 7682 7683@defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip}) 7684A C statement to output to the stdio stream @var{stream} an assembler 7685command to advance the location counter to a multiple of 2 to the 7686@var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to 7687satisfy the alignment request. @var{power} and @var{max_skip} will be 7688a C expression of type @code{int}. 7689@end defmac 7690 7691@need 3000 7692@node Debugging Info 7693@section Controlling Debugging Information Format 7694 7695@c prevent bad page break with this line 7696This describes how to specify debugging information. 7697 7698@menu 7699* All Debuggers:: Macros that affect all debugging formats uniformly. 7700* DBX Options:: Macros enabling specific options in DBX format. 7701* DBX Hooks:: Hook macros for varying DBX format. 7702* File Names and DBX:: Macros controlling output of file names in DBX format. 7703* SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. 7704* VMS Debug:: Macros for VMS debug format. 7705@end menu 7706 7707@node All Debuggers 7708@subsection Macros Affecting All Debugging Formats 7709 7710@c prevent bad page break with this line 7711These macros affect all debugging formats. 7712 7713@defmac DBX_REGISTER_NUMBER (@var{regno}) 7714A C expression that returns the DBX register number for the compiler 7715register number @var{regno}. In the default macro provided, the value 7716of this expression will be @var{regno} itself. But sometimes there are 7717some registers that the compiler knows about and DBX does not, or vice 7718versa. In such cases, some register may need to have one number in the 7719compiler and another for DBX@. 7720 7721If two registers have consecutive numbers inside GCC, and they can be 7722used as a pair to hold a multiword value, then they @emph{must} have 7723consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}. 7724Otherwise, debuggers will be unable to access such a pair, because they 7725expect register pairs to be consecutive in their own numbering scheme. 7726 7727If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that 7728does not preserve register pairs, then what you must do instead is 7729redefine the actual register numbering scheme. 7730@end defmac 7731 7732@defmac DEBUGGER_AUTO_OFFSET (@var{x}) 7733A C expression that returns the integer offset value for an automatic 7734variable having address @var{x} (an RTL expression). The default 7735computation assumes that @var{x} is based on the frame-pointer and 7736gives the offset from the frame-pointer. This is required for targets 7737that produce debugging output for DBX or COFF-style debugging output 7738for SDB and allow the frame-pointer to be eliminated when the 7739@option{-g} options is used. 7740@end defmac 7741 7742@defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) 7743A C expression that returns the integer offset value for an argument 7744having address @var{x} (an RTL expression). The nominal offset is 7745@var{offset}. 7746@end defmac 7747 7748@defmac PREFERRED_DEBUGGING_TYPE 7749A C expression that returns the type of debugging output GCC should 7750produce when the user specifies just @option{-g}. Define 7751this if you have arranged for GCC to support more than one format of 7752debugging output. Currently, the allowable values are @code{DBX_DEBUG}, 7753@code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, 7754@code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}. 7755 7756When the user specifies @option{-ggdb}, GCC normally also uses the 7757value of this macro to select the debugging output format, but with two 7758exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the 7759value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is 7760defined, GCC uses @code{DBX_DEBUG}. 7761 7762The value of this macro only affects the default debugging output; the 7763user can always get a specific type of output by using @option{-gstabs}, 7764@option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}. 7765@end defmac 7766 7767@node DBX Options 7768@subsection Specific Options for DBX Output 7769 7770@c prevent bad page break with this line 7771These are specific options for DBX output. 7772 7773@defmac DBX_DEBUGGING_INFO 7774Define this macro if GCC should produce debugging output for DBX 7775in response to the @option{-g} option. 7776@end defmac 7777 7778@defmac XCOFF_DEBUGGING_INFO 7779Define this macro if GCC should produce XCOFF format debugging output 7780in response to the @option{-g} option. This is a variant of DBX format. 7781@end defmac 7782 7783@defmac DEFAULT_GDB_EXTENSIONS 7784Define this macro to control whether GCC should by default generate 7785GDB's extended version of DBX debugging information (assuming DBX-format 7786debugging information is enabled at all). If you don't define the 7787macro, the default is 1: always generate the extended information 7788if there is any occasion to. 7789@end defmac 7790 7791@defmac DEBUG_SYMS_TEXT 7792Define this macro if all @code{.stabs} commands should be output while 7793in the text section. 7794@end defmac 7795 7796@defmac ASM_STABS_OP 7797A C string constant, including spacing, naming the assembler pseudo op to 7798use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol. 7799If you don't define this macro, @code{"\t.stabs\t"} is used. This macro 7800applies only to DBX debugging information format. 7801@end defmac 7802 7803@defmac ASM_STABD_OP 7804A C string constant, including spacing, naming the assembler pseudo op to 7805use instead of @code{"\t.stabd\t"} to define a debugging symbol whose 7806value is the current location. If you don't define this macro, 7807@code{"\t.stabd\t"} is used. This macro applies only to DBX debugging 7808information format. 7809@end defmac 7810 7811@defmac ASM_STABN_OP 7812A C string constant, including spacing, naming the assembler pseudo op to 7813use instead of @code{"\t.stabn\t"} to define a debugging symbol with no 7814name. If you don't define this macro, @code{"\t.stabn\t"} is used. This 7815macro applies only to DBX debugging information format. 7816@end defmac 7817 7818@defmac DBX_NO_XREFS 7819Define this macro if DBX on your system does not support the construct 7820@samp{xs@var{tagname}}. On some systems, this construct is used to 7821describe a forward reference to a structure named @var{tagname}. 7822On other systems, this construct is not supported at all. 7823@end defmac 7824 7825@defmac DBX_CONTIN_LENGTH 7826A symbol name in DBX-format debugging information is normally 7827continued (split into two separate @code{.stabs} directives) when it 7828exceeds a certain length (by default, 80 characters). On some 7829operating systems, DBX requires this splitting; on others, splitting 7830must not be done. You can inhibit splitting by defining this macro 7831with the value zero. You can override the default splitting-length by 7832defining this macro as an expression for the length you desire. 7833@end defmac 7834 7835@defmac DBX_CONTIN_CHAR 7836Normally continuation is indicated by adding a @samp{\} character to 7837the end of a @code{.stabs} string when a continuation follows. To use 7838a different character instead, define this macro as a character 7839constant for the character you want to use. Do not define this macro 7840if backslash is correct for your system. 7841@end defmac 7842 7843@defmac DBX_STATIC_STAB_DATA_SECTION 7844Define this macro if it is necessary to go to the data section before 7845outputting the @samp{.stabs} pseudo-op for a non-global static 7846variable. 7847@end defmac 7848 7849@defmac DBX_TYPE_DECL_STABS_CODE 7850The value to use in the ``code'' field of the @code{.stabs} directive 7851for a typedef. The default is @code{N_LSYM}. 7852@end defmac 7853 7854@defmac DBX_STATIC_CONST_VAR_CODE 7855The value to use in the ``code'' field of the @code{.stabs} directive 7856for a static variable located in the text section. DBX format does not 7857provide any ``right'' way to do this. The default is @code{N_FUN}. 7858@end defmac 7859 7860@defmac DBX_REGPARM_STABS_CODE 7861The value to use in the ``code'' field of the @code{.stabs} directive 7862for a parameter passed in registers. DBX format does not provide any 7863``right'' way to do this. The default is @code{N_RSYM}. 7864@end defmac 7865 7866@defmac DBX_REGPARM_STABS_LETTER 7867The letter to use in DBX symbol data to identify a symbol as a parameter 7868passed in registers. DBX format does not customarily provide any way to 7869do this. The default is @code{'P'}. 7870@end defmac 7871 7872@defmac DBX_MEMPARM_STABS_LETTER 7873The letter to use in DBX symbol data to identify a symbol as a stack 7874parameter. The default is @code{'p'}. 7875@end defmac 7876 7877@defmac DBX_FUNCTION_FIRST 7878Define this macro if the DBX information for a function and its 7879arguments should precede the assembler code for the function. Normally, 7880in DBX format, the debugging information entirely follows the assembler 7881code. 7882@end defmac 7883 7884@defmac DBX_BLOCKS_FUNCTION_RELATIVE 7885Define this macro if the value of a symbol describing the scope of a 7886block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start 7887of the enclosing function. Normally, GCC uses an absolute address. 7888@end defmac 7889 7890@defmac DBX_USE_BINCL 7891Define this macro if GCC should generate @code{N_BINCL} and 7892@code{N_EINCL} stabs for included header files, as on Sun systems. This 7893macro also directs GCC to output a type number as a pair of a file 7894number and a type number within the file. Normally, GCC does not 7895generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single 7896number for a type number. 7897@end defmac 7898 7899@node DBX Hooks 7900@subsection Open-Ended Hooks for DBX Format 7901 7902@c prevent bad page break with this line 7903These are hooks for DBX format. 7904 7905@defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name}) 7906Define this macro to say how to output to @var{stream} the debugging 7907information for the start of a scope level for variable names. The 7908argument @var{name} is the name of an assembler symbol (for use with 7909@code{assemble_name}) whose value is the address where the scope begins. 7910@end defmac 7911 7912@defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name}) 7913Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level. 7914@end defmac 7915 7916@defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl}) 7917Define this macro if the target machine requires special handling to 7918output an @code{N_FUN} entry for the function @var{decl}. 7919@end defmac 7920 7921@defmac DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function}) 7922Define this macro if the target machine requires special output at the 7923end of the debugging information for a function. The definition should 7924be a C statement (sans semicolon) to output the appropriate information 7925to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for 7926the function. 7927@end defmac 7928 7929@defmac DBX_OUTPUT_STANDARD_TYPES (@var{syms}) 7930Define this macro if you need to control the order of output of the 7931standard data types at the beginning of compilation. The argument 7932@var{syms} is a @code{tree} which is a chain of all the predefined 7933global symbols, including names of data types. 7934 7935Normally, DBX output starts with definitions of the types for integers 7936and characters, followed by all the other predefined types of the 7937particular language in no particular order. 7938 7939On some machines, it is necessary to output different particular types 7940first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output 7941those symbols in the necessary order. Any predefined types that you 7942don't explicitly output will be output afterward in no particular order. 7943 7944Be careful not to define this macro so that it works only for C@. There 7945are no global variables to access most of the built-in types, because 7946another language may have another set of types. The way to output a 7947particular type is to look through @var{syms} to see if you can find it. 7948Here is an example: 7949 7950@smallexample 7951@{ 7952 tree decl; 7953 for (decl = syms; decl; decl = TREE_CHAIN (decl)) 7954 if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)), 7955 "long int")) 7956 dbxout_symbol (decl); 7957 @dots{} 7958@} 7959@end smallexample 7960 7961@noindent 7962This does nothing if the expected type does not exist. 7963 7964See the function @code{init_decl_processing} in @file{c-decl.c} to find 7965the names to use for all the built-in C types. 7966 7967Here is another way of finding a particular type: 7968 7969@c this is still overfull. --mew 10feb93 7970@smallexample 7971@{ 7972 tree decl; 7973 for (decl = syms; decl; decl = TREE_CHAIN (decl)) 7974 if (TREE_CODE (decl) == TYPE_DECL 7975 && (TREE_CODE (TREE_TYPE (decl)) 7976 == INTEGER_CST) 7977 && TYPE_PRECISION (TREE_TYPE (decl)) == 16 7978 && TYPE_UNSIGNED (TREE_TYPE (decl))) 7979@group 7980 /* @r{This must be @code{unsigned short}.} */ 7981 dbxout_symbol (decl); 7982 @dots{} 7983@} 7984@end group 7985@end smallexample 7986@end defmac 7987 7988@defmac NO_DBX_FUNCTION_END 7989Some stabs encapsulation formats (in particular ECOFF), cannot handle the 7990@code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct. 7991On those machines, define this macro to turn this feature off without 7992disturbing the rest of the gdb extensions. 7993@end defmac 7994 7995@node File Names and DBX 7996@subsection File Names in DBX Format 7997 7998@c prevent bad page break with this line 7999This describes file names in DBX format. 8000 8001@defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name}) 8002A C statement to output DBX debugging information to the stdio stream 8003@var{stream} which indicates that file @var{name} is the main source 8004file---the file specified as the input file for compilation. 8005This macro is called only once, at the beginning of compilation. 8006 8007This macro need not be defined if the standard form of output 8008for DBX debugging information is appropriate. 8009@end defmac 8010 8011@defmac DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name}) 8012A C statement to output DBX debugging information to the stdio stream 8013@var{stream} which indicates that the current directory during 8014compilation is named @var{name}. 8015 8016This macro need not be defined if the standard form of output 8017for DBX debugging information is appropriate. 8018@end defmac 8019 8020@defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name}) 8021A C statement to output DBX debugging information at the end of 8022compilation of the main source file @var{name}. 8023 8024If you don't define this macro, nothing special is output at the end 8025of compilation, which is correct for most machines. 8026@end defmac 8027 8028@need 2000 8029@node SDB and DWARF 8030@subsection Macros for SDB and DWARF Output 8031 8032@c prevent bad page break with this line 8033Here are macros for SDB and DWARF output. 8034 8035@defmac SDB_DEBUGGING_INFO 8036Define this macro if GCC should produce COFF-style debugging output 8037for SDB in response to the @option{-g} option. 8038@end defmac 8039 8040@defmac DWARF2_DEBUGGING_INFO 8041Define this macro if GCC should produce dwarf version 2 format 8042debugging output in response to the @option{-g} option. 8043 8044To support optional call frame debugging information, you must also 8045define @code{INCOMING_RETURN_ADDR_RTX} and either set 8046@code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the 8047prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save} 8048as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't. 8049@end defmac 8050 8051@defmac DWARF2_FRAME_INFO 8052Define this macro to a nonzero value if GCC should always output 8053Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO} 8054(@pxref{Exception Region Output} is nonzero, GCC will output this 8055information not matter how you define @code{DWARF2_FRAME_INFO}. 8056@end defmac 8057 8058@defmac DWARF2_GENERATE_TEXT_SECTION_LABEL 8059By default, the Dwarf 2 debugging information generator will generate a 8060label to mark the beginning of the text section. If it is better simply 8061to use the name of the text section itself, rather than an explicit label, 8062to indicate the beginning of the text section, define this macro to zero. 8063@end defmac 8064 8065@defmac DWARF2_ASM_LINE_DEBUG_INFO 8066Define this macro to be a nonzero value if the assembler can generate Dwarf 2 8067line debug info sections. This will result in much more compact line number 8068tables, and hence is desirable if it works. 8069@end defmac 8070 8071@defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) 8072A C statement to issue assembly directives that create a difference 8073between the two given labels, using an integer of the given size. 8074@end defmac 8075 8076@defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}) 8077A C statement to issue assembly directives that create a 8078section-relative reference to the given label, using an integer of the 8079given size. 8080@end defmac 8081 8082@defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label}) 8083A C statement to issue assembly directives that create a self-relative 8084reference to the given label, using an integer of the given size. 8085@end defmac 8086 8087@defmac PUT_SDB_@dots{} 8088Define these macros to override the assembler syntax for the special 8089SDB assembler directives. See @file{sdbout.c} for a list of these 8090macros and their arguments. If the standard syntax is used, you need 8091not define them yourself. 8092@end defmac 8093 8094@defmac SDB_DELIM 8095Some assemblers do not support a semicolon as a delimiter, even between 8096SDB assembler directives. In that case, define this macro to be the 8097delimiter to use (usually @samp{\n}). It is not necessary to define 8098a new set of @code{PUT_SDB_@var{op}} macros if this is the only change 8099required. 8100@end defmac 8101 8102@defmac SDB_GENERATE_FAKE 8103Define this macro to override the usual method of constructing a dummy 8104name for anonymous structure and union types. See @file{sdbout.c} for 8105more information. 8106@end defmac 8107 8108@defmac SDB_ALLOW_UNKNOWN_REFERENCES 8109Define this macro to allow references to unknown structure, 8110union, or enumeration tags to be emitted. Standard COFF does not 8111allow handling of unknown references, MIPS ECOFF has support for 8112it. 8113@end defmac 8114 8115@defmac SDB_ALLOW_FORWARD_REFERENCES 8116Define this macro to allow references to structure, union, or 8117enumeration tags that have not yet been seen to be handled. Some 8118assemblers choke if forward tags are used, while some require it. 8119@end defmac 8120 8121@need 2000 8122@node VMS Debug 8123@subsection Macros for VMS Debug Format 8124 8125@c prevent bad page break with this line 8126Here are macros for VMS debug format. 8127 8128@defmac VMS_DEBUGGING_INFO 8129Define this macro if GCC should produce debugging output for VMS 8130in response to the @option{-g} option. The default behavior for VMS 8131is to generate minimal debug info for a traceback in the absence of 8132@option{-g} unless explicitly overridden with @option{-g0}. This 8133behavior is controlled by @code{OPTIMIZATION_OPTIONS} and 8134@code{OVERRIDE_OPTIONS}. 8135@end defmac 8136 8137@node Floating Point 8138@section Cross Compilation and Floating Point 8139@cindex cross compilation and floating point 8140@cindex floating point and cross compilation 8141 8142While all modern machines use twos-complement representation for integers, 8143there are a variety of representations for floating point numbers. This 8144means that in a cross-compiler the representation of floating point numbers 8145in the compiled program may be different from that used in the machine 8146doing the compilation. 8147 8148Because different representation systems may offer different amounts of 8149range and precision, all floating point constants must be represented in 8150the target machine's format. Therefore, the cross compiler cannot 8151safely use the host machine's floating point arithmetic; it must emulate 8152the target's arithmetic. To ensure consistency, GCC always uses 8153emulation to work with floating point values, even when the host and 8154target floating point formats are identical. 8155 8156The following macros are provided by @file{real.h} for the compiler to 8157use. All parts of the compiler which generate or optimize 8158floating-point calculations must use these macros. They may evaluate 8159their operands more than once, so operands must not have side effects. 8160 8161@defmac REAL_VALUE_TYPE 8162The C data type to be used to hold a floating point value in the target 8163machine's format. Typically this is a @code{struct} containing an 8164array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque 8165quantity. 8166@end defmac 8167 8168@deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 8169Compares for equality the two values, @var{x} and @var{y}. If the target 8170floating point format supports negative zeroes and/or NaNs, 8171@samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and 8172@samp{REAL_VALUES_EQUAL (NaN, NaN)} is false. 8173@end deftypefn 8174 8175@deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 8176Tests whether @var{x} is less than @var{y}. 8177@end deftypefn 8178 8179@deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x}) 8180Truncates @var{x} to a signed integer, rounding toward zero. 8181@end deftypefn 8182 8183@deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x}) 8184Truncates @var{x} to an unsigned integer, rounding toward zero. If 8185@var{x} is negative, returns zero. 8186@end deftypefn 8187 8188@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode}) 8189Converts @var{string} into a floating point number in the target machine's 8190representation for mode @var{mode}. This routine can handle both 8191decimal and hexadecimal floating point constants, using the syntax 8192defined by the C language for both. 8193@end deftypefn 8194 8195@deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x}) 8196Returns 1 if @var{x} is negative (including negative zero), 0 otherwise. 8197@end deftypefn 8198 8199@deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x}) 8200Determines whether @var{x} represents infinity (positive or negative). 8201@end deftypefn 8202 8203@deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x}) 8204Determines whether @var{x} represents a ``NaN'' (not-a-number). 8205@end deftypefn 8206 8207@deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 8208Calculates an arithmetic operation on the two floating point values 8209@var{x} and @var{y}, storing the result in @var{output} (which must be a 8210variable). 8211 8212The operation to be performed is specified by @var{code}. Only the 8213following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR}, 8214@code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}. 8215 8216If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the 8217target's floating point format cannot represent infinity, it will call 8218@code{abort}. Callers should check for this situation first, using 8219@code{MODE_HAS_INFINITIES}. @xref{Storage Layout}. 8220@end deftypefn 8221 8222@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x}) 8223Returns the negative of the floating point value @var{x}. 8224@end deftypefn 8225 8226@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x}) 8227Returns the absolute value of @var{x}. 8228@end deftypefn 8229 8230@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x}) 8231Truncates the floating point value @var{x} to fit in @var{mode}. The 8232return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an 8233appropriate bit pattern to be output asa floating constant whose 8234precision accords with mode @var{mode}. 8235@end deftypefn 8236 8237@deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x}) 8238Converts a floating point value @var{x} into a double-precision integer 8239which is then stored into @var{low} and @var{high}. If the value is not 8240integral, it is truncated. 8241@end deftypefn 8242 8243@deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode}) 8244Converts a double-precision integer found in @var{low} and @var{high}, 8245into a floating point value which is then stored into @var{x}. The 8246value is truncated to fit in mode @var{mode}. 8247@end deftypefn 8248 8249@node Mode Switching 8250@section Mode Switching Instructions 8251@cindex mode switching 8252The following macros control mode switching optimizations: 8253 8254@defmac OPTIMIZE_MODE_SWITCHING (@var{entity}) 8255Define this macro if the port needs extra instructions inserted for mode 8256switching in an optimizing compilation. 8257 8258For an example, the SH4 can perform both single and double precision 8259floating point operations, but to perform a single precision operation, 8260the FPSCR PR bit has to be cleared, while for a double precision 8261operation, this bit has to be set. Changing the PR bit requires a general 8262purpose register as a scratch register, hence these FPSCR sets have to 8263be inserted before reload, i.e.@: you can't put this into instruction emitting 8264or @code{TARGET_MACHINE_DEPENDENT_REORG}. 8265 8266You can have multiple entities that are mode-switched, and select at run time 8267which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should 8268return nonzero for any @var{entity} that needs mode-switching. 8269If you define this macro, you also have to define 8270@code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED}, 8271@code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}. 8272@code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT} 8273are optional. 8274@end defmac 8275 8276@defmac NUM_MODES_FOR_MODE_SWITCHING 8277If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as 8278initializer for an array of integers. Each initializer element 8279N refers to an entity that needs mode switching, and specifies the number 8280of different modes that might need to be set for this entity. 8281The position of the initializer in the initializer - starting counting at 8282zero - determines the integer that is used to refer to the mode-switched 8283entity in question. 8284In macros that take mode arguments / yield a mode result, modes are 8285represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode 8286switch is needed / supplied. 8287@end defmac 8288 8289@defmac MODE_NEEDED (@var{entity}, @var{insn}) 8290@var{entity} is an integer specifying a mode-switched entity. If 8291@code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to 8292return an integer value not larger than the corresponding element in 8293@code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must 8294be switched into prior to the execution of @var{insn}. 8295@end defmac 8296 8297@defmac MODE_AFTER (@var{mode}, @var{insn}) 8298If this macro is defined, it is evaluated for every @var{insn} during 8299mode switching. It determines the mode that an insn results in (if 8300different from the incoming mode). 8301@end defmac 8302 8303@defmac MODE_ENTRY (@var{entity}) 8304If this macro is defined, it is evaluated for every @var{entity} that needs 8305mode switching. It should evaluate to an integer, which is a mode that 8306@var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY} 8307is defined then @code{MODE_EXIT} must be defined. 8308@end defmac 8309 8310@defmac MODE_EXIT (@var{entity}) 8311If this macro is defined, it is evaluated for every @var{entity} that needs 8312mode switching. It should evaluate to an integer, which is a mode that 8313@var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT} 8314is defined then @code{MODE_ENTRY} must be defined. 8315@end defmac 8316 8317@defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n}) 8318This macro specifies the order in which modes for @var{entity} are processed. 83190 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the 8320lowest. The value of the macro should be an integer designating a mode 8321for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode} 8322(@var{entity}, @var{n}) shall be a bijection in 0 @dots{} 8323@code{num_modes_for_mode_switching[@var{entity}] - 1}. 8324@end defmac 8325 8326@defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live}) 8327Generate one or more insns to set @var{entity} to @var{mode}. 8328@var{hard_reg_live} is the set of hard registers live at the point where 8329the insn(s) are to be inserted. 8330@end defmac 8331 8332@node Target Attributes 8333@section Defining target-specific uses of @code{__attribute__} 8334@cindex target attributes 8335@cindex machine attributes 8336@cindex attributes, target-specific 8337 8338Target-specific attributes may be defined for functions, data and types. 8339These are described using the following target hooks; they also need to 8340be documented in @file{extend.texi}. 8341 8342@deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE 8343If defined, this target hook points to an array of @samp{struct 8344attribute_spec} (defined in @file{tree.h}) specifying the machine 8345specific attributes for this target and some of the restrictions on the 8346entities to which these attributes are applied and the arguments they 8347take. 8348@end deftypevr 8349 8350@deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) 8351If defined, this target hook is a function which returns zero if the attributes on 8352@var{type1} and @var{type2} are incompatible, one if they are compatible, 8353and two if they are nearly compatible (which causes a warning to be 8354generated). If this is not defined, machine-specific attributes are 8355supposed always to be compatible. 8356@end deftypefn 8357 8358@deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type}) 8359If defined, this target hook is a function which assigns default attributes to 8360newly defined @var{type}. 8361@end deftypefn 8362 8363@deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) 8364Define this target hook if the merging of type attributes needs special 8365handling. If defined, the result is a list of the combined 8366@code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed 8367that @code{comptypes} has already been called and returned 1. This 8368function may call @code{merge_attributes} to handle machine-independent 8369merging. 8370@end deftypefn 8371 8372@deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl}) 8373Define this target hook if the merging of decl attributes needs special 8374handling. If defined, the result is a list of the combined 8375@code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}. 8376@var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of 8377when this is needed are when one attribute overrides another, or when an 8378attribute is nullified by a subsequent definition. This function may 8379call @code{merge_attributes} to handle machine-independent merging. 8380 8381@findex TARGET_DLLIMPORT_DECL_ATTRIBUTES 8382If the only target-specific handling you require is @samp{dllimport} for 8383Microsoft Windows targets, you should define the macro 8384@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function 8385called @code{merge_dllimport_decl_attributes} which can then be defined 8386as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done 8387in @file{i386/cygwin.h} and @file{i386/i386.c}, for example. 8388@end deftypefn 8389 8390@deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr}) 8391Define this target hook if you want to be able to add attributes to a decl 8392when it is being created. This is normally useful for back ends which 8393wish to implement a pragma by using the attributes which correspond to 8394the pragma's effect. The @var{node} argument is the decl which is being 8395created. The @var{attr_ptr} argument is a pointer to the attribute list 8396for this decl. The list itself should not be modified, since it may be 8397shared with other decls, but attributes may be chained on the head of 8398the list and @code{*@var{attr_ptr}} modified to point to the new 8399attributes, or a copy of the list may be made if further changes are 8400needed. 8401@end deftypefn 8402 8403@deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl}) 8404@cindex inlining 8405This target hook returns @code{true} if it is ok to inline @var{fndecl} 8406into the current function, despite its having target-specific 8407attributes, @code{false} otherwise. By default, if a function has a 8408target specific attribute attached to it, it will not be inlined. 8409@end deftypefn 8410 8411@node MIPS Coprocessors 8412@section Defining coprocessor specifics for MIPS targets. 8413@cindex MIPS coprocessor-definition macros 8414 8415The MIPS specification allows MIPS implementations to have as many as 4 8416coprocessors, each with as many as 32 private registers. GCC supports 8417accessing these registers and transferring values between the registers 8418and memory using asm-ized variables. For example: 8419 8420@smallexample 8421 register unsigned int cp0count asm ("c0r1"); 8422 unsigned int d; 8423 8424 d = cp0count + 3; 8425@end smallexample 8426 8427(``c0r1'' is the default name of register 1 in coprocessor 0; alternate 8428names may be added as described below, or the default names may be 8429overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.) 8430 8431Coprocessor registers are assumed to be epilogue-used; sets to them will 8432be preserved even if it does not appear that the register is used again 8433later in the function. 8434 8435Another note: according to the MIPS spec, coprocessor 1 (if present) is 8436the FPU. One accesses COP1 registers through standard mips 8437floating-point support; they are not included in this mechanism. 8438 8439There is one macro used in defining the MIPS coprocessor interface which 8440you may want to override in subtargets; it is described below. 8441 8442@defmac ALL_COP_ADDITIONAL_REGISTER_NAMES 8443A comma-separated list (with leading comma) of pairs describing the 8444alternate names of coprocessor registers. The format of each entry should be 8445@smallexample 8446@{ @var{alternatename}, @var{register_number}@} 8447@end smallexample 8448Default: empty. 8449@end defmac 8450 8451@node PCH Target 8452@section Parameters for Precompiled Header Validity Checking 8453@cindex parameters, precompiled headers 8454 8455@deftypefn {Target Hook} void * TARGET_GET_PCH_VALIDITY (size_t * @var{sz}) 8456Define this hook if your target needs to check a different collection 8457of flags than the default, which is every flag defined by 8458@code{TARGET_SWITCHES} and @code{TARGET_OPTIONS}. It should return 8459some data which will be saved in the PCH file and presented to 8460@code{TARGET_PCH_VALID_P} later; it should set @code{SZ} to the size 8461of the data. 8462@end deftypefn 8463 8464@deftypefn {Target Hook} const char * TARGET_PCH_VALID_P (const void * @var{data}, size_t @var{sz}) 8465Define this hook if your target needs to check a different collection of 8466flags than the default, which is every flag defined by @code{TARGET_SWITCHES} 8467and @code{TARGET_OPTIONS}. It is given data which came from 8468@code{TARGET_GET_PCH_VALIDITY} (in this version of this compiler, so there 8469is no need for extensive validity checking). It returns @code{NULL} if 8470it is safe to load a PCH file with this data, or a suitable error message 8471if not. The error message will be presented to the user, so it should 8472be localized. 8473@end deftypefn 8474 8475@node Misc 8476@section Miscellaneous Parameters 8477@cindex parameters, miscellaneous 8478 8479@c prevent bad page break with this line 8480Here are several miscellaneous parameters. 8481 8482@defmac PREDICATE_CODES 8483Define this if you have defined special-purpose predicates in the file 8484@file{@var{machine}.c}. This macro is called within an initializer of an 8485array of structures. The first field in the structure is the name of a 8486predicate and the second field is an array of rtl codes. For each 8487predicate, list all rtl codes that can be in expressions matched by the 8488predicate. The list should have a trailing comma. Here is an example 8489of two entries in the list for a typical RISC machine: 8490 8491@smallexample 8492#define PREDICATE_CODES \ 8493 @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \ 8494 @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@}, 8495@end smallexample 8496 8497Defining this macro does not affect the generated code (however, 8498incorrect definitions that omit an rtl code that may be matched by the 8499predicate can cause the compiler to malfunction). Instead, it allows 8500the table built by @file{genrecog} to be more compact and efficient, 8501thus speeding up the compiler. The most important predicates to include 8502in the list specified by this macro are those used in the most insn 8503patterns. 8504 8505For each predicate function named in @code{PREDICATE_CODES}, a 8506declaration will be generated in @file{insn-codes.h}. 8507@end defmac 8508 8509@defmac SPECIAL_MODE_PREDICATES 8510Define this if you have special predicates that know special things 8511about modes. Genrecog will warn about certain forms of 8512@code{match_operand} without a mode; if the operand predicate is 8513listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be 8514suppressed. 8515 8516Here is an example from the IA-32 port (@code{ext_register_operand} 8517specially checks for @code{HImode} or @code{SImode} in preparation 8518for a byte extraction from @code{%ah} etc.). 8519 8520@smallexample 8521#define SPECIAL_MODE_PREDICATES \ 8522 "ext_register_operand", 8523@end smallexample 8524@end defmac 8525 8526@defmac CASE_VECTOR_MODE 8527An alias for a machine mode name. This is the machine mode that 8528elements of a jump-table should have. 8529@end defmac 8530 8531@defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body}) 8532Optional: return the preferred mode for an @code{addr_diff_vec} 8533when the minimum and maximum offset are known. If you define this, 8534it enables extra code in branch shortening to deal with @code{addr_diff_vec}. 8535To make this work, you also have to define @code{INSN_ALIGN} and 8536make the alignment for @code{addr_diff_vec} explicit. 8537The @var{body} argument is provided so that the offset_unsigned and scale 8538flags can be updated. 8539@end defmac 8540 8541@defmac CASE_VECTOR_PC_RELATIVE 8542Define this macro to be a C expression to indicate when jump-tables 8543should contain relative addresses. You need not define this macro if 8544jump-tables never contain relative addresses, or jump-tables should 8545contain relative addresses only when @option{-fPIC} or @option{-fPIC} 8546is in effect. 8547@end defmac 8548 8549@defmac CASE_DROPS_THROUGH 8550Define this if control falls through a @code{case} insn when the index 8551value is out of range. This means the specified default-label is 8552actually ignored by the @code{case} insn proper. 8553@end defmac 8554 8555@defmac CASE_VALUES_THRESHOLD 8556Define this to be the smallest number of different values for which it 8557is best to use a jump-table instead of a tree of conditional branches. 8558The default is four for machines with a @code{casesi} instruction and 8559five otherwise. This is best for most machines. 8560@end defmac 8561 8562@defmac CASE_USE_BIT_TESTS 8563Define this macro to be a C expression to indicate whether C switch 8564statements may be implemented by a sequence of bit tests. This is 8565advantageous on processors that can efficiently implement left shift 8566of 1 by the number of bits held in a register, but inappropriate on 8567targets that would require a loop. By default, this macro returns 8568@code{true} if the target defines an @code{ashlsi3} pattern, and 8569@code{false} otherwise. 8570@end defmac 8571 8572@defmac WORD_REGISTER_OPERATIONS 8573Define this macro if operations between registers with integral mode 8574smaller than a word are always performed on the entire register. 8575Most RISC machines have this property and most CISC machines do not. 8576@end defmac 8577 8578@defmac LOAD_EXTEND_OP (@var{mem_mode}) 8579Define this macro to be a C expression indicating when insns that read 8580memory in @var{mem_mode}, an integral mode narrower than a word, set the 8581bits outside of @var{mem_mode} to be either the sign-extension or the 8582zero-extension of the data read. Return @code{SIGN_EXTEND} for values 8583of @var{mem_mode} for which the 8584insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and 8585@code{NIL} for other modes. 8586 8587This macro is not called with @var{mem_mode} non-integral or with a width 8588greater than or equal to @code{BITS_PER_WORD}, so you may return any 8589value in this case. Do not define this macro if it would always return 8590@code{NIL}. On machines where this macro is defined, you will normally 8591define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. 8592 8593You may return a non-@code{NIL} value even if for some hard registers 8594the sign extension is not performed, if for the @code{REGNO_REG_CLASS} 8595of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero 8596when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any 8597integral mode larger than this but not larger than @code{word_mode}. 8598 8599You must return @code{NIL} if for some hard registers that allow this 8600mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to 8601@code{word_mode}, but that they can change to another integral mode that 8602is larger then @var{mem_mode} but still smaller than @code{word_mode}. 8603@end defmac 8604 8605@defmac SHORT_IMMEDIATES_SIGN_EXTEND 8606Define this macro if loading short immediate values into registers sign 8607extends. 8608@end defmac 8609 8610@defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC 8611Define this macro if the same instructions that convert a floating 8612point number to a signed fixed point number also convert validly to an 8613unsigned one. 8614@end defmac 8615 8616@defmac MOVE_MAX 8617The maximum number of bytes that a single instruction can move quickly 8618between memory and registers or between two memory locations. 8619@end defmac 8620 8621@defmac MAX_MOVE_MAX 8622The maximum number of bytes that a single instruction can move quickly 8623between memory and registers or between two memory locations. If this 8624is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the 8625constant value that is the largest value that @code{MOVE_MAX} can have 8626at run-time. 8627@end defmac 8628 8629@defmac SHIFT_COUNT_TRUNCATED 8630A C expression that is nonzero if on this machine the number of bits 8631actually used for the count of a shift operation is equal to the number 8632of bits needed to represent the size of the object being shifted. When 8633this macro is nonzero, the compiler will assume that it is safe to omit 8634a sign-extend, zero-extend, and certain bitwise `and' instructions that 8635truncates the count of a shift operation. On machines that have 8636instructions that act on bit-fields at variable positions, which may 8637include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} 8638also enables deletion of truncations of the values that serve as 8639arguments to bit-field instructions. 8640 8641If both types of instructions truncate the count (for shifts) and 8642position (for bit-field operations), or if no variable-position bit-field 8643instructions exist, you should define this macro. 8644 8645However, on some machines, such as the 80386 and the 680x0, truncation 8646only applies to shift operations and not the (real or pretended) 8647bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on 8648such machines. Instead, add patterns to the @file{md} file that include 8649the implied truncation of the shift instructions. 8650 8651You need not define this macro if it would always have the value of zero. 8652@end defmac 8653 8654@defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec}) 8655A C expression which is nonzero if on this machine it is safe to 8656``convert'' an integer of @var{inprec} bits to one of @var{outprec} 8657bits (where @var{outprec} is smaller than @var{inprec}) by merely 8658operating on it as if it had only @var{outprec} bits. 8659 8660On many machines, this expression can be 1. 8661 8662@c rearranged this, removed the phrase "it is reported that". this was 8663@c to fix an overfull hbox. --mew 10feb93 8664When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for 8665modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result. 8666If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in 8667such cases may improve things. 8668@end defmac 8669 8670@defmac STORE_FLAG_VALUE 8671A C expression describing the value returned by a comparison operator 8672with an integral mode and stored by a store-flag instruction 8673(@samp{s@var{cond}}) when the condition is true. This description must 8674apply to @emph{all} the @samp{s@var{cond}} patterns and all the 8675comparison operators whose results have a @code{MODE_INT} mode. 8676 8677A value of 1 or @minus{}1 means that the instruction implementing the 8678comparison operator returns exactly 1 or @minus{}1 when the comparison is true 8679and 0 when the comparison is false. Otherwise, the value indicates 8680which bits of the result are guaranteed to be 1 when the comparison is 8681true. This value is interpreted in the mode of the comparison 8682operation, which is given by the mode of the first operand in the 8683@samp{s@var{cond}} pattern. Either the low bit or the sign bit of 8684@code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by 8685the compiler. 8686 8687If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will 8688generate code that depends only on the specified bits. It can also 8689replace comparison operators with equivalent operations if they cause 8690the required bits to be set, even if the remaining bits are undefined. 8691For example, on a machine whose comparison operators return an 8692@code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as 8693@samp{0x80000000}, saying that just the sign bit is relevant, the 8694expression 8695 8696@smallexample 8697(ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) 8698@end smallexample 8699 8700@noindent 8701can be converted to 8702 8703@smallexample 8704(ashift:SI @var{x} (const_int @var{n})) 8705@end smallexample 8706 8707@noindent 8708where @var{n} is the appropriate shift count to move the bit being 8709tested into the sign bit. 8710 8711There is no way to describe a machine that always sets the low-order bit 8712for a true value, but does not guarantee the value of any other bits, 8713but we do not know of any machine that has such an instruction. If you 8714are trying to port GCC to such a machine, include an instruction to 8715perform a logical-and of the result with 1 in the pattern for the 8716comparison operators and let us know at @email{gcc@@gcc.gnu.org}. 8717 8718Often, a machine will have multiple instructions that obtain a value 8719from a comparison (or the condition codes). Here are rules to guide the 8720choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions 8721to be used: 8722 8723@itemize @bullet 8724@item 8725Use the shortest sequence that yields a valid definition for 8726@code{STORE_FLAG_VALUE}. It is more efficient for the compiler to 8727``normalize'' the value (convert it to, e.g., 1 or 0) than for the 8728comparison operators to do so because there may be opportunities to 8729combine the normalization with other operations. 8730 8731@item 8732For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being 8733slightly preferred on machines with expensive jumps and 1 preferred on 8734other machines. 8735 8736@item 8737As a second choice, choose a value of @samp{0x80000001} if instructions 8738exist that set both the sign and low-order bits but do not define the 8739others. 8740 8741@item 8742Otherwise, use a value of @samp{0x80000000}. 8743@end itemize 8744 8745Many machines can produce both the value chosen for 8746@code{STORE_FLAG_VALUE} and its negation in the same number of 8747instructions. On those machines, you should also define a pattern for 8748those cases, e.g., one matching 8749 8750@smallexample 8751(set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) 8752@end smallexample 8753 8754Some machines can also perform @code{and} or @code{plus} operations on 8755condition code values with less instructions than the corresponding 8756@samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those 8757machines, define the appropriate patterns. Use the names @code{incscc} 8758and @code{decscc}, respectively, for the patterns which perform 8759@code{plus} or @code{minus} operations on condition code values. See 8760@file{rs6000.md} for some examples. The GNU Superoptizer can be used to 8761find such instruction sequences on other machines. 8762 8763If this macro is not defined, the default value, 1, is used. You need 8764not define @code{STORE_FLAG_VALUE} if the machine has no store-flag 8765instructions, or if the value generated by these instructions is 1. 8766@end defmac 8767 8768@defmac FLOAT_STORE_FLAG_VALUE (@var{mode}) 8769A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is 8770returned when comparison operators with floating-point results are true. 8771Define this macro on machine that have comparison operations that return 8772floating-point values. If there are no such operations, do not define 8773this macro. 8774@end defmac 8775 8776@defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 8777@defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 8778A C expression that evaluates to true if the architecture defines a value 8779for @code{clz} or @code{ctz} with a zero operand. If so, @var{value} 8780should be set to this value. If this macro is not defined, the value of 8781@code{clz} or @code{ctz} is assumed to be undefined. 8782 8783This macro must be defined if the target's expansion for @code{ffs} 8784relies on a particular value to get correct results. Otherwise it 8785is not necessary, though it may be used to optimize some corner cases. 8786 8787Note that regardless of this macro the ``definedness'' of @code{clz} 8788and @code{ctz} at zero do @emph{not} extend to the builtin functions 8789visible to the user. Thus one may be free to adjust the value at will 8790to match the target expansion of these operations without fear of 8791breaking the API. 8792@end defmac 8793 8794@defmac Pmode 8795An alias for the machine mode for pointers. On most machines, define 8796this to be the integer mode corresponding to the width of a hardware 8797pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. 8798On some machines you must define this to be one of the partial integer 8799modes, such as @code{PSImode}. 8800 8801The width of @code{Pmode} must be at least as large as the value of 8802@code{POINTER_SIZE}. If it is not equal, you must define the macro 8803@code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended 8804to @code{Pmode}. 8805@end defmac 8806 8807@defmac FUNCTION_MODE 8808An alias for the machine mode used for memory references to functions 8809being called, in @code{call} RTL expressions. On most machines this 8810should be @code{QImode}. 8811@end defmac 8812 8813@defmac INTEGRATE_THRESHOLD (@var{decl}) 8814A C expression for the maximum number of instructions above which the 8815function @var{decl} should not be inlined. @var{decl} is a 8816@code{FUNCTION_DECL} node. 8817 8818The default definition of this macro is 64 plus 8 times the number of 8819arguments that the function accepts. Some people think a larger 8820threshold should be used on RISC machines. 8821@end defmac 8822 8823@defmac STDC_0_IN_SYSTEM_HEADERS 8824In normal operation, the preprocessor expands @code{__STDC__} to the 8825constant 1, to signify that GCC conforms to ISO Standard C@. On some 8826hosts, like Solaris, the system compiler uses a different convention, 8827where @code{__STDC__} is normally 0, but is 1 if the user specifies 8828strict conformance to the C Standard. 8829 8830Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host 8831convention when processing system header files, but when processing user 8832files @code{__STDC__} will always expand to 1. 8833@end defmac 8834 8835@defmac NO_IMPLICIT_EXTERN_C 8836Define this macro if the system header files support C++ as well as C@. 8837This macro inhibits the usual method of using system header files in 8838C++, which is to pretend that the file's contents are enclosed in 8839@samp{extern "C" @{@dots{}@}}. 8840@end defmac 8841 8842@findex #pragma 8843@findex pragma 8844@defmac REGISTER_TARGET_PRAGMAS () 8845Define this macro if you want to implement any target-specific pragmas. 8846If defined, it is a C expression which makes a series of calls to 8847@code{c_register_pragma} for each pragma. The macro may also do any 8848setup required for the pragmas. 8849 8850The primary reason to define this macro is to provide compatibility with 8851other compilers for the same target. In general, we discourage 8852definition of target-specific pragmas for GCC@. 8853 8854If the pragma can be implemented by attributes then you should consider 8855defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well. 8856 8857Preprocessor macros that appear on pragma lines are not expanded. All 8858@samp{#pragma} directives that do not match any registered pragma are 8859silently ignored, unless the user specifies @option{-Wunknown-pragmas}. 8860@end defmac 8861 8862@deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) 8863 8864Each call to @code{c_register_pragma} establishes one pragma. The 8865@var{callback} routine will be called when the preprocessor encounters a 8866pragma of the form 8867 8868@smallexample 8869#pragma [@var{space}] @var{name} @dots{} 8870@end smallexample 8871 8872@var{space} is the case-sensitive namespace of the pragma, or 8873@code{NULL} to put the pragma in the global namespace. The callback 8874routine receives @var{pfile} as its first argument, which can be passed 8875on to cpplib's functions if necessary. You can lex tokens after the 8876@var{name} by calling @code{c_lex}. Tokens that are not read by the 8877callback will be silently ignored. The end of the line is indicated by 8878a token of type @code{CPP_EOF} 8879 8880For an example use of this routine, see @file{c4x.h} and the callback 8881routines defined in @file{c4x-c.c}. 8882 8883Note that the use of @code{c_lex} is specific to the C and C++ 8884compilers. It will not work in the Java or Fortran compilers, or any 8885other language compilers for that matter. Thus if @code{c_lex} is going 8886to be called from target-specific code, it must only be done so when 8887building the C and C++ compilers. This can be done by defining the 8888variables @code{c_target_objs} and @code{cxx_target_objs} in the 8889target entry in the @file{config.gcc} file. These variables should name 8890the target-specific, language-specific object file which contains the 8891code that uses @code{c_lex}. Note it will also be necessary to add a 8892rule to the makefile fragment pointed to by @code{tmake_file} that shows 8893how to build this object file. 8894@end deftypefun 8895 8896@findex #pragma 8897@findex pragma 8898@defmac HANDLE_SYSV_PRAGMA 8899Define this macro (to a value of 1) if you want the System V style 8900pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name> 8901[=<value>]} to be supported by gcc. 8902 8903The pack pragma specifies the maximum alignment (in bytes) of fields 8904within a structure, in much the same way as the @samp{__aligned__} and 8905@samp{__packed__} @code{__attribute__}s do. A pack value of zero resets 8906the behavior to the default. 8907 8908A subtlety for Microsoft Visual C/C++ style bit-field packing 8909(e.g. -mms-bitfields) for targets that support it: 8910When a bit-field is inserted into a packed record, the whole size 8911of the underlying type is used by one or more same-size adjacent 8912bit-fields (that is, if its long:3, 32 bits is used in the record, 8913and any additional adjacent long bit-fields are packed into the same 8914chunk of 32 bits. However, if the size changes, a new field of that 8915size is allocated). 8916 8917If both MS bit-fields and @samp{__attribute__((packed))} are used, 8918the latter will take precedence. If @samp{__attribute__((packed))} is 8919used on a single field when MS bit-fields are in use, it will take 8920precedence for that field, but the alignment of the rest of the structure 8921may affect its placement. 8922 8923The weak pragma only works if @code{SUPPORTS_WEAK} and 8924@code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation 8925of specifically named weak labels, optionally with a value. 8926@end defmac 8927 8928@findex #pragma 8929@findex pragma 8930@defmac HANDLE_PRAGMA_PACK_PUSH_POP 8931Define this macro (to a value of 1) if you want to support the Win32 8932style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma 8933pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment 8934(in bytes) of fields within a structure, in much the same way as the 8935@samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A 8936pack value of zero resets the behavior to the default. Successive 8937invocations of this pragma cause the previous values to be stacked, so 8938that invocations of @samp{#pragma pack(pop)} will return to the previous 8939value. 8940@end defmac 8941 8942@defmac DOLLARS_IN_IDENTIFIERS 8943Define this macro to control use of the character @samp{$} in 8944identifier names for the C family of languages. 0 means @samp{$} is 8945not allowed by default; 1 means it is allowed. 1 is the default; 8946there is no need to define this macro in that case. 8947@end defmac 8948 8949@defmac NO_DOLLAR_IN_LABEL 8950Define this macro if the assembler does not accept the character 8951@samp{$} in label names. By default constructors and destructors in 8952G++ have @samp{$} in the identifiers. If this macro is defined, 8953@samp{.} is used instead. 8954@end defmac 8955 8956@defmac NO_DOT_IN_LABEL 8957Define this macro if the assembler does not accept the character 8958@samp{.} in label names. By default constructors and destructors in G++ 8959have names that use @samp{.}. If this macro is defined, these names 8960are rewritten to avoid @samp{.}. 8961@end defmac 8962 8963@defmac DEFAULT_MAIN_RETURN 8964Define this macro if the target system expects every program's @code{main} 8965function to return a standard ``success'' value by default (if no other 8966value is explicitly returned). 8967 8968The definition should be a C statement (sans semicolon) to generate the 8969appropriate rtl instructions. It is used only when compiling the end of 8970@code{main}. 8971@end defmac 8972 8973@defmac INSN_SETS_ARE_DELAYED (@var{insn}) 8974Define this macro as a C expression that is nonzero if it is safe for the 8975delay slot scheduler to place instructions in the delay slot of @var{insn}, 8976even if they appear to use a resource set or clobbered in @var{insn}. 8977@var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that 8978every @code{call_insn} has this behavior. On machines where some @code{insn} 8979or @code{jump_insn} is really a function call and hence has this behavior, 8980you should define this macro. 8981 8982You need not define this macro if it would always return zero. 8983@end defmac 8984 8985@defmac INSN_REFERENCES_ARE_DELAYED (@var{insn}) 8986Define this macro as a C expression that is nonzero if it is safe for the 8987delay slot scheduler to place instructions in the delay slot of @var{insn}, 8988even if they appear to set or clobber a resource referenced in @var{insn}. 8989@var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where 8990some @code{insn} or @code{jump_insn} is really a function call and its operands 8991are registers whose use is actually in the subroutine it calls, you should 8992define this macro. Doing so allows the delay slot scheduler to move 8993instructions which copy arguments into the argument registers into the delay 8994slot of @var{insn}. 8995 8996You need not define this macro if it would always return zero. 8997@end defmac 8998 8999@defmac MULTIPLE_SYMBOL_SPACES 9000Define this macro if in some cases global symbols from one translation 9001unit may not be bound to undefined symbols in another translation unit 9002without user intervention. For instance, under Microsoft Windows 9003symbols must be explicitly imported from shared libraries (DLLs). 9004@end defmac 9005 9006@defmac MD_ASM_CLOBBERS (@var{clobbers}) 9007A C statement that adds to @var{clobbers} @code{STRING_CST} trees for 9008any hard regs the port wishes to automatically clobber for all asms. 9009@end defmac 9010 9011@defmac MATH_LIBRARY 9012Define this macro as a C string constant for the linker argument to link 9013in the system math library, or @samp{""} if the target does not have a 9014separate math library. 9015 9016You need only define this macro if the default of @samp{"-lm"} is wrong. 9017@end defmac 9018 9019@defmac LIBRARY_PATH_ENV 9020Define this macro as a C string constant for the environment variable that 9021specifies where the linker should look for libraries. 9022 9023You need only define this macro if the default of @samp{"LIBRARY_PATH"} 9024is wrong. 9025@end defmac 9026 9027@defmac TARGET_HAS_F_SETLKW 9028Define this macro if the target supports file locking with fcntl / F_SETLKW@. 9029Note that this functionality is part of POSIX@. 9030Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code 9031to use file locking when exiting a program, which avoids race conditions 9032if the program has forked. 9033@end defmac 9034 9035@defmac MAX_CONDITIONAL_EXECUTE 9036 9037A C expression for the maximum number of instructions to execute via 9038conditional execution instructions instead of a branch. A value of 9039@code{BRANCH_COST}+1 is the default if the machine does not use cc0, and 90401 if it does use cc0. 9041@end defmac 9042 9043@defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr}) 9044Used if the target needs to perform machine-dependent modifications on the 9045conditionals used for turning basic blocks into conditionally executed code. 9046@var{ce_info} points to a data structure, @code{struct ce_if_block}, which 9047contains information about the currently processed blocks. @var{true_expr} 9048and @var{false_expr} are the tests that are used for converting the 9049then-block and the else-block, respectively. Set either @var{true_expr} or 9050@var{false_expr} to a null pointer if the tests cannot be converted. 9051@end defmac 9052 9053@defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr}) 9054Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated 9055if-statements into conditions combined by @code{and} and @code{or} operations. 9056@var{bb} contains the basic block that contains the test that is currently 9057being processed and about to be turned into a condition. 9058@end defmac 9059 9060@defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn}) 9061A C expression to modify the @var{PATTERN} of an @var{INSN} that is to 9062be converted to conditional execution format. @var{ce_info} points to 9063a data structure, @code{struct ce_if_block}, which contains information 9064about the currently processed blocks. 9065@end defmac 9066 9067@defmac IFCVT_MODIFY_FINAL (@var{ce_info}) 9068A C expression to perform any final machine dependent modifications in 9069converting code to conditional execution. The involved basic blocks 9070can be found in the @code{struct ce_if_block} structure that is pointed 9071to by @var{ce_info}. 9072@end defmac 9073 9074@defmac IFCVT_MODIFY_CANCEL (@var{ce_info}) 9075A C expression to cancel any machine dependent modifications in 9076converting code to conditional execution. The involved basic blocks 9077can be found in the @code{struct ce_if_block} structure that is pointed 9078to by @var{ce_info}. 9079@end defmac 9080 9081@defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info}) 9082A C expression to initialize any extra fields in a @code{struct ce_if_block} 9083structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro. 9084@end defmac 9085 9086@defmac IFCVT_EXTRA_FIELDS 9087If defined, it should expand to a set of field declarations that will be 9088added to the @code{struct ce_if_block} structure. These should be initialized 9089by the @code{IFCVT_INIT_EXTRA_FIELDS} macro. 9090@end defmac 9091 9092@deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG () 9093If non-null, this hook performs a target-specific pass over the 9094instruction stream. The compiler will run it at all optimization levels, 9095just before the point at which it normally does delayed-branch scheduling. 9096 9097The exact purpose of the hook varies from target to target. Some use 9098it to do transformations that are necessary for correctness, such as 9099laying out in-function constant pools or avoiding hardware hazards. 9100Others use it as an opportunity to do some machine-dependent optimizations. 9101 9102You need not implement the hook if it has nothing to do. The default 9103definition is null. 9104@end deftypefn 9105 9106@deftypefn {Target Hook} void TARGET_INIT_BUILTINS () 9107Define this hook if you have any machine-specific built-in functions 9108that need to be defined. It should be a function that performs the 9109necessary setup. 9110 9111Machine specific built-in functions can be useful to expand special machine 9112instructions that would otherwise not normally be generated because 9113they have no equivalent in the source language (for example, SIMD vector 9114instructions or prefetch instructions). 9115 9116To create a built-in function, call the function @code{builtin_function} 9117which is defined by the language front end. You can use any type nodes set 9118up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2}; 9119only language front ends that use those two functions will call 9120@samp{TARGET_INIT_BUILTINS}. 9121@end deftypefn 9122 9123@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore}) 9124 9125Expand a call to a machine specific built-in function that was set up by 9126@samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the 9127function call; the result should go to @var{target} if that is 9128convenient, and have mode @var{mode} if that is convenient. 9129@var{subtarget} may be used as the target for computing one of 9130@var{exp}'s operands. @var{ignore} is nonzero if the value is to be 9131ignored. This function should return the result of the call to the 9132built-in function. 9133@end deftypefn 9134 9135@defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2}) 9136 9137Take a branch insn in @var{branch1} and another in @var{branch2}. 9138Return true if redirecting @var{branch1} to the destination of 9139@var{branch2} is possible. 9140 9141On some targets, branches may have a limited range. Optimizing the 9142filling of delay slots can result in branches being redirected, and this 9143may in turn cause a branch offset to overflow. 9144@end defmac 9145 9146@defmac ALLOCATE_INITIAL_VALUE (@var{hard_reg}) 9147 9148When the initial value of a hard register has been copied in a pseudo 9149register, it is often not necessary to actually allocate another register 9150to this pseudo register, because the original hard register or a stack slot 9151it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if 9152defined, is called at the start of register allocation once for each 9153hard register that had its initial value copied by using 9154@code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}. 9155Possible values are @code{NULL_RTX}, if you don't want 9156to do any special allocation, a @code{REG} rtx---that would typically be 9157the hard register itself, if it is known not to be clobbered---or a 9158@code{MEM}. 9159If you are returning a @code{MEM}, this is only a hint for the allocator; 9160it might decide to use another register anyways. 9161You may use @code{current_function_leaf_function} in the definition of the 9162macro, functions that use @code{REG_N_SETS}, to determine if the hard 9163register in question will not be clobbered. 9164@end defmac 9165 9166@defmac TARGET_OBJECT_SUFFIX 9167Define this macro to be a C string representing the suffix for object 9168files on your target machine. If you do not define this macro, GCC will 9169use @samp{.o} as the suffix for object files. 9170@end defmac 9171 9172@defmac TARGET_EXECUTABLE_SUFFIX 9173Define this macro to be a C string representing the suffix to be 9174automatically added to executable files on your target machine. If you 9175do not define this macro, GCC will use the null string as the suffix for 9176executable files. 9177@end defmac 9178 9179@defmac COLLECT_EXPORT_LIST 9180If defined, @code{collect2} will scan the individual object files 9181specified on its command line and create an export list for the linker. 9182Define this macro for systems like AIX, where the linker discards 9183object files that are not referenced from @code{main} and uses export 9184lists. 9185@end defmac 9186 9187@defmac MODIFY_JNI_METHOD_CALL (@var{mdecl}) 9188Define this macro to a C expression representing a variant of the 9189method call @var{mdecl}, if Java Native Interface (JNI) methods 9190must be invoked differently from other methods on your target. 9191For example, on 32-bit Microsoft Windows, JNI methods must be invoked using 9192the @code{stdcall} calling convention and this macro is then 9193defined as this expression: 9194 9195@smallexample 9196build_type_attribute_variant (@var{mdecl}, 9197 build_tree_list 9198 (get_identifier ("stdcall"), 9199 NULL)) 9200@end smallexample 9201@end defmac 9202 9203@deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void) 9204This target hook returns @code{true} past the point in which new jump 9205instructions could be created. On machines that require a register for 9206every jump such as the SHmedia ISA of SH5, this point would typically be 9207reload, so this target hook should be defined to a function such as: 9208 9209@smallexample 9210static bool 9211cannot_modify_jumps_past_reload_p () 9212@{ 9213 return (reload_completed || reload_in_progress); 9214@} 9215@end smallexample 9216@end deftypefn 9217 9218@deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void) 9219This target hook returns a register class for which branch target register 9220optimizations should be applied. All registers in this class should be 9221usable interchangeably. After reload, registers in this class will be 9222re-allocated and loads will be hoisted out of loops and be subjected 9223to inter-block scheduling. 9224@end deftypefn 9225 9226@deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen}) 9227Branch target register optimization will by default exclude callee-saved 9228registers 9229that are not already live during the current function; if this target hook 9230returns true, they will be included. The target code must than make sure 9231that all target registers in the class returned by 9232@samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are 9233saved. @var{after_prologue_epilogue_gen} indicates if prologues and 9234epilogues have already been generated. Note, even if you only return 9235true when @var{after_prologue_epilogue_gen} is false, you still are likely 9236to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET} 9237to reserve space for caller-saved target registers. 9238@end deftypefn 9239 9240@defmac POWI_MAX_MULTS 9241If defined, this macro is interpreted as a signed integer C expression 9242that specifies the maximum number of floating point multiplications 9243that should be emitted when expanding exponentiation by an integer 9244constant inline. When this value is defined, exponentiation requiring 9245more than this number of multiplications is implemented by calling the 9246system library's @code{pow}, @code{powf} or @code{powl} routines. 9247The default value places no upper bound on the multiplication count. 9248@end defmac 9249