tm.texi revision 256281
1@c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001, 2@c 2002, 2003, 2004, 2005, 2006 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* Registers:: Naming and describing the hardware registers. 35* Register Classes:: Defining the classes of hardware registers. 36* Old Constraints:: The old way to define machine-specific constraints. 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* Anchored Addresses:: Defining how @option{-fsection-anchors} should work. 43* Condition Code:: Defining how insns update the condition code. 44* Costs:: Defining relative costs of different operations. 45* Scheduling:: Adjusting the behavior of the instruction scheduler. 46* Sections:: Dividing storage into text, data, and other sections. 47* PIC:: Macros for position independent code. 48* Assembler Format:: Defining how to write insns and pseudo-ops to output. 49* Debugging Info:: Defining the format of debugging output. 50* Floating Point:: Handling floating point for cross-compilers. 51* Mode Switching:: Insertion of mode-switching instructions. 52* Target Attributes:: Defining target-specific uses of @code{__attribute__}. 53* MIPS Coprocessors:: MIPS coprocessor support and how to customize it. 54* PCH Target:: Validity checking for precompiled headers. 55* C++ ABI:: Controlling C++ ABI changes. 56* Misc:: Everything else. 57@end menu 58 59@node Target Structure 60@section The Global @code{targetm} Variable 61@cindex target hooks 62@cindex target functions 63 64@deftypevar {struct gcc_target} targetm 65The target @file{.c} file must define the global @code{targetm} variable 66which contains pointers to functions and data relating to the target 67machine. The variable is declared in @file{target.h}; 68@file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is 69used to initialize the variable, and macros for the default initializers 70for elements of the structure. The @file{.c} file should override those 71macros for which the default definition is inappropriate. For example: 72@smallexample 73#include "target.h" 74#include "target-def.h" 75 76/* @r{Initialize the GCC target structure.} */ 77 78#undef TARGET_COMP_TYPE_ATTRIBUTES 79#define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes 80 81struct gcc_target targetm = TARGET_INITIALIZER; 82@end smallexample 83@end deftypevar 84 85Where a macro should be defined in the @file{.c} file in this manner to 86form part of the @code{targetm} structure, it is documented below as a 87``Target Hook'' with a prototype. Many macros will change in future 88from being defined in the @file{.h} file to being part of the 89@code{targetm} structure. 90 91@node Driver 92@section Controlling the Compilation Driver, @file{gcc} 93@cindex driver 94@cindex controlling the compilation driver 95 96@c prevent bad page break with this line 97You can control the compilation driver. 98 99@defmac SWITCH_TAKES_ARG (@var{char}) 100A C expression which determines whether the option @option{-@var{char}} 101takes arguments. The value should be the number of arguments that 102option takes--zero, for many options. 103 104By default, this macro is defined as 105@code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options 106properly. You need not define @code{SWITCH_TAKES_ARG} unless you 107wish to add additional options which take arguments. Any redefinition 108should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for 109additional options. 110@end defmac 111 112@defmac WORD_SWITCH_TAKES_ARG (@var{name}) 113A C expression which determines whether the option @option{-@var{name}} 114takes arguments. The value should be the number of arguments that 115option takes--zero, for many options. This macro rather than 116@code{SWITCH_TAKES_ARG} is used for multi-character option names. 117 118By default, this macro is defined as 119@code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options 120properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you 121wish to add additional options which take arguments. Any redefinition 122should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for 123additional options. 124@end defmac 125 126@defmac SWITCH_CURTAILS_COMPILATION (@var{char}) 127A C expression which determines whether the option @option{-@var{char}} 128stops compilation before the generation of an executable. The value is 129boolean, nonzero if the option does stop an executable from being 130generated, zero otherwise. 131 132By default, this macro is defined as 133@code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard 134options properly. You need not define 135@code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional 136options which affect the generation of an executable. Any redefinition 137should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check 138for additional options. 139@end defmac 140 141@defmac SWITCHES_NEED_SPACES 142A string-valued C expression which enumerates the options for which 143the linker needs a space between the option and its argument. 144 145If this macro is not defined, the default value is @code{""}. 146@end defmac 147 148@defmac TARGET_OPTION_TRANSLATE_TABLE 149If defined, a list of pairs of strings, the first of which is a 150potential command line target to the @file{gcc} driver program, and the 151second of which is a space-separated (tabs and other whitespace are not 152supported) list of options with which to replace the first option. The 153target defining this list is responsible for assuring that the results 154are valid. Replacement options may not be the @code{--opt} style, they 155must be the @code{-opt} style. It is the intention of this macro to 156provide a mechanism for substitution that affects the multilibs chosen, 157such as one option that enables many options, some of which select 158multilibs. Example nonsensical definition, where @option{-malt-abi}, 159@option{-EB}, and @option{-mspoo} cause different multilibs to be chosen: 160 161@smallexample 162#define TARGET_OPTION_TRANSLATE_TABLE \ 163@{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \ 164@{ "-compat", "-EB -malign=4 -mspoo" @} 165@end smallexample 166@end defmac 167 168@defmac DRIVER_SELF_SPECS 169A list of specs for the driver itself. It should be a suitable 170initializer for an array of strings, with no surrounding braces. 171 172The driver applies these specs to its own command line between loading 173default @file{specs} files (but not command-line specified ones) and 174choosing the multilib directory or running any subcommands. It 175applies them in the order given, so each spec can depend on the 176options added by earlier ones. It is also possible to remove options 177using @samp{%<@var{option}} in the usual way. 178 179This macro can be useful when a port has several interdependent target 180options. It provides a way of standardizing the command line so 181that the other specs are easier to write. 182 183Do not define this macro if it does not need to do anything. 184@end defmac 185 186@defmac OPTION_DEFAULT_SPECS 187A list of specs used to support configure-time default options (i.e.@: 188@option{--with} options) in the driver. It should be a suitable initializer 189for an array of structures, each containing two strings, without the 190outermost pair of surrounding braces. 191 192The first item in the pair is the name of the default. This must match 193the code in @file{config.gcc} for the target. The second item is a spec 194to apply if a default with this name was specified. The string 195@samp{%(VALUE)} in the spec will be replaced by the value of the default 196everywhere it occurs. 197 198The driver will apply these specs to its own command line between loading 199default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using 200the same mechanism as @code{DRIVER_SELF_SPECS}. 201 202Do not define this macro if it does not need to do anything. 203@end defmac 204 205@defmac CPP_SPEC 206A C string constant that tells the GCC driver program options to 207pass to CPP@. It can also specify how to translate options you 208give to GCC into options for GCC to pass to the CPP@. 209 210Do not define this macro if it does not need to do anything. 211@end defmac 212 213@defmac CPLUSPLUS_CPP_SPEC 214This macro is just like @code{CPP_SPEC}, but is used for C++, rather 215than C@. If you do not define this macro, then the value of 216@code{CPP_SPEC} (if any) will be used instead. 217@end defmac 218 219@defmac CC1_SPEC 220A C string constant that tells the GCC driver program options to 221pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language 222front ends. 223It can also specify how to translate options you give to GCC into options 224for GCC to pass to front ends. 225 226Do not define this macro if it does not need to do anything. 227@end defmac 228 229@defmac CC1PLUS_SPEC 230A C string constant that tells the GCC driver program options to 231pass to @code{cc1plus}. It can also specify how to translate options you 232give to GCC into options for GCC to pass to the @code{cc1plus}. 233 234Do not define this macro if it does not need to do anything. 235Note that everything defined in CC1_SPEC is already passed to 236@code{cc1plus} so there is no need to duplicate the contents of 237CC1_SPEC in CC1PLUS_SPEC@. 238@end defmac 239 240@defmac ASM_SPEC 241A C string constant that tells the GCC driver program options to 242pass to the assembler. It can also specify how to translate options 243you give to GCC into options for GCC to pass to the assembler. 244See the file @file{sun3.h} for an example of this. 245 246Do not define this macro if it does not need to do anything. 247@end defmac 248 249@defmac ASM_FINAL_SPEC 250A C string constant that tells the GCC driver program how to 251run any programs which cleanup after the normal assembler. 252Normally, this is not needed. See the file @file{mips.h} for 253an example of this. 254 255Do not define this macro if it does not need to do anything. 256@end defmac 257 258@defmac AS_NEEDS_DASH_FOR_PIPED_INPUT 259Define this macro, with no value, if the driver should give the assembler 260an argument consisting of a single dash, @option{-}, to instruct it to 261read from its standard input (which will be a pipe connected to the 262output of the compiler proper). This argument is given after any 263@option{-o} option specifying the name of the output file. 264 265If you do not define this macro, the assembler is assumed to read its 266standard input if given no non-option arguments. If your assembler 267cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct; 268see @file{mips.h} for instance. 269@end defmac 270 271@defmac LINK_SPEC 272A C string constant that tells the GCC driver program options to 273pass to the linker. It can also specify how to translate options you 274give to GCC into options for GCC to pass to the linker. 275 276Do not define this macro if it does not need to do anything. 277@end defmac 278 279@defmac LIB_SPEC 280Another C string constant used much like @code{LINK_SPEC}. The difference 281between the two is that @code{LIB_SPEC} is used at the end of the 282command given to the linker. 283 284If this macro is not defined, a default is provided that 285loads the standard C library from the usual place. See @file{gcc.c}. 286@end defmac 287 288@defmac LIBGCC_SPEC 289Another C string constant that tells the GCC driver program 290how and when to place a reference to @file{libgcc.a} into the 291linker command line. This constant is placed both before and after 292the value of @code{LIB_SPEC}. 293 294If this macro is not defined, the GCC driver provides a default that 295passes the string @option{-lgcc} to the linker. 296@end defmac 297 298@defmac REAL_LIBGCC_SPEC 299By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the 300@code{LIBGCC_SPEC} is not directly used by the driver program but is 301instead modified to refer to different versions of @file{libgcc.a} 302depending on the values of the command line flags @option{-static}, 303@option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On 304targets where these modifications are inappropriate, define 305@code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the 306driver how to place a reference to @file{libgcc} on the link command 307line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified. 308@end defmac 309 310@defmac USE_LD_AS_NEEDED 311A macro that controls the modifications to @code{LIBGCC_SPEC} 312mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be 313generated that uses --as-needed and the shared libgcc in place of the 314static exception handler library, when linking without any of 315@code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}. 316@end defmac 317 318@defmac LINK_EH_SPEC 319If defined, this C string constant is added to @code{LINK_SPEC}. 320When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects 321the modifications to @code{LIBGCC_SPEC} mentioned in 322@code{REAL_LIBGCC_SPEC}. 323@end defmac 324 325@defmac STARTFILE_SPEC 326Another C string constant used much like @code{LINK_SPEC}. The 327difference between the two is that @code{STARTFILE_SPEC} is used at 328the very beginning of the command given to the linker. 329 330If this macro is not defined, a default is provided that loads the 331standard C startup file from the usual place. See @file{gcc.c}. 332@end defmac 333 334@defmac ENDFILE_SPEC 335Another C string constant used much like @code{LINK_SPEC}. The 336difference between the two is that @code{ENDFILE_SPEC} is used at 337the very end of the command given to the linker. 338 339Do not define this macro if it does not need to do anything. 340@end defmac 341 342@defmac THREAD_MODEL_SPEC 343GCC @code{-v} will print the thread model GCC was configured to use. 344However, this doesn't work on platforms that are multilibbed on thread 345models, such as AIX 4.3. On such platforms, define 346@code{THREAD_MODEL_SPEC} such that it evaluates to a string without 347blanks that names one of the recognized thread models. @code{%*}, the 348default value of this macro, will expand to the value of 349@code{thread_file} set in @file{config.gcc}. 350@end defmac 351 352@defmac SYSROOT_SUFFIX_SPEC 353Define this macro to add a suffix to the target sysroot when GCC is 354configured with a sysroot. This will cause GCC to search for usr/lib, 355et al, within sysroot+suffix. 356@end defmac 357 358@defmac SYSROOT_HEADERS_SUFFIX_SPEC 359Define this macro to add a headers_suffix to the target sysroot when 360GCC is configured with a sysroot. This will cause GCC to pass the 361updated sysroot+headers_suffix to CPP, causing it to search for 362usr/include, et al, within sysroot+headers_suffix. 363@end defmac 364 365@defmac EXTRA_SPECS 366Define this macro to provide additional specifications to put in the 367@file{specs} file that can be used in various specifications like 368@code{CC1_SPEC}. 369 370The definition should be an initializer for an array of structures, 371containing a string constant, that defines the specification name, and a 372string constant that provides the specification. 373 374Do not define this macro if it does not need to do anything. 375 376@code{EXTRA_SPECS} is useful when an architecture contains several 377related targets, which have various @code{@dots{}_SPECS} which are similar 378to each other, and the maintainer would like one central place to keep 379these definitions. 380 381For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to 382define either @code{_CALL_SYSV} when the System V calling sequence is 383used or @code{_CALL_AIX} when the older AIX-based calling sequence is 384used. 385 386The @file{config/rs6000/rs6000.h} target file defines: 387 388@smallexample 389#define EXTRA_SPECS \ 390 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, 391 392#define CPP_SYS_DEFAULT "" 393@end smallexample 394 395The @file{config/rs6000/sysv.h} target file defines: 396@smallexample 397#undef CPP_SPEC 398#define CPP_SPEC \ 399"%@{posix: -D_POSIX_SOURCE @} \ 400%@{mcall-sysv: -D_CALL_SYSV @} \ 401%@{!mcall-sysv: %(cpp_sysv_default) @} \ 402%@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" 403 404#undef CPP_SYSV_DEFAULT 405#define CPP_SYSV_DEFAULT "-D_CALL_SYSV" 406@end smallexample 407 408while the @file{config/rs6000/eabiaix.h} target file defines 409@code{CPP_SYSV_DEFAULT} as: 410 411@smallexample 412#undef CPP_SYSV_DEFAULT 413#define CPP_SYSV_DEFAULT "-D_CALL_AIX" 414@end smallexample 415@end defmac 416 417@defmac LINK_LIBGCC_SPECIAL_1 418Define this macro if the driver program should find the library 419@file{libgcc.a}. If you do not define this macro, the driver program will pass 420the argument @option{-lgcc} to tell the linker to do the search. 421@end defmac 422 423@defmac LINK_GCC_C_SEQUENCE_SPEC 424The sequence in which libgcc and libc are specified to the linker. 425By default this is @code{%G %L %G}. 426@end defmac 427 428@defmac LINK_COMMAND_SPEC 429A C string constant giving the complete command line need to execute the 430linker. When you do this, you will need to update your port each time a 431change is made to the link command line within @file{gcc.c}. Therefore, 432define this macro only if you need to completely redefine the command 433line for invoking the linker and there is no other way to accomplish 434the effect you need. Overriding this macro may be avoidable by overriding 435@code{LINK_GCC_C_SEQUENCE_SPEC} instead. 436@end defmac 437 438@defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES 439A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search 440directories from linking commands. Do not give it a nonzero value if 441removing duplicate search directories changes the linker's semantics. 442@end defmac 443 444@defmac MULTILIB_DEFAULTS 445Define this macro as a C expression for the initializer of an array of 446string to tell the driver program which options are defaults for this 447target and thus do not need to be handled specially when using 448@code{MULTILIB_OPTIONS}. 449 450Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in 451the target makefile fragment or if none of the options listed in 452@code{MULTILIB_OPTIONS} are set by default. 453@xref{Target Fragment}. 454@end defmac 455 456@defmac RELATIVE_PREFIX_NOT_LINKDIR 457Define this macro to tell @command{gcc} that it should only translate 458a @option{-B} prefix into a @option{-L} linker option if the prefix 459indicates an absolute file name. 460@end defmac 461 462@defmac MD_EXEC_PREFIX 463If defined, this macro is an additional prefix to try after 464@code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched 465when the @option{-b} option is used, or the compiler is built as a cross 466compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it 467to the list of directories used to find the assembler in @file{configure.in}. 468@end defmac 469 470@defmac STANDARD_STARTFILE_PREFIX 471Define this macro as a C string constant if you wish to override the 472standard choice of @code{libdir} as the default prefix to 473try when searching for startup files such as @file{crt0.o}. 474@code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler 475is built as a cross compiler. 476@end defmac 477 478@defmac STANDARD_STARTFILE_PREFIX_1 479Define this macro as a C string constant if you wish to override the 480standard choice of @code{/lib} as a prefix to try after the default prefix 481when searching for startup files such as @file{crt0.o}. 482@code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler 483is built as a cross compiler. 484@end defmac 485 486@defmac STANDARD_STARTFILE_PREFIX_2 487Define this macro as a C string constant if you wish to override the 488standard choice of @code{/lib} as yet another prefix to try after the 489default prefix when searching for startup files such as @file{crt0.o}. 490@code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler 491is built as a cross compiler. 492@end defmac 493 494@defmac MD_STARTFILE_PREFIX 495If defined, this macro supplies an additional prefix to try after the 496standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the 497@option{-b} option is used, or when the compiler is built as a cross 498compiler. 499@end defmac 500 501@defmac MD_STARTFILE_PREFIX_1 502If defined, this macro supplies yet another prefix to try after the 503standard prefixes. It is not searched when the @option{-b} option is 504used, or when the compiler is built as a cross compiler. 505@end defmac 506 507@defmac INIT_ENVIRONMENT 508Define this macro as a C string constant if you wish to set environment 509variables for programs called by the driver, such as the assembler and 510loader. The driver passes the value of this macro to @code{putenv} to 511initialize the necessary environment variables. 512@end defmac 513 514@defmac LOCAL_INCLUDE_DIR 515Define this macro as a C string constant if you wish to override the 516standard choice of @file{/usr/local/include} as the default prefix to 517try when searching for local header files. @code{LOCAL_INCLUDE_DIR} 518comes before @code{SYSTEM_INCLUDE_DIR} in the search order. 519 520Cross compilers do not search either @file{/usr/local/include} or its 521replacement. 522@end defmac 523 524@defmac MODIFY_TARGET_NAME 525Define this macro if you wish to define command-line switches that 526modify the default target name. 527 528For each switch, you can include a string to be appended to the first 529part of the configuration name or a string to be deleted from the 530configuration name, if present. The definition should be an initializer 531for an array of structures. Each array element should have three 532elements: the switch name (a string constant, including the initial 533dash), one of the enumeration codes @code{ADD} or @code{DELETE} to 534indicate whether the string should be inserted or deleted, and the string 535to be inserted or deleted (a string constant). 536 537For example, on a machine where @samp{64} at the end of the 538configuration name denotes a 64-bit target and you want the @option{-32} 539and @option{-64} switches to select between 32- and 64-bit targets, you would 540code 541 542@smallexample 543#define MODIFY_TARGET_NAME \ 544 @{ @{ "-32", DELETE, "64"@}, \ 545 @{"-64", ADD, "64"@}@} 546@end smallexample 547@end defmac 548 549@defmac SYSTEM_INCLUDE_DIR 550Define this macro as a C string constant if you wish to specify a 551system-specific directory to search for header files before the standard 552directory. @code{SYSTEM_INCLUDE_DIR} comes before 553@code{STANDARD_INCLUDE_DIR} in the search order. 554 555Cross compilers do not use this macro and do not search the directory 556specified. 557@end defmac 558 559@defmac STANDARD_INCLUDE_DIR 560Define this macro as a C string constant if you wish to override the 561standard choice of @file{/usr/include} as the default prefix to 562try when searching for header files. 563 564Cross compilers ignore this macro and do not search either 565@file{/usr/include} or its replacement. 566@end defmac 567 568@defmac STANDARD_INCLUDE_COMPONENT 569The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}. 570See @code{INCLUDE_DEFAULTS}, below, for the description of components. 571If you do not define this macro, no component is used. 572@end defmac 573 574@defmac INCLUDE_DEFAULTS 575Define this macro if you wish to override the entire default search path 576for include files. For a native compiler, the default search path 577usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, 578@code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and 579@code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} 580and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, 581and specify private search areas for GCC@. The directory 582@code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. 583 584The definition should be an initializer for an array of structures. 585Each array element should have four elements: the directory name (a 586string constant), the component name (also a string constant), a flag 587for C++-only directories, 588and a flag showing that the includes in the directory don't need to be 589wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of 590the array with a null element. 591 592The component name denotes what GNU package the include file is part of, 593if any, in all uppercase letters. For example, it might be @samp{GCC} 594or @samp{BINUTILS}. If the package is part of a vendor-supplied 595operating system, code the component name as @samp{0}. 596 597For example, here is the definition used for VAX/VMS: 598 599@smallexample 600#define INCLUDE_DEFAULTS \ 601@{ \ 602 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \ 603 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \ 604 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \ 605 @{ ".", 0, 0, 0@}, \ 606 @{ 0, 0, 0, 0@} \ 607@} 608@end smallexample 609@end defmac 610 611Here is the order of prefixes tried for exec files: 612 613@enumerate 614@item 615Any prefixes specified by the user with @option{-B}. 616 617@item 618The environment variable @code{GCC_EXEC_PREFIX}, if any. 619 620@item 621The directories specified by the environment variable @code{COMPILER_PATH}. 622 623@item 624The macro @code{STANDARD_EXEC_PREFIX}. 625 626@item 627@file{/usr/lib/gcc/}. 628 629@item 630The macro @code{MD_EXEC_PREFIX}, if any. 631@end enumerate 632 633Here is the order of prefixes tried for startfiles: 634 635@enumerate 636@item 637Any prefixes specified by the user with @option{-B}. 638 639@item 640The environment variable @code{GCC_EXEC_PREFIX}, if any. 641 642@item 643The directories specified by the environment variable @code{LIBRARY_PATH} 644(or port-specific name; native only, cross compilers do not use this). 645 646@item 647The macro @code{STANDARD_EXEC_PREFIX}. 648 649@item 650@file{/usr/lib/gcc/}. 651 652@item 653The macro @code{MD_EXEC_PREFIX}, if any. 654 655@item 656The macro @code{MD_STARTFILE_PREFIX}, if any. 657 658@item 659The macro @code{STANDARD_STARTFILE_PREFIX}. 660 661@item 662@file{/lib/}. 663 664@item 665@file{/usr/lib/}. 666@end enumerate 667 668@node Run-time Target 669@section Run-time Target Specification 670@cindex run-time target specification 671@cindex predefined macros 672@cindex target specifications 673 674@c prevent bad page break with this line 675Here are run-time target specifications. 676 677@defmac TARGET_CPU_CPP_BUILTINS () 678This function-like macro expands to a block of code that defines 679built-in preprocessor macros and assertions for the target cpu, using 680the functions @code{builtin_define}, @code{builtin_define_std} and 681@code{builtin_assert}. When the front end 682calls this macro it provides a trailing semicolon, and since it has 683finished command line option processing your code can use those 684results freely. 685 686@code{builtin_assert} takes a string in the form you pass to the 687command-line option @option{-A}, such as @code{cpu=mips}, and creates 688the assertion. @code{builtin_define} takes a string in the form 689accepted by option @option{-D} and unconditionally defines the macro. 690 691@code{builtin_define_std} takes a string representing the name of an 692object-like macro. If it doesn't lie in the user's namespace, 693@code{builtin_define_std} defines it unconditionally. Otherwise, it 694defines a version with two leading underscores, and another version 695with two leading and trailing underscores, and defines the original 696only if an ISO standard was not requested on the command line. For 697example, passing @code{unix} defines @code{__unix}, @code{__unix__} 698and possibly @code{unix}; passing @code{_mips} defines @code{__mips}, 699@code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64} 700defines only @code{_ABI64}. 701 702You can also test for the C dialect being compiled. The variable 703@code{c_language} is set to one of @code{clk_c} or 704@code{clk_cplusplus}. Note that if we are preprocessing assembler, 705this variable will be @code{clk_c} but the function-like macro 706@code{preprocessing_asm_p()} will return true, so you might want to 707check for that first. If you need to check for strict ANSI, the 708variable @code{flag_iso} can be used. The function-like macro 709@code{preprocessing_trad_p()} can be used to check for traditional 710preprocessing. 711@end defmac 712 713@defmac TARGET_OS_CPP_BUILTINS () 714Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 715and is used for the target operating system instead. 716@end defmac 717 718@defmac TARGET_OBJFMT_CPP_BUILTINS () 719Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional 720and is used for the target object format. @file{elfos.h} uses this 721macro to define @code{__ELF__}, so you probably do not need to define 722it yourself. 723@end defmac 724 725@deftypevar {extern int} target_flags 726This variable is declared in @file{options.h}, which is included before 727any target-specific headers. 728@end deftypevar 729 730@deftypevar {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS 731This variable specifies the initial value of @code{target_flags}. 732Its default setting is 0. 733@end deftypevar 734 735@cindex optional hardware or system features 736@cindex features, optional, in system conventions 737 738@deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value}) 739This hook is called whenever the user specifies one of the 740target-specific options described by the @file{.opt} definition files 741(@pxref{Options}). It has the opportunity to do some option-specific 742processing and should return true if the option is valid. The default 743definition does nothing but return true. 744 745@var{code} specifies the @code{OPT_@var{name}} enumeration value 746associated with the selected option; @var{name} is just a rendering of 747the option name in which non-alphanumeric characters are replaced by 748underscores. @var{arg} specifies the string argument and is null if 749no argument was given. If the option is flagged as a @code{UInteger} 750(@pxref{Option properties}), @var{value} is the numeric value of the 751argument. Otherwise @var{value} is 1 if the positive form of the 752option was used and 0 if the ``no-'' form was. 753@end deftypefn 754 755@defmac TARGET_VERSION 756This macro is a C statement to print on @code{stderr} a string 757describing the particular machine description choice. Every machine 758description should define @code{TARGET_VERSION}. For example: 759 760@smallexample 761#ifdef MOTOROLA 762#define TARGET_VERSION \ 763 fprintf (stderr, " (68k, Motorola syntax)"); 764#else 765#define TARGET_VERSION \ 766 fprintf (stderr, " (68k, MIT syntax)"); 767#endif 768@end smallexample 769@end defmac 770 771@defmac OVERRIDE_OPTIONS 772Sometimes certain combinations of command options do not make sense on 773a particular target machine. You can define a macro 774@code{OVERRIDE_OPTIONS} to take account of this. This macro, if 775defined, is executed once just after all the command options have been 776parsed. 777 778Don't use this macro to turn on various extra optimizations for 779@option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for. 780@end defmac 781 782@defmac C_COMMON_OVERRIDE_OPTIONS 783This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C 784language frontends (C, C++) and so can be used to alter option flag 785variables which only exist in those frontends. 786@end defmac 787 788@defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size}) 789Some machines may desire to change what optimizations are performed for 790various optimization levels. This macro, if defined, is executed once 791just after the optimization level is determined and before the remainder 792of the command options have been parsed. Values set in this macro are 793used as the default values for the other command line options. 794 795@var{level} is the optimization level specified; 2 if @option{-O2} is 796specified, 1 if @option{-O} is specified, and 0 if neither is specified. 797 798@var{size} is nonzero if @option{-Os} is specified and zero otherwise. 799 800You should not use this macro to change options that are not 801machine-specific. These should uniformly selected by the same 802optimization level on all supported machines. Use this macro to enable 803machine-specific optimizations. 804 805@strong{Do not examine @code{write_symbols} in 806this macro!} The debugging options are not supposed to alter the 807generated code. 808@end defmac 809 810@defmac CAN_DEBUG_WITHOUT_FP 811Define this macro if debugging can be performed even without a frame 812pointer. If this macro is defined, GCC will turn on the 813@option{-fomit-frame-pointer} option whenever @option{-O} is specified. 814@end defmac 815 816@node Per-Function Data 817@section Defining data structures for per-function information. 818@cindex per-function data 819@cindex data structures 820 821If the target needs to store information on a per-function basis, GCC 822provides a macro and a couple of variables to allow this. Note, just 823using statics to store the information is a bad idea, since GCC supports 824nested functions, so you can be halfway through encoding one function 825when another one comes along. 826 827GCC defines a data structure called @code{struct function} which 828contains all of the data specific to an individual function. This 829structure contains a field called @code{machine} whose type is 830@code{struct machine_function *}, which can be used by targets to point 831to their own specific data. 832 833If a target needs per-function specific data it should define the type 834@code{struct machine_function} and also the macro @code{INIT_EXPANDERS}. 835This macro should be used to initialize the function pointer 836@code{init_machine_status}. This pointer is explained below. 837 838One typical use of per-function, target specific data is to create an 839RTX to hold the register containing the function's return address. This 840RTX can then be used to implement the @code{__builtin_return_address} 841function, for level 0. 842 843Note---earlier implementations of GCC used a single data area to hold 844all of the per-function information. Thus when processing of a nested 845function began the old per-function data had to be pushed onto a 846stack, and when the processing was finished, it had to be popped off the 847stack. GCC used to provide function pointers called 848@code{save_machine_status} and @code{restore_machine_status} to handle 849the saving and restoring of the target specific information. Since the 850single data area approach is no longer used, these pointers are no 851longer supported. 852 853@defmac INIT_EXPANDERS 854Macro called to initialize any target specific information. This macro 855is called once per function, before generation of any RTL has begun. 856The intention of this macro is to allow the initialization of the 857function pointer @code{init_machine_status}. 858@end defmac 859 860@deftypevar {void (*)(struct function *)} init_machine_status 861If this function pointer is non-@code{NULL} it will be called once per 862function, before function compilation starts, in order to allow the 863target to perform any target specific initialization of the 864@code{struct function} structure. It is intended that this would be 865used to initialize the @code{machine} of that structure. 866 867@code{struct machine_function} structures are expected to be freed by GC@. 868Generally, any memory that they reference must be allocated by using 869@code{ggc_alloc}, including the structure itself. 870@end deftypevar 871 872@node Storage Layout 873@section Storage Layout 874@cindex storage layout 875 876Note that the definitions of the macros in this table which are sizes or 877alignments measured in bits do not need to be constant. They can be C 878expressions that refer to static variables, such as the @code{target_flags}. 879@xref{Run-time Target}. 880 881@defmac BITS_BIG_ENDIAN 882Define this macro to have the value 1 if the most significant bit in a 883byte has the lowest number; otherwise define it to have the value zero. 884This means that bit-field instructions count from the most significant 885bit. If the machine has no bit-field instructions, then this must still 886be defined, but it doesn't matter which value it is defined to. This 887macro need not be a constant. 888 889This macro does not affect the way structure fields are packed into 890bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. 891@end defmac 892 893@defmac BYTES_BIG_ENDIAN 894Define this macro to have the value 1 if the most significant byte in a 895word has the lowest number. This macro need not be a constant. 896@end defmac 897 898@defmac WORDS_BIG_ENDIAN 899Define this macro to have the value 1 if, in a multiword object, the 900most significant word has the lowest number. This applies to both 901memory locations and registers; GCC fundamentally assumes that the 902order of words in memory is the same as the order in registers. This 903macro need not be a constant. 904@end defmac 905 906@defmac LIBGCC2_WORDS_BIG_ENDIAN 907Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a 908constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be 909used only when compiling @file{libgcc2.c}. Typically the value will be set 910based on preprocessor defines. 911@end defmac 912 913@defmac FLOAT_WORDS_BIG_ENDIAN 914Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or 915@code{TFmode} floating point numbers are stored in memory with the word 916containing the sign bit at the lowest address; otherwise define it to 917have the value 0. This macro need not be a constant. 918 919You need not define this macro if the ordering is the same as for 920multi-word integers. 921@end defmac 922 923@defmac BITS_PER_UNIT 924Define this macro to be the number of bits in an addressable storage 925unit (byte). If you do not define this macro the default is 8. 926@end defmac 927 928@defmac BITS_PER_WORD 929Number of bits in a word. If you do not define this macro, the default 930is @code{BITS_PER_UNIT * UNITS_PER_WORD}. 931@end defmac 932 933@defmac MAX_BITS_PER_WORD 934Maximum number of bits in a word. If this is undefined, the default is 935@code{BITS_PER_WORD}. Otherwise, it is the constant value that is the 936largest value that @code{BITS_PER_WORD} can have at run-time. 937@end defmac 938 939@defmac UNITS_PER_WORD 940Number of storage units in a word; normally the size of a general-purpose 941register, a power of two from 1 or 8. 942@end defmac 943 944@defmac MIN_UNITS_PER_WORD 945Minimum number of units in a word. If this is undefined, the default is 946@code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the 947smallest value that @code{UNITS_PER_WORD} can have at run-time. 948@end defmac 949 950@defmac UNITS_PER_SIMD_WORD 951Number of units in the vectors that the vectorizer can produce. 952The default is equal to @code{UNITS_PER_WORD}, because the vectorizer 953can do some transformations even in absence of specialized @acronym{SIMD} 954hardware. 955@end defmac 956 957@defmac POINTER_SIZE 958Width of a pointer, in bits. You must specify a value no wider than the 959width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, 960you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify 961a value the default is @code{BITS_PER_WORD}. 962@end defmac 963 964@defmac POINTERS_EXTEND_UNSIGNED 965A C expression whose value is greater than zero if pointers that need to be 966extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to 967be zero-extended and zero if they are to be sign-extended. If the value 968is less then zero then there must be an "ptr_extend" instruction that 969extends a pointer from @code{POINTER_SIZE} to @code{Pmode}. 970 971You need not define this macro if the @code{POINTER_SIZE} is equal 972to the width of @code{Pmode}. 973@end defmac 974 975@defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) 976A macro to update @var{m} and @var{unsignedp} when an object whose type 977is @var{type} and which has the specified mode and signedness is to be 978stored in a register. This macro is only called when @var{type} is a 979scalar type. 980 981On most RISC machines, which only have operations that operate on a full 982register, define this macro to set @var{m} to @code{word_mode} if 983@var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most 984cases, only integer modes should be widened because wider-precision 985floating-point operations are usually more expensive than their narrower 986counterparts. 987 988For most machines, the macro definition does not change @var{unsignedp}. 989However, some machines, have instructions that preferentially handle 990either signed or unsigned quantities of certain modes. For example, on 991the DEC Alpha, 32-bit loads from memory and 32-bit add instructions 992sign-extend the result to 64 bits. On such machines, set 993@var{unsignedp} according to which kind of extension is more efficient. 994 995Do not define this macro if it would never modify @var{m}. 996@end defmac 997 998@defmac PROMOTE_FUNCTION_MODE 999Like @code{PROMOTE_MODE}, but is applied to outgoing function arguments or 1000function return values, as specified by @code{TARGET_PROMOTE_FUNCTION_ARGS} 1001and @code{TARGET_PROMOTE_FUNCTION_RETURN}, respectively. 1002 1003The default is @code{PROMOTE_MODE}. 1004@end defmac 1005 1006@deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_ARGS (tree @var{fntype}) 1007This target hook should return @code{true} if the promotion described by 1008@code{PROMOTE_FUNCTION_MODE} should be done for outgoing function 1009arguments. 1010@end deftypefn 1011 1012@deftypefn {Target Hook} bool TARGET_PROMOTE_FUNCTION_RETURN (tree @var{fntype}) 1013This target hook should return @code{true} if the promotion described by 1014@code{PROMOTE_FUNCTION_MODE} should be done for the return value of 1015functions. 1016 1017If this target hook returns @code{true}, @code{TARGET_FUNCTION_VALUE} 1018must perform the same promotions done by @code{PROMOTE_FUNCTION_MODE}. 1019@end deftypefn 1020 1021@defmac PARM_BOUNDARY 1022Normal alignment required for function parameters on the stack, in 1023bits. All stack parameters receive at least this much alignment 1024regardless of data type. On most machines, this is the same as the 1025size of an integer. 1026@end defmac 1027 1028@defmac STACK_BOUNDARY 1029Define this macro to the minimum alignment enforced by hardware for the 1030stack pointer on this machine. The definition is a C expression for the 1031desired alignment (measured in bits). This value is used as a default 1032if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines, 1033this should be the same as @code{PARM_BOUNDARY}. 1034@end defmac 1035 1036@defmac PREFERRED_STACK_BOUNDARY 1037Define this macro if you wish to preserve a certain alignment for the 1038stack pointer, greater than what the hardware enforces. The definition 1039is a C expression for the desired alignment (measured in bits). This 1040macro must evaluate to a value equal to or larger than 1041@code{STACK_BOUNDARY}. 1042@end defmac 1043 1044@defmac FUNCTION_BOUNDARY 1045Alignment required for a function entry point, in bits. 1046@end defmac 1047 1048@defmac BIGGEST_ALIGNMENT 1049Biggest alignment that any data type can require on this machine, in bits. 1050@end defmac 1051 1052@defmac MINIMUM_ATOMIC_ALIGNMENT 1053If defined, the smallest alignment, in bits, that can be given to an 1054object that can be referenced in one operation, without disturbing any 1055nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger 1056on machines that don't have byte or half-word store operations. 1057@end defmac 1058 1059@defmac BIGGEST_FIELD_ALIGNMENT 1060Biggest alignment that any structure or union field can require on this 1061machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for 1062structure and union fields only, unless the field alignment has been set 1063by the @code{__attribute__ ((aligned (@var{n})))} construct. 1064@end defmac 1065 1066@defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed}) 1067An expression for the alignment of a structure field @var{field} if the 1068alignment computed in the usual way (including applying of 1069@code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the 1070alignment) is @var{computed}. It overrides alignment only if the 1071field alignment has not been set by the 1072@code{__attribute__ ((aligned (@var{n})))} construct. 1073@end defmac 1074 1075@defmac MAX_OFILE_ALIGNMENT 1076Biggest alignment supported by the object file format of this machine. 1077Use this macro to limit the alignment which can be specified using the 1078@code{__attribute__ ((aligned (@var{n})))} construct. If not defined, 1079the default value is @code{BIGGEST_ALIGNMENT}. 1080@end defmac 1081 1082@defmac DATA_ALIGNMENT (@var{type}, @var{basic-align}) 1083If defined, a C expression to compute the alignment for a variable in 1084the static store. @var{type} is the data type, and @var{basic-align} is 1085the alignment that the object would ordinarily have. The value of this 1086macro is used instead of that alignment to align the object. 1087 1088If this macro is not defined, then @var{basic-align} is used. 1089 1090@findex strcpy 1091One use of this macro is to increase alignment of medium-size data to 1092make it all fit in fewer cache lines. Another is to cause character 1093arrays to be word-aligned so that @code{strcpy} calls that copy 1094constants to character arrays can be done inline. 1095@end defmac 1096 1097@defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align}) 1098If defined, a C expression to compute the alignment given to a constant 1099that is being placed in memory. @var{constant} is the constant and 1100@var{basic-align} is the alignment that the object would ordinarily 1101have. The value of this macro is used instead of that alignment to 1102align the object. 1103 1104If this macro is not defined, then @var{basic-align} is used. 1105 1106The typical use of this macro is to increase alignment for string 1107constants to be word aligned so that @code{strcpy} calls that copy 1108constants can be done inline. 1109@end defmac 1110 1111@defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align}) 1112If defined, a C expression to compute the alignment for a variable in 1113the local store. @var{type} is the data type, and @var{basic-align} is 1114the alignment that the object would ordinarily have. The value of this 1115macro is used instead of that alignment to align the object. 1116 1117If this macro is not defined, then @var{basic-align} is used. 1118 1119One use of this macro is to increase alignment of medium-size data to 1120make it all fit in fewer cache lines. 1121@end defmac 1122 1123@defmac EMPTY_FIELD_BOUNDARY 1124Alignment in bits to be given to a structure bit-field that follows an 1125empty field such as @code{int : 0;}. 1126 1127If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro. 1128@end defmac 1129 1130@defmac STRUCTURE_SIZE_BOUNDARY 1131Number of bits which any structure or union's size must be a multiple of. 1132Each structure or union's size is rounded up to a multiple of this. 1133 1134If you do not define this macro, the default is the same as 1135@code{BITS_PER_UNIT}. 1136@end defmac 1137 1138@defmac STRICT_ALIGNMENT 1139Define this macro to be the value 1 if instructions will fail to work 1140if given data not on the nominal alignment. If instructions will merely 1141go slower in that case, define this macro as 0. 1142@end defmac 1143 1144@defmac PCC_BITFIELD_TYPE_MATTERS 1145Define this if you wish to imitate the way many other C compilers handle 1146alignment of bit-fields and the structures that contain them. 1147 1148The behavior is that the type written for a named bit-field (@code{int}, 1149@code{short}, or other integer type) imposes an alignment for the entire 1150structure, as if the structure really did contain an ordinary field of 1151that type. In addition, the bit-field is placed within the structure so 1152that it would fit within such a field, not crossing a boundary for it. 1153 1154Thus, on most machines, a named bit-field whose type is written as 1155@code{int} would not cross a four-byte boundary, and would force 1156four-byte alignment for the whole structure. (The alignment used may 1157not be four bytes; it is controlled by the other alignment parameters.) 1158 1159An unnamed bit-field will not affect the alignment of the containing 1160structure. 1161 1162If the macro is defined, its definition should be a C expression; 1163a nonzero value for the expression enables this behavior. 1164 1165Note that if this macro is not defined, or its value is zero, some 1166bit-fields may cross more than one alignment boundary. The compiler can 1167support such references if there are @samp{insv}, @samp{extv}, and 1168@samp{extzv} insns that can directly reference memory. 1169 1170The other known way of making bit-fields work is to define 1171@code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. 1172Then every structure can be accessed with fullwords. 1173 1174Unless the machine has bit-field instructions or you define 1175@code{STRUCTURE_SIZE_BOUNDARY} that way, you must define 1176@code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. 1177 1178If your aim is to make GCC use the same conventions for laying out 1179bit-fields as are used by another compiler, here is how to investigate 1180what the other compiler does. Compile and run this program: 1181 1182@smallexample 1183struct foo1 1184@{ 1185 char x; 1186 char :0; 1187 char y; 1188@}; 1189 1190struct foo2 1191@{ 1192 char x; 1193 int :0; 1194 char y; 1195@}; 1196 1197main () 1198@{ 1199 printf ("Size of foo1 is %d\n", 1200 sizeof (struct foo1)); 1201 printf ("Size of foo2 is %d\n", 1202 sizeof (struct foo2)); 1203 exit (0); 1204@} 1205@end smallexample 1206 1207If this prints 2 and 5, then the compiler's behavior is what you would 1208get from @code{PCC_BITFIELD_TYPE_MATTERS}. 1209@end defmac 1210 1211@defmac BITFIELD_NBYTES_LIMITED 1212Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited 1213to aligning a bit-field within the structure. 1214@end defmac 1215 1216@deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELDS (void) 1217When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine 1218whether unnamed bitfields affect the alignment of the containing 1219structure. The hook should return true if the structure should inherit 1220the alignment requirements of an unnamed bitfield's type. 1221@end deftypefn 1222 1223@deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELDS (void) 1224This target hook should return @code{true} if accesses to volatile bitfields 1225should use the narrowest mode possible. It should return @code{false} if 1226these accesses should use the bitfield container type. 1227 1228The default is @code{!TARGET_STRICT_ALIGN}. 1229@end deftypefn 1230 1231@defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode}) 1232Return 1 if a structure or array containing @var{field} should be accessed using 1233@code{BLKMODE}. 1234 1235If @var{field} is the only field in the structure, @var{mode} is its 1236mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the 1237case where structures of one field would require the structure's mode to 1238retain the field's mode. 1239 1240Normally, this is not needed. See the file @file{c4x.h} for an example 1241of how to use this macro to prevent a structure having a floating point 1242field from being accessed in an integer mode. 1243@end defmac 1244 1245@defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified}) 1246Define this macro as an expression for the alignment of a type (given 1247by @var{type} as a tree node) if the alignment computed in the usual 1248way is @var{computed} and the alignment explicitly specified was 1249@var{specified}. 1250 1251The default is to use @var{specified} if it is larger; otherwise, use 1252the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} 1253@end defmac 1254 1255@defmac MAX_FIXED_MODE_SIZE 1256An integer expression for the size in bits of the largest integer 1257machine mode that should actually be used. All integer machine modes of 1258this size or smaller can be used for structures and unions with the 1259appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE 1260(DImode)} is assumed. 1261@end defmac 1262 1263@defmac STACK_SAVEAREA_MODE (@var{save_level}) 1264If defined, an expression of type @code{enum machine_mode} that 1265specifies the mode of the save area operand of a 1266@code{save_stack_@var{level}} named pattern (@pxref{Standard Names}). 1267@var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or 1268@code{SAVE_NONLOCAL} and selects which of the three named patterns is 1269having its mode specified. 1270 1271You need not define this macro if it always returns @code{Pmode}. You 1272would most commonly define this macro if the 1273@code{save_stack_@var{level}} patterns need to support both a 32- and a 127464-bit mode. 1275@end defmac 1276 1277@defmac STACK_SIZE_MODE 1278If defined, an expression of type @code{enum machine_mode} that 1279specifies the mode of the size increment operand of an 1280@code{allocate_stack} named pattern (@pxref{Standard Names}). 1281 1282You need not define this macro if it always returns @code{word_mode}. 1283You would most commonly define this macro if the @code{allocate_stack} 1284pattern needs to support both a 32- and a 64-bit mode. 1285@end defmac 1286 1287@defmac TARGET_FLOAT_FORMAT 1288A code distinguishing the floating point format of the target machine. 1289There are four defined values: 1290 1291@ftable @code 1292@item IEEE_FLOAT_FORMAT 1293This code indicates IEEE floating point. It is the default; there is no 1294need to define @code{TARGET_FLOAT_FORMAT} when the format is IEEE@. 1295 1296@item VAX_FLOAT_FORMAT 1297This code indicates the ``F float'' (for @code{float}) and ``D float'' 1298or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@. 1299 1300@item IBM_FLOAT_FORMAT 1301This code indicates the format used on the IBM System/370. 1302 1303@item C4X_FLOAT_FORMAT 1304This code indicates the format used on the TMS320C3x/C4x. 1305@end ftable 1306 1307If your target uses a floating point format other than these, you must 1308define a new @var{name}_FLOAT_FORMAT code for it, and add support for 1309it to @file{real.c}. 1310 1311The ordering of the component words of floating point values stored in 1312memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}. 1313@end defmac 1314 1315@defmac MODE_HAS_NANS (@var{mode}) 1316When defined, this macro should be true if @var{mode} has a NaN 1317representation. The compiler assumes that NaNs are not equal to 1318anything (including themselves) and that addition, subtraction, 1319multiplication and division all return NaNs when one operand is 1320NaN@. 1321 1322By default, this macro is true if @var{mode} is a floating-point 1323mode and the target floating-point format is IEEE@. 1324@end defmac 1325 1326@defmac MODE_HAS_INFINITIES (@var{mode}) 1327This macro should be true if @var{mode} can represent infinity. At 1328present, the compiler uses this macro to decide whether @samp{x - x} 1329is always defined. By default, the macro is true when @var{mode} 1330is a floating-point mode and the target format is IEEE@. 1331@end defmac 1332 1333@defmac MODE_HAS_SIGNED_ZEROS (@var{mode}) 1334True if @var{mode} distinguishes between positive and negative zero. 1335The rules are expected to follow the IEEE standard: 1336 1337@itemize @bullet 1338@item 1339@samp{x + x} has the same sign as @samp{x}. 1340 1341@item 1342If the sum of two values with opposite sign is zero, the result is 1343positive for all rounding modes expect towards @minus{}infinity, for 1344which it is negative. 1345 1346@item 1347The sign of a product or quotient is negative when exactly one 1348of the operands is negative. 1349@end itemize 1350 1351The default definition is true if @var{mode} is a floating-point 1352mode and the target format is IEEE@. 1353@end defmac 1354 1355@defmac MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode}) 1356If defined, this macro should be true for @var{mode} if it has at 1357least one rounding mode in which @samp{x} and @samp{-x} can be 1358rounded to numbers of different magnitude. Two such modes are 1359towards @minus{}infinity and towards +infinity. 1360 1361The default definition of this macro is true if @var{mode} is 1362a floating-point mode and the target format is IEEE@. 1363@end defmac 1364 1365@defmac ROUND_TOWARDS_ZERO 1366If defined, this macro should be true if the prevailing rounding 1367mode is towards zero. A true value has the following effects: 1368 1369@itemize @bullet 1370@item 1371@code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes. 1372 1373@item 1374@file{libgcc.a}'s floating-point emulator will round towards zero 1375rather than towards nearest. 1376 1377@item 1378The compiler's floating-point emulator will round towards zero after 1379doing arithmetic, and when converting from the internal float format to 1380the target format. 1381@end itemize 1382 1383The macro does not affect the parsing of string literals. When the 1384primary rounding mode is towards zero, library functions like 1385@code{strtod} might still round towards nearest, and the compiler's 1386parser should behave like the target's @code{strtod} where possible. 1387 1388Not defining this macro is equivalent to returning zero. 1389@end defmac 1390 1391@defmac LARGEST_EXPONENT_IS_NORMAL (@var{size}) 1392This macro should return true if floats with @var{size} 1393bits do not have a NaN or infinity representation, but use the largest 1394exponent for normal numbers instead. 1395 1396Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS} 1397and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes. 1398It also affects the way @file{libgcc.a} and @file{real.c} emulate 1399floating-point arithmetic. 1400 1401The default definition of this macro returns false for all sizes. 1402@end defmac 1403 1404@deftypefn {Target Hook} bool TARGET_VECTOR_OPAQUE_P (tree @var{type}) 1405This target hook should return @code{true} a vector is opaque. That 1406is, if no cast is needed when copying a vector value of type 1407@var{type} into another vector lvalue of the same size. Vector opaque 1408types cannot be initialized. The default is that there are no such 1409types. 1410@end deftypefn 1411 1412@deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type}) 1413This target hook returns @code{true} if bit-fields in the given 1414@var{record_type} are to be laid out following the rules of Microsoft 1415Visual C/C++, namely: (i) a bit-field won't share the same storage 1416unit with the previous bit-field if their underlying types have 1417different sizes, and the bit-field will be aligned to the highest 1418alignment of the underlying types of itself and of the previous 1419bit-field; (ii) a zero-sized bit-field will affect the alignment of 1420the whole enclosing structure, even if it is unnamed; except that 1421(iii) a zero-sized bit-field will be disregarded unless it follows 1422another bit-field of nonzero size. If this hook returns @code{true}, 1423other macros that control bit-field layout are ignored. 1424 1425When a bit-field is inserted into a packed record, the whole size 1426of the underlying type is used by one or more same-size adjacent 1427bit-fields (that is, if its long:3, 32 bits is used in the record, 1428and any additional adjacent long bit-fields are packed into the same 1429chunk of 32 bits. However, if the size changes, a new field of that 1430size is allocated). In an unpacked record, this is the same as using 1431alignment, but not equivalent when packing. 1432 1433If both MS bit-fields and @samp{__attribute__((packed))} are used, 1434the latter will take precedence. If @samp{__attribute__((packed))} is 1435used on a single field when MS bit-fields are in use, it will take 1436precedence for that field, but the alignment of the rest of the structure 1437may affect its placement. 1438@end deftypefn 1439 1440@deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void) 1441Returns true if the target supports decimal floating point. 1442@end deftypefn 1443 1444@deftypefn {Target Hook} {const char *} TARGET_MANGLE_FUNDAMENTAL_TYPE (tree @var{type}) 1445If your target defines any fundamental types, define this hook to 1446return the appropriate encoding for these types as part of a C++ 1447mangled name. The @var{type} argument is the tree structure 1448representing the type to be mangled. The hook may be applied to trees 1449which are not target-specific fundamental types; it should return 1450@code{NULL} for all such types, as well as arguments it does not 1451recognize. If the return value is not @code{NULL}, it must point to 1452a statically-allocated string constant. 1453 1454Target-specific fundamental types might be new fundamental types or 1455qualified versions of ordinary fundamental types. Encode new 1456fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name} 1457is the name used for the type in source code, and @var{n} is the 1458length of @var{name} in decimal. Encode qualified versions of 1459ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where 1460@var{name} is the name used for the type qualifier in source code, 1461@var{n} is the length of @var{name} as above, and @var{code} is the 1462code used to represent the unqualified version of this type. (See 1463@code{write_builtin_type} in @file{cp/mangle.c} for the list of 1464codes.) In both cases the spaces are for clarity; do not include any 1465spaces in your string. 1466 1467The default version of this hook always returns @code{NULL}, which is 1468appropriate for a target that does not define any new fundamental 1469types. 1470@end deftypefn 1471 1472@node Type Layout 1473@section Layout of Source Language Data Types 1474 1475These macros define the sizes and other characteristics of the standard 1476basic data types used in programs being compiled. Unlike the macros in 1477the previous section, these apply to specific features of C and related 1478languages, rather than to fundamental aspects of storage layout. 1479 1480@defmac INT_TYPE_SIZE 1481A C expression for the size in bits of the type @code{int} on the 1482target machine. If you don't define this, the default is one word. 1483@end defmac 1484 1485@defmac SHORT_TYPE_SIZE 1486A C expression for the size in bits of the type @code{short} on the 1487target machine. If you don't define this, the default is half a word. 1488(If this would be less than one storage unit, it is rounded up to one 1489unit.) 1490@end defmac 1491 1492@defmac LONG_TYPE_SIZE 1493A C expression for the size in bits of the type @code{long} on the 1494target machine. If you don't define this, the default is one word. 1495@end defmac 1496 1497@defmac ADA_LONG_TYPE_SIZE 1498On some machines, the size used for the Ada equivalent of the type 1499@code{long} by a native Ada compiler differs from that used by C@. In 1500that situation, define this macro to be a C expression to be used for 1501the size of that type. If you don't define this, the default is the 1502value of @code{LONG_TYPE_SIZE}. 1503@end defmac 1504 1505@defmac LONG_LONG_TYPE_SIZE 1506A C expression for the size in bits of the type @code{long long} on the 1507target machine. If you don't define this, the default is two 1508words. If you want to support GNU Ada on your machine, the value of this 1509macro must be at least 64. 1510@end defmac 1511 1512@defmac CHAR_TYPE_SIZE 1513A C expression for the size in bits of the type @code{char} on the 1514target machine. If you don't define this, the default is 1515@code{BITS_PER_UNIT}. 1516@end defmac 1517 1518@defmac BOOL_TYPE_SIZE 1519A C expression for the size in bits of the C++ type @code{bool} and 1520C99 type @code{_Bool} on the target machine. If you don't define 1521this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}. 1522@end defmac 1523 1524@defmac FLOAT_TYPE_SIZE 1525A C expression for the size in bits of the type @code{float} on the 1526target machine. If you don't define this, the default is one word. 1527@end defmac 1528 1529@defmac DOUBLE_TYPE_SIZE 1530A C expression for the size in bits of the type @code{double} on the 1531target machine. If you don't define this, the default is two 1532words. 1533@end defmac 1534 1535@defmac LONG_DOUBLE_TYPE_SIZE 1536A C expression for the size in bits of the type @code{long double} on 1537the target machine. If you don't define this, the default is two 1538words. 1539@end defmac 1540 1541@defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE 1542Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or 1543if you want routines in @file{libgcc2.a} for a size other than 1544@code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the 1545default is @code{LONG_DOUBLE_TYPE_SIZE}. 1546@end defmac 1547 1548@defmac LIBGCC2_HAS_DF_MODE 1549Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor 1550@code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 1551@code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a} 1552anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE} 1553or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1, 1554otherwise it is 0. 1555@end defmac 1556 1557@defmac LIBGCC2_HAS_XF_MODE 1558Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not 1559@code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a} 1560anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} 1561is 80 then the default is 1, otherwise it is 0. 1562@end defmac 1563 1564@defmac LIBGCC2_HAS_TF_MODE 1565Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not 1566@code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a} 1567anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} 1568is 128 then the default is 1, otherwise it is 0. 1569@end defmac 1570 1571@defmac SF_SIZE 1572@defmacx DF_SIZE 1573@defmacx XF_SIZE 1574@defmacx TF_SIZE 1575Define these macros to be the size in bits of the mantissa of 1576@code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values, 1577if the defaults in @file{libgcc2.h} are inappropriate. By default, 1578@code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG} 1579for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or 1580@code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether 1581@code{LIBGCC2_DOUBLE_TYPE_SIZE} or 1582@code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64. 1583@end defmac 1584 1585@defmac TARGET_FLT_EVAL_METHOD 1586A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h}, 1587assuming, if applicable, that the floating-point control word is in its 1588default state. If you do not define this macro the value of 1589@code{FLT_EVAL_METHOD} will be zero. 1590@end defmac 1591 1592@defmac WIDEST_HARDWARE_FP_SIZE 1593A C expression for the size in bits of the widest floating-point format 1594supported by the hardware. If you define this macro, you must specify a 1595value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}. 1596If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE} 1597is the default. 1598@end defmac 1599 1600@defmac DEFAULT_SIGNED_CHAR 1601An expression whose value is 1 or 0, according to whether the type 1602@code{char} should be signed or unsigned by default. The user can 1603always override this default with the options @option{-fsigned-char} 1604and @option{-funsigned-char}. 1605@end defmac 1606 1607@deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void) 1608This target hook should return true if the compiler should give an 1609@code{enum} type only as many bytes as it takes to represent the range 1610of possible values of that type. It should return false if all 1611@code{enum} types should be allocated like @code{int}. 1612 1613The default is to return false. 1614@end deftypefn 1615 1616@defmac SIZE_TYPE 1617A C expression for a string describing the name of the data type to use 1618for size values. The typedef name @code{size_t} is defined using the 1619contents of the string. 1620 1621The string can contain more than one keyword. If so, separate them with 1622spaces, and write first any length keyword, then @code{unsigned} if 1623appropriate, and finally @code{int}. The string must exactly match one 1624of the data type names defined in the function 1625@code{init_decl_processing} in the file @file{c-decl.c}. You may not 1626omit @code{int} or change the order---that would cause the compiler to 1627crash on startup. 1628 1629If you don't define this macro, the default is @code{"long unsigned 1630int"}. 1631@end defmac 1632 1633@defmac PTRDIFF_TYPE 1634A C expression for a string describing the name of the data type to use 1635for the result of subtracting two pointers. The typedef name 1636@code{ptrdiff_t} is defined using the contents of the string. See 1637@code{SIZE_TYPE} above for more information. 1638 1639If you don't define this macro, the default is @code{"long int"}. 1640@end defmac 1641 1642@defmac WCHAR_TYPE 1643A C expression for a string describing the name of the data type to use 1644for wide characters. The typedef name @code{wchar_t} is defined using 1645the contents of the string. See @code{SIZE_TYPE} above for more 1646information. 1647 1648If you don't define this macro, the default is @code{"int"}. 1649@end defmac 1650 1651@defmac WCHAR_TYPE_SIZE 1652A C expression for the size in bits of the data type for wide 1653characters. This is used in @code{cpp}, which cannot make use of 1654@code{WCHAR_TYPE}. 1655@end defmac 1656 1657@defmac WINT_TYPE 1658A C expression for a string describing the name of the data type to 1659use for wide characters passed to @code{printf} and returned from 1660@code{getwc}. The typedef name @code{wint_t} is defined using the 1661contents of the string. See @code{SIZE_TYPE} above for more 1662information. 1663 1664If you don't define this macro, the default is @code{"unsigned int"}. 1665@end defmac 1666 1667@defmac INTMAX_TYPE 1668A C expression for a string describing the name of the data type that 1669can represent any value of any standard or extended signed integer type. 1670The typedef name @code{intmax_t} is defined using the contents of the 1671string. See @code{SIZE_TYPE} above for more information. 1672 1673If you don't define this macro, the default is the first of 1674@code{"int"}, @code{"long int"}, or @code{"long long int"} that has as 1675much precision as @code{long long int}. 1676@end defmac 1677 1678@defmac UINTMAX_TYPE 1679A C expression for a string describing the name of the data type that 1680can represent any value of any standard or extended unsigned integer 1681type. The typedef name @code{uintmax_t} is defined using the contents 1682of the string. See @code{SIZE_TYPE} above for more information. 1683 1684If you don't define this macro, the default is the first of 1685@code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long 1686unsigned int"} that has as much precision as @code{long long unsigned 1687int}. 1688@end defmac 1689 1690@defmac TARGET_PTRMEMFUNC_VBIT_LOCATION 1691The C++ compiler represents a pointer-to-member-function with a struct 1692that looks like: 1693 1694@smallexample 1695 struct @{ 1696 union @{ 1697 void (*fn)(); 1698 ptrdiff_t vtable_index; 1699 @}; 1700 ptrdiff_t delta; 1701 @}; 1702@end smallexample 1703 1704@noindent 1705The C++ compiler must use one bit to indicate whether the function that 1706will be called through a pointer-to-member-function is virtual. 1707Normally, we assume that the low-order bit of a function pointer must 1708always be zero. Then, by ensuring that the vtable_index is odd, we can 1709distinguish which variant of the union is in use. But, on some 1710platforms function pointers can be odd, and so this doesn't work. In 1711that case, we use the low-order bit of the @code{delta} field, and shift 1712the remainder of the @code{delta} field to the left. 1713 1714GCC will automatically make the right selection about where to store 1715this bit using the @code{FUNCTION_BOUNDARY} setting for your platform. 1716However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY} 1717set such that functions always start at even addresses, but the lowest 1718bit of pointers to functions indicate whether the function at that 1719address is in ARM or Thumb mode. If this is the case of your 1720architecture, you should define this macro to 1721@code{ptrmemfunc_vbit_in_delta}. 1722 1723In general, you should not have to define this macro. On architectures 1724in which function addresses are always even, according to 1725@code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to 1726@code{ptrmemfunc_vbit_in_pfn}. 1727@end defmac 1728 1729@defmac TARGET_VTABLE_USES_DESCRIPTORS 1730Normally, the C++ compiler uses function pointers in vtables. This 1731macro allows the target to change to use ``function descriptors'' 1732instead. Function descriptors are found on targets for whom a 1733function pointer is actually a small data structure. Normally the 1734data structure consists of the actual code address plus a data 1735pointer to which the function's data is relative. 1736 1737If vtables are used, the value of this macro should be the number 1738of words that the function descriptor occupies. 1739@end defmac 1740 1741@defmac TARGET_VTABLE_ENTRY_ALIGN 1742By default, the vtable entries are void pointers, the so the alignment 1743is the same as pointer alignment. The value of this macro specifies 1744the alignment of the vtable entry in bits. It should be defined only 1745when special alignment is necessary. */ 1746@end defmac 1747 1748@defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE 1749There are a few non-descriptor entries in the vtable at offsets below 1750zero. If these entries must be padded (say, to preserve the alignment 1751specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number 1752of words in each data entry. 1753@end defmac 1754 1755@node Registers 1756@section Register Usage 1757@cindex register usage 1758 1759This section explains how to describe what registers the target machine 1760has, and how (in general) they can be used. 1761 1762The description of which registers a specific instruction can use is 1763done with register classes; see @ref{Register Classes}. For information 1764on using registers to access a stack frame, see @ref{Frame Registers}. 1765For passing values in registers, see @ref{Register Arguments}. 1766For returning values in registers, see @ref{Scalar Return}. 1767 1768@menu 1769* Register Basics:: Number and kinds of registers. 1770* Allocation Order:: Order in which registers are allocated. 1771* Values in Registers:: What kinds of values each reg can hold. 1772* Leaf Functions:: Renumbering registers for leaf functions. 1773* Stack Registers:: Handling a register stack such as 80387. 1774@end menu 1775 1776@node Register Basics 1777@subsection Basic Characteristics of Registers 1778 1779@c prevent bad page break with this line 1780Registers have various characteristics. 1781 1782@defmac FIRST_PSEUDO_REGISTER 1783Number of hardware registers known to the compiler. They receive 1784numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first 1785pseudo register's number really is assigned the number 1786@code{FIRST_PSEUDO_REGISTER}. 1787@end defmac 1788 1789@defmac FIXED_REGISTERS 1790@cindex fixed register 1791An initializer that says which registers are used for fixed purposes 1792all throughout the compiled code and are therefore not available for 1793general allocation. These would include the stack pointer, the frame 1794pointer (except on machines where that can be used as a general 1795register when no frame pointer is needed), the program counter on 1796machines where that is considered one of the addressable registers, 1797and any other numbered register with a standard use. 1798 1799This information is expressed as a sequence of numbers, separated by 1800commas and surrounded by braces. The @var{n}th number is 1 if 1801register @var{n} is fixed, 0 otherwise. 1802 1803The table initialized from this macro, and the table initialized by 1804the following one, may be overridden at run time either automatically, 1805by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by 1806the user with the command options @option{-ffixed-@var{reg}}, 1807@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}. 1808@end defmac 1809 1810@defmac CALL_USED_REGISTERS 1811@cindex call-used register 1812@cindex call-clobbered register 1813@cindex call-saved register 1814Like @code{FIXED_REGISTERS} but has 1 for each register that is 1815clobbered (in general) by function calls as well as for fixed 1816registers. This macro therefore identifies the registers that are not 1817available for general allocation of values that must live across 1818function calls. 1819 1820If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler 1821automatically saves it on function entry and restores it on function 1822exit, if the register is used within the function. 1823@end defmac 1824 1825@defmac CALL_REALLY_USED_REGISTERS 1826@cindex call-used register 1827@cindex call-clobbered register 1828@cindex call-saved register 1829Like @code{CALL_USED_REGISTERS} except this macro doesn't require 1830that the entire set of @code{FIXED_REGISTERS} be included. 1831(@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}). 1832This macro is optional. If not specified, it defaults to the value 1833of @code{CALL_USED_REGISTERS}. 1834@end defmac 1835 1836@defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode}) 1837@cindex call-used register 1838@cindex call-clobbered register 1839@cindex call-saved register 1840A C expression that is nonzero if it is not permissible to store a 1841value of mode @var{mode} in hard register number @var{regno} across a 1842call without some part of it being clobbered. For most machines this 1843macro need not be defined. It is only required for machines that do not 1844preserve the entire contents of a register across a call. 1845@end defmac 1846 1847@findex fixed_regs 1848@findex call_used_regs 1849@findex global_regs 1850@findex reg_names 1851@findex reg_class_contents 1852@defmac CONDITIONAL_REGISTER_USAGE 1853Zero or more C statements that may conditionally modify five variables 1854@code{fixed_regs}, @code{call_used_regs}, @code{global_regs}, 1855@code{reg_names}, and @code{reg_class_contents}, to take into account 1856any dependence of these register sets on target flags. The first three 1857of these are of type @code{char []} (interpreted as Boolean vectors). 1858@code{global_regs} is a @code{const char *[]}, and 1859@code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is 1860called, @code{fixed_regs}, @code{call_used_regs}, 1861@code{reg_class_contents}, and @code{reg_names} have been initialized 1862from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS}, 1863@code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively. 1864@code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}}, 1865@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}} 1866command options have been applied. 1867 1868You need not define this macro if it has no work to do. 1869 1870@cindex disabling certain registers 1871@cindex controlling register usage 1872If the usage of an entire class of registers depends on the target 1873flags, you may indicate this to GCC by using this macro to modify 1874@code{fixed_regs} and @code{call_used_regs} to 1 for each of the 1875registers in the classes which should not be used by GCC@. Also define 1876the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT} 1877to return @code{NO_REGS} if it 1878is called with a letter for a class that shouldn't be used. 1879 1880(However, if this class is not included in @code{GENERAL_REGS} and all 1881of the insn patterns whose constraints permit this class are 1882controlled by target switches, then GCC will automatically avoid using 1883these registers when the target switches are opposed to them.) 1884@end defmac 1885 1886@defmac INCOMING_REGNO (@var{out}) 1887Define this macro if the target machine has register windows. This C 1888expression returns the register number as seen by the called function 1889corresponding to the register number @var{out} as seen by the calling 1890function. Return @var{out} if register number @var{out} is not an 1891outbound register. 1892@end defmac 1893 1894@defmac OUTGOING_REGNO (@var{in}) 1895Define this macro if the target machine has register windows. This C 1896expression returns the register number as seen by the calling function 1897corresponding to the register number @var{in} as seen by the called 1898function. Return @var{in} if register number @var{in} is not an inbound 1899register. 1900@end defmac 1901 1902@defmac LOCAL_REGNO (@var{regno}) 1903Define this macro if the target machine has register windows. This C 1904expression returns true if the register is call-saved but is in the 1905register window. Unlike most call-saved registers, such registers 1906need not be explicitly restored on function exit or during non-local 1907gotos. 1908@end defmac 1909 1910@defmac PC_REGNUM 1911If the program counter has a register number, define this as that 1912register number. Otherwise, do not define it. 1913@end defmac 1914 1915@node Allocation Order 1916@subsection Order of Allocation of Registers 1917@cindex order of register allocation 1918@cindex register allocation order 1919 1920@c prevent bad page break with this line 1921Registers are allocated in order. 1922 1923@defmac REG_ALLOC_ORDER 1924If defined, an initializer for a vector of integers, containing the 1925numbers of hard registers in the order in which GCC should prefer 1926to use them (from most preferred to least). 1927 1928If this macro is not defined, registers are used lowest numbered first 1929(all else being equal). 1930 1931One use of this macro is on machines where the highest numbered 1932registers must always be saved and the save-multiple-registers 1933instruction supports only sequences of consecutive registers. On such 1934machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists 1935the highest numbered allocable register first. 1936@end defmac 1937 1938@defmac ORDER_REGS_FOR_LOCAL_ALLOC 1939A C statement (sans semicolon) to choose the order in which to allocate 1940hard registers for pseudo-registers local to a basic block. 1941 1942Store the desired register order in the array @code{reg_alloc_order}. 1943Element 0 should be the register to allocate first; element 1, the next 1944register; and so on. 1945 1946The macro body should not assume anything about the contents of 1947@code{reg_alloc_order} before execution of the macro. 1948 1949On most machines, it is not necessary to define this macro. 1950@end defmac 1951 1952@node Values in Registers 1953@subsection How Values Fit in Registers 1954 1955This section discusses the macros that describe which kinds of values 1956(specifically, which machine modes) each register can hold, and how many 1957consecutive registers are needed for a given mode. 1958 1959@defmac HARD_REGNO_NREGS (@var{regno}, @var{mode}) 1960A C expression for the number of consecutive hard registers, starting 1961at register number @var{regno}, required to hold a value of mode 1962@var{mode}. 1963 1964On a machine where all registers are exactly one word, a suitable 1965definition of this macro is 1966 1967@smallexample 1968#define HARD_REGNO_NREGS(REGNO, MODE) \ 1969 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ 1970 / UNITS_PER_WORD) 1971@end smallexample 1972@end defmac 1973 1974@defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode}) 1975A C expression that is nonzero if a value of mode @var{mode}, stored 1976in memory, ends with padding that causes it to take up more space than 1977in registers starting at register number @var{regno} (as determined by 1978multiplying GCC's notion of the size of the register when containing 1979this mode by the number of registers returned by 1980@code{HARD_REGNO_NREGS}). By default this is zero. 1981 1982For example, if a floating-point value is stored in three 32-bit 1983registers but takes up 128 bits in memory, then this would be 1984nonzero. 1985 1986This macros only needs to be defined if there are cases where 1987@code{subreg_regno_offset} and @code{subreg_offset_representable_p} 1988would otherwise wrongly determine that a @code{subreg} can be 1989represented by an offset to the register number, when in fact such a 1990@code{subreg} would contain some of the padding not stored in 1991registers and so not be representable. 1992@end defmac 1993 1994@defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode}) 1995For values of @var{regno} and @var{mode} for which 1996@code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression 1997returning the greater number of registers required to hold the value 1998including any padding. In the example above, the value would be four. 1999@end defmac 2000 2001@defmac REGMODE_NATURAL_SIZE (@var{mode}) 2002Define this macro if the natural size of registers that hold values 2003of mode @var{mode} is not the word size. It is a C expression that 2004should give the natural size in bytes for the specified mode. It is 2005used by the register allocator to try to optimize its results. This 2006happens for example on SPARC 64-bit where the natural size of 2007floating-point registers is still 32-bit. 2008@end defmac 2009 2010@defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode}) 2011A C expression that is nonzero if it is permissible to store a value 2012of mode @var{mode} in hard register number @var{regno} (or in several 2013registers starting with that one). For a machine where all registers 2014are equivalent, a suitable definition is 2015 2016@smallexample 2017#define HARD_REGNO_MODE_OK(REGNO, MODE) 1 2018@end smallexample 2019 2020You need not include code to check for the numbers of fixed registers, 2021because the allocation mechanism considers them to be always occupied. 2022 2023@cindex register pairs 2024On some machines, double-precision values must be kept in even/odd 2025register pairs. You can implement that by defining this macro to reject 2026odd register numbers for such modes. 2027 2028The minimum requirement for a mode to be OK in a register is that the 2029@samp{mov@var{mode}} instruction pattern support moves between the 2030register and other hard register in the same class and that moving a 2031value into the register and back out not alter it. 2032 2033Since the same instruction used to move @code{word_mode} will work for 2034all narrower integer modes, it is not necessary on any machine for 2035@code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided 2036you define patterns @samp{movhi}, etc., to take advantage of this. This 2037is useful because of the interaction between @code{HARD_REGNO_MODE_OK} 2038and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes 2039to be tieable. 2040 2041Many machines have special registers for floating point arithmetic. 2042Often people assume that floating point machine modes are allowed only 2043in floating point registers. This is not true. Any registers that 2044can hold integers can safely @emph{hold} a floating point machine 2045mode, whether or not floating arithmetic can be done on it in those 2046registers. Integer move instructions can be used to move the values. 2047 2048On some machines, though, the converse is true: fixed-point machine 2049modes may not go in floating registers. This is true if the floating 2050registers normalize any value stored in them, because storing a 2051non-floating value there would garble it. In this case, 2052@code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in 2053floating registers. But if the floating registers do not automatically 2054normalize, if you can store any bit pattern in one and retrieve it 2055unchanged without a trap, then any machine mode may go in a floating 2056register, so you can define this macro to say so. 2057 2058The primary significance of special floating registers is rather that 2059they are the registers acceptable in floating point arithmetic 2060instructions. However, this is of no concern to 2061@code{HARD_REGNO_MODE_OK}. You handle it by writing the proper 2062constraints for those instructions. 2063 2064On some machines, the floating registers are especially slow to access, 2065so that it is better to store a value in a stack frame than in such a 2066register if floating point arithmetic is not being done. As long as the 2067floating registers are not in class @code{GENERAL_REGS}, they will not 2068be used unless some pattern's constraint asks for one. 2069@end defmac 2070 2071@defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to}) 2072A C expression that is nonzero if it is OK to rename a hard register 2073@var{from} to another hard register @var{to}. 2074 2075One common use of this macro is to prevent renaming of a register to 2076another register that is not saved by a prologue in an interrupt 2077handler. 2078 2079The default is always nonzero. 2080@end defmac 2081 2082@defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2}) 2083A C expression that is nonzero if a value of mode 2084@var{mode1} is accessible in mode @var{mode2} without copying. 2085 2086If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and 2087@code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for 2088any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})} 2089should be nonzero. If they differ for any @var{r}, you should define 2090this macro to return zero unless some other mechanism ensures the 2091accessibility of the value in a narrower mode. 2092 2093You should define this macro to return nonzero in as many cases as 2094possible since doing so will allow GCC to perform better register 2095allocation. 2096@end defmac 2097 2098@defmac AVOID_CCMODE_COPIES 2099Define this macro if the compiler should avoid copies to/from @code{CCmode} 2100registers. You should only define this macro if support for copying to/from 2101@code{CCmode} is incomplete. 2102@end defmac 2103 2104@node Leaf Functions 2105@subsection Handling Leaf Functions 2106 2107@cindex leaf functions 2108@cindex functions, leaf 2109On some machines, a leaf function (i.e., one which makes no calls) can run 2110more efficiently if it does not make its own register window. Often this 2111means it is required to receive its arguments in the registers where they 2112are passed by the caller, instead of the registers where they would 2113normally arrive. 2114 2115The special treatment for leaf functions generally applies only when 2116other conditions are met; for example, often they may use only those 2117registers for its own variables and temporaries. We use the term ``leaf 2118function'' to mean a function that is suitable for this special 2119handling, so that functions with no calls are not necessarily ``leaf 2120functions''. 2121 2122GCC assigns register numbers before it knows whether the function is 2123suitable for leaf function treatment. So it needs to renumber the 2124registers in order to output a leaf function. The following macros 2125accomplish this. 2126 2127@defmac LEAF_REGISTERS 2128Name of a char vector, indexed by hard register number, which 2129contains 1 for a register that is allowable in a candidate for leaf 2130function treatment. 2131 2132If leaf function treatment involves renumbering the registers, then the 2133registers marked here should be the ones before renumbering---those that 2134GCC would ordinarily allocate. The registers which will actually be 2135used in the assembler code, after renumbering, should not be marked with 1 2136in this vector. 2137 2138Define this macro only if the target machine offers a way to optimize 2139the treatment of leaf functions. 2140@end defmac 2141 2142@defmac LEAF_REG_REMAP (@var{regno}) 2143A C expression whose value is the register number to which @var{regno} 2144should be renumbered, when a function is treated as a leaf function. 2145 2146If @var{regno} is a register number which should not appear in a leaf 2147function before renumbering, then the expression should yield @minus{}1, which 2148will cause the compiler to abort. 2149 2150Define this macro only if the target machine offers a way to optimize the 2151treatment of leaf functions, and registers need to be renumbered to do 2152this. 2153@end defmac 2154 2155@findex current_function_is_leaf 2156@findex current_function_uses_only_leaf_regs 2157@code{TARGET_ASM_FUNCTION_PROLOGUE} and 2158@code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions 2159specially. They can test the C variable @code{current_function_is_leaf} 2160which is nonzero for leaf functions. @code{current_function_is_leaf} is 2161set prior to local register allocation and is valid for the remaining 2162compiler passes. They can also test the C variable 2163@code{current_function_uses_only_leaf_regs} which is nonzero for leaf 2164functions which only use leaf registers. 2165@code{current_function_uses_only_leaf_regs} is valid after all passes 2166that modify the instructions have been run and is only useful if 2167@code{LEAF_REGISTERS} is defined. 2168@c changed this to fix overfull. ALSO: why the "it" at the beginning 2169@c of the next paragraph?! --mew 2feb93 2170 2171@node Stack Registers 2172@subsection Registers That Form a Stack 2173 2174There are special features to handle computers where some of the 2175``registers'' form a stack. Stack registers are normally written by 2176pushing onto the stack, and are numbered relative to the top of the 2177stack. 2178 2179Currently, GCC can only handle one group of stack-like registers, and 2180they must be consecutively numbered. Furthermore, the existing 2181support for stack-like registers is specific to the 80387 floating 2182point coprocessor. If you have a new architecture that uses 2183stack-like registers, you will need to do substantial work on 2184@file{reg-stack.c} and write your machine description to cooperate 2185with it, as well as defining these macros. 2186 2187@defmac STACK_REGS 2188Define this if the machine has any stack-like registers. 2189@end defmac 2190 2191@defmac FIRST_STACK_REG 2192The number of the first stack-like register. This one is the top 2193of the stack. 2194@end defmac 2195 2196@defmac LAST_STACK_REG 2197The number of the last stack-like register. This one is the bottom of 2198the stack. 2199@end defmac 2200 2201@node Register Classes 2202@section Register Classes 2203@cindex register class definitions 2204@cindex class definitions, register 2205 2206On many machines, the numbered registers are not all equivalent. 2207For example, certain registers may not be allowed for indexed addressing; 2208certain registers may not be allowed in some instructions. These machine 2209restrictions are described to the compiler using @dfn{register classes}. 2210 2211You define a number of register classes, giving each one a name and saying 2212which of the registers belong to it. Then you can specify register classes 2213that are allowed as operands to particular instruction patterns. 2214 2215@findex ALL_REGS 2216@findex NO_REGS 2217In general, each register will belong to several classes. In fact, one 2218class must be named @code{ALL_REGS} and contain all the registers. Another 2219class must be named @code{NO_REGS} and contain no registers. Often the 2220union of two classes will be another class; however, this is not required. 2221 2222@findex GENERAL_REGS 2223One of the classes must be named @code{GENERAL_REGS}. There is nothing 2224terribly special about the name, but the operand constraint letters 2225@samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is 2226the same as @code{ALL_REGS}, just define it as a macro which expands 2227to @code{ALL_REGS}. 2228 2229Order the classes so that if class @var{x} is contained in class @var{y} 2230then @var{x} has a lower class number than @var{y}. 2231 2232The way classes other than @code{GENERAL_REGS} are specified in operand 2233constraints is through machine-dependent operand constraint letters. 2234You can define such letters to correspond to various classes, then use 2235them in operand constraints. 2236 2237You should define a class for the union of two classes whenever some 2238instruction allows both classes. For example, if an instruction allows 2239either a floating point (coprocessor) register or a general register for a 2240certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} 2241which includes both of them. Otherwise you will get suboptimal code. 2242 2243You must also specify certain redundant information about the register 2244classes: for each class, which classes contain it and which ones are 2245contained in it; for each pair of classes, the largest class contained 2246in their union. 2247 2248When a value occupying several consecutive registers is expected in a 2249certain class, all the registers used must belong to that class. 2250Therefore, register classes cannot be used to enforce a requirement for 2251a register pair to start with an even-numbered register. The way to 2252specify this requirement is with @code{HARD_REGNO_MODE_OK}. 2253 2254Register classes used for input-operands of bitwise-and or shift 2255instructions have a special requirement: each such class must have, for 2256each fixed-point machine mode, a subclass whose registers can transfer that 2257mode to or from memory. For example, on some machines, the operations for 2258single-byte values (@code{QImode}) are limited to certain registers. When 2259this is so, each register class that is used in a bitwise-and or shift 2260instruction must have a subclass consisting of registers from which 2261single-byte values can be loaded or stored. This is so that 2262@code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. 2263 2264@deftp {Data type} {enum reg_class} 2265An enumerated type that must be defined with all the register class names 2266as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS} 2267must be the last register class, followed by one more enumerated value, 2268@code{LIM_REG_CLASSES}, which is not a register class but rather 2269tells how many classes there are. 2270 2271Each register class has a number, which is the value of casting 2272the class name to type @code{int}. The number serves as an index 2273in many of the tables described below. 2274@end deftp 2275 2276@defmac N_REG_CLASSES 2277The number of distinct register classes, defined as follows: 2278 2279@smallexample 2280#define N_REG_CLASSES (int) LIM_REG_CLASSES 2281@end smallexample 2282@end defmac 2283 2284@defmac REG_CLASS_NAMES 2285An initializer containing the names of the register classes as C string 2286constants. These names are used in writing some of the debugging dumps. 2287@end defmac 2288 2289@defmac REG_CLASS_CONTENTS 2290An initializer containing the contents of the register classes, as integers 2291which are bit masks. The @var{n}th integer specifies the contents of class 2292@var{n}. The way the integer @var{mask} is interpreted is that 2293register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. 2294 2295When the machine has more than 32 registers, an integer does not suffice. 2296Then the integers are replaced by sub-initializers, braced groupings containing 2297several integers. Each sub-initializer must be suitable as an initializer 2298for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. 2299In this situation, the first integer in each sub-initializer corresponds to 2300registers 0 through 31, the second integer to registers 32 through 63, and 2301so on. 2302@end defmac 2303 2304@defmac REGNO_REG_CLASS (@var{regno}) 2305A C expression whose value is a register class containing hard register 2306@var{regno}. In general there is more than one such class; choose a class 2307which is @dfn{minimal}, meaning that no smaller class also contains the 2308register. 2309@end defmac 2310 2311@defmac BASE_REG_CLASS 2312A macro whose definition is the name of the class to which a valid 2313base register must belong. A base register is one used in an address 2314which is the register value plus a displacement. 2315@end defmac 2316 2317@defmac MODE_BASE_REG_CLASS (@var{mode}) 2318This is a variation of the @code{BASE_REG_CLASS} macro which allows 2319the selection of a base register in a mode dependent manner. If 2320@var{mode} is VOIDmode then it should return the same value as 2321@code{BASE_REG_CLASS}. 2322@end defmac 2323 2324@defmac MODE_BASE_REG_REG_CLASS (@var{mode}) 2325A C expression whose value is the register class to which a valid 2326base register must belong in order to be used in a base plus index 2327register address. You should define this macro if base plus index 2328addresses have different requirements than other base register uses. 2329@end defmac 2330 2331@defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code}) 2332A C expression whose value is the register class to which a valid 2333base register must belong. @var{outer_code} and @var{index_code} define the 2334context in which the base register occurs. @var{outer_code} is the code of 2335the immediately enclosing expression (@code{MEM} for the top level of an 2336address, @code{ADDRESS} for something that occurs in an 2337@code{address_operand}). @var{index_code} is the code of the corresponding 2338index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise. 2339@end defmac 2340 2341@defmac INDEX_REG_CLASS 2342A macro whose definition is the name of the class to which a valid 2343index register must belong. An index register is one used in an 2344address where its value is either multiplied by a scale factor or 2345added to another register (as well as added to a displacement). 2346@end defmac 2347 2348@defmac REGNO_OK_FOR_BASE_P (@var{num}) 2349A C expression which is nonzero if register number @var{num} is 2350suitable for use as a base register in operand addresses. It may be 2351either a suitable hard register or a pseudo register that has been 2352allocated such a hard register. 2353@end defmac 2354 2355@defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode}) 2356A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that 2357that expression may examine the mode of the memory reference in 2358@var{mode}. You should define this macro if the mode of the memory 2359reference affects whether a register may be used as a base register. If 2360you define this macro, the compiler will use it instead of 2361@code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for addresses 2362that appear outside a @code{MEM}, i.e. as an @code{address_operand}. 2363 2364@end defmac 2365 2366@defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode}) 2367A C expression which is nonzero if register number @var{num} is suitable for 2368use as a base register in base plus index operand addresses, accessing 2369memory in mode @var{mode}. It may be either a suitable hard register or a 2370pseudo register that has been allocated such a hard register. You should 2371define this macro if base plus index addresses have different requirements 2372than other base register uses. 2373 2374Use of this macro is deprecated; please use the more general 2375@code{REGNO_MODE_CODE_OK_FOR_BASE_P}. 2376@end defmac 2377 2378@defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code}) 2379A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except that 2380that expression may examine the context in which the register appears in the 2381memory reference. @var{outer_code} is the code of the immediately enclosing 2382expression (@code{MEM} if at the top level of the address, @code{ADDRESS} for 2383something that occurs in an @code{address_operand}). @var{index_code} is the 2384code of the corresponding index expression if @var{outer_code} is @code{PLUS}; 2385@code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses 2386that appear outside a @code{MEM}, i.e. as an @code{address_operand}. 2387@end defmac 2388 2389@defmac REGNO_OK_FOR_INDEX_P (@var{num}) 2390A C expression which is nonzero if register number @var{num} is 2391suitable for use as an index register in operand addresses. It may be 2392either a suitable hard register or a pseudo register that has been 2393allocated such a hard register. 2394 2395The difference between an index register and a base register is that 2396the index register may be scaled. If an address involves the sum of 2397two registers, neither one of them scaled, then either one may be 2398labeled the ``base'' and the other the ``index''; but whichever 2399labeling is used must fit the machine's constraints of which registers 2400may serve in each capacity. The compiler will try both labelings, 2401looking for one that is valid, and will reload one or both registers 2402only if neither labeling works. 2403@end defmac 2404 2405@defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) 2406A C expression that places additional restrictions on the register class 2407to use when it is necessary to copy value @var{x} into a register in class 2408@var{class}. The value is a register class; perhaps @var{class}, or perhaps 2409another, smaller class. On many machines, the following definition is 2410safe: 2411 2412@smallexample 2413#define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS 2414@end smallexample 2415 2416Sometimes returning a more restrictive class makes better code. For 2417example, on the 68000, when @var{x} is an integer constant that is in range 2418for a @samp{moveq} instruction, the value of this macro is always 2419@code{DATA_REGS} as long as @var{class} includes the data registers. 2420Requiring a data register guarantees that a @samp{moveq} will be used. 2421 2422One case where @code{PREFERRED_RELOAD_CLASS} must not return 2423@var{class} is if @var{x} is a legitimate constant which cannot be 2424loaded into some register class. By returning @code{NO_REGS} you can 2425force @var{x} into a memory location. For example, rs6000 can load 2426immediate values into general-purpose registers, but does not have an 2427instruction for loading an immediate value into a floating-point 2428register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when 2429@var{x} is a floating-point constant. If the constant can't be loaded 2430into any kind of register, code generation will be better if 2431@code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead 2432of using @code{PREFERRED_RELOAD_CLASS}. 2433 2434If an insn has pseudos in it after register allocation, reload will go 2435through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS} 2436to find the best one. Returning @code{NO_REGS}, in this case, makes 2437reload add a @code{!} in front of the constraint: the x86 back-end uses 2438this feature to discourage usage of 387 registers when math is done in 2439the SSE registers (and vice versa). 2440@end defmac 2441 2442@defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class}) 2443Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of 2444input reloads. If you don't define this macro, the default is to use 2445@var{class}, unchanged. 2446 2447You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage 2448reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}. 2449@end defmac 2450 2451@defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) 2452A C expression that places additional restrictions on the register class 2453to use when it is necessary to be able to hold a value of mode 2454@var{mode} in a reload register for which class @var{class} would 2455ordinarily be used. 2456 2457Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when 2458there are certain modes that simply can't go in certain reload classes. 2459 2460The value is a register class; perhaps @var{class}, or perhaps another, 2461smaller class. 2462 2463Don't define this macro unless the target machine has limitations which 2464require the macro to do something nontrivial. 2465@end defmac 2466 2467@deftypefn {Target Hook} enum reg_class TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, enum reg_class @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri}) 2468Many machines have some registers that cannot be copied directly to or 2469from memory or even from other types of registers. An example is the 2470@samp{MQ} register, which on most machines, can only be copied to or 2471from general registers, but not memory. Below, we shall be using the 2472term 'intermediate register' when a move operation cannot be performed 2473directly, but has to be done by copying the source into the intermediate 2474register first, and then copying the intermediate register to the 2475destination. An intermediate register always has the same mode as 2476source and destination. Since it holds the actual value being copied, 2477reload might apply optimizations to re-use an intermediate register 2478and eliding the copy from the source when it can determine that the 2479intermediate register still holds the required value. 2480 2481Another kind of secondary reload is required on some machines which 2482allow copying all registers to and from memory, but require a scratch 2483register for stores to some memory locations (e.g., those with symbolic 2484address on the RT, and those with certain symbolic address on the SPARC 2485when compiling PIC)@. Scratch registers need not have the same mode 2486as the value being copied, and usually hold a different value that 2487that being copied. Special patterns in the md file are needed to 2488describe how the copy is performed with the help of the scratch register; 2489these patterns also describe the number, register class(es) and mode(s) 2490of the scratch register(s). 2491 2492In some cases, both an intermediate and a scratch register are required. 2493 2494For input reloads, this target hook is called with nonzero @var{in_p}, 2495and @var{x} is an rtx that needs to be copied to a register in of class 2496@var{reload_class} in @var{reload_mode}. For output reloads, this target 2497hook is called with zero @var{in_p}, and a register of class @var{reload_mode} 2498needs to be copied to rtx @var{x} in @var{reload_mode}. 2499 2500If copying a register of @var{reload_class} from/to @var{x} requires 2501an intermediate register, the hook @code{secondary_reload} should 2502return the register class required for this intermediate register. 2503If no intermediate register is required, it should return NO_REGS. 2504If more than one intermediate register is required, describe the one 2505that is closest in the copy chain to the reload register. 2506 2507If scratch registers are needed, you also have to describe how to 2508perform the copy from/to the reload register to/from this 2509closest intermediate register. Or if no intermediate register is 2510required, but still a scratch register is needed, describe the 2511copy from/to the reload register to/from the reload operand @var{x}. 2512 2513You do this by setting @code{sri->icode} to the instruction code of a pattern 2514in the md file which performs the move. Operands 0 and 1 are the output 2515and input of this copy, respectively. Operands from operand 2 onward are 2516for scratch operands. These scratch operands must have a mode, and a 2517single-register-class 2518@c [later: or memory] 2519output constraint. 2520 2521When an intermediate register is used, the @code{secondary_reload} 2522hook will be called again to determine how to copy the intermediate 2523register to/from the reload operand @var{x}, so your hook must also 2524have code to handle the register class of the intermediate operand. 2525 2526@c [For later: maybe we'll allow multi-alternative reload patterns - 2527@c the port maintainer could name a mov<mode> pattern that has clobbers - 2528@c and match the constraints of input and output to determine the required 2529@c alternative. A restriction would be that constraints used to match 2530@c against reloads registers would have to be written as register class 2531@c constraints, or we need a new target macro / hook that tells us if an 2532@c arbitrary constraint can match an unknown register of a given class. 2533@c Such a macro / hook would also be useful in other places.] 2534 2535 2536@var{x} might be a pseudo-register or a @code{subreg} of a 2537pseudo-register, which could either be in a hard register or in memory. 2538Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is 2539in memory and the hard register number if it is in a register. 2540 2541Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are 2542currently not supported. For the time being, you will have to continue 2543to use @code{SECONDARY_MEMORY_NEEDED} for that purpose. 2544 2545@code{copy_cost} also uses this target hook to find out how values are 2546copied. If you want it to include some extra cost for the need to allocate 2547(a) scratch register(s), set @code{sri->extra_cost} to the additional cost. 2548Or if two dependent moves are supposed to have a lower cost than the sum 2549of the individual moves due to expected fortuitous scheduling and/or special 2550forwarding logic, you can set @code{sri->extra_cost} to a negative amount. 2551@end deftypefn 2552 2553@defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2554@defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2555@defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) 2556These macros are obsolete, new ports should use the target hook 2557@code{TARGET_SECONDARY_RELOAD} instead. 2558 2559These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD} 2560target hook. Older ports still define these macros to indicate to the 2561reload phase that it may 2562need to allocate at least one register for a reload in addition to the 2563register to contain the data. Specifically, if copying @var{x} to a 2564register @var{class} in @var{mode} requires an intermediate register, 2565you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the 2566largest register class all of whose registers can be used as 2567intermediate registers or scratch registers. 2568 2569If copying a register @var{class} in @var{mode} to @var{x} requires an 2570intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} 2571was supposed to be defined be defined to return the largest register 2572class required. If the 2573requirements for input and output reloads were the same, the macro 2574@code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both 2575macros identically. 2576 2577The values returned by these macros are often @code{GENERAL_REGS}. 2578Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} 2579can be directly copied to or from a register of @var{class} in 2580@var{mode} without requiring a scratch register. Do not define this 2581macro if it would always return @code{NO_REGS}. 2582 2583If a scratch register is required (either with or without an 2584intermediate register), you were supposed to define patterns for 2585@samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required 2586(@pxref{Standard Names}. These patterns, which were normally 2587implemented with a @code{define_expand}, should be similar to the 2588@samp{mov@var{m}} patterns, except that operand 2 is the scratch 2589register. 2590 2591These patterns need constraints for the reload register and scratch 2592register that 2593contain a single register class. If the original reload register (whose 2594class is @var{class}) can meet the constraint given in the pattern, the 2595value returned by these macros is used for the class of the scratch 2596register. Otherwise, two additional reload registers are required. 2597Their classes are obtained from the constraints in the insn pattern. 2598 2599@var{x} might be a pseudo-register or a @code{subreg} of a 2600pseudo-register, which could either be in a hard register or in memory. 2601Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is 2602in memory and the hard register number if it is in a register. 2603 2604These macros should not be used in the case where a particular class of 2605registers can only be copied to memory and not to another class of 2606registers. In that case, secondary reload registers are not needed and 2607would not be helpful. Instead, a stack location must be used to perform 2608the copy and the @code{mov@var{m}} pattern should use memory as an 2609intermediate storage. This case often occurs between floating-point and 2610general registers. 2611@end defmac 2612 2613@defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m}) 2614Certain machines have the property that some registers cannot be copied 2615to some other registers without using memory. Define this macro on 2616those machines to be a C expression that is nonzero if objects of mode 2617@var{m} in registers of @var{class1} can only be copied to registers of 2618class @var{class2} by storing a register of @var{class1} into memory 2619and loading that memory location into a register of @var{class2}. 2620 2621Do not define this macro if its value would always be zero. 2622@end defmac 2623 2624@defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) 2625Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler 2626allocates a stack slot for a memory location needed for register copies. 2627If this macro is defined, the compiler instead uses the memory location 2628defined by this macro. 2629 2630Do not define this macro if you do not define 2631@code{SECONDARY_MEMORY_NEEDED}. 2632@end defmac 2633 2634@defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode}) 2635When the compiler needs a secondary memory location to copy between two 2636registers of mode @var{mode}, it normally allocates sufficient memory to 2637hold a quantity of @code{BITS_PER_WORD} bits and performs the store and 2638load operations in a mode that many bits wide and whose class is the 2639same as that of @var{mode}. 2640 2641This is right thing to do on most machines because it ensures that all 2642bits of the register are copied and prevents accesses to the registers 2643in a narrower mode, which some machines prohibit for floating-point 2644registers. 2645 2646However, this default behavior is not correct on some machines, such as 2647the DEC Alpha, that store short integers in floating-point registers 2648differently than in integer registers. On those machines, the default 2649widening will not work correctly and you must define this macro to 2650suppress that widening in some cases. See the file @file{alpha.h} for 2651details. 2652 2653Do not define this macro if you do not define 2654@code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that 2655is @code{BITS_PER_WORD} bits wide is correct for your machine. 2656@end defmac 2657 2658@defmac SMALL_REGISTER_CLASSES 2659On some machines, it is risky to let hard registers live across arbitrary 2660insns. Typically, these machines have instructions that require values 2661to be in specific registers (like an accumulator), and reload will fail 2662if the required hard register is used for another purpose across such an 2663insn. 2664 2665Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero 2666value on these machines. When this macro has a nonzero value, the 2667compiler will try to minimize the lifetime of hard registers. 2668 2669It is always safe to define this macro with a nonzero value, but if you 2670unnecessarily define it, you will reduce the amount of optimizations 2671that can be performed in some cases. If you do not define this macro 2672with a nonzero value when it is required, the compiler will run out of 2673spill registers and print a fatal error message. For most machines, you 2674should not define this macro at all. 2675@end defmac 2676 2677@defmac CLASS_LIKELY_SPILLED_P (@var{class}) 2678A C expression whose value is nonzero if pseudos that have been assigned 2679to registers of class @var{class} would likely be spilled because 2680registers of @var{class} are needed for spill registers. 2681 2682The default value of this macro returns 1 if @var{class} has exactly one 2683register and zero otherwise. On most machines, this default should be 2684used. Only define this macro to some other expression if pseudos 2685allocated by @file{local-alloc.c} end up in memory because their hard 2686registers were needed for spill registers. If this macro returns nonzero 2687for those classes, those pseudos will only be allocated by 2688@file{global.c}, which knows how to reallocate the pseudo to another 2689register. If there would not be another register available for 2690reallocation, you should not change the definition of this macro since 2691the only effect of such a definition would be to slow down register 2692allocation. 2693@end defmac 2694 2695@defmac CLASS_MAX_NREGS (@var{class}, @var{mode}) 2696A C expression for the maximum number of consecutive registers 2697of class @var{class} needed to hold a value of mode @var{mode}. 2698 2699This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, 2700the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} 2701should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, 2702@var{mode})} for all @var{regno} values in the class @var{class}. 2703 2704This macro helps control the handling of multiple-word values 2705in the reload pass. 2706@end defmac 2707 2708@defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class}) 2709If defined, a C expression that returns nonzero for a @var{class} for which 2710a change from mode @var{from} to mode @var{to} is invalid. 2711 2712For the example, loading 32-bit integer or floating-point objects into 2713floating-point registers on the Alpha extends them to 64 bits. 2714Therefore loading a 64-bit object and then storing it as a 32-bit object 2715does not store the low-order 32 bits, as would be the case for a normal 2716register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS} 2717as below: 2718 2719@smallexample 2720#define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \ 2721 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \ 2722 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0) 2723@end smallexample 2724@end defmac 2725 2726@node Old Constraints 2727@section Obsolete Macros for Defining Constraints 2728@cindex defining constraints, obsolete method 2729@cindex constraints, defining, obsolete method 2730 2731Machine-specific constraints can be defined with these macros instead 2732of the machine description constructs described in @ref{Define 2733Constraints}. This mechanism is obsolete. New ports should not use 2734it; old ports should convert to the new mechanism. 2735 2736@defmac CONSTRAINT_LEN (@var{char}, @var{str}) 2737For the constraint at the start of @var{str}, which starts with the letter 2738@var{c}, return the length. This allows you to have register class / 2739constant / extra constraints that are longer than a single letter; 2740you don't need to define this macro if you can do with single-letter 2741constraints only. The definition of this macro should use 2742DEFAULT_CONSTRAINT_LEN for all the characters that you don't want 2743to handle specially. 2744There are some sanity checks in genoutput.c that check the constraint lengths 2745for the md file, so you can also use this macro to help you while you are 2746transitioning from a byzantine single-letter-constraint scheme: when you 2747return a negative length for a constraint you want to re-use, genoutput 2748will complain about every instance where it is used in the md file. 2749@end defmac 2750 2751@defmac REG_CLASS_FROM_LETTER (@var{char}) 2752A C expression which defines the machine-dependent operand constraint 2753letters for register classes. If @var{char} is such a letter, the 2754value should be the register class corresponding to it. Otherwise, 2755the value should be @code{NO_REGS}. The register letter @samp{r}, 2756corresponding to class @code{GENERAL_REGS}, will not be passed 2757to this macro; you do not need to handle it. 2758@end defmac 2759 2760@defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str}) 2761Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string 2762passed in @var{str}, so that you can use suffixes to distinguish between 2763different variants. 2764@end defmac 2765 2766@defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c}) 2767A C expression that defines the machine-dependent operand constraint 2768letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify 2769particular ranges of integer values. If @var{c} is one of those 2770letters, the expression should check that @var{value}, an integer, is in 2771the appropriate range and return 1 if so, 0 otherwise. If @var{c} is 2772not one of those letters, the value should be 0 regardless of 2773@var{value}. 2774@end defmac 2775 2776@defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str}) 2777Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint 2778string passed in @var{str}, so that you can use suffixes to distinguish 2779between different variants. 2780@end defmac 2781 2782@defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c}) 2783A C expression that defines the machine-dependent operand constraint 2784letters that specify particular ranges of @code{const_double} values 2785(@samp{G} or @samp{H}). 2786 2787If @var{c} is one of those letters, the expression should check that 2788@var{value}, an RTX of code @code{const_double}, is in the appropriate 2789range and return 1 if so, 0 otherwise. If @var{c} is not one of those 2790letters, the value should be 0 regardless of @var{value}. 2791 2792@code{const_double} is used for all floating-point constants and for 2793@code{DImode} fixed-point constants. A given letter can accept either 2794or both kinds of values. It can use @code{GET_MODE} to distinguish 2795between these kinds. 2796@end defmac 2797 2798@defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str}) 2799Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint 2800string passed in @var{str}, so that you can use suffixes to distinguish 2801between different variants. 2802@end defmac 2803 2804@defmac EXTRA_CONSTRAINT (@var{value}, @var{c}) 2805A C expression that defines the optional machine-dependent constraint 2806letters that can be used to segregate specific types of operands, usually 2807memory references, for the target machine. Any letter that is not 2808elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} / 2809@code{REG_CLASS_FROM_CONSTRAINT} 2810may be used. Normally this macro will not be defined. 2811 2812If it is required for a particular target machine, it should return 1 2813if @var{value} corresponds to the operand type represented by the 2814constraint letter @var{c}. If @var{c} is not defined as an extra 2815constraint, the value returned should be 0 regardless of @var{value}. 2816 2817For example, on the ROMP, load instructions cannot have their output 2818in r0 if the memory reference contains a symbolic address. Constraint 2819letter @samp{Q} is defined as representing a memory address that does 2820@emph{not} contain a symbolic address. An alternative is specified with 2821a @samp{Q} constraint on the input and @samp{r} on the output. The next 2822alternative specifies @samp{m} on the input and a register class that 2823does not include r0 on the output. 2824@end defmac 2825 2826@defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str}) 2827Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed 2828in @var{str}, so that you can use suffixes to distinguish between different 2829variants. 2830@end defmac 2831 2832@defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str}) 2833A C expression that defines the optional machine-dependent constraint 2834letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should 2835be treated like memory constraints by the reload pass. 2836 2837It should return 1 if the operand type represented by the constraint 2838at the start of @var{str}, the first letter of which is the letter @var{c}, 2839 comprises a subset of all memory references including 2840all those whose address is simply a base register. This allows the reload 2841pass to reload an operand, if it does not directly correspond to the operand 2842type of @var{c}, by copying its address into a base register. 2843 2844For example, on the S/390, some instructions do not accept arbitrary 2845memory references, but only those that do not make use of an index 2846register. The constraint letter @samp{Q} is defined via 2847@code{EXTRA_CONSTRAINT} as representing a memory address of this type. 2848If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT}, 2849a @samp{Q} constraint can handle any memory operand, because the 2850reload pass knows it can be reloaded by copying the memory address 2851into a base register if required. This is analogous to the way 2852a @samp{o} constraint can handle any memory operand. 2853@end defmac 2854 2855@defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str}) 2856A C expression that defines the optional machine-dependent constraint 2857letters, amongst those accepted by @code{EXTRA_CONSTRAINT} / 2858@code{EXTRA_CONSTRAINT_STR}, that should 2859be treated like address constraints by the reload pass. 2860 2861It should return 1 if the operand type represented by the constraint 2862at the start of @var{str}, which starts with the letter @var{c}, comprises 2863a subset of all memory addresses including 2864all those that consist of just a base register. This allows the reload 2865pass to reload an operand, if it does not directly correspond to the operand 2866type of @var{str}, by copying it into a base register. 2867 2868Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only 2869be used with the @code{address_operand} predicate. It is treated 2870analogously to the @samp{p} constraint. 2871@end defmac 2872 2873@node Stack and Calling 2874@section Stack Layout and Calling Conventions 2875@cindex calling conventions 2876 2877@c prevent bad page break with this line 2878This describes the stack layout and calling conventions. 2879 2880@menu 2881* Frame Layout:: 2882* Exception Handling:: 2883* Stack Checking:: 2884* Frame Registers:: 2885* Elimination:: 2886* Stack Arguments:: 2887* Register Arguments:: 2888* Scalar Return:: 2889* Aggregate Return:: 2890* Caller Saves:: 2891* Function Entry:: 2892* Profiling:: 2893* Tail Calls:: 2894* Stack Smashing Protection:: 2895@end menu 2896 2897@node Frame Layout 2898@subsection Basic Stack Layout 2899@cindex stack frame layout 2900@cindex frame layout 2901 2902@c prevent bad page break with this line 2903Here is the basic stack layout. 2904 2905@defmac STACK_GROWS_DOWNWARD 2906Define this macro if pushing a word onto the stack moves the stack 2907pointer to a smaller address. 2908 2909When we say, ``define this macro if @dots{}'', it means that the 2910compiler checks this macro only with @code{#ifdef} so the precise 2911definition used does not matter. 2912@end defmac 2913 2914@defmac STACK_PUSH_CODE 2915This macro defines the operation used when something is pushed 2916on the stack. In RTL, a push operation will be 2917@code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})} 2918 2919The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC}, 2920and @code{POST_INC}. Which of these is correct depends on 2921the stack direction and on whether the stack pointer points 2922to the last item on the stack or whether it points to the 2923space for the next item on the stack. 2924 2925The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is 2926defined, which is almost always right, and @code{PRE_INC} otherwise, 2927which is often wrong. 2928@end defmac 2929 2930@defmac FRAME_GROWS_DOWNWARD 2931Define this macro to nonzero value if the addresses of local variable slots 2932are at negative offsets from the frame pointer. 2933@end defmac 2934 2935@defmac ARGS_GROW_DOWNWARD 2936Define this macro if successive arguments to a function occupy decreasing 2937addresses on the stack. 2938@end defmac 2939 2940@defmac STARTING_FRAME_OFFSET 2941Offset from the frame pointer to the first local variable slot to be allocated. 2942 2943If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by 2944subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}. 2945Otherwise, it is found by adding the length of the first slot to the 2946value @code{STARTING_FRAME_OFFSET}. 2947@c i'm not sure if the above is still correct.. had to change it to get 2948@c rid of an overfull. --mew 2feb93 2949@end defmac 2950 2951@defmac STACK_ALIGNMENT_NEEDED 2952Define to zero to disable final alignment of the stack during reload. 2953The nonzero default for this macro is suitable for most ports. 2954 2955On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there 2956is a register save block following the local block that doesn't require 2957alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable 2958stack alignment and do it in the backend. 2959@end defmac 2960 2961@defmac STACK_POINTER_OFFSET 2962Offset from the stack pointer register to the first location at which 2963outgoing arguments are placed. If not specified, the default value of 2964zero is used. This is the proper value for most machines. 2965 2966If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2967the first location at which outgoing arguments are placed. 2968@end defmac 2969 2970@defmac FIRST_PARM_OFFSET (@var{fundecl}) 2971Offset from the argument pointer register to the first argument's 2972address. On some machines it may depend on the data type of the 2973function. 2974 2975If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above 2976the first argument's address. 2977@end defmac 2978 2979@defmac STACK_DYNAMIC_OFFSET (@var{fundecl}) 2980Offset from the stack pointer register to an item dynamically allocated 2981on the stack, e.g., by @code{alloca}. 2982 2983The default value for this macro is @code{STACK_POINTER_OFFSET} plus the 2984length of the outgoing arguments. The default is correct for most 2985machines. See @file{function.c} for details. 2986@end defmac 2987 2988@defmac INITIAL_FRAME_ADDRESS_RTX 2989A C expression whose value is RTL representing the address of the initial 2990stack frame. This address is passed to @code{RETURN_ADDR_RTX} and 2991@code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable 2992default value will be used. Define this macro in order to make frame pointer 2993elimination work in the presence of @code{__builtin_frame_address (count)} and 2994@code{__builtin_return_address (count)} for @code{count} not equal to zero. 2995@end defmac 2996 2997@defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) 2998A C expression whose value is RTL representing the address in a stack 2999frame where the pointer to the caller's frame is stored. Assume that 3000@var{frameaddr} is an RTL expression for the address of the stack frame 3001itself. 3002 3003If you don't define this macro, the default is to return the value 3004of @var{frameaddr}---that is, the stack frame address is also the 3005address of the stack word that points to the previous frame. 3006@end defmac 3007 3008@defmac SETUP_FRAME_ADDRESSES 3009If defined, a C expression that produces the machine-specific code to 3010setup the stack so that arbitrary frames can be accessed. For example, 3011on the SPARC, we must flush all of the register windows to the stack 3012before we can access arbitrary stack frames. You will seldom need to 3013define this macro. 3014@end defmac 3015 3016@deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE () 3017This target hook should return an rtx that is used to store 3018the address of the current frame into the built in @code{setjmp} buffer. 3019The default value, @code{virtual_stack_vars_rtx}, is correct for most 3020machines. One reason you may need to define this target hook is if 3021@code{hard_frame_pointer_rtx} is the appropriate value on your machine. 3022@end deftypefn 3023 3024@defmac FRAME_ADDR_RTX (@var{frameaddr}) 3025A C expression whose value is RTL representing the value of the frame 3026address for the current frame. @var{frameaddr} is the frame pointer 3027of the current frame. This is used for __builtin_frame_address. 3028You need only define this macro if the frame address is not the same 3029as the frame pointer. Most machines do not need to define it. 3030@end defmac 3031 3032@defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) 3033A C expression whose value is RTL representing the value of the return 3034address for the frame @var{count} steps up from the current frame, after 3035the prologue. @var{frameaddr} is the frame pointer of the @var{count} 3036frame, or the frame pointer of the @var{count} @minus{} 1 frame if 3037@code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined. 3038 3039The value of the expression must always be the correct address when 3040@var{count} is zero, but may be @code{NULL_RTX} if there is not way to 3041determine the return address of other frames. 3042@end defmac 3043 3044@defmac RETURN_ADDR_IN_PREVIOUS_FRAME 3045Define this if the return address of a particular stack frame is accessed 3046from the frame pointer of the previous stack frame. 3047@end defmac 3048 3049@defmac INCOMING_RETURN_ADDR_RTX 3050A C expression whose value is RTL representing the location of the 3051incoming return address at the beginning of any function, before the 3052prologue. This RTL is either a @code{REG}, indicating that the return 3053value is saved in @samp{REG}, or a @code{MEM} representing a location in 3054the stack. 3055 3056You only need to define this macro if you want to support call frame 3057debugging information like that provided by DWARF 2. 3058 3059If this RTL is a @code{REG}, you should also define 3060@code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}. 3061@end defmac 3062 3063@defmac DWARF_ALT_FRAME_RETURN_COLUMN 3064A C expression whose value is an integer giving a DWARF 2 column 3065number that may be used as an alternate return column. This should 3066be defined only if @code{DWARF_FRAME_RETURN_COLUMN} is set to a 3067general register, but an alternate column needs to be used for 3068signal frames. 3069@end defmac 3070 3071@defmac DWARF_ZERO_REG 3072A C expression whose value is an integer giving a DWARF 2 register 3073number that is considered to always have the value zero. This should 3074only be defined if the target has an architected zero register, and 3075someone decided it was a good idea to use that register number to 3076terminate the stack backtrace. New ports should avoid this. 3077@end defmac 3078 3079@deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index}) 3080This target hook allows the backend to emit frame-related insns that 3081contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging 3082info engine will invoke it on insns of the form 3083@smallexample 3084(set (reg) (unspec [...] UNSPEC_INDEX)) 3085@end smallexample 3086and 3087@smallexample 3088(set (reg) (unspec_volatile [...] UNSPECV_INDEX)). 3089@end smallexample 3090to let the backend emit the call frame instructions. @var{label} is 3091the CFI label attached to the insn, @var{pattern} is the pattern of 3092the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}. 3093@end deftypefn 3094 3095@defmac INCOMING_FRAME_SP_OFFSET 3096A C expression whose value is an integer giving the offset, in bytes, 3097from the value of the stack pointer register to the top of the stack 3098frame at the beginning of any function, before the prologue. The top of 3099the frame is defined to be the value of the stack pointer in the 3100previous frame, just before the call instruction. 3101 3102You only need to define this macro if you want to support call frame 3103debugging information like that provided by DWARF 2. 3104@end defmac 3105 3106@defmac ARG_POINTER_CFA_OFFSET (@var{fundecl}) 3107A C expression whose value is an integer giving the offset, in bytes, 3108from the argument pointer to the canonical frame address (cfa). The 3109final value should coincide with that calculated by 3110@code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable 3111during virtual register instantiation. 3112 3113The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)}, 3114which is correct for most machines; in general, the arguments are found 3115immediately before the stack frame. Note that this is not the case on 3116some targets that save registers into the caller's frame, such as SPARC 3117and rs6000, and so such targets need to define this macro. 3118 3119You only need to define this macro if the default is incorrect, and you 3120want to support call frame debugging information like that provided by 3121DWARF 2. 3122@end defmac 3123 3124@defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl}) 3125If defined, a C expression whose value is an integer giving the offset 3126in bytes from the frame pointer to the canonical frame address (cfa). 3127The final value should coincide with that calculated by 3128@code{INCOMING_FRAME_SP_OFFSET}. 3129 3130Normally the CFA is calculated as an offset from the argument pointer, 3131via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is 3132variable due to the ABI, this may not be possible. If this macro is 3133defined, it implies that the virtual register instantiation should be 3134based on the frame pointer instead of the argument pointer. Only one 3135of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET} 3136should be defined. 3137@end defmac 3138 3139@defmac CFA_FRAME_BASE_OFFSET (@var{fundecl}) 3140If defined, a C expression whose value is an integer giving the offset 3141in bytes from the canonical frame address (cfa) to the frame base used 3142in DWARF 2 debug information. The default is zero. A different value 3143may reduce the size of debug information on some ports. 3144@end defmac 3145 3146@node Exception Handling 3147@subsection Exception Handling Support 3148@cindex exception handling 3149 3150@defmac EH_RETURN_DATA_REGNO (@var{N}) 3151A C expression whose value is the @var{N}th register number used for 3152data by exception handlers, or @code{INVALID_REGNUM} if fewer than 3153@var{N} registers are usable. 3154 3155The exception handling library routines communicate with the exception 3156handlers via a set of agreed upon registers. Ideally these registers 3157should be call-clobbered; it is possible to use call-saved registers, 3158but may negatively impact code size. The target must support at least 31592 data registers, but should define 4 if there are enough free registers. 3160 3161You must define this macro if you want to support call frame exception 3162handling like that provided by DWARF 2. 3163@end defmac 3164 3165@defmac EH_RETURN_STACKADJ_RTX 3166A C expression whose value is RTL representing a location in which 3167to store a stack adjustment to be applied before function return. 3168This is used to unwind the stack to an exception handler's call frame. 3169It will be assigned zero on code paths that return normally. 3170 3171Typically this is a call-clobbered hard register that is otherwise 3172untouched by the epilogue, but could also be a stack slot. 3173 3174Do not define this macro if the stack pointer is saved and restored 3175by the regular prolog and epilog code in the call frame itself; in 3176this case, the exception handling library routines will update the 3177stack location to be restored in place. Otherwise, you must define 3178this macro if you want to support call frame exception handling like 3179that provided by DWARF 2. 3180@end defmac 3181 3182@defmac EH_RETURN_HANDLER_RTX 3183A C expression whose value is RTL representing a location in which 3184to store the address of an exception handler to which we should 3185return. It will not be assigned on code paths that return normally. 3186 3187Typically this is the location in the call frame at which the normal 3188return address is stored. For targets that return by popping an 3189address off the stack, this might be a memory address just below 3190the @emph{target} call frame rather than inside the current call 3191frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already 3192been assigned, so it may be used to calculate the location of the 3193target call frame. 3194 3195Some targets have more complex requirements than storing to an 3196address calculable during initial code generation. In that case 3197the @code{eh_return} instruction pattern should be used instead. 3198 3199If you want to support call frame exception handling, you must 3200define either this macro or the @code{eh_return} instruction pattern. 3201@end defmac 3202 3203@defmac RETURN_ADDR_OFFSET 3204If defined, an integer-valued C expression for which rtl will be generated 3205to add it to the exception handler address before it is searched in the 3206exception handling tables, and to subtract it again from the address before 3207using it to return to the exception handler. 3208@end defmac 3209 3210@defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global}) 3211This macro chooses the encoding of pointers embedded in the exception 3212handling sections. If at all possible, this should be defined such 3213that the exception handling section will not require dynamic relocations, 3214and so may be read-only. 3215 3216@var{code} is 0 for data, 1 for code labels, 2 for function pointers. 3217@var{global} is true if the symbol may be affected by dynamic relocations. 3218The macro should return a combination of the @code{DW_EH_PE_*} defines 3219as found in @file{dwarf2.h}. 3220 3221If this macro is not defined, pointers will not be encoded but 3222represented directly. 3223@end defmac 3224 3225@defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done}) 3226This macro allows the target to emit whatever special magic is required 3227to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}. 3228Generic code takes care of pc-relative and indirect encodings; this must 3229be defined if the target uses text-relative or data-relative encodings. 3230 3231This is a C statement that branches to @var{done} if the format was 3232handled. @var{encoding} is the format chosen, @var{size} is the number 3233of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF} 3234to be emitted. 3235@end defmac 3236 3237@defmac MD_UNWIND_SUPPORT 3238A string specifying a file to be #include'd in unwind-dw2.c. The file 3239so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}. 3240@end defmac 3241 3242@defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}) 3243This macro allows the target to add cpu and operating system specific 3244code to the call-frame unwinder for use when there is no unwind data 3245available. The most common reason to implement this macro is to unwind 3246through signal frames. 3247 3248This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c} 3249and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; 3250@var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra} 3251for the address of the code being executed and @code{context->cfa} for 3252the stack pointer value. If the frame can be decoded, the register save 3253addresses should be updated in @var{fs} and the macro should evaluate to 3254@code{_URC_NO_REASON}. If the frame cannot be decoded, the macro should 3255evaluate to @code{_URC_END_OF_STACK}. 3256 3257For proper signal handling in Java this macro is accompanied by 3258@code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers. 3259@end defmac 3260 3261@defmac MD_HANDLE_UNWABI (@var{context}, @var{fs}) 3262This macro allows the target to add operating system specific code to the 3263call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive, 3264usually used for signal or interrupt frames. 3265 3266This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}. 3267@var{context} is an @code{_Unwind_Context}; 3268@var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi} 3269for the abi and context in the @code{.unwabi} directive. If the 3270@code{.unwabi} directive can be handled, the register save addresses should 3271be updated in @var{fs}. 3272@end defmac 3273 3274@defmac TARGET_USES_WEAK_UNWIND_INFO 3275A C expression that evaluates to true if the target requires unwind 3276info to be given comdat linkage. Define it to be @code{1} if comdat 3277linkage is necessary. The default is @code{0}. 3278@end defmac 3279 3280@node Stack Checking 3281@subsection Specifying How Stack Checking is Done 3282 3283GCC will check that stack references are within the boundaries of 3284the stack, if the @option{-fstack-check} is specified, in one of three ways: 3285 3286@enumerate 3287@item 3288If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC 3289will assume that you have arranged for stack checking to be done at 3290appropriate places in the configuration files, e.g., in 3291@code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special 3292processing. 3293 3294@item 3295If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern 3296called @code{check_stack} in your @file{md} file, GCC will call that 3297pattern with one argument which is the address to compare the stack 3298value against. You must arrange for this pattern to report an error if 3299the stack pointer is out of range. 3300 3301@item 3302If neither of the above are true, GCC will generate code to periodically 3303``probe'' the stack pointer using the values of the macros defined below. 3304@end enumerate 3305 3306Normally, you will use the default values of these macros, so GCC 3307will use the third approach. 3308 3309@defmac STACK_CHECK_BUILTIN 3310A nonzero value if stack checking is done by the configuration files in a 3311machine-dependent manner. You should define this macro if stack checking 3312is require by the ABI of your machine or if you would like to have to stack 3313checking in some more efficient way than GCC's portable approach. 3314The default value of this macro is zero. 3315@end defmac 3316 3317@defmac STACK_CHECK_PROBE_INTERVAL 3318An integer representing the interval at which GCC must generate stack 3319probe instructions. You will normally define this macro to be no larger 3320than the size of the ``guard pages'' at the end of a stack area. The 3321default value of 4096 is suitable for most systems. 3322@end defmac 3323 3324@defmac STACK_CHECK_PROBE_LOAD 3325A integer which is nonzero if GCC should perform the stack probe 3326as a load instruction and zero if GCC should use a store instruction. 3327The default is zero, which is the most efficient choice on most systems. 3328@end defmac 3329 3330@defmac STACK_CHECK_PROTECT 3331The number of bytes of stack needed to recover from a stack overflow, 3332for languages where such a recovery is supported. The default value of 333375 words should be adequate for most machines. 3334@end defmac 3335 3336@defmac STACK_CHECK_MAX_FRAME_SIZE 3337The maximum size of a stack frame, in bytes. GCC will generate probe 3338instructions in non-leaf functions to ensure at least this many bytes of 3339stack are available. If a stack frame is larger than this size, stack 3340checking will not be reliable and GCC will issue a warning. The 3341default is chosen so that GCC only generates one instruction on most 3342systems. You should normally not change the default value of this macro. 3343@end defmac 3344 3345@defmac STACK_CHECK_FIXED_FRAME_SIZE 3346GCC uses this value to generate the above warning message. It 3347represents the amount of fixed frame used by a function, not including 3348space for any callee-saved registers, temporaries and user variables. 3349You need only specify an upper bound for this amount and will normally 3350use the default of four words. 3351@end defmac 3352 3353@defmac STACK_CHECK_MAX_VAR_SIZE 3354The maximum size, in bytes, of an object that GCC will place in the 3355fixed area of the stack frame when the user specifies 3356@option{-fstack-check}. 3357GCC computed the default from the values of the above macros and you will 3358normally not need to override that default. 3359@end defmac 3360 3361@need 2000 3362@node Frame Registers 3363@subsection Registers That Address the Stack Frame 3364 3365@c prevent bad page break with this line 3366This discusses registers that address the stack frame. 3367 3368@defmac STACK_POINTER_REGNUM 3369The register number of the stack pointer register, which must also be a 3370fixed register according to @code{FIXED_REGISTERS}. On most machines, 3371the hardware determines which register this is. 3372@end defmac 3373 3374@defmac FRAME_POINTER_REGNUM 3375The register number of the frame pointer register, which is used to 3376access automatic variables in the stack frame. On some machines, the 3377hardware determines which register this is. On other machines, you can 3378choose any register you wish for this purpose. 3379@end defmac 3380 3381@defmac HARD_FRAME_POINTER_REGNUM 3382On some machines the offset between the frame pointer and starting 3383offset of the automatic variables is not known until after register 3384allocation has been done (for example, because the saved registers are 3385between these two locations). On those machines, define 3386@code{FRAME_POINTER_REGNUM} the number of a special, fixed register to 3387be used internally until the offset is known, and define 3388@code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number 3389used for the frame pointer. 3390 3391You should define this macro only in the very rare circumstances when it 3392is not possible to calculate the offset between the frame pointer and 3393the automatic variables until after register allocation has been 3394completed. When this macro is defined, you must also indicate in your 3395definition of @code{ELIMINABLE_REGS} how to eliminate 3396@code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} 3397or @code{STACK_POINTER_REGNUM}. 3398 3399Do not define this macro if it would be the same as 3400@code{FRAME_POINTER_REGNUM}. 3401@end defmac 3402 3403@defmac ARG_POINTER_REGNUM 3404The register number of the arg pointer register, which is used to access 3405the function's argument list. On some machines, this is the same as the 3406frame pointer register. On some machines, the hardware determines which 3407register this is. On other machines, you can choose any register you 3408wish for this purpose. If this is not the same register as the frame 3409pointer register, then you must mark it as a fixed register according to 3410@code{FIXED_REGISTERS}, or arrange to be able to eliminate it 3411(@pxref{Elimination}). 3412@end defmac 3413 3414@defmac RETURN_ADDRESS_POINTER_REGNUM 3415The register number of the return address pointer register, which is used to 3416access the current function's return address from the stack. On some 3417machines, the return address is not at a fixed offset from the frame 3418pointer or stack pointer or argument pointer. This register can be defined 3419to point to the return address on the stack, and then be converted by 3420@code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. 3421 3422Do not define this macro unless there is no other way to get the return 3423address from the stack. 3424@end defmac 3425 3426@defmac STATIC_CHAIN_REGNUM 3427@defmacx STATIC_CHAIN_INCOMING_REGNUM 3428Register numbers used for passing a function's static chain pointer. If 3429register windows are used, the register number as seen by the called 3430function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register 3431number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If 3432these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need 3433not be defined. 3434 3435The static chain register need not be a fixed register. 3436 3437If the static chain is passed in memory, these macros should not be 3438defined; instead, the next two macros should be defined. 3439@end defmac 3440 3441@defmac STATIC_CHAIN 3442@defmacx STATIC_CHAIN_INCOMING 3443If the static chain is passed in memory, these macros provide rtx giving 3444@code{mem} expressions that denote where they are stored. 3445@code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations 3446as seen by the calling and called functions, respectively. Often the former 3447will be at an offset from the stack pointer and the latter at an offset from 3448the frame pointer. 3449 3450@findex stack_pointer_rtx 3451@findex frame_pointer_rtx 3452@findex arg_pointer_rtx 3453The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and 3454@code{arg_pointer_rtx} will have been initialized prior to the use of these 3455macros and should be used to refer to those items. 3456 3457If the static chain is passed in a register, the two previous macros should 3458be defined instead. 3459@end defmac 3460 3461@defmac DWARF_FRAME_REGISTERS 3462This macro specifies the maximum number of hard registers that can be 3463saved in a call frame. This is used to size data structures used in 3464DWARF2 exception handling. 3465 3466Prior to GCC 3.0, this macro was needed in order to establish a stable 3467exception handling ABI in the face of adding new hard registers for ISA 3468extensions. In GCC 3.0 and later, the EH ABI is insulated from changes 3469in the number of hard registers. Nevertheless, this macro can still be 3470used to reduce the runtime memory requirements of the exception handling 3471routines, which can be substantial if the ISA contains a lot of 3472registers that are not call-saved. 3473 3474If this macro is not defined, it defaults to 3475@code{FIRST_PSEUDO_REGISTER}. 3476@end defmac 3477 3478@defmac PRE_GCC3_DWARF_FRAME_REGISTERS 3479 3480This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided 3481for backward compatibility in pre GCC 3.0 compiled code. 3482 3483If this macro is not defined, it defaults to 3484@code{DWARF_FRAME_REGISTERS}. 3485@end defmac 3486 3487@defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno}) 3488 3489Define this macro if the target's representation for dwarf registers 3490is different than the internal representation for unwind column. 3491Given a dwarf register, this macro should return the internal unwind 3492column number to use instead. 3493 3494See the PowerPC's SPE target for an example. 3495@end defmac 3496 3497@defmac DWARF_FRAME_REGNUM (@var{regno}) 3498 3499Define this macro if the target's representation for dwarf registers 3500used in .eh_frame or .debug_frame is different from that used in other 3501debug info sections. Given a GCC hard register number, this macro 3502should return the .eh_frame register number. The default is 3503@code{DBX_REGISTER_NUMBER (@var{regno})}. 3504 3505@end defmac 3506 3507@defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh}) 3508 3509Define this macro to map register numbers held in the call frame info 3510that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that 3511should be output in .debug_frame (@code{@var{for_eh}} is zero) and 3512.eh_frame (@code{@var{for_eh}} is nonzero). The default is to 3513return @code{@var{regno}}. 3514 3515@end defmac 3516 3517@node Elimination 3518@subsection Eliminating Frame Pointer and Arg Pointer 3519 3520@c prevent bad page break with this line 3521This is about eliminating the frame pointer and arg pointer. 3522 3523@defmac FRAME_POINTER_REQUIRED 3524A C expression which is nonzero if a function must have and use a frame 3525pointer. This expression is evaluated in the reload pass. If its value is 3526nonzero the function will have a frame pointer. 3527 3528The expression can in principle examine the current function and decide 3529according to the facts, but on most machines the constant 0 or the 3530constant 1 suffices. Use 0 when the machine allows code to be generated 3531with no frame pointer, and doing so saves some time or space. Use 1 3532when there is no possible advantage to avoiding a frame pointer. 3533 3534In certain cases, the compiler does not know how to produce valid code 3535without a frame pointer. The compiler recognizes those cases and 3536automatically gives the function a frame pointer regardless of what 3537@code{FRAME_POINTER_REQUIRED} says. You don't need to worry about 3538them. 3539 3540In a function that does not require a frame pointer, the frame pointer 3541register can be allocated for ordinary usage, unless you mark it as a 3542fixed register. See @code{FIXED_REGISTERS} for more information. 3543@end defmac 3544 3545@findex get_frame_size 3546@defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var}) 3547A C statement to store in the variable @var{depth-var} the difference 3548between the frame pointer and the stack pointer values immediately after 3549the function prologue. The value would be computed from information 3550such as the result of @code{get_frame_size ()} and the tables of 3551registers @code{regs_ever_live} and @code{call_used_regs}. 3552 3553If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and 3554need not be defined. Otherwise, it must be defined even if 3555@code{FRAME_POINTER_REQUIRED} is defined to always be true; in that 3556case, you may set @var{depth-var} to anything. 3557@end defmac 3558 3559@defmac ELIMINABLE_REGS 3560If defined, this macro specifies a table of register pairs used to 3561eliminate unneeded registers that point into the stack frame. If it is not 3562defined, the only elimination attempted by the compiler is to replace 3563references to the frame pointer with references to the stack pointer. 3564 3565The definition of this macro is a list of structure initializations, each 3566of which specifies an original and replacement register. 3567 3568On some machines, the position of the argument pointer is not known until 3569the compilation is completed. In such a case, a separate hard register 3570must be used for the argument pointer. This register can be eliminated by 3571replacing it with either the frame pointer or the argument pointer, 3572depending on whether or not the frame pointer has been eliminated. 3573 3574In this case, you might specify: 3575@smallexample 3576#define ELIMINABLE_REGS \ 3577@{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ 3578 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ 3579 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} 3580@end smallexample 3581 3582Note that the elimination of the argument pointer with the stack pointer is 3583specified first since that is the preferred elimination. 3584@end defmac 3585 3586@defmac CAN_ELIMINATE (@var{from-reg}, @var{to-reg}) 3587A C expression that returns nonzero if the compiler is allowed to try 3588to replace register number @var{from-reg} with register number 3589@var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS} 3590is defined, and will usually be the constant 1, since most of the cases 3591preventing register elimination are things that the compiler already 3592knows about. 3593@end defmac 3594 3595@defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) 3596This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It 3597specifies the initial difference between the specified pair of 3598registers. This macro must be defined if @code{ELIMINABLE_REGS} is 3599defined. 3600@end defmac 3601 3602@node Stack Arguments 3603@subsection Passing Function Arguments on the Stack 3604@cindex arguments on stack 3605@cindex stack arguments 3606 3607The macros in this section control how arguments are passed 3608on the stack. See the following section for other macros that 3609control passing certain arguments in registers. 3610 3611@deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype}) 3612This target hook returns @code{true} if an argument declared in a 3613prototype as an integral type smaller than @code{int} should actually be 3614passed as an @code{int}. In addition to avoiding errors in certain 3615cases of mismatch, it also makes for better code on certain machines. 3616The default is to not promote prototypes. 3617@end deftypefn 3618 3619@defmac PUSH_ARGS 3620A C expression. If nonzero, push insns will be used to pass 3621outgoing arguments. 3622If the target machine does not have a push instruction, set it to zero. 3623That directs GCC to use an alternate strategy: to 3624allocate the entire argument block and then store the arguments into 3625it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too. 3626@end defmac 3627 3628@defmac PUSH_ARGS_REVERSED 3629A C expression. If nonzero, function arguments will be evaluated from 3630last to first, rather than from first to last. If this macro is not 3631defined, it defaults to @code{PUSH_ARGS} on targets where the stack 3632and args grow in opposite directions, and 0 otherwise. 3633@end defmac 3634 3635@defmac PUSH_ROUNDING (@var{npushed}) 3636A C expression that is the number of bytes actually pushed onto the 3637stack when an instruction attempts to push @var{npushed} bytes. 3638 3639On some machines, the definition 3640 3641@smallexample 3642#define PUSH_ROUNDING(BYTES) (BYTES) 3643@end smallexample 3644 3645@noindent 3646will suffice. But on other machines, instructions that appear 3647to push one byte actually push two bytes in an attempt to maintain 3648alignment. Then the definition should be 3649 3650@smallexample 3651#define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) 3652@end smallexample 3653@end defmac 3654 3655@findex current_function_outgoing_args_size 3656@defmac ACCUMULATE_OUTGOING_ARGS 3657A C expression. If nonzero, the maximum amount of space required for outgoing arguments 3658will be computed and placed into the variable 3659@code{current_function_outgoing_args_size}. No space will be pushed 3660onto the stack for each call; instead, the function prologue should 3661increase the stack frame size by this amount. 3662 3663Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS} 3664is not proper. 3665@end defmac 3666 3667@defmac REG_PARM_STACK_SPACE (@var{fndecl}) 3668Define this macro if functions should assume that stack space has been 3669allocated for arguments even when their values are passed in 3670registers. 3671 3672The value of this macro is the size, in bytes, of the area reserved for 3673arguments passed in registers for the function represented by @var{fndecl}, 3674which can be zero if GCC is calling a library function. 3675 3676This space can be allocated by the caller, or be a part of the 3677machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says 3678which. 3679@end defmac 3680@c above is overfull. not sure what to do. --mew 5feb93 did 3681@c something, not sure if it looks good. --mew 10feb93 3682 3683@defmac OUTGOING_REG_PARM_STACK_SPACE 3684Define this if it is the responsibility of the caller to allocate the area 3685reserved for arguments passed in registers. 3686 3687If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls 3688whether the space for these arguments counts in the value of 3689@code{current_function_outgoing_args_size}. 3690@end defmac 3691 3692@defmac STACK_PARMS_IN_REG_PARM_AREA 3693Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the 3694stack parameters don't skip the area specified by it. 3695@c i changed this, makes more sens and it should have taken care of the 3696@c overfull.. not as specific, tho. --mew 5feb93 3697 3698Normally, when a parameter is not passed in registers, it is placed on the 3699stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro 3700suppresses this behavior and causes the parameter to be passed on the 3701stack in its natural location. 3702@end defmac 3703 3704@defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size}) 3705A C expression that should indicate the number of bytes of its own 3706arguments that a function pops on returning, or 0 if the 3707function pops no arguments and the caller must therefore pop them all 3708after the function returns. 3709 3710@var{fundecl} is a C variable whose value is a tree node that describes 3711the function in question. Normally it is a node of type 3712@code{FUNCTION_DECL} that describes the declaration of the function. 3713From this you can obtain the @code{DECL_ATTRIBUTES} of the function. 3714 3715@var{funtype} is a C variable whose value is a tree node that 3716describes the function in question. Normally it is a node of type 3717@code{FUNCTION_TYPE} that describes the data type of the function. 3718From this it is possible to obtain the data types of the value and 3719arguments (if known). 3720 3721When a call to a library function is being considered, @var{fundecl} 3722will contain an identifier node for the library function. Thus, if 3723you need to distinguish among various library functions, you can do so 3724by their names. Note that ``library function'' in this context means 3725a function used to perform arithmetic, whose name is known specially 3726in the compiler and was not mentioned in the C code being compiled. 3727 3728@var{stack-size} is the number of bytes of arguments passed on the 3729stack. If a variable number of bytes is passed, it is zero, and 3730argument popping will always be the responsibility of the calling function. 3731 3732On the VAX, all functions always pop their arguments, so the definition 3733of this macro is @var{stack-size}. On the 68000, using the standard 3734calling convention, no functions pop their arguments, so the value of 3735the macro is always 0 in this case. But an alternative calling 3736convention is available in which functions that take a fixed number of 3737arguments pop them but other functions (such as @code{printf}) pop 3738nothing (the caller pops all). When this convention is in use, 3739@var{funtype} is examined to determine whether a function takes a fixed 3740number of arguments. 3741@end defmac 3742 3743@defmac CALL_POPS_ARGS (@var{cum}) 3744A C expression that should indicate the number of bytes a call sequence 3745pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS} 3746when compiling a function call. 3747 3748@var{cum} is the variable in which all arguments to the called function 3749have been accumulated. 3750 3751On certain architectures, such as the SH5, a call trampoline is used 3752that pops certain registers off the stack, depending on the arguments 3753that have been passed to the function. Since this is a property of the 3754call site, not of the called function, @code{RETURN_POPS_ARGS} is not 3755appropriate. 3756@end defmac 3757 3758@node Register Arguments 3759@subsection Passing Arguments in Registers 3760@cindex arguments in registers 3761@cindex registers arguments 3762 3763This section describes the macros which let you control how various 3764types of arguments are passed in registers or how they are arranged in 3765the stack. 3766 3767@defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) 3768A C expression that controls whether a function argument is passed 3769in a register, and which register. 3770 3771The arguments are @var{cum}, which summarizes all the previous 3772arguments; @var{mode}, the machine mode of the argument; @var{type}, 3773the data type of the argument as a tree node or 0 if that is not known 3774(which happens for C support library functions); and @var{named}, 3775which is 1 for an ordinary argument and 0 for nameless arguments that 3776correspond to @samp{@dots{}} in the called function's prototype. 3777@var{type} can be an incomplete type if a syntax error has previously 3778occurred. 3779 3780The value of the expression is usually either a @code{reg} RTX for the 3781hard register in which to pass the argument, or zero to pass the 3782argument on the stack. 3783 3784For machines like the VAX and 68000, where normally all arguments are 3785pushed, zero suffices as a definition. 3786 3787The value of the expression can also be a @code{parallel} RTX@. This is 3788used when an argument is passed in multiple locations. The mode of the 3789@code{parallel} should be the mode of the entire argument. The 3790@code{parallel} holds any number of @code{expr_list} pairs; each one 3791describes where part of the argument is passed. In each 3792@code{expr_list} the first operand must be a @code{reg} RTX for the hard 3793register in which to pass this part of the argument, and the mode of the 3794register RTX indicates how large this part of the argument is. The 3795second operand of the @code{expr_list} is a @code{const_int} which gives 3796the offset in bytes into the entire argument of where this part starts. 3797As a special exception the first @code{expr_list} in the @code{parallel} 3798RTX may have a first operand of zero. This indicates that the entire 3799argument is also stored on the stack. 3800 3801The last time this macro is called, it is called with @code{MODE == 3802VOIDmode}, and its result is passed to the @code{call} or @code{call_value} 3803pattern as operands 2 and 3 respectively. 3804 3805@cindex @file{stdarg.h} and register arguments 3806The usual way to make the ISO library @file{stdarg.h} work on a machine 3807where some arguments are usually passed in registers, is to cause 3808nameless arguments to be passed on the stack instead. This is done 3809by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0. 3810 3811@cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG} 3812@cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG} 3813You may use the hook @code{targetm.calls.must_pass_in_stack} 3814in the definition of this macro to determine if this argument is of a 3815type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE} 3816is not defined and @code{FUNCTION_ARG} returns nonzero for such an 3817argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is 3818defined, the argument will be computed in the stack and then loaded into 3819a register. 3820@end defmac 3821 3822@deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type}) 3823This target hook should return @code{true} if we should not pass @var{type} 3824solely in registers. The file @file{expr.h} defines a 3825definition that is usually appropriate, refer to @file{expr.h} for additional 3826documentation. 3827@end deftypefn 3828 3829@defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) 3830Define this macro if the target machine has ``register windows'', so 3831that the register in which a function sees an arguments is not 3832necessarily the same as the one in which the caller passed the 3833argument. 3834 3835For such machines, @code{FUNCTION_ARG} computes the register in which 3836the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should 3837be defined in a similar fashion to tell the function being called 3838where the arguments will arrive. 3839 3840If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG} 3841serves both purposes. 3842@end defmac 3843 3844@deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named}) 3845This target hook returns the number of bytes at the beginning of an 3846argument that must be put in registers. The value must be zero for 3847arguments that are passed entirely in registers or that are entirely 3848pushed on the stack. 3849 3850On some machines, certain arguments must be passed partially in 3851registers and partially in memory. On these machines, typically the 3852first few words of arguments are passed in registers, and the rest 3853on the stack. If a multi-word argument (a @code{double} or a 3854structure) crosses that boundary, its first few words must be passed 3855in registers and the rest must be pushed. This macro tells the 3856compiler when this occurs, and how many bytes should go in registers. 3857 3858@code{FUNCTION_ARG} for these arguments should return the first 3859register to be used by the caller for this argument; likewise 3860@code{FUNCTION_INCOMING_ARG}, for the called function. 3861@end deftypefn 3862 3863@deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named}) 3864This target hook should return @code{true} if an argument at the 3865position indicated by @var{cum} should be passed by reference. This 3866predicate is queried after target independent reasons for being 3867passed by reference, such as @code{TREE_ADDRESSABLE (type)}. 3868 3869If the hook returns true, a copy of that argument is made in memory and a 3870pointer to the argument is passed instead of the argument itself. 3871The pointer is passed in whatever way is appropriate for passing a pointer 3872to that type. 3873@end deftypefn 3874 3875@deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named}) 3876The function argument described by the parameters to this hook is 3877known to be passed by reference. The hook should return true if the 3878function argument should be copied by the callee instead of copied 3879by the caller. 3880 3881For any argument for which the hook returns true, if it can be 3882determined that the argument is not modified, then a copy need 3883not be generated. 3884 3885The default version of this hook always returns false. 3886@end deftypefn 3887 3888@defmac CUMULATIVE_ARGS 3889A C type for declaring a variable that is used as the first argument of 3890@code{FUNCTION_ARG} and other related values. For some target machines, 3891the type @code{int} suffices and can hold the number of bytes of 3892argument so far. 3893 3894There is no need to record in @code{CUMULATIVE_ARGS} anything about the 3895arguments that have been passed on the stack. The compiler has other 3896variables to keep track of that. For target machines on which all 3897arguments are passed on the stack, there is no need to store anything in 3898@code{CUMULATIVE_ARGS}; however, the data structure must exist and 3899should not be empty, so use @code{int}. 3900@end defmac 3901 3902@defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args}) 3903A C statement (sans semicolon) for initializing the variable 3904@var{cum} for the state at the beginning of the argument list. The 3905variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype} 3906is the tree node for the data type of the function which will receive 3907the args, or 0 if the args are to a compiler support library function. 3908For direct calls that are not libcalls, @var{fndecl} contain the 3909declaration node of the function. @var{fndecl} is also set when 3910@code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function 3911being compiled. @var{n_named_args} is set to the number of named 3912arguments, including a structure return address if it is passed as a 3913parameter, when making a call. When processing incoming arguments, 3914@var{n_named_args} is set to @minus{}1. 3915 3916When processing a call to a compiler support library function, 3917@var{libname} identifies which one. It is a @code{symbol_ref} rtx which 3918contains the name of the function, as a string. @var{libname} is 0 when 3919an ordinary C function call is being processed. Thus, each time this 3920macro is called, either @var{libname} or @var{fntype} is nonzero, but 3921never both of them at once. 3922@end defmac 3923 3924@defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname}) 3925Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls, 3926it gets a @code{MODE} argument instead of @var{fntype}, that would be 3927@code{NULL}. @var{indirect} would always be zero, too. If this macro 3928is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, 39290)} is used instead. 3930@end defmac 3931 3932@defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) 3933Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of 3934finding the arguments for the function being compiled. If this macro is 3935undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. 3936 3937The value passed for @var{libname} is always 0, since library routines 3938with special calling conventions are never compiled with GCC@. The 3939argument @var{libname} exists for symmetry with 3940@code{INIT_CUMULATIVE_ARGS}. 3941@c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. 3942@c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 3943@end defmac 3944 3945@defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named}) 3946A C statement (sans semicolon) to update the summarizer variable 3947@var{cum} to advance past an argument in the argument list. The 3948values @var{mode}, @var{type} and @var{named} describe that argument. 3949Once this is done, the variable @var{cum} is suitable for analyzing 3950the @emph{following} argument with @code{FUNCTION_ARG}, etc. 3951 3952This macro need not do anything if the argument in question was passed 3953on the stack. The compiler knows how to track the amount of stack space 3954used for arguments without any special help. 3955@end defmac 3956 3957@defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type}) 3958If defined, a C expression which determines whether, and in which direction, 3959to pad out an argument with extra space. The value should be of type 3960@code{enum direction}: either @code{upward} to pad above the argument, 3961@code{downward} to pad below, or @code{none} to inhibit padding. 3962 3963The @emph{amount} of padding is always just enough to reach the next 3964multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control 3965it. 3966 3967This macro has a default definition which is right for most systems. 3968For little-endian machines, the default is to pad upward. For 3969big-endian machines, the default is to pad downward for an argument of 3970constant size shorter than an @code{int}, and upward otherwise. 3971@end defmac 3972 3973@defmac PAD_VARARGS_DOWN 3974If defined, a C expression which determines whether the default 3975implementation of va_arg will attempt to pad down before reading the 3976next argument, if that argument is smaller than its aligned space as 3977controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such 3978arguments are padded down if @code{BYTES_BIG_ENDIAN} is true. 3979@end defmac 3980 3981@defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first}) 3982Specify padding for the last element of a block move between registers and 3983memory. @var{first} is nonzero if this is the only element. Defining this 3984macro allows better control of register function parameters on big-endian 3985machines, without using @code{PARALLEL} rtl. In particular, 3986@code{MUST_PASS_IN_STACK} need not test padding and mode of types in 3987registers, as there is no longer a "wrong" part of a register; For example, 3988a three byte aggregate may be passed in the high part of a register if so 3989required. 3990@end defmac 3991 3992@defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type}) 3993If defined, a C expression that gives the alignment boundary, in bits, 3994of an argument with the specified mode and type. If it is not defined, 3995@code{PARM_BOUNDARY} is used for all arguments. 3996@end defmac 3997 3998@defmac FUNCTION_ARG_REGNO_P (@var{regno}) 3999A C expression that is nonzero if @var{regno} is the number of a hard 4000register in which function arguments are sometimes passed. This does 4001@emph{not} include implicit arguments such as the static chain and 4002the structure-value address. On many machines, no registers can be 4003used for this purpose since all function arguments are pushed on the 4004stack. 4005@end defmac 4006 4007@deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type}) 4008This hook should return true if parameter of type @var{type} are passed 4009as two scalar parameters. By default, GCC will attempt to pack complex 4010arguments into the target's word size. Some ABIs require complex arguments 4011to be split and treated as their individual components. For example, on 4012AIX64, complex floats should be passed in a pair of floating point 4013registers, even though a complex float would fit in one 64-bit floating 4014point register. 4015 4016The default value of this hook is @code{NULL}, which is treated as always 4017false. 4018@end deftypefn 4019 4020@deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void) 4021This hook returns a type node for @code{va_list} for the target. 4022The default version of the hook returns @code{void*}. 4023@end deftypefn 4024 4025@deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p}) 4026This hook performs target-specific gimplification of 4027@code{VA_ARG_EXPR}. The first two parameters correspond to the 4028arguments to @code{va_arg}; the latter two are as in 4029@code{gimplify.c:gimplify_expr}. 4030@end deftypefn 4031 4032@deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode}) 4033Define this to return nonzero if the port can handle pointers 4034with machine mode @var{mode}. The default version of this 4035hook returns true for both @code{ptr_mode} and @code{Pmode}. 4036@end deftypefn 4037 4038@deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode}) 4039Define this to return nonzero if the port is prepared to handle 4040insns involving scalar mode @var{mode}. For a scalar mode to be 4041considered supported, all the basic arithmetic and comparisons 4042must work. 4043 4044The default version of this hook returns true for any mode 4045required to handle the basic C types (as defined by the port). 4046Included here are the double-word arithmetic supported by the 4047code in @file{optabs.c}. 4048@end deftypefn 4049 4050@deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode}) 4051Define this to return nonzero if the port is prepared to handle 4052insns involving vector mode @var{mode}. At the very least, it 4053must have move patterns for this mode. 4054@end deftypefn 4055 4056@node Scalar Return 4057@subsection How Scalar Function Values Are Returned 4058@cindex return values in registers 4059@cindex values, returned by functions 4060@cindex scalars, returned as values 4061 4062This section discusses the macros that control returning scalars as 4063values---values that can fit in registers. 4064 4065@deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing}) 4066 4067Define this to return an RTX representing the place where a function 4068returns or receives a value of data type @var{ret_type}, a tree node 4069node representing a data type. @var{fn_decl_or_type} is a tree node 4070representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a 4071function being called. If @var{outgoing} is false, the hook should 4072compute the register in which the caller will see the return value. 4073Otherwise, the hook should return an RTX representing the place where 4074a function returns a value. 4075 4076On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant. 4077(Actually, on most machines, scalar values are returned in the same 4078place regardless of mode.) The value of the expression is usually a 4079@code{reg} RTX for the hard register where the return value is stored. 4080The value can also be a @code{parallel} RTX, if the return value is in 4081multiple places. See @code{FUNCTION_ARG} for an explanation of the 4082@code{parallel} form. 4083 4084If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply 4085the same promotion rules specified in @code{PROMOTE_MODE} if 4086@var{valtype} is a scalar type. 4087 4088If the precise function being called is known, @var{func} is a tree 4089node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null 4090pointer. This makes it possible to use a different value-returning 4091convention for specific functions when all their calls are 4092known. 4093 4094Some target machines have ``register windows'' so that the register in 4095which a function returns its value is not the same as the one in which 4096the caller sees the value. For such machines, you should return 4097different RTX depending on @var{outgoing}. 4098 4099@code{TARGET_FUNCTION_VALUE} is not used for return values with 4100aggregate data types, because these are returned in another way. See 4101@code{TARGET_STRUCT_VALUE_RTX} and related macros, below. 4102@end deftypefn 4103 4104@defmac FUNCTION_VALUE (@var{valtype}, @var{func}) 4105This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for 4106a new target instead. 4107@end defmac 4108 4109@defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func}) 4110This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for 4111a new target instead. 4112@end defmac 4113 4114@defmac LIBCALL_VALUE (@var{mode}) 4115A C expression to create an RTX representing the place where a library 4116function returns a value of mode @var{mode}. If the precise function 4117being called is known, @var{func} is a tree node 4118(@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null 4119pointer. This makes it possible to use a different value-returning 4120convention for specific functions when all their calls are 4121known. 4122 4123Note that ``library function'' in this context means a compiler 4124support routine, used to perform arithmetic, whose name is known 4125specially by the compiler and was not mentioned in the C code being 4126compiled. 4127 4128The definition of @code{LIBRARY_VALUE} need not be concerned aggregate 4129data types, because none of the library functions returns such types. 4130@end defmac 4131 4132@defmac FUNCTION_VALUE_REGNO_P (@var{regno}) 4133A C expression that is nonzero if @var{regno} is the number of a hard 4134register in which the values of called function may come back. 4135 4136A register whose use for returning values is limited to serving as the 4137second of a pair (for a value of type @code{double}, say) need not be 4138recognized by this macro. So for most machines, this definition 4139suffices: 4140 4141@smallexample 4142#define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) 4143@end smallexample 4144 4145If the machine has register windows, so that the caller and the called 4146function use different registers for the return value, this macro 4147should recognize only the caller's register numbers. 4148@end defmac 4149 4150@defmac APPLY_RESULT_SIZE 4151Define this macro if @samp{untyped_call} and @samp{untyped_return} 4152need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for 4153saving and restoring an arbitrary return value. 4154@end defmac 4155 4156@deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type}) 4157This hook should return true if values of type @var{type} are returned 4158at the most significant end of a register (in other words, if they are 4159padded at the least significant end). You can assume that @var{type} 4160is returned in a register; the caller is required to check this. 4161 4162Note that the register provided by @code{TARGET_FUNCTION_VALUE} must 4163be able to hold the complete return value. For example, if a 1-, 2- 4164or 3-byte structure is returned at the most significant end of a 41654-byte register, @code{TARGET_FUNCTION_VALUE} should provide an 4166@code{SImode} rtx. 4167@end deftypefn 4168 4169@node Aggregate Return 4170@subsection How Large Values Are Returned 4171@cindex aggregates as return values 4172@cindex large return values 4173@cindex returning aggregate values 4174@cindex structure value address 4175 4176When a function value's mode is @code{BLKmode} (and in some other 4177cases), the value is not returned according to 4178@code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the 4179caller passes the address of a block of memory in which the value 4180should be stored. This address is called the @dfn{structure value 4181address}. 4182 4183This section describes how to control returning structure values in 4184memory. 4185 4186@deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype}) 4187This target hook should return a nonzero value to say to return the 4188function value in memory, just as large structures are always returned. 4189Here @var{type} will be the data type of the value, and @var{fntype} 4190will be the type of the function doing the returning, or @code{NULL} for 4191libcalls. 4192 4193Note that values of mode @code{BLKmode} must be explicitly handled 4194by this function. Also, the option @option{-fpcc-struct-return} 4195takes effect regardless of this macro. On most systems, it is 4196possible to leave the hook undefined; this causes a default 4197definition to be used, whose value is the constant 1 for @code{BLKmode} 4198values, and 0 otherwise. 4199 4200Do not use this hook to indicate that structures and unions should always 4201be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN} 4202to indicate this. 4203@end deftypefn 4204 4205@defmac DEFAULT_PCC_STRUCT_RETURN 4206Define this macro to be 1 if all structure and union return values must be 4207in memory. Since this results in slower code, this should be defined 4208only if needed for compatibility with other compilers or with an ABI@. 4209If you define this macro to be 0, then the conventions used for structure 4210and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY} 4211target hook. 4212 4213If not defined, this defaults to the value 1. 4214@end defmac 4215 4216@deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming}) 4217This target hook should return the location of the structure value 4218address (normally a @code{mem} or @code{reg}), or 0 if the address is 4219passed as an ``invisible'' first argument. Note that @var{fndecl} may 4220be @code{NULL}, for libcalls. You do not need to define this target 4221hook if the address is always passed as an ``invisible'' first 4222argument. 4223 4224On some architectures the place where the structure value address 4225is found by the called function is not the same place that the 4226caller put it. This can be due to register windows, or it could 4227be because the function prologue moves it to a different place. 4228@var{incoming} is @code{1} or @code{2} when the location is needed in 4229the context of the called function, and @code{0} in the context of 4230the caller. 4231 4232If @var{incoming} is nonzero and the address is to be found on the 4233stack, return a @code{mem} which refers to the frame pointer. If 4234@var{incoming} is @code{2}, the result is being used to fetch the 4235structure value address at the beginning of a function. If you need 4236to emit adjusting code, you should do it at this point. 4237@end deftypefn 4238 4239@defmac PCC_STATIC_STRUCT_RETURN 4240Define this macro if the usual system convention on the target machine 4241for returning structures and unions is for the called function to return 4242the address of a static variable containing the value. 4243 4244Do not define this if the usual system convention is for the caller to 4245pass an address to the subroutine. 4246 4247This macro has effect in @option{-fpcc-struct-return} mode, but it does 4248nothing when you use @option{-freg-struct-return} mode. 4249@end defmac 4250 4251@node Caller Saves 4252@subsection Caller-Saves Register Allocation 4253 4254If you enable it, GCC can save registers around function calls. This 4255makes it possible to use call-clobbered registers to hold variables that 4256must live across calls. 4257 4258@defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls}) 4259A C expression to determine whether it is worthwhile to consider placing 4260a pseudo-register in a call-clobbered hard register and saving and 4261restoring it around each function call. The expression should be 1 when 4262this is worth doing, and 0 otherwise. 4263 4264If you don't define this macro, a default is used which is good on most 4265machines: @code{4 * @var{calls} < @var{refs}}. 4266@end defmac 4267 4268@defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs}) 4269A C expression specifying which mode is required for saving @var{nregs} 4270of a pseudo-register in call-clobbered hard register @var{regno}. If 4271@var{regno} is unsuitable for caller save, @code{VOIDmode} should be 4272returned. For most machines this macro need not be defined since GCC 4273will select the smallest suitable mode. 4274@end defmac 4275 4276@node Function Entry 4277@subsection Function Entry and Exit 4278@cindex function entry and exit 4279@cindex prologue 4280@cindex epilogue 4281 4282This section describes the macros that output function entry 4283(@dfn{prologue}) and exit (@dfn{epilogue}) code. 4284 4285@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4286If defined, a function that outputs the assembler code for entry to a 4287function. The prologue is responsible for setting up the stack frame, 4288initializing the frame pointer register, saving registers that must be 4289saved, and allocating @var{size} additional bytes of storage for the 4290local variables. @var{size} is an integer. @var{file} is a stdio 4291stream to which the assembler code should be output. 4292 4293The label for the beginning of the function need not be output by this 4294macro. That has already been done when the macro is run. 4295 4296@findex regs_ever_live 4297To determine which registers to save, the macro can refer to the array 4298@code{regs_ever_live}: element @var{r} is nonzero if hard register 4299@var{r} is used anywhere within the function. This implies the function 4300prologue should save register @var{r}, provided it is not one of the 4301call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use 4302@code{regs_ever_live}.) 4303 4304On machines that have ``register windows'', the function entry code does 4305not save on the stack the registers that are in the windows, even if 4306they are supposed to be preserved by function calls; instead it takes 4307appropriate steps to ``push'' the register stack, if any non-call-used 4308registers are used in the function. 4309 4310@findex frame_pointer_needed 4311On machines where functions may or may not have frame-pointers, the 4312function entry code must vary accordingly; it must set up the frame 4313pointer if one is wanted, and not otherwise. To determine whether a 4314frame pointer is in wanted, the macro can refer to the variable 4315@code{frame_pointer_needed}. The variable's value will be 1 at run 4316time in a function that needs a frame pointer. @xref{Elimination}. 4317 4318The function entry code is responsible for allocating any stack space 4319required for the function. This stack space consists of the regions 4320listed below. In most cases, these regions are allocated in the 4321order listed, with the last listed region closest to the top of the 4322stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and 4323the highest address if it is not defined). You can use a different order 4324for a machine if doing so is more convenient or required for 4325compatibility reasons. Except in cases where required by standard 4326or by a debugger, there is no reason why the stack layout used by GCC 4327need agree with that used by other compilers for a machine. 4328@end deftypefn 4329 4330@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file}) 4331If defined, a function that outputs assembler code at the end of a 4332prologue. This should be used when the function prologue is being 4333emitted as RTL, and you have some extra assembler that needs to be 4334emitted. @xref{prologue instruction pattern}. 4335@end deftypefn 4336 4337@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file}) 4338If defined, a function that outputs assembler code at the start of an 4339epilogue. This should be used when the function epilogue is being 4340emitted as RTL, and you have some extra assembler that needs to be 4341emitted. @xref{epilogue instruction pattern}. 4342@end deftypefn 4343 4344@deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) 4345If defined, a function that outputs the assembler code for exit from a 4346function. The epilogue is responsible for restoring the saved 4347registers and stack pointer to their values when the function was 4348called, and returning control to the caller. This macro takes the 4349same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the 4350registers to restore are determined from @code{regs_ever_live} and 4351@code{CALL_USED_REGISTERS} in the same way. 4352 4353On some machines, there is a single instruction that does all the work 4354of returning from the function. On these machines, give that 4355instruction the name @samp{return} and do not define the macro 4356@code{TARGET_ASM_FUNCTION_EPILOGUE} at all. 4357 4358Do not define a pattern named @samp{return} if you want the 4359@code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target 4360switches to control whether return instructions or epilogues are used, 4361define a @samp{return} pattern with a validity condition that tests the 4362target switches appropriately. If the @samp{return} pattern's validity 4363condition is false, epilogues will be used. 4364 4365On machines where functions may or may not have frame-pointers, the 4366function exit code must vary accordingly. Sometimes the code for these 4367two cases is completely different. To determine whether a frame pointer 4368is wanted, the macro can refer to the variable 4369@code{frame_pointer_needed}. The variable's value will be 1 when compiling 4370a function that needs a frame pointer. 4371 4372Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and 4373@code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially. 4374The C variable @code{current_function_is_leaf} is nonzero for such a 4375function. @xref{Leaf Functions}. 4376 4377On some machines, some functions pop their arguments on exit while 4378others leave that for the caller to do. For example, the 68020 when 4379given @option{-mrtd} pops arguments in functions that take a fixed 4380number of arguments. 4381 4382@findex current_function_pops_args 4383Your definition of the macro @code{RETURN_POPS_ARGS} decides which 4384functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE} 4385needs to know what was decided. The variable that is called 4386@code{current_function_pops_args} is the number of bytes of its 4387arguments that a function should pop. @xref{Scalar Return}. 4388@c what is the "its arguments" in the above sentence referring to, pray 4389@c tell? --mew 5feb93 4390@end deftypefn 4391 4392@itemize @bullet 4393@item 4394@findex current_function_pretend_args_size 4395A region of @code{current_function_pretend_args_size} bytes of 4396uninitialized space just underneath the first argument arriving on the 4397stack. (This may not be at the very start of the allocated stack region 4398if the calling sequence has pushed anything else since pushing the stack 4399arguments. But usually, on such machines, nothing else has been pushed 4400yet, because the function prologue itself does all the pushing.) This 4401region is used on machines where an argument may be passed partly in 4402registers and partly in memory, and, in some cases to support the 4403features in @code{<stdarg.h>}. 4404 4405@item 4406An area of memory used to save certain registers used by the function. 4407The size of this area, which may also include space for such things as 4408the return address and pointers to previous stack frames, is 4409machine-specific and usually depends on which registers have been used 4410in the function. Machines with register windows often do not require 4411a save area. 4412 4413@item 4414A region of at least @var{size} bytes, possibly rounded up to an allocation 4415boundary, to contain the local variables of the function. On some machines, 4416this region and the save area may occur in the opposite order, with the 4417save area closer to the top of the stack. 4418 4419@item 4420@cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames 4421Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of 4422@code{current_function_outgoing_args_size} bytes to be used for outgoing 4423argument lists of the function. @xref{Stack Arguments}. 4424@end itemize 4425 4426@defmac EXIT_IGNORE_STACK 4427Define this macro as a C expression that is nonzero if the return 4428instruction or the function epilogue ignores the value of the stack 4429pointer; in other words, if it is safe to delete an instruction to 4430adjust the stack pointer before a return from the function. The 4431default is 0. 4432 4433Note that this macro's value is relevant only for functions for which 4434frame pointers are maintained. It is never safe to delete a final 4435stack adjustment in a function that has no frame pointer, and the 4436compiler knows this regardless of @code{EXIT_IGNORE_STACK}. 4437@end defmac 4438 4439@defmac EPILOGUE_USES (@var{regno}) 4440Define this macro as a C expression that is nonzero for registers that are 4441used by the epilogue or the @samp{return} pattern. The stack and frame 4442pointer registers are already assumed to be used as needed. 4443@end defmac 4444 4445@defmac EH_USES (@var{regno}) 4446Define this macro as a C expression that is nonzero for registers that are 4447used by the exception handling mechanism, and so should be considered live 4448on entry to an exception edge. 4449@end defmac 4450 4451@defmac DELAY_SLOTS_FOR_EPILOGUE 4452Define this macro if the function epilogue contains delay slots to which 4453instructions from the rest of the function can be ``moved''. The 4454definition should be a C expression whose value is an integer 4455representing the number of delay slots there. 4456@end defmac 4457 4458@defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n}) 4459A C expression that returns 1 if @var{insn} can be placed in delay 4460slot number @var{n} of the epilogue. 4461 4462The argument @var{n} is an integer which identifies the delay slot now 4463being considered (since different slots may have different rules of 4464eligibility). It is never negative and is always less than the number 4465of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns). 4466If you reject a particular insn for a given delay slot, in principle, it 4467may be reconsidered for a subsequent delay slot. Also, other insns may 4468(at least in principle) be considered for the so far unfilled delay 4469slot. 4470 4471@findex current_function_epilogue_delay_list 4472@findex final_scan_insn 4473The insns accepted to fill the epilogue delay slots are put in an RTL 4474list made with @code{insn_list} objects, stored in the variable 4475@code{current_function_epilogue_delay_list}. The insn for the first 4476delay slot comes first in the list. Your definition of the macro 4477@code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by 4478outputting the insns in this list, usually by calling 4479@code{final_scan_insn}. 4480 4481You need not define this macro if you did not define 4482@code{DELAY_SLOTS_FOR_EPILOGUE}. 4483@end defmac 4484 4485@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function}) 4486A function that outputs the assembler code for a thunk 4487function, used to implement C++ virtual function calls with multiple 4488inheritance. The thunk acts as a wrapper around a virtual function, 4489adjusting the implicit object parameter before handing control off to 4490the real function. 4491 4492First, emit code to add the integer @var{delta} to the location that 4493contains the incoming first argument. Assume that this argument 4494contains a pointer, and is the one used to pass the @code{this} pointer 4495in C++. This is the incoming argument @emph{before} the function prologue, 4496e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of 4497all other incoming arguments. 4498 4499Then, if @var{vcall_offset} is nonzero, an additional adjustment should be 4500made after adding @code{delta}. In particular, if @var{p} is the 4501adjusted pointer, the following adjustment should be made: 4502 4503@smallexample 4504p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)] 4505@end smallexample 4506 4507After the additions, emit code to jump to @var{function}, which is a 4508@code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does 4509not touch the return address. Hence returning from @var{FUNCTION} will 4510return to whoever called the current @samp{thunk}. 4511 4512The effect must be as if @var{function} had been called directly with 4513the adjusted first argument. This macro is responsible for emitting all 4514of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE} 4515and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked. 4516 4517The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function} 4518have already been extracted from it.) It might possibly be useful on 4519some targets, but probably not. 4520 4521If you do not define this macro, the target-independent code in the C++ 4522front end will generate a less efficient heavyweight thunk that calls 4523@var{function} instead of jumping to it. The generic approach does 4524not support varargs. 4525@end deftypefn 4526 4527@deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function}) 4528A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able 4529to output the assembler code for the thunk function specified by the 4530arguments it is passed, and false otherwise. In the latter case, the 4531generic approach will be used by the C++ front end, with the limitations 4532previously exposed. 4533@end deftypefn 4534 4535@node Profiling 4536@subsection Generating Code for Profiling 4537@cindex profiling, code generation 4538 4539These macros will help you generate code for profiling. 4540 4541@defmac FUNCTION_PROFILER (@var{file}, @var{labelno}) 4542A C statement or compound statement to output to @var{file} some 4543assembler code to call the profiling subroutine @code{mcount}. 4544 4545@findex mcount 4546The details of how @code{mcount} expects to be called are determined by 4547your operating system environment, not by GCC@. To figure them out, 4548compile a small program for profiling using the system's installed C 4549compiler and look at the assembler code that results. 4550 4551Older implementations of @code{mcount} expect the address of a counter 4552variable to be loaded into some register. The name of this variable is 4553@samp{LP} followed by the number @var{labelno}, so you would generate 4554the name using @samp{LP%d} in a @code{fprintf}. 4555@end defmac 4556 4557@defmac PROFILE_HOOK 4558A C statement or compound statement to output to @var{file} some assembly 4559code to call the profiling subroutine @code{mcount} even the target does 4560not support profiling. 4561@end defmac 4562 4563@defmac NO_PROFILE_COUNTERS 4564Define this macro to be an expression with a nonzero value if the 4565@code{mcount} subroutine on your system does not need a counter variable 4566allocated for each function. This is true for almost all modern 4567implementations. If you define this macro, you must not use the 4568@var{labelno} argument to @code{FUNCTION_PROFILER}. 4569@end defmac 4570 4571@defmac PROFILE_BEFORE_PROLOGUE 4572Define this macro if the code for function profiling should come before 4573the function prologue. Normally, the profiling code comes after. 4574@end defmac 4575 4576@node Tail Calls 4577@subsection Permitting tail calls 4578@cindex tail calls 4579 4580@deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp}) 4581True if it is ok to do sibling call optimization for the specified 4582call expression @var{exp}. @var{decl} will be the called function, 4583or @code{NULL} if this is an indirect call. 4584 4585It is not uncommon for limitations of calling conventions to prevent 4586tail calls to functions outside the current unit of translation, or 4587during PIC compilation. The hook is used to enforce these restrictions, 4588as the @code{sibcall} md pattern can not fail, or fall over to a 4589``normal'' call. The criteria for successful sibling call optimization 4590may vary greatly between different architectures. 4591@end deftypefn 4592 4593@deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs}) 4594Add any hard registers to @var{regs} that are live on entry to the 4595function. This hook only needs to be defined to provide registers that 4596cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved 4597registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM, 4598TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES, 4599FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM. 4600@end deftypefn 4601 4602@node Stack Smashing Protection 4603@subsection Stack smashing protection 4604@cindex stack smashing protection 4605 4606@deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void) 4607This hook returns a @code{DECL} node for the external variable to use 4608for the stack protection guard. This variable is initialized by the 4609runtime to some random value and is used to initialize the guard value 4610that is placed at the top of the local stack frame. The type of this 4611variable must be @code{ptr_type_node}. 4612 4613The default version of this hook creates a variable called 4614@samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}. 4615@end deftypefn 4616 4617@deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void) 4618This hook returns a tree expression that alerts the runtime that the 4619stack protect guard variable has been modified. This expression should 4620involve a call to a @code{noreturn} function. 4621 4622The default version of this hook invokes a function called 4623@samp{__stack_chk_fail}, taking no arguments. This function is 4624normally defined in @file{libgcc2.c}. 4625@end deftypefn 4626 4627@node Varargs 4628@section Implementing the Varargs Macros 4629@cindex varargs implementation 4630 4631GCC comes with an implementation of @code{<varargs.h>} and 4632@code{<stdarg.h>} that work without change on machines that pass arguments 4633on the stack. Other machines require their own implementations of 4634varargs, and the two machine independent header files must have 4635conditionals to include it. 4636 4637ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in 4638the calling convention for @code{va_start}. The traditional 4639implementation takes just one argument, which is the variable in which 4640to store the argument pointer. The ISO implementation of 4641@code{va_start} takes an additional second argument. The user is 4642supposed to write the last named argument of the function here. 4643 4644However, @code{va_start} should not use this argument. The way to find 4645the end of the named arguments is with the built-in functions described 4646below. 4647 4648@defmac __builtin_saveregs () 4649Use this built-in function to save the argument registers in memory so 4650that the varargs mechanism can access them. Both ISO and traditional 4651versions of @code{va_start} must use @code{__builtin_saveregs}, unless 4652you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead. 4653 4654On some machines, @code{__builtin_saveregs} is open-coded under the 4655control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On 4656other machines, it calls a routine written in assembler language, 4657found in @file{libgcc2.c}. 4658 4659Code generated for the call to @code{__builtin_saveregs} appears at the 4660beginning of the function, as opposed to where the call to 4661@code{__builtin_saveregs} is written, regardless of what the code is. 4662This is because the registers must be saved before the function starts 4663to use them for its own purposes. 4664@c i rewrote the first sentence above to fix an overfull hbox. --mew 4665@c 10feb93 4666@end defmac 4667 4668@defmac __builtin_args_info (@var{category}) 4669Use this built-in function to find the first anonymous arguments in 4670registers. 4671 4672In general, a machine may have several categories of registers used for 4673arguments, each for a particular category of data types. (For example, 4674on some machines, floating-point registers are used for floating-point 4675arguments while other arguments are passed in the general registers.) 4676To make non-varargs functions use the proper calling convention, you 4677have defined the @code{CUMULATIVE_ARGS} data type to record how many 4678registers in each category have been used so far 4679 4680@code{__builtin_args_info} accesses the same data structure of type 4681@code{CUMULATIVE_ARGS} after the ordinary argument layout is finished 4682with it, with @var{category} specifying which word to access. Thus, the 4683value indicates the first unused register in a given category. 4684 4685Normally, you would use @code{__builtin_args_info} in the implementation 4686of @code{va_start}, accessing each category just once and storing the 4687value in the @code{va_list} object. This is because @code{va_list} will 4688have to update the values, and there is no way to alter the 4689values accessed by @code{__builtin_args_info}. 4690@end defmac 4691 4692@defmac __builtin_next_arg (@var{lastarg}) 4693This is the equivalent of @code{__builtin_args_info}, for stack 4694arguments. It returns the address of the first anonymous stack 4695argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it 4696returns the address of the location above the first anonymous stack 4697argument. Use it in @code{va_start} to initialize the pointer for 4698fetching arguments from the stack. Also use it in @code{va_start} to 4699verify that the second parameter @var{lastarg} is the last named argument 4700of the current function. 4701@end defmac 4702 4703@defmac __builtin_classify_type (@var{object}) 4704Since each machine has its own conventions for which data types are 4705passed in which kind of register, your implementation of @code{va_arg} 4706has to embody these conventions. The easiest way to categorize the 4707specified data type is to use @code{__builtin_classify_type} together 4708with @code{sizeof} and @code{__alignof__}. 4709 4710@code{__builtin_classify_type} ignores the value of @var{object}, 4711considering only its data type. It returns an integer describing what 4712kind of type that is---integer, floating, pointer, structure, and so on. 4713 4714The file @file{typeclass.h} defines an enumeration that you can use to 4715interpret the values of @code{__builtin_classify_type}. 4716@end defmac 4717 4718These machine description macros help implement varargs: 4719 4720@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void) 4721If defined, this hook produces the machine-specific code for a call to 4722@code{__builtin_saveregs}. This code will be moved to the very 4723beginning of the function, before any parameter access are made. The 4724return value of this function should be an RTX that contains the value 4725to use as the return of @code{__builtin_saveregs}. 4726@end deftypefn 4727 4728@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}) 4729This target hook offers an alternative to using 4730@code{__builtin_saveregs} and defining the hook 4731@code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous 4732register arguments into the stack so that all the arguments appear to 4733have been passed consecutively on the stack. Once this is done, you can 4734use the standard implementation of varargs that works for machines that 4735pass all their arguments on the stack. 4736 4737The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data 4738structure, containing the values that are obtained after processing the 4739named arguments. The arguments @var{mode} and @var{type} describe the 4740last named argument---its machine mode and its data type as a tree node. 4741 4742The target hook should do two things: first, push onto the stack all the 4743argument registers @emph{not} used for the named arguments, and second, 4744store the size of the data thus pushed into the @code{int}-valued 4745variable pointed to by @var{pretend_args_size}. The value that you 4746store here will serve as additional offset for setting up the stack 4747frame. 4748 4749Because you must generate code to push the anonymous arguments at 4750compile time without knowing their data types, 4751@code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that 4752have just a single category of argument register and use it uniformly 4753for all data types. 4754 4755If the argument @var{second_time} is nonzero, it means that the 4756arguments of the function are being analyzed for the second time. This 4757happens for an inline function, which is not actually compiled until the 4758end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should 4759not generate any instructions in this case. 4760@end deftypefn 4761 4762@deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca}) 4763Define this hook to return @code{true} if the location where a function 4764argument is passed depends on whether or not it is a named argument. 4765 4766This hook controls how the @var{named} argument to @code{FUNCTION_ARG} 4767is set for varargs and stdarg functions. If this hook returns 4768@code{true}, the @var{named} argument is always true for named 4769arguments, and false for unnamed arguments. If it returns @code{false}, 4770but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true}, 4771then all arguments are treated as named. Otherwise, all named arguments 4772except the last are treated as named. 4773 4774You need not define this hook if it always returns zero. 4775@end deftypefn 4776 4777@deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED 4778If you need to conditionally change ABIs so that one works with 4779@code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither 4780@code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was 4781defined, then define this hook to return @code{true} if 4782@code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise. 4783Otherwise, you should not define this hook. 4784@end deftypefn 4785 4786@node Trampolines 4787@section Trampolines for Nested Functions 4788@cindex trampolines for nested functions 4789@cindex nested functions, trampolines for 4790 4791A @dfn{trampoline} is a small piece of code that is created at run time 4792when the address of a nested function is taken. It normally resides on 4793the stack, in the stack frame of the containing function. These macros 4794tell GCC how to generate code to allocate and initialize a 4795trampoline. 4796 4797The instructions in the trampoline must do two things: load a constant 4798address into the static chain register, and jump to the real address of 4799the nested function. On CISC machines such as the m68k, this requires 4800two instructions, a move immediate and a jump. Then the two addresses 4801exist in the trampoline as word-long immediate operands. On RISC 4802machines, it is often necessary to load each address into a register in 4803two parts. Then pieces of each address form separate immediate 4804operands. 4805 4806The code generated to initialize the trampoline must store the variable 4807parts---the static chain value and the function address---into the 4808immediate operands of the instructions. On a CISC machine, this is 4809simply a matter of copying each address to a memory reference at the 4810proper offset from the start of the trampoline. On a RISC machine, it 4811may be necessary to take out pieces of the address and store them 4812separately. 4813 4814@defmac TRAMPOLINE_TEMPLATE (@var{file}) 4815A C statement to output, on the stream @var{file}, assembler code for a 4816block of data that contains the constant parts of a trampoline. This 4817code should not include a label---the label is taken care of 4818automatically. 4819 4820If you do not define this macro, it means no template is needed 4821for the target. Do not define this macro on systems where the block move 4822code to copy the trampoline into place would be larger than the code 4823to generate it on the spot. 4824@end defmac 4825 4826@defmac TRAMPOLINE_SECTION 4827Return the section into which the trampoline template is to be placed 4828(@pxref{Sections}). The default value is @code{readonly_data_section}. 4829@end defmac 4830 4831@defmac TRAMPOLINE_SIZE 4832A C expression for the size in bytes of the trampoline, as an integer. 4833@end defmac 4834 4835@defmac TRAMPOLINE_ALIGNMENT 4836Alignment required for trampolines, in bits. 4837 4838If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT} 4839is used for aligning trampolines. 4840@end defmac 4841 4842@defmac INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain}) 4843A C statement to initialize the variable parts of a trampoline. 4844@var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is 4845an RTX for the address of the nested function; @var{static_chain} is an 4846RTX for the static chain value that should be passed to the function 4847when it is called. 4848@end defmac 4849 4850@defmac TRAMPOLINE_ADJUST_ADDRESS (@var{addr}) 4851A C statement that should perform any machine-specific adjustment in 4852the address of the trampoline. Its argument contains the address that 4853was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be 4854used for a function call should be different from the address in which 4855the template was stored, the different address should be assigned to 4856@var{addr}. If this macro is not defined, @var{addr} will be used for 4857function calls. 4858 4859@cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines 4860@cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines 4861If this macro is not defined, by default the trampoline is allocated as 4862a stack slot. This default is right for most machines. The exceptions 4863are machines where it is impossible to execute instructions in the stack 4864area. On such machines, you may have to implement a separate stack, 4865using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE} 4866and @code{TARGET_ASM_FUNCTION_EPILOGUE}. 4867 4868@var{fp} points to a data structure, a @code{struct function}, which 4869describes the compilation status of the immediate containing function of 4870the function which the trampoline is for. The stack slot for the 4871trampoline is in the stack frame of this containing function. Other 4872allocation strategies probably must do something analogous with this 4873information. 4874@end defmac 4875 4876Implementing trampolines is difficult on many machines because they have 4877separate instruction and data caches. Writing into a stack location 4878fails to clear the memory in the instruction cache, so when the program 4879jumps to that location, it executes the old contents. 4880 4881Here are two possible solutions. One is to clear the relevant parts of 4882the instruction cache whenever a trampoline is set up. The other is to 4883make all trampolines identical, by having them jump to a standard 4884subroutine. The former technique makes trampoline execution faster; the 4885latter makes initialization faster. 4886 4887To clear the instruction cache when a trampoline is initialized, define 4888the following macro. 4889 4890@defmac CLEAR_INSN_CACHE (@var{beg}, @var{end}) 4891If defined, expands to a C expression clearing the @emph{instruction 4892cache} in the specified interval. The definition of this macro would 4893typically be a series of @code{asm} statements. Both @var{beg} and 4894@var{end} are both pointer expressions. 4895@end defmac 4896 4897The operating system may also require the stack to be made executable 4898before calling the trampoline. To implement this requirement, define 4899the following macro. 4900 4901@defmac ENABLE_EXECUTE_STACK 4902Define this macro if certain operations must be performed before executing 4903code located on the stack. The macro should expand to a series of C 4904file-scope constructs (e.g.@: functions) and provide a unique entry point 4905named @code{__enable_execute_stack}. The target is responsible for 4906emitting calls to the entry point in the code, for example from the 4907@code{INITIALIZE_TRAMPOLINE} macro. 4908@end defmac 4909 4910To use a standard subroutine, define the following macro. In addition, 4911you must make sure that the instructions in a trampoline fill an entire 4912cache line with identical instructions, or else ensure that the 4913beginning of the trampoline code is always aligned at the same point in 4914its cache line. Look in @file{m68k.h} as a guide. 4915 4916@defmac TRANSFER_FROM_TRAMPOLINE 4917Define this macro if trampolines need a special subroutine to do their 4918work. The macro should expand to a series of @code{asm} statements 4919which will be compiled with GCC@. They go in a library function named 4920@code{__transfer_from_trampoline}. 4921 4922If you need to avoid executing the ordinary prologue code of a compiled 4923C function when you jump to the subroutine, you can do so by placing a 4924special label of your own in the assembler code. Use one @code{asm} 4925statement to generate an assembler label, and another to make the label 4926global. Then trampolines can use that label to jump directly to your 4927special assembler code. 4928@end defmac 4929 4930@node Library Calls 4931@section Implicit Calls to Library Routines 4932@cindex library subroutine names 4933@cindex @file{libgcc.a} 4934 4935@c prevent bad page break with this line 4936Here is an explanation of implicit calls to library routines. 4937 4938@defmac DECLARE_LIBRARY_RENAMES 4939This macro, if defined, should expand to a piece of C code that will get 4940expanded when compiling functions for libgcc.a. It can be used to 4941provide alternate names for GCC's internal library functions if there 4942are ABI-mandated names that the compiler should provide. 4943@end defmac 4944 4945@findex init_one_libfunc 4946@findex set_optab_libfunc 4947@deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void) 4948This hook should declare additional library routines or rename 4949existing ones, using the functions @code{set_optab_libfunc} and 4950@code{init_one_libfunc} defined in @file{optabs.c}. 4951@code{init_optabs} calls this macro after initializing all the normal 4952library routines. 4953 4954The default is to do nothing. Most ports don't need to define this hook. 4955@end deftypefn 4956 4957@defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison}) 4958This macro should return @code{true} if the library routine that 4959implements the floating point comparison operator @var{comparison} in 4960mode @var{mode} will return a boolean, and @var{false} if it will 4961return a tristate. 4962 4963GCC's own floating point libraries return tristates from the 4964comparison operators, so the default returns false always. Most ports 4965don't need to define this macro. 4966@end defmac 4967 4968@defmac TARGET_LIB_INT_CMP_BIASED 4969This macro should evaluate to @code{true} if the integer comparison 4970functions (like @code{__cmpdi2}) return 0 to indicate that the first 4971operand is smaller than the second, 1 to indicate that they are equal, 4972and 2 to indicate that the first operand is greater than the second. 4973If this macro evaluates to @code{false} the comparison functions return 4974@minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines 4975in @file{libgcc.a}, you do not need to define this macro. 4976@end defmac 4977 4978@cindex US Software GOFAST, floating point emulation library 4979@cindex floating point emulation library, US Software GOFAST 4980@cindex GOFAST, floating point emulation library 4981@findex gofast_maybe_init_libfuncs 4982@defmac US_SOFTWARE_GOFAST 4983Define this macro if your system C library uses the US Software GOFAST 4984library to provide floating point emulation. 4985 4986In addition to defining this macro, your architecture must set 4987@code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or 4988else call that function from its version of that hook. It is defined 4989in @file{config/gofast.h}, which must be included by your 4990architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for 4991an example. 4992 4993If this macro is defined, the 4994@code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return 4995false for @code{SFmode} and @code{DFmode} comparisons. 4996@end defmac 4997 4998@cindex @code{EDOM}, implicit usage 4999@findex matherr 5000@defmac TARGET_EDOM 5001The value of @code{EDOM} on the target machine, as a C integer constant 5002expression. If you don't define this macro, GCC does not attempt to 5003deposit the value of @code{EDOM} into @code{errno} directly. Look in 5004@file{/usr/include/errno.h} to find the value of @code{EDOM} on your 5005system. 5006 5007If you do not define @code{TARGET_EDOM}, then compiled code reports 5008domain errors by calling the library function and letting it report the 5009error. If mathematical functions on your system use @code{matherr} when 5010there is an error, then you should leave @code{TARGET_EDOM} undefined so 5011that @code{matherr} is used normally. 5012@end defmac 5013 5014@cindex @code{errno}, implicit usage 5015@defmac GEN_ERRNO_RTX 5016Define this macro as a C expression to create an rtl expression that 5017refers to the global ``variable'' @code{errno}. (On certain systems, 5018@code{errno} may not actually be a variable.) If you don't define this 5019macro, a reasonable default is used. 5020@end defmac 5021 5022@cindex C99 math functions, implicit usage 5023@defmac TARGET_C99_FUNCTIONS 5024When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into 5025@code{sinf} and similarly for other functions defined by C99 standard. The 5026default is nonzero that should be proper value for most modern systems, however 5027number of existing systems lacks support for these functions in the runtime so 5028they needs this macro to be redefined to 0. 5029@end defmac 5030 5031@node Addressing Modes 5032@section Addressing Modes 5033@cindex addressing modes 5034 5035@c prevent bad page break with this line 5036This is about addressing modes. 5037 5038@defmac HAVE_PRE_INCREMENT 5039@defmacx HAVE_PRE_DECREMENT 5040@defmacx HAVE_POST_INCREMENT 5041@defmacx HAVE_POST_DECREMENT 5042A C expression that is nonzero if the machine supports pre-increment, 5043pre-decrement, post-increment, or post-decrement addressing respectively. 5044@end defmac 5045 5046@defmac HAVE_PRE_MODIFY_DISP 5047@defmacx HAVE_POST_MODIFY_DISP 5048A C expression that is nonzero if the machine supports pre- or 5049post-address side-effect generation involving constants other than 5050the size of the memory operand. 5051@end defmac 5052 5053@defmac HAVE_PRE_MODIFY_REG 5054@defmacx HAVE_POST_MODIFY_REG 5055A C expression that is nonzero if the machine supports pre- or 5056post-address side-effect generation involving a register displacement. 5057@end defmac 5058 5059@defmac CONSTANT_ADDRESS_P (@var{x}) 5060A C expression that is 1 if the RTX @var{x} is a constant which 5061is a valid address. On most machines, this can be defined as 5062@code{CONSTANT_P (@var{x})}, but a few machines are more restrictive 5063in which constant addresses are supported. 5064@end defmac 5065 5066@defmac CONSTANT_P (@var{x}) 5067@code{CONSTANT_P}, which is defined by target-independent code, 5068accepts integer-values expressions whose values are not explicitly 5069known, such as @code{symbol_ref}, @code{label_ref}, and @code{high} 5070expressions and @code{const} arithmetic expressions, in addition to 5071@code{const_int} and @code{const_double} expressions. 5072@end defmac 5073 5074@defmac MAX_REGS_PER_ADDRESS 5075A number, the maximum number of registers that can appear in a valid 5076memory address. Note that it is up to you to specify a value equal to 5077the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever 5078accept. 5079@end defmac 5080 5081@defmac GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label}) 5082A C compound statement with a conditional @code{goto @var{label};} 5083executed if @var{x} (an RTX) is a legitimate memory address on the 5084target machine for a memory operand of mode @var{mode}. 5085 5086It usually pays to define several simpler macros to serve as 5087subroutines for this one. Otherwise it may be too complicated to 5088understand. 5089 5090This macro must exist in two variants: a strict variant and a 5091non-strict one. The strict variant is used in the reload pass. It 5092must be defined so that any pseudo-register that has not been 5093allocated a hard register is considered a memory reference. In 5094contexts where some kind of register is required, a pseudo-register 5095with no hard register must be rejected. 5096 5097The non-strict variant is used in other passes. It must be defined to 5098accept all pseudo-registers in every context where some kind of 5099register is required. 5100 5101@findex REG_OK_STRICT 5102Compiler source files that want to use the strict variant of this 5103macro define the macro @code{REG_OK_STRICT}. You should use an 5104@code{#ifdef REG_OK_STRICT} conditional to define the strict variant 5105in that case and the non-strict variant otherwise. 5106 5107Subroutines to check for acceptable registers for various purposes (one 5108for base registers, one for index registers, and so on) are typically 5109among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}. 5110Then only these subroutine macros need have two variants; the higher 5111levels of macros may be the same whether strict or not. 5112 5113Normally, constant addresses which are the sum of a @code{symbol_ref} 5114and an integer are stored inside a @code{const} RTX to mark them as 5115constant. Therefore, there is no need to recognize such sums 5116specifically as legitimate addresses. Normally you would simply 5117recognize any @code{const} as legitimate. 5118 5119Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant 5120sums that are not marked with @code{const}. It assumes that a naked 5121@code{plus} indicates indexing. If so, then you @emph{must} reject such 5122naked constant sums as illegitimate addresses, so that none of them will 5123be given to @code{PRINT_OPERAND_ADDRESS}. 5124 5125@cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation 5126On some machines, whether a symbolic address is legitimate depends on 5127the section that the address refers to. On these machines, define the 5128target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information 5129into the @code{symbol_ref}, and then check for it here. When you see a 5130@code{const}, you will have to look inside it to find the 5131@code{symbol_ref} in order to determine the section. @xref{Assembler 5132Format}. 5133@end defmac 5134 5135@defmac FIND_BASE_TERM (@var{x}) 5136A C expression to determine the base term of address @var{x}. 5137This macro is used in only one place: `find_base_term' in alias.c. 5138 5139It is always safe for this macro to not be defined. It exists so 5140that alias analysis can understand machine-dependent addresses. 5141 5142The typical use of this macro is to handle addresses containing 5143a label_ref or symbol_ref within an UNSPEC@. 5144@end defmac 5145 5146@defmac LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win}) 5147A C compound statement that attempts to replace @var{x} with a valid 5148memory address for an operand of mode @var{mode}. @var{win} will be a 5149C statement label elsewhere in the code; the macro definition may use 5150 5151@smallexample 5152GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win}); 5153@end smallexample 5154 5155@noindent 5156to avoid further processing if the address has become legitimate. 5157 5158@findex break_out_memory_refs 5159@var{x} will always be the result of a call to @code{break_out_memory_refs}, 5160and @var{oldx} will be the operand that was given to that function to produce 5161@var{x}. 5162 5163The code generated by this macro should not alter the substructure of 5164@var{x}. If it transforms @var{x} into a more legitimate form, it 5165should assign @var{x} (which will always be a C variable) a new value. 5166 5167It is not necessary for this macro to come up with a legitimate 5168address. The compiler has standard ways of doing so in all cases. In 5169fact, it is safe to omit this macro. But often a 5170machine-dependent strategy can generate better code. 5171@end defmac 5172 5173@defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win}) 5174A C compound statement that attempts to replace @var{x}, which is an address 5175that needs reloading, with a valid memory address for an operand of mode 5176@var{mode}. @var{win} will be a C statement label elsewhere in the code. 5177It is not necessary to define this macro, but it might be useful for 5178performance reasons. 5179 5180For example, on the i386, it is sometimes possible to use a single 5181reload register instead of two by reloading a sum of two pseudo 5182registers into a register. On the other hand, for number of RISC 5183processors offsets are limited so that often an intermediate address 5184needs to be generated in order to address a stack slot. By defining 5185@code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses 5186generated for adjacent some stack slots can be made identical, and thus 5187be shared. 5188 5189@emph{Note}: This macro should be used with caution. It is necessary 5190to know something of how reload works in order to effectively use this, 5191and it is quite easy to produce macros that build in too much knowledge 5192of reload internals. 5193 5194@emph{Note}: This macro must be able to reload an address created by a 5195previous invocation of this macro. If it fails to handle such addresses 5196then the compiler may generate incorrect code or abort. 5197 5198@findex push_reload 5199The macro definition should use @code{push_reload} to indicate parts that 5200need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually 5201suitable to be passed unaltered to @code{push_reload}. 5202 5203The code generated by this macro must not alter the substructure of 5204@var{x}. If it transforms @var{x} into a more legitimate form, it 5205should assign @var{x} (which will always be a C variable) a new value. 5206This also applies to parts that you change indirectly by calling 5207@code{push_reload}. 5208 5209@findex strict_memory_address_p 5210The macro definition may use @code{strict_memory_address_p} to test if 5211the address has become legitimate. 5212 5213@findex copy_rtx 5214If you want to change only a part of @var{x}, one standard way of doing 5215this is to use @code{copy_rtx}. Note, however, that is unshares only a 5216single level of rtl. Thus, if the part to be changed is not at the 5217top level, you'll need to replace first the top level. 5218It is not necessary for this macro to come up with a legitimate 5219address; but often a machine-dependent strategy can generate better code. 5220@end defmac 5221 5222@defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label}) 5223A C statement or compound statement with a conditional @code{goto 5224@var{label};} executed if memory address @var{x} (an RTX) can have 5225different meanings depending on the machine mode of the memory 5226reference it is used for or if the address is valid for some modes 5227but not others. 5228 5229Autoincrement and autodecrement addresses typically have mode-dependent 5230effects because the amount of the increment or decrement is the size 5231of the operand being addressed. Some machines have other mode-dependent 5232addresses. Many RISC machines have no mode-dependent addresses. 5233 5234You may assume that @var{addr} is a valid address for the machine. 5235@end defmac 5236 5237@defmac LEGITIMATE_CONSTANT_P (@var{x}) 5238A C expression that is nonzero if @var{x} is a legitimate constant for 5239an immediate operand on the target machine. You can assume that 5240@var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact, 5241@samp{1} is a suitable definition for this macro on machines where 5242anything @code{CONSTANT_P} is valid. 5243@end defmac 5244 5245@deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x}) 5246This hook is used to undo the possibly obfuscating effects of the 5247@code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target 5248macros. Some backend implementations of these macros wrap symbol 5249references inside an @code{UNSPEC} rtx to represent PIC or similar 5250addressing modes. This target hook allows GCC's optimizers to understand 5251the semantics of these opaque @code{UNSPEC}s by converting them back 5252into their original form. 5253@end deftypefn 5254 5255@deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x}) 5256This hook should return true if @var{x} is of a form that cannot (or 5257should not) be spilled to the constant pool. The default version of 5258this hook returns false. 5259 5260The primary reason to define this hook is to prevent reload from 5261deciding that a non-legitimate constant would be better reloaded 5262from the constant pool instead of spilling and reloading a register 5263holding the constant. This restriction is often true of addresses 5264of TLS symbols for various targets. 5265@end deftypefn 5266 5267@deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x}) 5268This hook should return true if pool entries for constant @var{x} can 5269be placed in an @code{object_block} structure. @var{mode} is the mode 5270of @var{x}. 5271 5272The default version returns false for all constants. 5273@end deftypefn 5274 5275@deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void) 5276This hook should return the DECL of a function @var{f} that given an 5277address @var{addr} as an argument returns a mask @var{m} that can be 5278used to extract from two vectors the relevant data that resides in 5279@var{addr} in case @var{addr} is not properly aligned. 5280 5281The autovectrizer, when vectorizing a load operation from an address 5282@var{addr} that may be unaligned, will generate two vector loads from 5283the two aligned addresses around @var{addr}. It then generates a 5284@code{REALIGN_LOAD} operation to extract the relevant data from the 5285two loaded vectors. The first two arguments to @code{REALIGN_LOAD}, 5286@var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and 5287the third argument, @var{OFF}, defines how the data will be extracted 5288from these two vectors: if @var{OFF} is 0, then the returned vector is 5289@var{v2}; otherwise, the returned vector is composed from the last 5290@var{VS}-@var{OFF} elements of @var{v1} concatenated to the first 5291@var{OFF} elements of @var{v2}. 5292 5293If this hook is defined, the autovectorizer will generate a call 5294to @var{f} (using the DECL tree that this hook returns) and will 5295use the return value of @var{f} as the argument @var{OFF} to 5296@code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f} 5297should comply with the semantics expected by @code{REALIGN_LOAD} 5298described above. 5299If this hook is not defined, then @var{addr} will be used as 5300the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low 5301log2(@var{VS})-1 bits of @var{addr} will be considered. 5302@end deftypefn 5303 5304@node Anchored Addresses 5305@section Anchored Addresses 5306@cindex anchored addresses 5307@cindex @option{-fsection-anchors} 5308 5309GCC usually addresses every static object as a separate entity. 5310For example, if we have: 5311 5312@smallexample 5313static int a, b, c; 5314int foo (void) @{ return a + b + c; @} 5315@end smallexample 5316 5317the code for @code{foo} will usually calculate three separate symbolic 5318addresses: those of @code{a}, @code{b} and @code{c}. On some targets, 5319it would be better to calculate just one symbolic address and access 5320the three variables relative to it. The equivalent pseudocode would 5321be something like: 5322 5323@smallexample 5324int foo (void) 5325@{ 5326 register int *xr = &x; 5327 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; 5328@} 5329@end smallexample 5330 5331(which isn't valid C). We refer to shared addresses like @code{x} as 5332``section anchors''. Their use is controlled by @option{-fsection-anchors}. 5333 5334The hooks below describe the target properties that GCC needs to know 5335in order to make effective use of section anchors. It won't use 5336section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET} 5337or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value. 5338 5339@deftypevar {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET 5340The minimum offset that should be applied to a section anchor. 5341On most targets, it should be the smallest offset that can be 5342applied to a base register while still giving a legitimate address 5343for every mode. The default value is 0. 5344@end deftypevar 5345 5346@deftypevar {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET 5347Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive) 5348offset that should be applied to section anchors. The default 5349value is 0. 5350@end deftypevar 5351 5352@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x}) 5353Write the assembly code to define section anchor @var{x}, which is a 5354@code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true. 5355The hook is called with the assembly output position set to the beginning 5356of @code{SYMBOL_REF_BLOCK (@var{x})}. 5357 5358If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses 5359it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}. 5360If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition 5361is @code{NULL}, which disables the use of section anchors altogether. 5362@end deftypefn 5363 5364@deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x}) 5365Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF} 5366@var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and 5367@samp{!SYMBOL_REF_ANCHOR_P (@var{x})}. 5368 5369The default version is correct for most targets, but you might need to 5370intercept this hook to handle things like target-specific attributes 5371or target-specific sections. 5372@end deftypefn 5373 5374@node Condition Code 5375@section Condition Code Status 5376@cindex condition code status 5377 5378@c prevent bad page break with this line 5379This describes the condition code status. 5380 5381@findex cc_status 5382The file @file{conditions.h} defines a variable @code{cc_status} to 5383describe how the condition code was computed (in case the interpretation of 5384the condition code depends on the instruction that it was set by). This 5385variable contains the RTL expressions on which the condition code is 5386currently based, and several standard flags. 5387 5388Sometimes additional machine-specific flags must be defined in the machine 5389description header file. It can also add additional machine-specific 5390information by defining @code{CC_STATUS_MDEP}. 5391 5392@defmac CC_STATUS_MDEP 5393C code for a data type which is used for declaring the @code{mdep} 5394component of @code{cc_status}. It defaults to @code{int}. 5395 5396This macro is not used on machines that do not use @code{cc0}. 5397@end defmac 5398 5399@defmac CC_STATUS_MDEP_INIT 5400A C expression to initialize the @code{mdep} field to ``empty''. 5401The default definition does nothing, since most machines don't use 5402the field anyway. If you want to use the field, you should probably 5403define this macro to initialize it. 5404 5405This macro is not used on machines that do not use @code{cc0}. 5406@end defmac 5407 5408@defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn}) 5409A C compound statement to set the components of @code{cc_status} 5410appropriately for an insn @var{insn} whose body is @var{exp}. It is 5411this macro's responsibility to recognize insns that set the condition 5412code as a byproduct of other activity as well as those that explicitly 5413set @code{(cc0)}. 5414 5415This macro is not used on machines that do not use @code{cc0}. 5416 5417If there are insns that do not set the condition code but do alter 5418other machine registers, this macro must check to see whether they 5419invalidate the expressions that the condition code is recorded as 5420reflecting. For example, on the 68000, insns that store in address 5421registers do not set the condition code, which means that usually 5422@code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such 5423insns. But suppose that the previous insn set the condition code 5424based on location @samp{a4@@(102)} and the current insn stores a new 5425value in @samp{a4}. Although the condition code is not changed by 5426this, it will no longer be true that it reflects the contents of 5427@samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter 5428@code{cc_status} in this case to say that nothing is known about the 5429condition code value. 5430 5431The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal 5432with the results of peephole optimization: insns whose patterns are 5433@code{parallel} RTXs containing various @code{reg}, @code{mem} or 5434constants which are just the operands. The RTL structure of these 5435insns is not sufficient to indicate what the insns actually do. What 5436@code{NOTICE_UPDATE_CC} should do when it sees one is just to run 5437@code{CC_STATUS_INIT}. 5438 5439A possible definition of @code{NOTICE_UPDATE_CC} is to call a function 5440that looks at an attribute (@pxref{Insn Attributes}) named, for example, 5441@samp{cc}. This avoids having detailed information about patterns in 5442two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}. 5443@end defmac 5444 5445@defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) 5446Returns a mode from class @code{MODE_CC} to be used when comparison 5447operation code @var{op} is applied to rtx @var{x} and @var{y}. For 5448example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see 5449@pxref{Jump Patterns} for a description of the reason for this 5450definition) 5451 5452@smallexample 5453#define SELECT_CC_MODE(OP,X,Y) \ 5454 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ 5455 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ 5456 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ 5457 || GET_CODE (X) == NEG) \ 5458 ? CC_NOOVmode : CCmode)) 5459@end smallexample 5460 5461You should define this macro if and only if you define extra CC modes 5462in @file{@var{machine}-modes.def}. 5463@end defmac 5464 5465@defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1}) 5466On some machines not all possible comparisons are defined, but you can 5467convert an invalid comparison into a valid one. For example, the Alpha 5468does not have a @code{GT} comparison, but you can use an @code{LT} 5469comparison instead and swap the order of the operands. 5470 5471On such machines, define this macro to be a C statement to do any 5472required conversions. @var{code} is the initial comparison code 5473and @var{op0} and @var{op1} are the left and right operands of the 5474comparison, respectively. You should modify @var{code}, @var{op0}, and 5475@var{op1} as required. 5476 5477GCC will not assume that the comparison resulting from this macro is 5478valid but will see if the resulting insn matches a pattern in the 5479@file{md} file. 5480 5481You need not define this macro if it would never change the comparison 5482code or operands. 5483@end defmac 5484 5485@defmac REVERSIBLE_CC_MODE (@var{mode}) 5486A C expression whose value is one if it is always safe to reverse a 5487comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} 5488can ever return @var{mode} for a floating-point inequality comparison, 5489then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. 5490 5491You need not define this macro if it would always returns zero or if the 5492floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. 5493For example, here is the definition used on the SPARC, where floating-point 5494inequality comparisons are always given @code{CCFPEmode}: 5495 5496@smallexample 5497#define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) 5498@end smallexample 5499@end defmac 5500 5501@defmac REVERSE_CONDITION (@var{code}, @var{mode}) 5502A C expression whose value is reversed condition code of the @var{code} for 5503comparison done in CC_MODE @var{mode}. The macro is used only in case 5504@code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case 5505machine has some non-standard way how to reverse certain conditionals. For 5506instance in case all floating point conditions are non-trapping, compiler may 5507freely convert unordered compares to ordered one. Then definition may look 5508like: 5509 5510@smallexample 5511#define REVERSE_CONDITION(CODE, MODE) \ 5512 ((MODE) != CCFPmode ? reverse_condition (CODE) \ 5513 : reverse_condition_maybe_unordered (CODE)) 5514@end smallexample 5515@end defmac 5516 5517@defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2}) 5518A C expression that returns true if the conditional execution predicate 5519@var{op1}, a comparison operation, is the inverse of @var{op2} and vice 5520versa. Define this to return 0 if the target has conditional execution 5521predicates that cannot be reversed safely. There is no need to validate 5522that the arguments of op1 and op2 are the same, this is done separately. 5523If no expansion is specified, this macro is defined as follows: 5524 5525@smallexample 5526#define REVERSE_CONDEXEC_PREDICATES_P (x, y) \ 5527 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL)) 5528@end smallexample 5529@end defmac 5530 5531@deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *) 5532On targets which do not use @code{(cc0)}, and which use a hard 5533register rather than a pseudo-register to hold condition codes, the 5534regular CSE passes are often not able to identify cases in which the 5535hard register is set to a common value. Use this hook to enable a 5536small pass which optimizes such cases. This hook should return true 5537to enable this pass, and it should set the integers to which its 5538arguments point to the hard register numbers used for condition codes. 5539When there is only one such register, as is true on most systems, the 5540integer pointed to by the second argument should be set to 5541@code{INVALID_REGNUM}. 5542 5543The default version of this hook returns false. 5544@end deftypefn 5545 5546@deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode) 5547On targets which use multiple condition code modes in class 5548@code{MODE_CC}, it is sometimes the case that a comparison can be 5549validly done in more than one mode. On such a system, define this 5550target hook to take two mode arguments and to return a mode in which 5551both comparisons may be validly done. If there is no such mode, 5552return @code{VOIDmode}. 5553 5554The default version of this hook checks whether the modes are the 5555same. If they are, it returns that mode. If they are different, it 5556returns @code{VOIDmode}. 5557@end deftypefn 5558 5559@node Costs 5560@section Describing Relative Costs of Operations 5561@cindex costs of instructions 5562@cindex relative costs 5563@cindex speed of instructions 5564 5565These macros let you describe the relative speed of various operations 5566on the target machine. 5567 5568@defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to}) 5569A C expression for the cost of moving data of mode @var{mode} from a 5570register in class @var{from} to one in class @var{to}. The classes are 5571expressed using the enumeration values such as @code{GENERAL_REGS}. A 5572value of 2 is the default; other values are interpreted relative to 5573that. 5574 5575It is not required that the cost always equal 2 when @var{from} is the 5576same as @var{to}; on some machines it is expensive to move between 5577registers if they are not general registers. 5578 5579If reload sees an insn consisting of a single @code{set} between two 5580hard registers, and if @code{REGISTER_MOVE_COST} applied to their 5581classes returns a value of 2, reload does not check to ensure that the 5582constraints of the insn are met. Setting a cost of other than 2 will 5583allow reload to verify that the constraints are met. You should do this 5584if the @samp{mov@var{m}} pattern's constraints do not allow such copying. 5585@end defmac 5586 5587@defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in}) 5588A C expression for the cost of moving data of mode @var{mode} between a 5589register of class @var{class} and memory; @var{in} is zero if the value 5590is to be written to memory, nonzero if it is to be read in. This cost 5591is relative to those in @code{REGISTER_MOVE_COST}. If moving between 5592registers and memory is more expensive than between two registers, you 5593should define this macro to express the relative cost. 5594 5595If you do not define this macro, GCC uses a default cost of 4 plus 5596the cost of copying via a secondary reload register, if one is 5597needed. If your machine requires a secondary reload register to copy 5598between memory and a register of @var{class} but the reload mechanism is 5599more complex than copying via an intermediate, define this macro to 5600reflect the actual cost of the move. 5601 5602GCC defines the function @code{memory_move_secondary_cost} if 5603secondary reloads are needed. It computes the costs due to copying via 5604a secondary register. If your machine copies from memory using a 5605secondary register in the conventional way but the default base value of 56064 is not correct for your machine, define this macro to add some other 5607value to the result of that function. The arguments to that function 5608are the same as to this macro. 5609@end defmac 5610 5611@defmac BRANCH_COST 5612A C expression for the cost of a branch instruction. A value of 1 is 5613the default; other values are interpreted relative to that. 5614@end defmac 5615 5616Here are additional macros which do not specify precise relative costs, 5617but only that certain actions are more expensive than GCC would 5618ordinarily expect. 5619 5620@defmac SLOW_BYTE_ACCESS 5621Define this macro as a C expression which is nonzero if accessing less 5622than a word of memory (i.e.@: a @code{char} or a @code{short}) is no 5623faster than accessing a word of memory, i.e., if such access 5624require more than one instruction or if there is no difference in cost 5625between byte and (aligned) word loads. 5626 5627When this macro is not defined, the compiler will access a field by 5628finding the smallest containing object; when it is defined, a fullword 5629load will be used if alignment permits. Unless bytes accesses are 5630faster than word accesses, using word accesses is preferable since it 5631may eliminate subsequent memory access if subsequent accesses occur to 5632other fields in the same word of the structure, but to different bytes. 5633@end defmac 5634 5635@defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment}) 5636Define this macro to be the value 1 if memory accesses described by the 5637@var{mode} and @var{alignment} parameters have a cost many times greater 5638than aligned accesses, for example if they are emulated in a trap 5639handler. 5640 5641When this macro is nonzero, the compiler will act as if 5642@code{STRICT_ALIGNMENT} were nonzero when generating code for block 5643moves. This can cause significantly more instructions to be produced. 5644Therefore, do not set this macro nonzero if unaligned accesses only add a 5645cycle or two to the time for a memory access. 5646 5647If the value of this macro is always zero, it need not be defined. If 5648this macro is defined, it should produce a nonzero value when 5649@code{STRICT_ALIGNMENT} is nonzero. 5650@end defmac 5651 5652@defmac MOVE_RATIO 5653The threshold of number of scalar memory-to-memory move insns, @emph{below} 5654which a sequence of insns should be generated instead of a 5655string move insn or a library call. Increasing the value will always 5656make code faster, but eventually incurs high cost in increased code size. 5657 5658Note that on machines where the corresponding move insn is a 5659@code{define_expand} that emits a sequence of insns, this macro counts 5660the number of such sequences. 5661 5662If you don't define this, a reasonable default is used. 5663@end defmac 5664 5665@defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment}) 5666A C expression used to determine whether @code{move_by_pieces} will be used to 5667copy a chunk of memory, or whether some other block move mechanism 5668will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less 5669than @code{MOVE_RATIO}. 5670@end defmac 5671 5672@defmac MOVE_MAX_PIECES 5673A C expression used by @code{move_by_pieces} to determine the largest unit 5674a load or store used to copy memory is. Defaults to @code{MOVE_MAX}. 5675@end defmac 5676 5677@defmac CLEAR_RATIO 5678The threshold of number of scalar move insns, @emph{below} which a sequence 5679of insns should be generated to clear memory instead of a string clear insn 5680or a library call. Increasing the value will always make code faster, but 5681eventually incurs high cost in increased code size. 5682 5683If you don't define this, a reasonable default is used. 5684@end defmac 5685 5686@defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment}) 5687A C expression used to determine whether @code{clear_by_pieces} will be used 5688to clear a chunk of memory, or whether some other block clear mechanism 5689will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less 5690than @code{CLEAR_RATIO}. 5691@end defmac 5692 5693@defmac STORE_BY_PIECES_P (@var{size}, @var{alignment}) 5694A C expression used to determine whether @code{store_by_pieces} will be 5695used to set a chunk of memory to a constant value, or whether some other 5696mechanism will be used. Used by @code{__builtin_memset} when storing 5697values other than constant zero and by @code{__builtin_strcpy} when 5698when called with a constant source string. 5699Defaults to 1 if @code{move_by_pieces_ninsns} returns less 5700than @code{MOVE_RATIO}. 5701@end defmac 5702 5703@defmac USE_LOAD_POST_INCREMENT (@var{mode}) 5704A C expression used to determine whether a load postincrement is a good 5705thing to use for a given mode. Defaults to the value of 5706@code{HAVE_POST_INCREMENT}. 5707@end defmac 5708 5709@defmac USE_LOAD_POST_DECREMENT (@var{mode}) 5710A C expression used to determine whether a load postdecrement is a good 5711thing to use for a given mode. Defaults to the value of 5712@code{HAVE_POST_DECREMENT}. 5713@end defmac 5714 5715@defmac USE_LOAD_PRE_INCREMENT (@var{mode}) 5716A C expression used to determine whether a load preincrement is a good 5717thing to use for a given mode. Defaults to the value of 5718@code{HAVE_PRE_INCREMENT}. 5719@end defmac 5720 5721@defmac USE_LOAD_PRE_DECREMENT (@var{mode}) 5722A C expression used to determine whether a load predecrement is a good 5723thing to use for a given mode. Defaults to the value of 5724@code{HAVE_PRE_DECREMENT}. 5725@end defmac 5726 5727@defmac USE_STORE_POST_INCREMENT (@var{mode}) 5728A C expression used to determine whether a store postincrement is a good 5729thing to use for a given mode. Defaults to the value of 5730@code{HAVE_POST_INCREMENT}. 5731@end defmac 5732 5733@defmac USE_STORE_POST_DECREMENT (@var{mode}) 5734A C expression used to determine whether a store postdecrement is a good 5735thing to use for a given mode. Defaults to the value of 5736@code{HAVE_POST_DECREMENT}. 5737@end defmac 5738 5739@defmac USE_STORE_PRE_INCREMENT (@var{mode}) 5740This macro is used to determine whether a store preincrement is a good 5741thing to use for a given mode. Defaults to the value of 5742@code{HAVE_PRE_INCREMENT}. 5743@end defmac 5744 5745@defmac USE_STORE_PRE_DECREMENT (@var{mode}) 5746This macro is used to determine whether a store predecrement is a good 5747thing to use for a given mode. Defaults to the value of 5748@code{HAVE_PRE_DECREMENT}. 5749@end defmac 5750 5751@defmac NO_FUNCTION_CSE 5752Define this macro if it is as good or better to call a constant 5753function address than to call an address kept in a register. 5754@end defmac 5755 5756@defmac RANGE_TEST_NON_SHORT_CIRCUIT 5757Define this macro if a non-short-circuit operation produced by 5758@samp{fold_range_test ()} is optimal. This macro defaults to true if 5759@code{BRANCH_COST} is greater than or equal to the value 2. 5760@end defmac 5761 5762@deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total}) 5763This target hook describes the relative costs of RTL expressions. 5764 5765The cost may depend on the precise form of the expression, which is 5766available for examination in @var{x}, and the rtx code of the expression 5767in which it is contained, found in @var{outer_code}. @var{code} is the 5768expression code---redundant, since it can be obtained with 5769@code{GET_CODE (@var{x})}. 5770 5771In implementing this hook, you can use the construct 5772@code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast 5773instructions. 5774 5775On entry to the hook, @code{*@var{total}} contains a default estimate 5776for the cost of the expression. The hook should modify this value as 5777necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)} 5778for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus 5779operations, and @code{COSTS_N_INSNS (1)} for all other operations. 5780 5781When optimizing for code size, i.e.@: when @code{optimize_size} is 5782nonzero, this target hook should be used to estimate the relative 5783size cost of an expression, again relative to @code{COSTS_N_INSNS}. 5784 5785The hook returns true when all subexpressions of @var{x} have been 5786processed, and false when @code{rtx_cost} should recurse. 5787@end deftypefn 5788 5789@deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}) 5790This hook computes the cost of an addressing mode that contains 5791@var{address}. If not defined, the cost is computed from 5792the @var{address} expression and the @code{TARGET_RTX_COST} hook. 5793 5794For most CISC machines, the default cost is a good approximation of the 5795true cost of the addressing mode. However, on RISC machines, all 5796instructions normally have the same length and execution time. Hence 5797all addresses will have equal costs. 5798 5799In cases where more than one form of an address is known, the form with 5800the lowest cost will be used. If multiple forms have the same, lowest, 5801cost, the one that is the most complex will be used. 5802 5803For example, suppose an address that is equal to the sum of a register 5804and a constant is used twice in the same basic block. When this macro 5805is not defined, the address will be computed in a register and memory 5806references will be indirect through that register. On machines where 5807the cost of the addressing mode containing the sum is no higher than 5808that of a simple indirect reference, this will produce an additional 5809instruction and possibly require an additional register. Proper 5810specification of this macro eliminates this overhead for such machines. 5811 5812This hook is never called with an invalid address. 5813 5814On machines where an address involving more than one register is as 5815cheap as an address computation involving only one register, defining 5816@code{TARGET_ADDRESS_COST} to reflect this can cause two registers to 5817be live over a region of code where only one would have been if 5818@code{TARGET_ADDRESS_COST} were not defined in that manner. This effect 5819should be considered in the definition of this macro. Equivalent costs 5820should probably only be given to addresses with different numbers of 5821registers on machines with lots of registers. 5822@end deftypefn 5823 5824@node Scheduling 5825@section Adjusting the Instruction Scheduler 5826 5827The instruction scheduler may need a fair amount of machine-specific 5828adjustment in order to produce good code. GCC provides several target 5829hooks for this purpose. It is usually enough to define just a few of 5830them: try the first ones in this list first. 5831 5832@deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void) 5833This hook returns the maximum number of instructions that can ever 5834issue at the same time on the target machine. The default is one. 5835Although the insn scheduler can define itself the possibility of issue 5836an insn on the same cycle, the value can serve as an additional 5837constraint to issue insns on the same simulated processor cycle (see 5838hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}). 5839This value must be constant over the entire compilation. If you need 5840it to vary depending on what the instructions are, you must use 5841@samp{TARGET_SCHED_VARIABLE_ISSUE}. 5842@end deftypefn 5843 5844@deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more}) 5845This hook is executed by the scheduler after it has scheduled an insn 5846from the ready list. It should return the number of insns which can 5847still be issued in the current cycle. The default is 5848@samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and 5849@code{USE}, which normally are not counted against the issue rate. 5850You should define this hook if some insns take more machine resources 5851than others, so that fewer insns can follow them in the same cycle. 5852@var{file} is either a null pointer, or a stdio stream to write any 5853debug output to. @var{verbose} is the verbose level provided by 5854@option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that 5855was scheduled. 5856@end deftypefn 5857 5858@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost}) 5859This function corrects the value of @var{cost} based on the 5860relationship between @var{insn} and @var{dep_insn} through the 5861dependence @var{link}. It should return the new value. The default 5862is to make no adjustment to @var{cost}. This can be used for example 5863to specify to the scheduler using the traditional pipeline description 5864that an output- or anti-dependence does not incur the same cost as a 5865data-dependence. If the scheduler using the automaton based pipeline 5866description, the cost of anti-dependence is zero and the cost of 5867output-dependence is maximum of one and the difference of latency 5868times of the first and the second insns. If these values are not 5869acceptable, you could use the hook to modify them too. See also 5870@pxref{Processor pipeline description}. 5871@end deftypefn 5872 5873@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority}) 5874This hook adjusts the integer scheduling priority @var{priority} of 5875@var{insn}. It should return the new priority. Increase the priority to 5876execute @var{insn} earlier, reduce the priority to execute @var{insn} 5877later. Do not define this hook if you do not need to adjust the 5878scheduling priorities of insns. 5879@end deftypefn 5880 5881@deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock}) 5882This hook is executed by the scheduler after it has scheduled the ready 5883list, to allow the machine description to reorder it (for example to 5884combine two small instructions together on @samp{VLIW} machines). 5885@var{file} is either a null pointer, or a stdio stream to write any 5886debug output to. @var{verbose} is the verbose level provided by 5887@option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready 5888list of instructions that are ready to be scheduled. @var{n_readyp} is 5889a pointer to the number of elements in the ready list. The scheduler 5890reads the ready list in reverse order, starting with 5891@var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock} 5892is the timer tick of the scheduler. You may modify the ready list and 5893the number of ready insns. The return value is the number of insns that 5894can issue this cycle; normally this is just @code{issue_rate}. See also 5895@samp{TARGET_SCHED_REORDER2}. 5896@end deftypefn 5897 5898@deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock}) 5899Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That 5900function is called whenever the scheduler starts a new cycle. This one 5901is called once per iteration over a cycle, immediately after 5902@samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and 5903return the number of insns to be scheduled in the same cycle. Defining 5904this hook can be useful if there are frequent situations where 5905scheduling one insn causes other insns to become ready in the same 5906cycle. These other insns can then be taken into account properly. 5907@end deftypefn 5908 5909@deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail}) 5910This hook is called after evaluation forward dependencies of insns in 5911chain given by two parameter values (@var{head} and @var{tail} 5912correspondingly) but before insns scheduling of the insn chain. For 5913example, it can be used for better insn classification if it requires 5914analysis of dependencies. This hook can use backward and forward 5915dependencies of the insn scheduler because they are already 5916calculated. 5917@end deftypefn 5918 5919@deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready}) 5920This hook is executed by the scheduler at the beginning of each block of 5921instructions that are to be scheduled. @var{file} is either a null 5922pointer, or a stdio stream to write any debug output to. @var{verbose} 5923is the verbose level provided by @option{-fsched-verbose-@var{n}}. 5924@var{max_ready} is the maximum number of insns in the current scheduling 5925region that can be live at the same time. This can be used to allocate 5926scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}. 5927@end deftypefn 5928 5929@deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose}) 5930This hook is executed by the scheduler at the end of each block of 5931instructions that are to be scheduled. It can be used to perform 5932cleanup of any actions done by the other scheduling hooks. @var{file} 5933is either a null pointer, or a stdio stream to write any debug output 5934to. @var{verbose} is the verbose level provided by 5935@option{-fsched-verbose-@var{n}}. 5936@end deftypefn 5937 5938@deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid}) 5939This hook is executed by the scheduler after function level initializations. 5940@var{file} is either a null pointer, or a stdio stream to write any debug output to. 5941@var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}. 5942@var{old_max_uid} is the maximum insn uid when scheduling begins. 5943@end deftypefn 5944 5945@deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose}) 5946This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}. 5947@var{file} is either a null pointer, or a stdio stream to write any debug output to. 5948@var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}. 5949@end deftypefn 5950 5951@deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) 5952The hook returns an RTL insn. The automaton state used in the 5953pipeline hazard recognizer is changed as if the insn were scheduled 5954when the new simulated processor cycle starts. Usage of the hook may 5955simplify the automaton pipeline description for some @acronym{VLIW} 5956processors. If the hook is defined, it is used only for the automaton 5957based pipeline description. The default is not to change the state 5958when the new simulated processor cycle starts. 5959@end deftypefn 5960 5961@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) 5962The hook can be used to initialize data used by the previous hook. 5963@end deftypefn 5964 5965@deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void) 5966The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used 5967to changed the state as if the insn were scheduled when the new 5968simulated processor cycle finishes. 5969@end deftypefn 5970 5971@deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) 5972The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but 5973used to initialize data used by the previous hook. 5974@end deftypefn 5975 5976@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void) 5977This hook controls better choosing an insn from the ready insn queue 5978for the @acronym{DFA}-based insn scheduler. Usually the scheduler 5979chooses the first insn from the queue. If the hook returns a positive 5980value, an additional scheduler code tries all permutations of 5981@samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()} 5982subsequent ready insns to choose an insn whose issue will result in 5983maximal number of issued insns on the same cycle. For the 5984@acronym{VLIW} processor, the code could actually solve the problem of 5985packing simple insns into the @acronym{VLIW} insn. Of course, if the 5986rules of @acronym{VLIW} packing are described in the automaton. 5987 5988This code also could be used for superscalar @acronym{RISC} 5989processors. Let us consider a superscalar @acronym{RISC} processor 5990with 3 pipelines. Some insns can be executed in pipelines @var{A} or 5991@var{B}, some insns can be executed only in pipelines @var{B} or 5992@var{C}, and one insn can be executed in pipeline @var{B}. The 5993processor may issue the 1st insn into @var{A} and the 2nd one into 5994@var{B}. In this case, the 3rd insn will wait for freeing @var{B} 5995until the next cycle. If the scheduler issues the 3rd insn the first, 5996the processor could issue all 3 insns per cycle. 5997 5998Actually this code demonstrates advantages of the automaton based 5999pipeline hazard recognizer. We try quickly and easy many insn 6000schedules to choose the best one. 6001 6002The default is no multipass scheduling. 6003@end deftypefn 6004 6005@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx) 6006 6007This hook controls what insns from the ready insn queue will be 6008considered for the multipass insn scheduling. If the hook returns 6009zero for insn passed as the parameter, the insn will be not chosen to 6010be issued. 6011 6012The default is that any ready insns can be chosen to be issued. 6013@end deftypefn 6014 6015@deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *) 6016 6017This hook is called by the insn scheduler before issuing insn passed 6018as the third parameter on given cycle. If the hook returns nonzero, 6019the insn is not issued on given processors cycle. Instead of that, 6020the processor cycle is advanced. If the value passed through the last 6021parameter is zero, the insn ready queue is not sorted on the new cycle 6022start as usually. The first parameter passes file for debugging 6023output. The second one passes the scheduler verbose level of the 6024debugging output. The forth and the fifth parameter values are 6025correspondingly processor cycle on which the previous insn has been 6026issued and the current processor cycle. 6027@end deftypefn 6028 6029@deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (rtx @var{insn1}, rtx @var{insn2}, rtx @var{dep_link}, int @var{dep_cost}, int @var{distance}) 6030This hook is used to define which dependences are considered costly by 6031the target, so costly that it is not advisable to schedule the insns that 6032are involved in the dependence too close to one another. The parameters 6033to this hook are as follows: The second parameter @var{insn2} is dependent 6034upon the first parameter @var{insn1}. The dependence between @var{insn1} 6035and @var{insn2} is represented by the third parameter @var{dep_link}. The 6036fourth parameter @var{cost} is the cost of the dependence, and the fifth 6037parameter @var{distance} is the distance in cycles between the two insns. 6038The hook returns @code{true} if considering the distance between the two 6039insns the dependence between them is considered costly by the target, 6040and @code{false} otherwise. 6041 6042Defining this hook can be useful in multiple-issue out-of-order machines, 6043where (a) it's practically hopeless to predict the actual data/resource 6044delays, however: (b) there's a better chance to predict the actual grouping 6045that will be formed, and (c) correctly emulating the grouping can be very 6046important. In such targets one may want to allow issuing dependent insns 6047closer to one another---i.e., closer than the dependence distance; however, 6048not in cases of "costly dependences", which this hooks allows to define. 6049@end deftypefn 6050 6051@deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST_2 (rtx @var{insn}, int @var{dep_type}, rtx @var{dep_insn}, int @var{cost}) 6052This hook is a modified version of @samp{TARGET_SCHED_ADJUST_COST}. Instead 6053of passing dependence as a second parameter, it passes a type of that 6054dependence. This is useful to calculate cost of dependence between insns 6055not having the corresponding link. If @samp{TARGET_SCHED_ADJUST_COST_2} is 6056defined it is used instead of @samp{TARGET_SCHED_ADJUST_COST}. 6057@end deftypefn 6058 6059@deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void) 6060This hook is called by the insn scheduler after emitting a new instruction to 6061the instruction stream. The hook notifies a target backend to extend its 6062per instruction data structures. 6063@end deftypefn 6064 6065@deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat}) 6066This hook is called by the insn scheduler when @var{insn} has only 6067speculative dependencies and therefore can be scheduled speculatively. 6068The hook is used to check if the pattern of @var{insn} has a speculative 6069version and, in case of successful check, to generate that speculative 6070pattern. The hook should return 1, if the instruction has a speculative form, 6071or -1, if it doesn't. @var{request} describes the type of requested 6072speculation. If the return value equals 1 then @var{new_pat} is assigned 6073the generated speculative pattern. 6074@end deftypefn 6075 6076@deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn}) 6077This hook is called by the insn scheduler during generation of recovery code 6078for @var{insn}. It should return nonzero, if the corresponding check 6079instruction should branch to recovery code, or zero otherwise. 6080@end deftypefn 6081 6082@deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p}) 6083This hook is called by the insn scheduler to generate a pattern for recovery 6084check instruction. If @var{mutate_p} is zero, then @var{insn} is a 6085speculative instruction for which the check should be generated. 6086@var{label} is either a label of a basic block, where recovery code should 6087be emitted, or a null pointer, when requested check doesn't branch to 6088recovery code (a simple check). If @var{mutate_p} is nonzero, then 6089a pattern for a branchy check corresponding to a simple check denoted by 6090@var{insn} should be generated. In this case @var{label} can't be null. 6091@end deftypefn 6092 6093@deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn}) 6094This hook is used as a workaround for 6095@samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being 6096called on the first instruction of the ready list. The hook is used to 6097discard speculative instruction that stand first in the ready list from 6098being scheduled on the current cycle. For non-speculative instructions, 6099the hook should always return nonzero. For example, in the ia64 backend 6100the hook is used to cancel data speculative insns when the ALAT table 6101is nearly full. 6102@end deftypefn 6103 6104@deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info}) 6105This hook is used by the insn scheduler to find out what features should be 6106enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB 6107bit set. This denotes the scheduler pass for which the data should be 6108provided. The target backend should modify @var{flags} by modifying 6109the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT, 6110DETACH_LIFE_INFO, and DO_SPECULATION. For the DO_SPECULATION feature 6111an additional structure @var{spec_info} should be filled by the target. 6112The structure describes speculation types that can be used in the scheduler. 6113@end deftypefn 6114 6115@node Sections 6116@section Dividing the Output into Sections (Texts, Data, @dots{}) 6117@c the above section title is WAY too long. maybe cut the part between 6118@c the (...)? --mew 10feb93 6119 6120An object file is divided into sections containing different types of 6121data. In the most common case, there are three sections: the @dfn{text 6122section}, which holds instructions and read-only data; the @dfn{data 6123section}, which holds initialized writable data; and the @dfn{bss 6124section}, which holds uninitialized data. Some systems have other kinds 6125of sections. 6126 6127@file{varasm.c} provides several well-known sections, such as 6128@code{text_section}, @code{data_section} and @code{bss_section}. 6129The normal way of controlling a @code{@var{foo}_section} variable 6130is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro, 6131as described below. The macros are only read once, when @file{varasm.c} 6132initializes itself, so their values must be run-time constants. 6133They may however depend on command-line flags. 6134 6135@emph{Note:} Some run-time files, such @file{crtstuff.c}, also make 6136use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them 6137to be string literals. 6138 6139Some assemblers require a different string to be written every time a 6140section is selected. If your assembler falls into this category, you 6141should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use 6142@code{get_unnamed_section} to set up the sections. 6143 6144You must always create a @code{text_section}, either by defining 6145@code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section} 6146in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of 6147@code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not 6148create a distinct @code{readonly_data_section}, the default is to 6149reuse @code{text_section}. 6150 6151All the other @file{varasm.c} sections are optional, and are null 6152if the target does not provide them. 6153 6154@defmac TEXT_SECTION_ASM_OP 6155A C expression whose value is a string, including spacing, containing the 6156assembler operation that should precede instructions and read-only data. 6157Normally @code{"\t.text"} is right. 6158@end defmac 6159 6160@defmac HOT_TEXT_SECTION_NAME 6161If defined, a C string constant for the name of the section containing most 6162frequently executed functions of the program. If not defined, GCC will provide 6163a default definition if the target supports named sections. 6164@end defmac 6165 6166@defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME 6167If defined, a C string constant for the name of the section containing unlikely 6168executed functions in the program. 6169@end defmac 6170 6171@defmac DATA_SECTION_ASM_OP 6172A C expression whose value is a string, including spacing, containing the 6173assembler operation to identify the following data as writable initialized 6174data. Normally @code{"\t.data"} is right. 6175@end defmac 6176 6177@defmac SDATA_SECTION_ASM_OP 6178If defined, a C expression whose value is a string, including spacing, 6179containing the assembler operation to identify the following data as 6180initialized, writable small data. 6181@end defmac 6182 6183@defmac READONLY_DATA_SECTION_ASM_OP 6184A C expression whose value is a string, including spacing, containing the 6185assembler operation to identify the following data as read-only initialized 6186data. 6187@end defmac 6188 6189@defmac BSS_SECTION_ASM_OP 6190If defined, a C expression whose value is a string, including spacing, 6191containing the assembler operation to identify the following data as 6192uninitialized global data. If not defined, and neither 6193@code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined, 6194uninitialized global data will be output in the data section if 6195@option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be 6196used. 6197@end defmac 6198 6199@defmac SBSS_SECTION_ASM_OP 6200If defined, a C expression whose value is a string, including spacing, 6201containing the assembler operation to identify the following data as 6202uninitialized, writable small data. 6203@end defmac 6204 6205@defmac INIT_SECTION_ASM_OP 6206If defined, a C expression whose value is a string, including spacing, 6207containing the assembler operation to identify the following data as 6208initialization code. If not defined, GCC will assume such a section does 6209not exist. This section has no corresponding @code{init_section} 6210variable; it is used entirely in runtime code. 6211@end defmac 6212 6213@defmac FINI_SECTION_ASM_OP 6214If defined, a C expression whose value is a string, including spacing, 6215containing the assembler operation to identify the following data as 6216finalization code. If not defined, GCC will assume such a section does 6217not exist. This section has no corresponding @code{fini_section} 6218variable; it is used entirely in runtime code. 6219@end defmac 6220 6221@defmac INIT_ARRAY_SECTION_ASM_OP 6222If defined, a C expression whose value is a string, including spacing, 6223containing the assembler operation to identify the following data as 6224part of the @code{.init_array} (or equivalent) section. If not 6225defined, GCC will assume such a section does not exist. Do not define 6226both this macro and @code{INIT_SECTION_ASM_OP}. 6227@end defmac 6228 6229@defmac FINI_ARRAY_SECTION_ASM_OP 6230If defined, a C expression whose value is a string, including spacing, 6231containing the assembler operation to identify the following data as 6232part of the @code{.fini_array} (or equivalent) section. If not 6233defined, GCC will assume such a section does not exist. Do not define 6234both this macro and @code{FINI_SECTION_ASM_OP}. 6235@end defmac 6236 6237@defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function}) 6238If defined, an ASM statement that switches to a different section 6239via @var{section_op}, calls @var{function}, and switches back to 6240the text section. This is used in @file{crtstuff.c} if 6241@code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls 6242to initialization and finalization functions from the init and fini 6243sections. By default, this macro uses a simple function call. Some 6244ports need hand-crafted assembly code to avoid dependencies on 6245registers initialized in the function prologue or to ensure that 6246constant pools don't end up too far way in the text section. 6247@end defmac 6248 6249@defmac TARGET_LIBGCC_SDATA_SECTION 6250If defined, a string which names the section into which small 6251variables defined in crtstuff and libgcc should go. This is useful 6252when the target has options for optimizing access to small data, and 6253you want the crtstuff and libgcc routines to be conservative in what 6254they expect of your application yet liberal in what your application 6255expects. For example, for targets with a @code{.sdata} section (like 6256MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't 6257require small data support from your application, but use this macro 6258to put small data into @code{.sdata} so that your application can 6259access these variables whether it uses small data or not. 6260@end defmac 6261 6262@defmac FORCE_CODE_SECTION_ALIGN 6263If defined, an ASM statement that aligns a code section to some 6264arbitrary boundary. This is used to force all fragments of the 6265@code{.init} and @code{.fini} sections to have to same alignment 6266and thus prevent the linker from having to add any padding. 6267@end defmac 6268 6269@defmac JUMP_TABLES_IN_TEXT_SECTION 6270Define this macro to be an expression with a nonzero value if jump 6271tables (for @code{tablejump} insns) should be output in the text 6272section, along with the assembler instructions. Otherwise, the 6273readonly data section is used. 6274 6275This macro is irrelevant if there is no separate readonly data section. 6276@end defmac 6277 6278@deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void) 6279Define this hook if you need to do something special to set up the 6280@file{varasm.c} sections, or if your target has some special sections 6281of its own that you need to create. 6282 6283GCC calls this hook after processing the command line, but before writing 6284any assembly code, and before calling any of the section-returning hooks 6285described below. 6286@end deftypefn 6287 6288@deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void) 6289Return a mask describing how relocations should be treated when 6290selecting sections. Bit 1 should be set if global relocations 6291should be placed in a read-write section; bit 0 should be set if 6292local relocations should be placed in a read-write section. 6293 6294The default version of this function returns 3 when @option{-fpic} 6295is in effect, and 0 otherwise. The hook is typically redefined 6296when the target cannot support (some kinds of) dynamic relocations 6297in read-only sections even in executables. 6298@end deftypefn 6299 6300@deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align}) 6301Return the section into which @var{exp} should be placed. You can 6302assume that @var{exp} is either a @code{VAR_DECL} node or a constant of 6303some sort. @var{reloc} indicates whether the initial value of @var{exp} 6304requires link-time relocations. Bit 0 is set when variable contains 6305local relocations only, while bit 1 is set for global relocations. 6306@var{align} is the constant alignment in bits. 6307 6308The default version of this function takes care of putting read-only 6309variables in @code{readonly_data_section}. 6310 6311See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}. 6312@end deftypefn 6313 6314@defmac USE_SELECT_SECTION_FOR_FUNCTIONS 6315Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called 6316for @code{FUNCTION_DECL}s as well as for variables and constants. 6317 6318In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the 6319function has been determined to be likely to be called, and nonzero if 6320it is unlikely to be called. 6321@end defmac 6322 6323@deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc}) 6324Build up a unique section name, expressed as a @code{STRING_CST} node, 6325and assign it to @samp{DECL_SECTION_NAME (@var{decl})}. 6326As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether 6327the initial value of @var{exp} requires link-time relocations. 6328 6329The default version of this function appends the symbol name to the 6330ELF section name that would normally be used for the symbol. For 6331example, the function @code{foo} would be placed in @code{.text.foo}. 6332Whatever the actual target object format, this is often good enough. 6333@end deftypefn 6334 6335@deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl}) 6336Return the readonly data section associated with 6337@samp{DECL_SECTION_NAME (@var{decl})}. 6338The default version of this function selects @code{.gnu.linkonce.r.name} if 6339the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name} 6340if function is in @code{.text.name}, and the normal readonly-data section 6341otherwise. 6342@end deftypefn 6343 6344@deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align}) 6345Return the section into which a constant @var{x}, of mode @var{mode}, 6346should be placed. You can assume that @var{x} is some kind of 6347constant in RTL@. The argument @var{mode} is redundant except in the 6348case of a @code{const_int} rtx. @var{align} is the constant alignment 6349in bits. 6350 6351The default version of this function takes care of putting symbolic 6352constants in @code{flag_pic} mode in @code{data_section} and everything 6353else in @code{readonly_data_section}. 6354@end deftypefn 6355 6356@deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p}) 6357Define this hook if references to a symbol or a constant must be 6358treated differently depending on something about the variable or 6359function named by the symbol (such as what section it is in). 6360 6361The hook is executed immediately after rtl has been created for 6362@var{decl}, which may be a variable or function declaration or 6363an entry in the constant pool. In either case, @var{rtl} is the 6364rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})} 6365in this hook; that field may not have been initialized yet. 6366 6367In the case of a constant, it is safe to assume that the rtl is 6368a @code{mem} whose address is a @code{symbol_ref}. Most decls 6369will also have this form, but that is not guaranteed. Global 6370register variables, for instance, will have a @code{reg} for their 6371rtl. (Normally the right thing to do with such unusual rtl is 6372leave it alone.) 6373 6374The @var{new_decl_p} argument will be true if this is the first time 6375that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will 6376be false for subsequent invocations, which will happen for duplicate 6377declarations. Whether or not anything must be done for the duplicate 6378declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}. 6379@var{new_decl_p} is always true when the hook is called for a constant. 6380 6381@cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO} 6382The usual thing for this hook to do is to record flags in the 6383@code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}. 6384Historically, the name string was modified if it was necessary to 6385encode more than one bit of information, but this practice is now 6386discouraged; use @code{SYMBOL_REF_FLAGS}. 6387 6388The default definition of this hook, @code{default_encode_section_info} 6389in @file{varasm.c}, sets a number of commonly-useful bits in 6390@code{SYMBOL_REF_FLAGS}. Check whether the default does what you need 6391before overriding it. 6392@end deftypefn 6393 6394@deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name) 6395Decode @var{name} and return the real name part, sans 6396the characters that @code{TARGET_ENCODE_SECTION_INFO} 6397may have added. 6398@end deftypefn 6399 6400@deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp}) 6401Returns true if @var{exp} should be placed into a ``small data'' section. 6402The default version of this hook always returns false. 6403@end deftypefn 6404 6405@deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION 6406Contains the value true if the target places read-only 6407``small data'' into a separate section. The default value is false. 6408@end deftypevar 6409 6410@deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp}) 6411Returns true if @var{exp} names an object for which name resolution 6412rules must resolve to the current ``module'' (dynamic shared library 6413or executable image). 6414 6415The default version of this hook implements the name resolution rules 6416for ELF, which has a looser model of global name binding than other 6417currently supported object file formats. 6418@end deftypefn 6419 6420@deftypevar {Target Hook} bool TARGET_HAVE_TLS 6421Contains the value true if the target supports thread-local storage. 6422The default value is false. 6423@end deftypevar 6424 6425 6426@node PIC 6427@section Position Independent Code 6428@cindex position independent code 6429@cindex PIC 6430 6431This section describes macros that help implement generation of position 6432independent code. Simply defining these macros is not enough to 6433generate valid PIC; you must also add support to the macros 6434@code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as 6435well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of 6436@samp{movsi} to do something appropriate when the source operand 6437contains a symbolic address. You may also need to alter the handling of 6438switch statements so that they use relative addresses. 6439@c i rearranged the order of the macros above to try to force one of 6440@c them to the next line, to eliminate an overfull hbox. --mew 10feb93 6441 6442@defmac PIC_OFFSET_TABLE_REGNUM 6443The register number of the register used to address a table of static 6444data addresses in memory. In some cases this register is defined by a 6445processor's ``application binary interface'' (ABI)@. When this macro 6446is defined, RTL is generated for this register once, as with the stack 6447pointer and frame pointer registers. If this macro is not defined, it 6448is up to the machine-dependent files to allocate such a register (if 6449necessary). Note that this register must be fixed when in use (e.g.@: 6450when @code{flag_pic} is true). 6451@end defmac 6452 6453@defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED 6454Define this macro if the register defined by 6455@code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define 6456this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined. 6457@end defmac 6458 6459@defmac LEGITIMATE_PIC_OPERAND_P (@var{x}) 6460A C expression that is nonzero if @var{x} is a legitimate immediate 6461operand on the target machine when generating position independent code. 6462You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not 6463check this. You can also assume @var{flag_pic} is true, so you need not 6464check it either. You need not define this macro if all constants 6465(including @code{SYMBOL_REF}) can be immediate operands when generating 6466position independent code. 6467@end defmac 6468 6469@node Assembler Format 6470@section Defining the Output Assembler Language 6471 6472This section describes macros whose principal purpose is to describe how 6473to write instructions in assembler language---rather than what the 6474instructions do. 6475 6476@menu 6477* File Framework:: Structural information for the assembler file. 6478* Data Output:: Output of constants (numbers, strings, addresses). 6479* Uninitialized Data:: Output of uninitialized variables. 6480* Label Output:: Output and generation of labels. 6481* Initialization:: General principles of initialization 6482 and termination routines. 6483* Macros for Initialization:: 6484 Specific macros that control the handling of 6485 initialization and termination routines. 6486* Instruction Output:: Output of actual instructions. 6487* Dispatch Tables:: Output of jump tables. 6488* Exception Region Output:: Output of exception region code. 6489* Alignment Output:: Pseudo ops for alignment and skipping data. 6490@end menu 6491 6492@node File Framework 6493@subsection The Overall Framework of an Assembler File 6494@cindex assembler format 6495@cindex output of assembler code 6496 6497@c prevent bad page break with this line 6498This describes the overall framework of an assembly file. 6499 6500@deftypefn {Target Hook} void TARGET_ASM_FILE_START () 6501@findex default_file_start 6502Output to @code{asm_out_file} any text which the assembler expects to 6503find at the beginning of a file. The default behavior is controlled 6504by two flags, documented below. Unless your target's assembler is 6505quite unusual, if you override the default, you should call 6506@code{default_file_start} at some point in your target hook. This 6507lets other target files rely on these variables. 6508@end deftypefn 6509 6510@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF 6511If this flag is true, the text of the macro @code{ASM_APP_OFF} will be 6512printed as the very first line in the assembly file, unless 6513@option{-fverbose-asm} is in effect. (If that macro has been defined 6514to the empty string, this variable has no effect.) With the normal 6515definition of @code{ASM_APP_OFF}, the effect is to notify the GNU 6516assembler that it need not bother stripping comments or extra 6517whitespace from its input. This allows it to work a bit faster. 6518 6519The default is false. You should not set it to true unless you have 6520verified that your port does not generate any extra whitespace or 6521comments that will cause GAS to issue errors in NO_APP mode. 6522@end deftypevr 6523 6524@deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE 6525If this flag is true, @code{output_file_directive} will be called 6526for the primary source file, immediately after printing 6527@code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect 6528this to be done. The default is false. 6529@end deftypevr 6530 6531@deftypefn {Target Hook} void TARGET_ASM_FILE_END () 6532Output to @code{asm_out_file} any text which the assembler expects 6533to find at the end of a file. The default is to output nothing. 6534@end deftypefn 6535 6536@deftypefun void file_end_indicate_exec_stack () 6537Some systems use a common convention, the @samp{.note.GNU-stack} 6538special section, to indicate whether or not an object file relies on 6539the stack being executable. If your system uses this convention, you 6540should define @code{TARGET_ASM_FILE_END} to this function. If you 6541need to do other things in that hook, have your hook function call 6542this function. 6543@end deftypefun 6544 6545@defmac ASM_COMMENT_START 6546A C string constant describing how to begin a comment in the target 6547assembler language. The compiler assumes that the comment will end at 6548the end of the line. 6549@end defmac 6550 6551@defmac ASM_APP_ON 6552A C string constant for text to be output before each @code{asm} 6553statement or group of consecutive ones. Normally this is 6554@code{"#APP"}, which is a comment that has no effect on most 6555assemblers but tells the GNU assembler that it must check the lines 6556that follow for all valid assembler constructs. 6557@end defmac 6558 6559@defmac ASM_APP_OFF 6560A C string constant for text to be output after each @code{asm} 6561statement or group of consecutive ones. Normally this is 6562@code{"#NO_APP"}, which tells the GNU assembler to resume making the 6563time-saving assumptions that are valid for ordinary compiler output. 6564@end defmac 6565 6566@defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) 6567A C statement to output COFF information or DWARF debugging information 6568which indicates that filename @var{name} is the current source file to 6569the stdio stream @var{stream}. 6570 6571This macro need not be defined if the standard form of output 6572for the file format in use is appropriate. 6573@end defmac 6574 6575@defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string}) 6576A C statement to output the string @var{string} to the stdio stream 6577@var{stream}. If you do not call the function @code{output_quoted_string} 6578in your config files, GCC will only call it to output filenames to 6579the assembler source. So you can use it to canonicalize the format 6580of the filename using this macro. 6581@end defmac 6582 6583@defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string}) 6584A C statement to output something to the assembler file to handle a 6585@samp{#ident} directive containing the text @var{string}. If this 6586macro is not defined, nothing is output for a @samp{#ident} directive. 6587@end defmac 6588 6589@deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align}) 6590Output assembly directives to switch to section @var{name}. The section 6591should have attributes as specified by @var{flags}, which is a bit mask 6592of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align} 6593is nonzero, it contains an alignment in bytes to be used for the section, 6594otherwise some target default should be used. Only targets that must 6595specify an alignment within the section directive need pay attention to 6596@var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}. 6597@end deftypefn 6598 6599@deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS 6600This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}. 6601@end deftypefn 6602 6603@anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS} 6604@deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS 6605This flag is true if we can create zeroed data by switching to a BSS 6606section and then using @code{ASM_OUTPUT_SKIP} to allocate the space. 6607This is true on most ELF targets. 6608@end deftypefn 6609 6610@deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc}) 6611Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION} 6612based on a variable or function decl, a section name, and whether or not the 6613declaration's initializer may contain runtime relocations. @var{decl} may be 6614 null, in which case read-write data should be assumed. 6615 6616The default version of this function handles choosing code vs data, 6617read-only vs read-write data, and @code{flag_pic}. You should only 6618need to override this if your target has special flags that might be 6619set via @code{__attribute__}. 6620@end deftypefn 6621 6622@need 2000 6623@node Data Output 6624@subsection Output of Data 6625 6626 6627@deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP 6628@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP 6629@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP 6630@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP 6631@deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP 6632@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP 6633@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP 6634@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP 6635@deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP 6636These hooks specify assembly directives for creating certain kinds 6637of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a 6638byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an 6639aligned two-byte object, and so on. Any of the hooks may be 6640@code{NULL}, indicating that no suitable directive is available. 6641 6642The compiler will print these strings at the start of a new line, 6643followed immediately by the object's initial value. In most cases, 6644the string should contain a tab, a pseudo-op, and then another tab. 6645@end deftypevr 6646 6647@deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p}) 6648The @code{assemble_integer} function uses this hook to output an 6649integer object. @var{x} is the object's value, @var{size} is its size 6650in bytes and @var{aligned_p} indicates whether it is aligned. The 6651function should return @code{true} if it was able to output the 6652object. If it returns false, @code{assemble_integer} will try to 6653split the object into smaller parts. 6654 6655The default implementation of this hook will use the 6656@code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false} 6657when the relevant string is @code{NULL}. 6658@end deftypefn 6659 6660@defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail}) 6661A C statement to recognize @var{rtx} patterns that 6662@code{output_addr_const} can't deal with, and output assembly code to 6663@var{stream} corresponding to the pattern @var{x}. This may be used to 6664allow machine-dependent @code{UNSPEC}s to appear within constants. 6665 6666If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must 6667@code{goto fail}, so that a standard error message is printed. If it 6668prints an error message itself, by calling, for example, 6669@code{output_operand_lossage}, it may just complete normally. 6670@end defmac 6671 6672@defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) 6673A C statement to output to the stdio stream @var{stream} an assembler 6674instruction to assemble a string constant containing the @var{len} 6675bytes at @var{ptr}. @var{ptr} will be a C expression of type 6676@code{char *} and @var{len} a C expression of type @code{int}. 6677 6678If the assembler has a @code{.ascii} pseudo-op as found in the 6679Berkeley Unix assembler, do not define the macro 6680@code{ASM_OUTPUT_ASCII}. 6681@end defmac 6682 6683@defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n}) 6684A C statement to output word @var{n} of a function descriptor for 6685@var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS} 6686is defined, and is otherwise unused. 6687@end defmac 6688 6689@defmac CONSTANT_POOL_BEFORE_FUNCTION 6690You may define this macro as a C expression. You should define the 6691expression to have a nonzero value if GCC should output the constant 6692pool for a function before the code for the function, or a zero value if 6693GCC should output the constant pool after the function. If you do 6694not define this macro, the usual case, GCC will output the constant 6695pool before the function. 6696@end defmac 6697 6698@defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size}) 6699A C statement to output assembler commands to define the start of the 6700constant pool for a function. @var{funname} is a string giving 6701the name of the function. Should the return type of the function 6702be required, it can be obtained via @var{fundecl}. @var{size} 6703is the size, in bytes, of the constant pool that will be written 6704immediately after this call. 6705 6706If no constant-pool prefix is required, the usual case, this macro need 6707not be defined. 6708@end defmac 6709 6710@defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) 6711A C statement (with or without semicolon) to output a constant in the 6712constant pool, if it needs special treatment. (This macro need not do 6713anything for RTL expressions that can be output normally.) 6714 6715The argument @var{file} is the standard I/O stream to output the 6716assembler code on. @var{x} is the RTL expression for the constant to 6717output, and @var{mode} is the machine mode (in case @var{x} is a 6718@samp{const_int}). @var{align} is the required alignment for the value 6719@var{x}; you should output an assembler directive to force this much 6720alignment. 6721 6722The argument @var{labelno} is a number to use in an internal label for 6723the address of this pool entry. The definition of this macro is 6724responsible for outputting the label definition at the proper place. 6725Here is how to do this: 6726 6727@smallexample 6728@code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno}); 6729@end smallexample 6730 6731When you output a pool entry specially, you should end with a 6732@code{goto} to the label @var{jumpto}. This will prevent the same pool 6733entry from being output a second time in the usual manner. 6734 6735You need not define this macro if it would do nothing. 6736@end defmac 6737 6738@defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) 6739A C statement to output assembler commands to at the end of the constant 6740pool for a function. @var{funname} is a string giving the name of the 6741function. Should the return type of the function be required, you can 6742obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the 6743constant pool that GCC wrote immediately before this call. 6744 6745If no constant-pool epilogue is required, the usual case, you need not 6746define this macro. 6747@end defmac 6748 6749@defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}) 6750Define this macro as a C expression which is nonzero if @var{C} is 6751used as a logical line separator by the assembler. 6752 6753If you do not define this macro, the default is that only 6754the character @samp{;} is treated as a logical line separator. 6755@end defmac 6756 6757@deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN 6758@deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN 6759These target hooks are C string constants, describing the syntax in the 6760assembler for grouping arithmetic expressions. If not overridden, they 6761default to normal parentheses, which is correct for most assemblers. 6762@end deftypevr 6763 6764 These macros are provided by @file{real.h} for writing the definitions 6765of @code{ASM_OUTPUT_DOUBLE} and the like: 6766 6767@defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) 6768@defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) 6769@defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) 6770@defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l}) 6771@defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l}) 6772@defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l}) 6773These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the 6774target's floating point representation, and store its bit pattern in 6775the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and 6776@code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a 6777simple @code{long int}. For the others, it should be an array of 6778@code{long int}. The number of elements in this array is determined 6779by the size of the desired target floating point data type: 32 bits of 6780it go in each @code{long int} array element. Each array element holds 678132 bits of the result, even if @code{long int} is wider than 32 bits 6782on the host machine. 6783 6784The array element values are designed so that you can print them out 6785using @code{fprintf} in the order they should appear in the target 6786machine's memory. 6787@end defmac 6788 6789@node Uninitialized Data 6790@subsection Output of Uninitialized Variables 6791 6792Each of the macros in this section is used to do the whole job of 6793outputting a single uninitialized variable. 6794 6795@defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6796A C statement (sans semicolon) to output to the stdio stream 6797@var{stream} the assembler definition of a common-label named 6798@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6799is the size rounded up to whatever alignment the caller wants. 6800 6801Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6802output the name itself; before and after that, output the additional 6803assembler syntax for defining the name, and a newline. 6804 6805This macro controls how the assembler definitions of uninitialized 6806common global variables are output. 6807@end defmac 6808 6809@defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) 6810Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a 6811separate, explicit argument. If you define this macro, it is used in 6812place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in 6813handling the required alignment of the variable. The alignment is specified 6814as the number of bits. 6815@end defmac 6816 6817@defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6818Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the 6819variable to be output, if there is one, or @code{NULL_TREE} if there 6820is no corresponding variable. If you define this macro, GCC will use it 6821in place of both @code{ASM_OUTPUT_COMMON} and 6822@code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see 6823the variable's decl in order to chose what to output. 6824@end defmac 6825 6826@defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) 6827A C statement (sans semicolon) to output to the stdio stream 6828@var{stream} the assembler definition of uninitialized global @var{decl} named 6829@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6830is the size rounded up to whatever alignment the caller wants. 6831 6832Try to use function @code{asm_output_bss} defined in @file{varasm.c} when 6833defining this macro. If unable, use the expression 6834@code{assemble_name (@var{stream}, @var{name})} to output the name itself; 6835before and after that, output the additional assembler syntax for defining 6836the name, and a newline. 6837 6838There are two ways of handling global BSS. One is to define either 6839this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}. 6840The other is to have @code{TARGET_ASM_SELECT_SECTION} return a 6841switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}). 6842You do not need to do both. 6843 6844Some languages do not have @code{common} data, and require a 6845non-common form of global BSS in order to handle uninitialized globals 6846efficiently. C++ is one example of this. However, if the target does 6847not support global BSS, the front end may choose to make globals 6848common in order to save space in the object file. 6849@end defmac 6850 6851@defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6852Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a 6853separate, explicit argument. If you define this macro, it is used in 6854place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in 6855handling the required alignment of the variable. The alignment is specified 6856as the number of bits. 6857 6858Try to use function @code{asm_output_aligned_bss} defined in file 6859@file{varasm.c} when defining this macro. 6860@end defmac 6861 6862@defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) 6863A C statement (sans semicolon) to output to the stdio stream 6864@var{stream} the assembler definition of a local-common-label named 6865@var{name} whose size is @var{size} bytes. The variable @var{rounded} 6866is the size rounded up to whatever alignment the caller wants. 6867 6868Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6869output the name itself; before and after that, output the additional 6870assembler syntax for defining the name, and a newline. 6871 6872This macro controls how the assembler definitions of uninitialized 6873static variables are output. 6874@end defmac 6875 6876@defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) 6877Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a 6878separate, explicit argument. If you define this macro, it is used in 6879place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in 6880handling the required alignment of the variable. The alignment is specified 6881as the number of bits. 6882@end defmac 6883 6884@defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) 6885Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the 6886variable to be output, if there is one, or @code{NULL_TREE} if there 6887is no corresponding variable. If you define this macro, GCC will use it 6888in place of both @code{ASM_OUTPUT_DECL} and 6889@code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see 6890the variable's decl in order to chose what to output. 6891@end defmac 6892 6893@node Label Output 6894@subsection Output and Generation of Labels 6895 6896@c prevent bad page break with this line 6897This is about outputting labels. 6898 6899@findex assemble_name 6900@defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name}) 6901A C statement (sans semicolon) to output to the stdio stream 6902@var{stream} the assembler definition of a label named @var{name}. 6903Use the expression @code{assemble_name (@var{stream}, @var{name})} to 6904output the name itself; before and after that, output the additional 6905assembler syntax for defining the name, and a newline. A default 6906definition of this macro is provided which is correct for most systems. 6907@end defmac 6908 6909@findex assemble_name_raw 6910@defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name}) 6911Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known 6912to refer to a compiler-generated label. The default definition uses 6913@code{assemble_name_raw}, which is like @code{assemble_name} except 6914that it is more efficient. 6915@end defmac 6916 6917@defmac SIZE_ASM_OP 6918A C string containing the appropriate assembler directive to specify the 6919size of a symbol, without any arguments. On systems that use ELF, the 6920default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other 6921systems, the default is not to define this macro. 6922 6923Define this macro only if it is correct to use the default definitions 6924of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE} 6925for your system. If you need your own custom definitions of those 6926macros, or if you do not need explicit symbol sizes at all, do not 6927define this macro. 6928@end defmac 6929 6930@defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size}) 6931A C statement (sans semicolon) to output to the stdio stream 6932@var{stream} a directive telling the assembler that the size of the 6933symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}. 6934If you define @code{SIZE_ASM_OP}, a default definition of this macro is 6935provided. 6936@end defmac 6937 6938@defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name}) 6939A C statement (sans semicolon) to output to the stdio stream 6940@var{stream} a directive telling the assembler to calculate the size of 6941the symbol @var{name} by subtracting its address from the current 6942address. 6943 6944If you define @code{SIZE_ASM_OP}, a default definition of this macro is 6945provided. The default assumes that the assembler recognizes a special 6946@samp{.} symbol as referring to the current address, and can calculate 6947the difference between this and another symbol. If your assembler does 6948not recognize @samp{.} or cannot do calculations with it, you will need 6949to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique. 6950@end defmac 6951 6952@defmac TYPE_ASM_OP 6953A C string containing the appropriate assembler directive to specify the 6954type of a symbol, without any arguments. On systems that use ELF, the 6955default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other 6956systems, the default is not to define this macro. 6957 6958Define this macro only if it is correct to use the default definition of 6959@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 6960custom definition of this macro, or if you do not need explicit symbol 6961types at all, do not define this macro. 6962@end defmac 6963 6964@defmac TYPE_OPERAND_FMT 6965A C string which specifies (using @code{printf} syntax) the format of 6966the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the 6967default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems, 6968the default is not to define this macro. 6969 6970Define this macro only if it is correct to use the default definition of 6971@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own 6972custom definition of this macro, or if you do not need explicit symbol 6973types at all, do not define this macro. 6974@end defmac 6975 6976@defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type}) 6977A C statement (sans semicolon) to output to the stdio stream 6978@var{stream} a directive telling the assembler that the type of the 6979symbol @var{name} is @var{type}. @var{type} is a C string; currently, 6980that string is always either @samp{"function"} or @samp{"object"}, but 6981you should not count on this. 6982 6983If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default 6984definition of this macro is provided. 6985@end defmac 6986 6987@defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) 6988A C statement (sans semicolon) to output to the stdio stream 6989@var{stream} any text necessary for declaring the name @var{name} of a 6990function which is being defined. This macro is responsible for 6991outputting the label definition (perhaps using 6992@code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the 6993@code{FUNCTION_DECL} tree node representing the function. 6994 6995If this macro is not defined, then the function name is defined in the 6996usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 6997 6998You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition 6999of this macro. 7000@end defmac 7001 7002@defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) 7003A C statement (sans semicolon) to output to the stdio stream 7004@var{stream} any text necessary for declaring the size of a function 7005which is being defined. The argument @var{name} is the name of the 7006function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node 7007representing the function. 7008 7009If this macro is not defined, then the function size is not defined. 7010 7011You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition 7012of this macro. 7013@end defmac 7014 7015@defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) 7016A C statement (sans semicolon) to output to the stdio stream 7017@var{stream} any text necessary for declaring the name @var{name} of an 7018initialized variable which is being defined. This macro must output the 7019label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument 7020@var{decl} is the @code{VAR_DECL} tree node representing the variable. 7021 7022If this macro is not defined, then the variable name is defined in the 7023usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 7024 7025You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or 7026@code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro. 7027@end defmac 7028 7029@defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size}) 7030A C statement (sans semicolon) to output to the stdio stream 7031@var{stream} any text necessary for declaring the name @var{name} of a 7032constant which is being defined. This macro is responsible for 7033outputting the label definition (perhaps using 7034@code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the 7035value of the constant, and @var{size} is the size of the constant 7036in bytes. @var{name} will be an internal label. 7037 7038If this macro is not defined, then the @var{name} is defined in the 7039usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). 7040 7041You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition 7042of this macro. 7043@end defmac 7044 7045@defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name}) 7046A C statement (sans semicolon) to output to the stdio stream 7047@var{stream} any text necessary for claiming a register @var{regno} 7048for a global variable @var{decl} with name @var{name}. 7049 7050If you don't define this macro, that is equivalent to defining it to do 7051nothing. 7052@end defmac 7053 7054@defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) 7055A C statement (sans semicolon) to finish up declaring a variable name 7056once the compiler has processed its initializer fully and thus has had a 7057chance to determine the size of an array when controlled by an 7058initializer. This is used on systems where it's necessary to declare 7059something about the size of the object. 7060 7061If you don't define this macro, that is equivalent to defining it to do 7062nothing. 7063 7064You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or 7065@code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro. 7066@end defmac 7067 7068@deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name}) 7069This target hook is a function to output to the stdio stream 7070@var{stream} some commands that will make the label @var{name} global; 7071that is, available for reference from other files. 7072 7073The default implementation relies on a proper definition of 7074@code{GLOBAL_ASM_OP}. 7075@end deftypefn 7076 7077@defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name}) 7078A C statement (sans semicolon) to output to the stdio stream 7079@var{stream} some commands that will make the label @var{name} weak; 7080that is, available for reference from other files but only used if 7081no other definition is available. Use the expression 7082@code{assemble_name (@var{stream}, @var{name})} to output the name 7083itself; before and after that, output the additional assembler syntax 7084for making that name weak, and a newline. 7085 7086If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not 7087support weak symbols and you should not define the @code{SUPPORTS_WEAK} 7088macro. 7089@end defmac 7090 7091@defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value}) 7092Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and 7093@code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function 7094or variable decl. If @var{value} is not @code{NULL}, this C statement 7095should output to the stdio stream @var{stream} assembler code which 7096defines (equates) the weak symbol @var{name} to have the value 7097@var{value}. If @var{value} is @code{NULL}, it should output commands 7098to make @var{name} weak. 7099@end defmac 7100 7101@defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value}) 7102Outputs a directive that enables @var{name} to be used to refer to 7103symbol @var{value} with weak-symbol semantics. @code{decl} is the 7104declaration of @code{name}. 7105@end defmac 7106 7107@defmac SUPPORTS_WEAK 7108A C expression which evaluates to true if the target supports weak symbols. 7109 7110If you don't define this macro, @file{defaults.h} provides a default 7111definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL} 7112is defined, the default definition is @samp{1}; otherwise, it is 7113@samp{0}. Define this macro if you want to control weak symbol support 7114with a compiler flag such as @option{-melf}. 7115@end defmac 7116 7117@defmac MAKE_DECL_ONE_ONLY (@var{decl}) 7118A C statement (sans semicolon) to mark @var{decl} to be emitted as a 7119public symbol such that extra copies in multiple translation units will 7120be discarded by the linker. Define this macro if your object file 7121format provides support for this concept, such as the @samp{COMDAT} 7122section flags in the Microsoft Windows PE/COFF format, and this support 7123requires changes to @var{decl}, such as putting it in a separate section. 7124@end defmac 7125 7126@defmac SUPPORTS_ONE_ONLY 7127A C expression which evaluates to true if the target supports one-only 7128semantics. 7129 7130If you don't define this macro, @file{varasm.c} provides a default 7131definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default 7132definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if 7133you want to control one-only symbol support with a compiler flag, or if 7134setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to 7135be emitted as one-only. 7136@end defmac 7137 7138@deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility}) 7139This target hook is a function to output to @var{asm_out_file} some 7140commands that will make the symbol(s) associated with @var{decl} have 7141hidden, protected or internal visibility as specified by @var{visibility}. 7142@end deftypefn 7143 7144@defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC 7145A C expression that evaluates to true if the target's linker expects 7146that weak symbols do not appear in a static archive's table of contents. 7147The default is @code{0}. 7148 7149Leaving weak symbols out of an archive's table of contents means that, 7150if a symbol will only have a definition in one translation unit and 7151will have undefined references from other translation units, that 7152symbol should not be weak. Defining this macro to be nonzero will 7153thus have the effect that certain symbols that would normally be weak 7154(explicit template instantiations, and vtables for polymorphic classes 7155with noninline key methods) will instead be nonweak. 7156 7157The C++ ABI requires this macro to be zero. Define this macro for 7158targets where full C++ ABI compliance is impossible and where linker 7159restrictions require weak symbols to be left out of a static archive's 7160table of contents. 7161@end defmac 7162 7163@defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) 7164A C statement (sans semicolon) to output to the stdio stream 7165@var{stream} any text necessary for declaring the name of an external 7166symbol named @var{name} which is referenced in this compilation but 7167not defined. The value of @var{decl} is the tree node for the 7168declaration. 7169 7170This macro need not be defined if it does not need to output anything. 7171The GNU assembler and most Unix assemblers don't require anything. 7172@end defmac 7173 7174@deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref}) 7175This target hook is a function to output to @var{asm_out_file} an assembler 7176pseudo-op to declare a library function name external. The name of the 7177library function is given by @var{symref}, which is a @code{symbol_ref}. 7178@end deftypefn 7179 7180@deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl}) 7181This target hook is a function to output to @var{asm_out_file} an assembler 7182directive to annotate used symbol. Darwin target use .no_dead_code_strip 7183directive. 7184@end deftypefn 7185 7186@defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) 7187A C statement (sans semicolon) to output to the stdio stream 7188@var{stream} a reference in assembler syntax to a label named 7189@var{name}. This should add @samp{_} to the front of the name, if that 7190is customary on your operating system, as it is in most Berkeley Unix 7191systems. This macro is used in @code{assemble_name}. 7192@end defmac 7193 7194@defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym}) 7195A C statement (sans semicolon) to output a reference to 7196@code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name} 7197will be used to output the name of the symbol. This macro may be used 7198to modify the way a symbol is referenced depending on information 7199encoded by @code{TARGET_ENCODE_SECTION_INFO}. 7200@end defmac 7201 7202@defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf}) 7203A C statement (sans semicolon) to output a reference to @var{buf}, the 7204result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined, 7205@code{assemble_name} will be used to output the name of the symbol. 7206This macro is not used by @code{output_asm_label}, or the @code{%l} 7207specifier that calls it; the intention is that this macro should be set 7208when it is necessary to output a label differently when its address is 7209being taken. 7210@end defmac 7211 7212@deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno}) 7213A function to output to the stdio stream @var{stream} a label whose 7214name is made from the string @var{prefix} and the number @var{labelno}. 7215 7216It is absolutely essential that these labels be distinct from the labels 7217used for user-level functions and variables. Otherwise, certain programs 7218will have name conflicts with internal labels. 7219 7220It is desirable to exclude internal labels from the symbol table of the 7221object file. Most assemblers have a naming convention for labels that 7222should be excluded; on many systems, the letter @samp{L} at the 7223beginning of a label has this effect. You should find out what 7224convention your system uses, and follow it. 7225 7226The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}. 7227@end deftypefn 7228 7229@defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num}) 7230A C statement to output to the stdio stream @var{stream} a debug info 7231label whose name is made from the string @var{prefix} and the number 7232@var{num}. This is useful for VLIW targets, where debug info labels 7233may need to be treated differently than branch target labels. On some 7234systems, branch target labels must be at the beginning of instruction 7235bundles, but debug info labels can occur in the middle of instruction 7236bundles. 7237 7238If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be 7239used. 7240@end defmac 7241 7242@defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) 7243A C statement to store into the string @var{string} a label whose name 7244is made from the string @var{prefix} and the number @var{num}. 7245 7246This string, when output subsequently by @code{assemble_name}, should 7247produce the output that @code{(*targetm.asm_out.internal_label)} would produce 7248with the same @var{prefix} and @var{num}. 7249 7250If the string begins with @samp{*}, then @code{assemble_name} will 7251output the rest of the string unchanged. It is often convenient for 7252@code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the 7253string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets 7254to output the string, and may change it. (Of course, 7255@code{ASM_OUTPUT_LABELREF} is also part of your machine description, so 7256you should know what it does on your machine.) 7257@end defmac 7258 7259@defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) 7260A C expression to assign to @var{outvar} (which is a variable of type 7261@code{char *}) a newly allocated string made from the string 7262@var{name} and the number @var{number}, with some suitable punctuation 7263added. Use @code{alloca} to get space for the string. 7264 7265The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to 7266produce an assembler label for an internal static variable whose name is 7267@var{name}. Therefore, the string must be such as to result in valid 7268assembler code. The argument @var{number} is different each time this 7269macro is executed; it prevents conflicts between similarly-named 7270internal static variables in different scopes. 7271 7272Ideally this string should not be a valid C identifier, to prevent any 7273conflict with the user's own symbols. Most assemblers allow periods 7274or percent signs in assembler symbols; putting at least one of these 7275between the name and the number will suffice. 7276 7277If this macro is not defined, a default definition will be provided 7278which is correct for most systems. 7279@end defmac 7280 7281@defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) 7282A C statement to output to the stdio stream @var{stream} assembler code 7283which defines (equates) the symbol @var{name} to have the value @var{value}. 7284 7285@findex SET_ASM_OP 7286If @code{SET_ASM_OP} is defined, a default definition is provided which is 7287correct for most systems. 7288@end defmac 7289 7290@defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value}) 7291A C statement to output to the stdio stream @var{stream} assembler code 7292which defines (equates) the symbol whose tree node is @var{decl_of_name} 7293to have the value of the tree node @var{decl_of_value}. This macro will 7294be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if 7295the tree nodes are available. 7296 7297@findex SET_ASM_OP 7298If @code{SET_ASM_OP} is defined, a default definition is provided which is 7299correct for most systems. 7300@end defmac 7301 7302@defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value}) 7303A C statement that evaluates to true if the assembler code which defines 7304(equates) the symbol whose tree node is @var{decl_of_name} to have the value 7305of the tree node @var{decl_of_value} should be emitted near the end of the 7306current compilation unit. The default is to not defer output of defines. 7307This macro affects defines output by @samp{ASM_OUTPUT_DEF} and 7308@samp{ASM_OUTPUT_DEF_FROM_DECLS}. 7309@end defmac 7310 7311@defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value}) 7312A C statement to output to the stdio stream @var{stream} assembler code 7313which defines (equates) the weak symbol @var{name} to have the value 7314@var{value}. If @var{value} is @code{NULL}, it defines @var{name} as 7315an undefined weak symbol. 7316 7317Define this macro if the target only supports weak aliases; define 7318@code{ASM_OUTPUT_DEF} instead if possible. 7319@end defmac 7320 7321@node Initialization 7322@subsection How Initialization Functions Are Handled 7323@cindex initialization routines 7324@cindex termination routines 7325@cindex constructors, output of 7326@cindex destructors, output of 7327 7328The compiled code for certain languages includes @dfn{constructors} 7329(also called @dfn{initialization routines})---functions to initialize 7330data in the program when the program is started. These functions need 7331to be called before the program is ``started''---that is to say, before 7332@code{main} is called. 7333 7334Compiling some languages generates @dfn{destructors} (also called 7335@dfn{termination routines}) that should be called when the program 7336terminates. 7337 7338To make the initialization and termination functions work, the compiler 7339must output something in the assembler code to cause those functions to 7340be called at the appropriate time. When you port the compiler to a new 7341system, you need to specify how to do this. 7342 7343There are two major ways that GCC currently supports the execution of 7344initialization and termination functions. Each way has two variants. 7345Much of the structure is common to all four variations. 7346 7347@findex __CTOR_LIST__ 7348@findex __DTOR_LIST__ 7349The linker must build two lists of these functions---a list of 7350initialization functions, called @code{__CTOR_LIST__}, and a list of 7351termination functions, called @code{__DTOR_LIST__}. 7352 7353Each list always begins with an ignored function pointer (which may hold 73540, @minus{}1, or a count of the function pointers after it, depending on 7355the environment). This is followed by a series of zero or more function 7356pointers to constructors (or destructors), followed by a function 7357pointer containing zero. 7358 7359Depending on the operating system and its executable file format, either 7360@file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup 7361time and exit time. Constructors are called in reverse order of the 7362list; destructors in forward order. 7363 7364The best way to handle static constructors works only for object file 7365formats which provide arbitrarily-named sections. A section is set 7366aside for a list of constructors, and another for a list of destructors. 7367Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each 7368object file that defines an initialization function also puts a word in 7369the constructor section to point to that function. The linker 7370accumulates all these words into one contiguous @samp{.ctors} section. 7371Termination functions are handled similarly. 7372 7373This method will be chosen as the default by @file{target-def.h} if 7374@code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not 7375support arbitrary sections, but does support special designated 7376constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP} 7377and @code{DTORS_SECTION_ASM_OP} to achieve the same effect. 7378 7379When arbitrary sections are available, there are two variants, depending 7380upon how the code in @file{crtstuff.c} is called. On systems that 7381support a @dfn{.init} section which is executed at program startup, 7382parts of @file{crtstuff.c} are compiled into that section. The 7383program is linked by the @command{gcc} driver like this: 7384 7385@smallexample 7386ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o 7387@end smallexample 7388 7389The prologue of a function (@code{__init}) appears in the @code{.init} 7390section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise 7391for the function @code{__fini} in the @dfn{.fini} section. Normally these 7392files are provided by the operating system or by the GNU C library, but 7393are provided by GCC for a few targets. 7394 7395The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets) 7396compiled from @file{crtstuff.c}. They contain, among other things, code 7397fragments within the @code{.init} and @code{.fini} sections that branch 7398to routines in the @code{.text} section. The linker will pull all parts 7399of a section together, which results in a complete @code{__init} function 7400that invokes the routines we need at startup. 7401 7402To use this variant, you must define the @code{INIT_SECTION_ASM_OP} 7403macro properly. 7404 7405If no init section is available, when GCC compiles any function called 7406@code{main} (or more accurately, any function designated as a program 7407entry point by the language front end calling @code{expand_main_function}), 7408it inserts a procedure call to @code{__main} as the first executable code 7409after the function prologue. The @code{__main} function is defined 7410in @file{libgcc2.c} and runs the global constructors. 7411 7412In file formats that don't support arbitrary sections, there are again 7413two variants. In the simplest variant, the GNU linker (GNU @code{ld}) 7414and an `a.out' format must be used. In this case, 7415@code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs} 7416entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, 7417and with the address of the void function containing the initialization 7418code as its value. The GNU linker recognizes this as a request to add 7419the value to a @dfn{set}; the values are accumulated, and are eventually 7420placed in the executable as a vector in the format described above, with 7421a leading (ignored) count and a trailing zero element. 7422@code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init 7423section is available, the absence of @code{INIT_SECTION_ASM_OP} causes 7424the compilation of @code{main} to call @code{__main} as above, starting 7425the initialization process. 7426 7427The last variant uses neither arbitrary sections nor the GNU linker. 7428This is preferable when you want to do dynamic linking and when using 7429file formats which the GNU linker does not support, such as `ECOFF'@. In 7430this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and 7431termination functions are recognized simply by their names. This requires 7432an extra program in the linkage step, called @command{collect2}. This program 7433pretends to be the linker, for use with GCC; it does its job by running 7434the ordinary linker, but also arranges to include the vectors of 7435initialization and termination functions. These functions are called 7436via @code{__main} as described above. In order to use this method, 7437@code{use_collect2} must be defined in the target in @file{config.gcc}. 7438 7439@ifinfo 7440The following section describes the specific macros that control and 7441customize the handling of initialization and termination functions. 7442@end ifinfo 7443 7444@node Macros for Initialization 7445@subsection Macros Controlling Initialization Routines 7446 7447Here are the macros that control how the compiler handles initialization 7448and termination functions: 7449 7450@defmac INIT_SECTION_ASM_OP 7451If defined, a C string constant, including spacing, for the assembler 7452operation to identify the following data as initialization code. If not 7453defined, GCC will assume such a section does not exist. When you are 7454using special sections for initialization and termination functions, this 7455macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to 7456run the initialization functions. 7457@end defmac 7458 7459@defmac HAS_INIT_SECTION 7460If defined, @code{main} will not call @code{__main} as described above. 7461This macro should be defined for systems that control start-up code 7462on a symbol-by-symbol basis, such as OSF/1, and should not 7463be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}. 7464@end defmac 7465 7466@defmac LD_INIT_SWITCH 7467If defined, a C string constant for a switch that tells the linker that 7468the following symbol is an initialization routine. 7469@end defmac 7470 7471@defmac LD_FINI_SWITCH 7472If defined, a C string constant for a switch that tells the linker that 7473the following symbol is a finalization routine. 7474@end defmac 7475 7476@defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func}) 7477If defined, a C statement that will write a function that can be 7478automatically called when a shared library is loaded. The function 7479should call @var{func}, which takes no arguments. If not defined, and 7480the object format requires an explicit initialization function, then a 7481function called @code{_GLOBAL__DI} will be generated. 7482 7483This function and the following one are used by collect2 when linking a 7484shared library that needs constructors or destructors, or has DWARF2 7485exception tables embedded in the code. 7486@end defmac 7487 7488@defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func}) 7489If defined, a C statement that will write a function that can be 7490automatically called when a shared library is unloaded. The function 7491should call @var{func}, which takes no arguments. If not defined, and 7492the object format requires an explicit finalization function, then a 7493function called @code{_GLOBAL__DD} will be generated. 7494@end defmac 7495 7496@defmac INVOKE__main 7497If defined, @code{main} will call @code{__main} despite the presence of 7498@code{INIT_SECTION_ASM_OP}. This macro should be defined for systems 7499where the init section is not actually run automatically, but is still 7500useful for collecting the lists of constructors and destructors. 7501@end defmac 7502 7503@defmac SUPPORTS_INIT_PRIORITY 7504If nonzero, the C++ @code{init_priority} attribute is supported and the 7505compiler should emit instructions to control the order of initialization 7506of objects. If zero, the compiler will issue an error message upon 7507encountering an @code{init_priority} attribute. 7508@end defmac 7509 7510@deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS 7511This value is true if the target supports some ``native'' method of 7512collecting constructors and destructors to be run at startup and exit. 7513It is false if we must use @command{collect2}. 7514@end deftypefn 7515 7516@deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority}) 7517If defined, a function that outputs assembler code to arrange to call 7518the function referenced by @var{symbol} at initialization time. 7519 7520Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking 7521no arguments and with no return value. If the target supports initialization 7522priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY}; 7523otherwise it must be @code{DEFAULT_INIT_PRIORITY}. 7524 7525If this macro is not defined by the target, a suitable default will 7526be chosen if (1) the target supports arbitrary section names, (2) the 7527target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2} 7528is not defined. 7529@end deftypefn 7530 7531@deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority}) 7532This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination 7533functions rather than initialization functions. 7534@end deftypefn 7535 7536If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine 7537generated for the generated object file will have static linkage. 7538 7539If your system uses @command{collect2} as the means of processing 7540constructors, then that program normally uses @command{nm} to scan 7541an object file for constructor functions to be called. 7542 7543On certain kinds of systems, you can define this macro to make 7544@command{collect2} work faster (and, in some cases, make it work at all): 7545 7546@defmac OBJECT_FORMAT_COFF 7547Define this macro if the system uses COFF (Common Object File Format) 7548object files, so that @command{collect2} can assume this format and scan 7549object files directly for dynamic constructor/destructor functions. 7550 7551This macro is effective only in a native compiler; @command{collect2} as 7552part of a cross compiler always uses @command{nm} for the target machine. 7553@end defmac 7554 7555@defmac REAL_NM_FILE_NAME 7556Define this macro as a C string constant containing the file name to use 7557to execute @command{nm}. The default is to search the path normally for 7558@command{nm}. 7559 7560If your system supports shared libraries and has a program to list the 7561dynamic dependencies of a given library or executable, you can define 7562these macros to enable support for running initialization and 7563termination functions in shared libraries: 7564@end defmac 7565 7566@defmac LDD_SUFFIX 7567Define this macro to a C string constant containing the name of the program 7568which lists dynamic dependencies, like @command{"ldd"} under SunOS 4. 7569@end defmac 7570 7571@defmac PARSE_LDD_OUTPUT (@var{ptr}) 7572Define this macro to be C code that extracts filenames from the output 7573of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable 7574of type @code{char *} that points to the beginning of a line of output 7575from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the 7576code must advance @var{ptr} to the beginning of the filename on that 7577line. Otherwise, it must set @var{ptr} to @code{NULL}. 7578@end defmac 7579 7580@node Instruction Output 7581@subsection Output of Assembler Instructions 7582 7583@c prevent bad page break with this line 7584This describes assembler instruction output. 7585 7586@defmac REGISTER_NAMES 7587A C initializer containing the assembler's names for the machine 7588registers, each one as a C string constant. This is what translates 7589register numbers in the compiler into assembler language. 7590@end defmac 7591 7592@defmac ADDITIONAL_REGISTER_NAMES 7593If defined, a C initializer for an array of structures containing a name 7594and a register number. This macro defines additional names for hard 7595registers, thus allowing the @code{asm} option in declarations to refer 7596to registers using alternate names. 7597@end defmac 7598 7599@defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) 7600Define this macro if you are using an unusual assembler that 7601requires different names for the machine instructions. 7602 7603The definition is a C statement or statements which output an 7604assembler instruction opcode to the stdio stream @var{stream}. The 7605macro-operand @var{ptr} is a variable of type @code{char *} which 7606points to the opcode name in its ``internal'' form---the form that is 7607written in the machine description. The definition should output the 7608opcode name to @var{stream}, performing any translation you desire, and 7609increment the variable @var{ptr} to point at the end of the opcode 7610so that it will not be output twice. 7611 7612In fact, your macro definition may process less than the entire opcode 7613name, or more than the opcode name; but if you want to process text 7614that includes @samp{%}-sequences to substitute operands, you must take 7615care of the substitution yourself. Just be sure to increment 7616@var{ptr} over whatever text should not be output normally. 7617 7618@findex recog_data.operand 7619If you need to look at the operand values, they can be found as the 7620elements of @code{recog_data.operand}. 7621 7622If the macro definition does nothing, the instruction is output 7623in the usual way. 7624@end defmac 7625 7626@defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) 7627If defined, a C statement to be executed just prior to the output of 7628assembler code for @var{insn}, to modify the extracted operands so 7629they will be output differently. 7630 7631Here the argument @var{opvec} is the vector containing the operands 7632extracted from @var{insn}, and @var{noperands} is the number of 7633elements of the vector which contain meaningful data for this insn. 7634The contents of this vector are what will be used to convert the insn 7635template into assembler code, so you can change the assembler output 7636by changing the contents of the vector. 7637 7638This macro is useful when various assembler syntaxes share a single 7639file of instruction patterns; by defining this macro differently, you 7640can cause a large class of instructions to be output differently (such 7641as with rearranged operands). Naturally, variations in assembler 7642syntax affecting individual insn patterns ought to be handled by 7643writing conditional output routines in those patterns. 7644 7645If this macro is not defined, it is equivalent to a null statement. 7646@end defmac 7647 7648@defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) 7649A C compound statement to output to stdio stream @var{stream} the 7650assembler syntax for an instruction operand @var{x}. @var{x} is an 7651RTL expression. 7652 7653@var{code} is a value that can be used to specify one of several ways 7654of printing the operand. It is used when identical operands must be 7655printed differently depending on the context. @var{code} comes from 7656the @samp{%} specification that was used to request printing of the 7657operand. If the specification was just @samp{%@var{digit}} then 7658@var{code} is 0; if the specification was @samp{%@var{ltr} 7659@var{digit}} then @var{code} is the ASCII code for @var{ltr}. 7660 7661@findex reg_names 7662If @var{x} is a register, this macro should print the register's name. 7663The names can be found in an array @code{reg_names} whose type is 7664@code{char *[]}. @code{reg_names} is initialized from 7665@code{REGISTER_NAMES}. 7666 7667When the machine description has a specification @samp{%@var{punct}} 7668(a @samp{%} followed by a punctuation character), this macro is called 7669with a null pointer for @var{x} and the punctuation character for 7670@var{code}. 7671@end defmac 7672 7673@defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code}) 7674A C expression which evaluates to true if @var{code} is a valid 7675punctuation character for use in the @code{PRINT_OPERAND} macro. If 7676@code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no 7677punctuation characters (except for the standard one, @samp{%}) are used 7678in this way. 7679@end defmac 7680 7681@defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) 7682A C compound statement to output to stdio stream @var{stream} the 7683assembler syntax for an instruction operand that is a memory reference 7684whose address is @var{x}. @var{x} is an RTL expression. 7685 7686@cindex @code{TARGET_ENCODE_SECTION_INFO} usage 7687On some machines, the syntax for a symbolic address depends on the 7688section that the address refers to. On these machines, define the hook 7689@code{TARGET_ENCODE_SECTION_INFO} to store the information into the 7690@code{symbol_ref}, and then check for it here. @xref{Assembler 7691Format}. 7692@end defmac 7693 7694@findex dbr_sequence_length 7695@defmac DBR_OUTPUT_SEQEND (@var{file}) 7696A C statement, to be executed after all slot-filler instructions have 7697been output. If necessary, call @code{dbr_sequence_length} to 7698determine the number of slots filled in a sequence (zero if not 7699currently outputting a sequence), to decide how many no-ops to output, 7700or whatever. 7701 7702Don't define this macro if it has nothing to do, but it is helpful in 7703reading assembly output if the extent of the delay sequence is made 7704explicit (e.g.@: with white space). 7705@end defmac 7706 7707@findex final_sequence 7708Note that output routines for instructions with delay slots must be 7709prepared to deal with not being output as part of a sequence 7710(i.e.@: when the scheduling pass is not run, or when no slot fillers could be 7711found.) The variable @code{final_sequence} is null when not 7712processing a sequence, otherwise it contains the @code{sequence} rtx 7713being output. 7714 7715@findex asm_fprintf 7716@defmac REGISTER_PREFIX 7717@defmacx LOCAL_LABEL_PREFIX 7718@defmacx USER_LABEL_PREFIX 7719@defmacx IMMEDIATE_PREFIX 7720If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, 7721@samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see 7722@file{final.c}). These are useful when a single @file{md} file must 7723support multiple assembler formats. In that case, the various @file{tm.h} 7724files can define these macros differently. 7725@end defmac 7726 7727@defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format}) 7728If defined this macro should expand to a series of @code{case} 7729statements which will be parsed inside the @code{switch} statement of 7730the @code{asm_fprintf} function. This allows targets to define extra 7731printf formats which may useful when generating their assembler 7732statements. Note that uppercase letters are reserved for future 7733generic extensions to asm_fprintf, and so are not available to target 7734specific code. The output file is given by the parameter @var{file}. 7735The varargs input pointer is @var{argptr} and the rest of the format 7736string, starting the character after the one that is being switched 7737upon, is pointed to by @var{format}. 7738@end defmac 7739 7740@defmac ASSEMBLER_DIALECT 7741If your target supports multiple dialects of assembler language (such as 7742different opcodes), define this macro as a C expression that gives the 7743numeric index of the assembler language dialect to use, with zero as the 7744first variant. 7745 7746If this macro is defined, you may use constructs of the form 7747@smallexample 7748@samp{@{option0|option1|option2@dots{}@}} 7749@end smallexample 7750@noindent 7751in the output templates of patterns (@pxref{Output Template}) or in the 7752first argument of @code{asm_fprintf}. This construct outputs 7753@samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of 7754@code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters 7755within these strings retain their usual meaning. If there are fewer 7756alternatives within the braces than the value of 7757@code{ASSEMBLER_DIALECT}, the construct outputs nothing. 7758 7759If you do not define this macro, the characters @samp{@{}, @samp{|} and 7760@samp{@}} do not have any special meaning when used in templates or 7761operands to @code{asm_fprintf}. 7762 7763Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, 7764@code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express 7765the variations in assembler language syntax with that mechanism. Define 7766@code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax 7767if the syntax variant are larger and involve such things as different 7768opcodes or operand order. 7769@end defmac 7770 7771@defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) 7772A C expression to output to @var{stream} some assembler code 7773which will push hard register number @var{regno} onto the stack. 7774The code need not be optimal, since this macro is used only when 7775profiling. 7776@end defmac 7777 7778@defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) 7779A C expression to output to @var{stream} some assembler code 7780which will pop hard register number @var{regno} off of the stack. 7781The code need not be optimal, since this macro is used only when 7782profiling. 7783@end defmac 7784 7785@node Dispatch Tables 7786@subsection Output of Dispatch Tables 7787 7788@c prevent bad page break with this line 7789This concerns dispatch tables. 7790 7791@cindex dispatch table 7792@defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel}) 7793A C statement to output to the stdio stream @var{stream} an assembler 7794pseudo-instruction to generate a difference between two labels. 7795@var{value} and @var{rel} are the numbers of two internal labels. The 7796definitions of these labels are output using 7797@code{(*targetm.asm_out.internal_label)}, and they must be printed in the same 7798way here. For example, 7799 7800@smallexample 7801fprintf (@var{stream}, "\t.word L%d-L%d\n", 7802 @var{value}, @var{rel}) 7803@end smallexample 7804 7805You must provide this macro on machines where the addresses in a 7806dispatch table are relative to the table's own address. If defined, GCC 7807will also use this macro on all machines when producing PIC@. 7808@var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the 7809mode and flags can be read. 7810@end defmac 7811 7812@defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) 7813This macro should be provided on machines where the addresses 7814in a dispatch table are absolute. 7815 7816The definition should be a C statement to output to the stdio stream 7817@var{stream} an assembler pseudo-instruction to generate a reference to 7818a label. @var{value} is the number of an internal label whose 7819definition is output using @code{(*targetm.asm_out.internal_label)}. 7820For example, 7821 7822@smallexample 7823fprintf (@var{stream}, "\t.word L%d\n", @var{value}) 7824@end smallexample 7825@end defmac 7826 7827@defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) 7828Define this if the label before a jump-table needs to be output 7829specially. The first three arguments are the same as for 7830@code{(*targetm.asm_out.internal_label)}; the fourth argument is the 7831jump-table which follows (a @code{jump_insn} containing an 7832@code{addr_vec} or @code{addr_diff_vec}). 7833 7834This feature is used on system V to output a @code{swbeg} statement 7835for the table. 7836 7837If this macro is not defined, these labels are output with 7838@code{(*targetm.asm_out.internal_label)}. 7839@end defmac 7840 7841@defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) 7842Define this if something special must be output at the end of a 7843jump-table. The definition should be a C statement to be executed 7844after the assembler code for the table is written. It should write 7845the appropriate code to stdio stream @var{stream}. The argument 7846@var{table} is the jump-table insn, and @var{num} is the label-number 7847of the preceding label. 7848 7849If this macro is not defined, nothing special is output at the end of 7850the jump-table. 7851@end defmac 7852 7853@deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty}) 7854This target hook emits a label at the beginning of each FDE@. It 7855should be defined on targets where FDEs need special labels, and it 7856should write the appropriate label, for the FDE associated with the 7857function declaration @var{decl}, to the stdio stream @var{stream}. 7858The third argument, @var{for_eh}, is a boolean: true if this is for an 7859exception table. The fourth argument, @var{empty}, is a boolean: 7860true if this is a placeholder label for an omitted FDE@. 7861 7862The default is that FDEs are not given nonlocal labels. 7863@end deftypefn 7864 7865@deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream}) 7866This target hook emits a label at the beginning of the exception table. 7867It should be defined on targets where it is desirable for the table 7868to be broken up according to function. 7869 7870The default is that no label is emitted. 7871@end deftypefn 7872 7873@deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn}) 7874This target hook emits and assembly directives required to unwind the 7875given instruction. This is only used when TARGET_UNWIND_INFO is set. 7876@end deftypefn 7877 7878@node Exception Region Output 7879@subsection Assembler Commands for Exception Regions 7880 7881@c prevent bad page break with this line 7882 7883This describes commands marking the start and the end of an exception 7884region. 7885 7886@defmac EH_FRAME_SECTION_NAME 7887If defined, a C string constant for the name of the section containing 7888exception handling frame unwind information. If not defined, GCC will 7889provide a default definition if the target supports named sections. 7890@file{crtstuff.c} uses this macro to switch to the appropriate section. 7891 7892You should define this symbol if your target supports DWARF 2 frame 7893unwind information and the default definition does not work. 7894@end defmac 7895 7896@defmac EH_FRAME_IN_DATA_SECTION 7897If defined, DWARF 2 frame unwind information will be placed in the 7898data section even though the target supports named sections. This 7899might be necessary, for instance, if the system linker does garbage 7900collection and sections cannot be marked as not to be collected. 7901 7902Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is 7903also defined. 7904@end defmac 7905 7906@defmac EH_TABLES_CAN_BE_READ_ONLY 7907Define this macro to 1 if your target is such that no frame unwind 7908information encoding used with non-PIC code will ever require a 7909runtime relocation, but the linker may not support merging read-only 7910and read-write sections into a single read-write section. 7911@end defmac 7912 7913@defmac MASK_RETURN_ADDR 7914An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so 7915that it does not contain any extraneous set bits in it. 7916@end defmac 7917 7918@defmac DWARF2_UNWIND_INFO 7919Define this macro to 0 if your target supports DWARF 2 frame unwind 7920information, but it does not yet work with exception handling. 7921Otherwise, if your target supports this information (if it defines 7922@samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP} 7923or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1. 7924 7925If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder 7926will be used in all cases. Defining this macro will enable the generation 7927of DWARF 2 frame debugging information. 7928 7929If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1, 7930the DWARF 2 unwinder will be the default exception handling mechanism; 7931otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by 7932default. 7933@end defmac 7934 7935@defmac TARGET_UNWIND_INFO 7936Define this macro if your target has ABI specified unwind tables. Usually 7937these will be output by @code{TARGET_UNWIND_EMIT}. 7938@end defmac 7939 7940@deftypevar {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT 7941This variable should be set to @code{true} if the target ABI requires unwinding 7942tables even when exceptions are not used. 7943@end deftypevar 7944 7945@defmac MUST_USE_SJLJ_EXCEPTIONS 7946This macro need only be defined if @code{DWARF2_UNWIND_INFO} is 7947runtime-variable. In that case, @file{except.h} cannot correctly 7948determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS}, 7949so the target must provide it directly. 7950@end defmac 7951 7952@defmac DONT_USE_BUILTIN_SETJMP 7953Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme 7954should use the @code{setjmp}/@code{longjmp} functions from the C library 7955instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery. 7956@end defmac 7957 7958@defmac DWARF_CIE_DATA_ALIGNMENT 7959This macro need only be defined if the target might save registers in the 7960function prologue at an offset to the stack pointer that is not aligned to 7961@code{UNITS_PER_WORD}. The definition should be the negative minimum 7962alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive 7963minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if 7964the target supports DWARF 2 frame unwind information. 7965@end defmac 7966 7967@deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO 7968Contains the value true if the target should add a zero word onto the 7969end of a Dwarf-2 frame info section when used for exception handling. 7970Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and 7971true otherwise. 7972@end deftypevar 7973 7974@deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg}) 7975Given a register, this hook should return a parallel of registers to 7976represent where to find the register pieces. Define this hook if the 7977register and its mode are represented in Dwarf in non-contiguous 7978locations, or if the register should be represented in more than one 7979register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}. 7980If not defined, the default is to return @code{NULL_RTX}. 7981@end deftypefn 7982 7983@deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym}) 7984This hook is used to output a reference from a frame unwinding table to 7985the type_info object identified by @var{sym}. It should return @code{true} 7986if the reference was output. Returning @code{false} will cause the 7987reference to be output using the normal Dwarf2 routines. 7988@end deftypefn 7989 7990@deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER 7991This hook should be set to @code{true} on targets that use an ARM EABI 7992based unwinding library, and @code{false} on other targets. This effects 7993the format of unwinding tables, and how the unwinder in entered after 7994running a cleanup. The default is @code{false}. 7995@end deftypefn 7996 7997@node Alignment Output 7998@subsection Assembler Commands for Alignment 7999 8000@c prevent bad page break with this line 8001This describes commands for alignment. 8002 8003@defmac JUMP_ALIGN (@var{label}) 8004The alignment (log base 2) to put in front of @var{label}, which is 8005a common destination of jumps and has no fallthru incoming edge. 8006 8007This macro need not be defined if you don't want any special alignment 8008to be done at such a time. Most machine descriptions do not currently 8009define the macro. 8010 8011Unless it's necessary to inspect the @var{label} parameter, it is better 8012to set the variable @var{align_jumps} in the target's 8013@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 8014selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation. 8015@end defmac 8016 8017@defmac LABEL_ALIGN_AFTER_BARRIER (@var{label}) 8018The alignment (log base 2) to put in front of @var{label}, which follows 8019a @code{BARRIER}. 8020 8021This macro need not be defined if you don't want any special alignment 8022to be done at such a time. Most machine descriptions do not currently 8023define the macro. 8024@end defmac 8025 8026@defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP 8027The maximum number of bytes to skip when applying 8028@code{LABEL_ALIGN_AFTER_BARRIER}. This works only if 8029@code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 8030@end defmac 8031 8032@defmac LOOP_ALIGN (@var{label}) 8033The alignment (log base 2) to put in front of @var{label}, which follows 8034a @code{NOTE_INSN_LOOP_BEG} note. 8035 8036This macro need not be defined if you don't want any special alignment 8037to be done at such a time. Most machine descriptions do not currently 8038define the macro. 8039 8040Unless it's necessary to inspect the @var{label} parameter, it is better 8041to set the variable @code{align_loops} in the target's 8042@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 8043selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation. 8044@end defmac 8045 8046@defmac LOOP_ALIGN_MAX_SKIP 8047The maximum number of bytes to skip when applying @code{LOOP_ALIGN}. 8048This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 8049@end defmac 8050 8051@defmac LABEL_ALIGN (@var{label}) 8052The alignment (log base 2) to put in front of @var{label}. 8053If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment, 8054the maximum of the specified values is used. 8055 8056Unless it's necessary to inspect the @var{label} parameter, it is better 8057to set the variable @code{align_labels} in the target's 8058@code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's 8059selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation. 8060@end defmac 8061 8062@defmac LABEL_ALIGN_MAX_SKIP 8063The maximum number of bytes to skip when applying @code{LABEL_ALIGN}. 8064This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. 8065@end defmac 8066 8067@defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) 8068A C statement to output to the stdio stream @var{stream} an assembler 8069instruction to advance the location counter by @var{nbytes} bytes. 8070Those bytes should be zero when loaded. @var{nbytes} will be a C 8071expression of type @code{int}. 8072@end defmac 8073 8074@defmac ASM_NO_SKIP_IN_TEXT 8075Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the 8076text section because it fails to put zeros in the bytes that are skipped. 8077This is true on many Unix systems, where the pseudo--op to skip bytes 8078produces no-op instructions rather than zeros when used in the text 8079section. 8080@end defmac 8081 8082@defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) 8083A C statement to output to the stdio stream @var{stream} an assembler 8084command to advance the location counter to a multiple of 2 to the 8085@var{power} bytes. @var{power} will be a C expression of type @code{int}. 8086@end defmac 8087 8088@defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power}) 8089Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used 8090for padding, if necessary. 8091@end defmac 8092 8093@defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip}) 8094A C statement to output to the stdio stream @var{stream} an assembler 8095command to advance the location counter to a multiple of 2 to the 8096@var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to 8097satisfy the alignment request. @var{power} and @var{max_skip} will be 8098a C expression of type @code{int}. 8099@end defmac 8100 8101@need 3000 8102@node Debugging Info 8103@section Controlling Debugging Information Format 8104 8105@c prevent bad page break with this line 8106This describes how to specify debugging information. 8107 8108@menu 8109* All Debuggers:: Macros that affect all debugging formats uniformly. 8110* DBX Options:: Macros enabling specific options in DBX format. 8111* DBX Hooks:: Hook macros for varying DBX format. 8112* File Names and DBX:: Macros controlling output of file names in DBX format. 8113* SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. 8114* VMS Debug:: Macros for VMS debug format. 8115@end menu 8116 8117@node All Debuggers 8118@subsection Macros Affecting All Debugging Formats 8119 8120@c prevent bad page break with this line 8121These macros affect all debugging formats. 8122 8123@defmac DBX_REGISTER_NUMBER (@var{regno}) 8124A C expression that returns the DBX register number for the compiler 8125register number @var{regno}. In the default macro provided, the value 8126of this expression will be @var{regno} itself. But sometimes there are 8127some registers that the compiler knows about and DBX does not, or vice 8128versa. In such cases, some register may need to have one number in the 8129compiler and another for DBX@. 8130 8131If two registers have consecutive numbers inside GCC, and they can be 8132used as a pair to hold a multiword value, then they @emph{must} have 8133consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}. 8134Otherwise, debuggers will be unable to access such a pair, because they 8135expect register pairs to be consecutive in their own numbering scheme. 8136 8137If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that 8138does not preserve register pairs, then what you must do instead is 8139redefine the actual register numbering scheme. 8140@end defmac 8141 8142@defmac DEBUGGER_AUTO_OFFSET (@var{x}) 8143A C expression that returns the integer offset value for an automatic 8144variable having address @var{x} (an RTL expression). The default 8145computation assumes that @var{x} is based on the frame-pointer and 8146gives the offset from the frame-pointer. This is required for targets 8147that produce debugging output for DBX or COFF-style debugging output 8148for SDB and allow the frame-pointer to be eliminated when the 8149@option{-g} options is used. 8150@end defmac 8151 8152@defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) 8153A C expression that returns the integer offset value for an argument 8154having address @var{x} (an RTL expression). The nominal offset is 8155@var{offset}. 8156@end defmac 8157 8158@defmac PREFERRED_DEBUGGING_TYPE 8159A C expression that returns the type of debugging output GCC should 8160produce when the user specifies just @option{-g}. Define 8161this if you have arranged for GCC to support more than one format of 8162debugging output. Currently, the allowable values are @code{DBX_DEBUG}, 8163@code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, 8164@code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}. 8165 8166When the user specifies @option{-ggdb}, GCC normally also uses the 8167value of this macro to select the debugging output format, but with two 8168exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the 8169value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is 8170defined, GCC uses @code{DBX_DEBUG}. 8171 8172The value of this macro only affects the default debugging output; the 8173user can always get a specific type of output by using @option{-gstabs}, 8174@option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}. 8175@end defmac 8176 8177@node DBX Options 8178@subsection Specific Options for DBX Output 8179 8180@c prevent bad page break with this line 8181These are specific options for DBX output. 8182 8183@defmac DBX_DEBUGGING_INFO 8184Define this macro if GCC should produce debugging output for DBX 8185in response to the @option{-g} option. 8186@end defmac 8187 8188@defmac XCOFF_DEBUGGING_INFO 8189Define this macro if GCC should produce XCOFF format debugging output 8190in response to the @option{-g} option. This is a variant of DBX format. 8191@end defmac 8192 8193@defmac DEFAULT_GDB_EXTENSIONS 8194Define this macro to control whether GCC should by default generate 8195GDB's extended version of DBX debugging information (assuming DBX-format 8196debugging information is enabled at all). If you don't define the 8197macro, the default is 1: always generate the extended information 8198if there is any occasion to. 8199@end defmac 8200 8201@defmac DEBUG_SYMS_TEXT 8202Define this macro if all @code{.stabs} commands should be output while 8203in the text section. 8204@end defmac 8205 8206@defmac ASM_STABS_OP 8207A C string constant, including spacing, naming the assembler pseudo op to 8208use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol. 8209If you don't define this macro, @code{"\t.stabs\t"} is used. This macro 8210applies only to DBX debugging information format. 8211@end defmac 8212 8213@defmac ASM_STABD_OP 8214A C string constant, including spacing, naming the assembler pseudo op to 8215use instead of @code{"\t.stabd\t"} to define a debugging symbol whose 8216value is the current location. If you don't define this macro, 8217@code{"\t.stabd\t"} is used. This macro applies only to DBX debugging 8218information format. 8219@end defmac 8220 8221@defmac ASM_STABN_OP 8222A C string constant, including spacing, naming the assembler pseudo op to 8223use instead of @code{"\t.stabn\t"} to define a debugging symbol with no 8224name. If you don't define this macro, @code{"\t.stabn\t"} is used. This 8225macro applies only to DBX debugging information format. 8226@end defmac 8227 8228@defmac DBX_NO_XREFS 8229Define this macro if DBX on your system does not support the construct 8230@samp{xs@var{tagname}}. On some systems, this construct is used to 8231describe a forward reference to a structure named @var{tagname}. 8232On other systems, this construct is not supported at all. 8233@end defmac 8234 8235@defmac DBX_CONTIN_LENGTH 8236A symbol name in DBX-format debugging information is normally 8237continued (split into two separate @code{.stabs} directives) when it 8238exceeds a certain length (by default, 80 characters). On some 8239operating systems, DBX requires this splitting; on others, splitting 8240must not be done. You can inhibit splitting by defining this macro 8241with the value zero. You can override the default splitting-length by 8242defining this macro as an expression for the length you desire. 8243@end defmac 8244 8245@defmac DBX_CONTIN_CHAR 8246Normally continuation is indicated by adding a @samp{\} character to 8247the end of a @code{.stabs} string when a continuation follows. To use 8248a different character instead, define this macro as a character 8249constant for the character you want to use. Do not define this macro 8250if backslash is correct for your system. 8251@end defmac 8252 8253@defmac DBX_STATIC_STAB_DATA_SECTION 8254Define this macro if it is necessary to go to the data section before 8255outputting the @samp{.stabs} pseudo-op for a non-global static 8256variable. 8257@end defmac 8258 8259@defmac DBX_TYPE_DECL_STABS_CODE 8260The value to use in the ``code'' field of the @code{.stabs} directive 8261for a typedef. The default is @code{N_LSYM}. 8262@end defmac 8263 8264@defmac DBX_STATIC_CONST_VAR_CODE 8265The value to use in the ``code'' field of the @code{.stabs} directive 8266for a static variable located in the text section. DBX format does not 8267provide any ``right'' way to do this. The default is @code{N_FUN}. 8268@end defmac 8269 8270@defmac DBX_REGPARM_STABS_CODE 8271The value to use in the ``code'' field of the @code{.stabs} directive 8272for a parameter passed in registers. DBX format does not provide any 8273``right'' way to do this. The default is @code{N_RSYM}. 8274@end defmac 8275 8276@defmac DBX_REGPARM_STABS_LETTER 8277The letter to use in DBX symbol data to identify a symbol as a parameter 8278passed in registers. DBX format does not customarily provide any way to 8279do this. The default is @code{'P'}. 8280@end defmac 8281 8282@defmac DBX_FUNCTION_FIRST 8283Define this macro if the DBX information for a function and its 8284arguments should precede the assembler code for the function. Normally, 8285in DBX format, the debugging information entirely follows the assembler 8286code. 8287@end defmac 8288 8289@defmac DBX_BLOCKS_FUNCTION_RELATIVE 8290Define this macro, with value 1, if the value of a symbol describing 8291the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be 8292relative to the start of the enclosing function. Normally, GCC uses 8293an absolute address. 8294@end defmac 8295 8296@defmac DBX_LINES_FUNCTION_RELATIVE 8297Define this macro, with value 1, if the value of a symbol indicating 8298the current line number (@code{N_SLINE}) should be relative to the 8299start of the enclosing function. Normally, GCC uses an absolute address. 8300@end defmac 8301 8302@defmac DBX_USE_BINCL 8303Define this macro if GCC should generate @code{N_BINCL} and 8304@code{N_EINCL} stabs for included header files, as on Sun systems. This 8305macro also directs GCC to output a type number as a pair of a file 8306number and a type number within the file. Normally, GCC does not 8307generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single 8308number for a type number. 8309@end defmac 8310 8311@node DBX Hooks 8312@subsection Open-Ended Hooks for DBX Format 8313 8314@c prevent bad page break with this line 8315These are hooks for DBX format. 8316 8317@defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name}) 8318Define this macro to say how to output to @var{stream} the debugging 8319information for the start of a scope level for variable names. The 8320argument @var{name} is the name of an assembler symbol (for use with 8321@code{assemble_name}) whose value is the address where the scope begins. 8322@end defmac 8323 8324@defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name}) 8325Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level. 8326@end defmac 8327 8328@defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl}) 8329Define this macro if the target machine requires special handling to 8330output an @code{N_FUN} entry for the function @var{decl}. 8331@end defmac 8332 8333@defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter}) 8334A C statement to output DBX debugging information before code for line 8335number @var{line} of the current source file to the stdio stream 8336@var{stream}. @var{counter} is the number of time the macro was 8337invoked, including the current invocation; it is intended to generate 8338unique labels in the assembly output. 8339 8340This macro should not be defined if the default output is correct, or 8341if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}. 8342@end defmac 8343 8344@defmac NO_DBX_FUNCTION_END 8345Some stabs encapsulation formats (in particular ECOFF), cannot handle the 8346@code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct. 8347On those machines, define this macro to turn this feature off without 8348disturbing the rest of the gdb extensions. 8349@end defmac 8350 8351@defmac NO_DBX_BNSYM_ENSYM 8352Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx 8353extension construct. On those machines, define this macro to turn this 8354feature off without disturbing the rest of the gdb extensions. 8355@end defmac 8356 8357@node File Names and DBX 8358@subsection File Names in DBX Format 8359 8360@c prevent bad page break with this line 8361This describes file names in DBX format. 8362 8363@defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name}) 8364A C statement to output DBX debugging information to the stdio stream 8365@var{stream}, which indicates that file @var{name} is the main source 8366file---the file specified as the input file for compilation. 8367This macro is called only once, at the beginning of compilation. 8368 8369This macro need not be defined if the standard form of output 8370for DBX debugging information is appropriate. 8371 8372It may be necessary to refer to a label equal to the beginning of the 8373text section. You can use @samp{assemble_name (stream, ltext_label_name)} 8374to do so. If you do this, you must also set the variable 8375@var{used_ltext_label_name} to @code{true}. 8376@end defmac 8377 8378@defmac NO_DBX_MAIN_SOURCE_DIRECTORY 8379Define this macro, with value 1, if GCC should not emit an indication 8380of the current directory for compilation and current source language at 8381the beginning of the file. 8382@end defmac 8383 8384@defmac NO_DBX_GCC_MARKER 8385Define this macro, with value 1, if GCC should not emit an indication 8386that this object file was compiled by GCC@. The default is to emit 8387an @code{N_OPT} stab at the beginning of every source file, with 8388@samp{gcc2_compiled.} for the string and value 0. 8389@end defmac 8390 8391@defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name}) 8392A C statement to output DBX debugging information at the end of 8393compilation of the main source file @var{name}. Output should be 8394written to the stdio stream @var{stream}. 8395 8396If you don't define this macro, nothing special is output at the end 8397of compilation, which is correct for most machines. 8398@end defmac 8399 8400@defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END 8401Define this macro @emph{instead of} defining 8402@code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at 8403the end of compilation is a @code{N_SO} stab with an empty string, 8404whose value is the highest absolute text address in the file. 8405@end defmac 8406 8407@need 2000 8408@node SDB and DWARF 8409@subsection Macros for SDB and DWARF Output 8410 8411@c prevent bad page break with this line 8412Here are macros for SDB and DWARF output. 8413 8414@defmac SDB_DEBUGGING_INFO 8415Define this macro if GCC should produce COFF-style debugging output 8416for SDB in response to the @option{-g} option. 8417@end defmac 8418 8419@defmac DWARF2_DEBUGGING_INFO 8420Define this macro if GCC should produce dwarf version 2 format 8421debugging output in response to the @option{-g} option. 8422 8423@deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function}) 8424Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to 8425be emitted for each function. Instead of an integer return the enum 8426value for the @code{DW_CC_} tag. 8427@end deftypefn 8428 8429To support optional call frame debugging information, you must also 8430define @code{INCOMING_RETURN_ADDR_RTX} and either set 8431@code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the 8432prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save} 8433as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't. 8434@end defmac 8435 8436@defmac DWARF2_FRAME_INFO 8437Define this macro to a nonzero value if GCC should always output 8438Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO} 8439(@pxref{Exception Region Output} is nonzero, GCC will output this 8440information not matter how you define @code{DWARF2_FRAME_INFO}. 8441@end defmac 8442 8443@defmac DWARF2_ASM_LINE_DEBUG_INFO 8444Define this macro to be a nonzero value if the assembler can generate Dwarf 2 8445line debug info sections. This will result in much more compact line number 8446tables, and hence is desirable if it works. 8447@end defmac 8448 8449@defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) 8450A C statement to issue assembly directives that create a difference 8451@var{lab1} minus @var{lab2}, using an integer of the given @var{size}. 8452@end defmac 8453 8454@defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section}) 8455A C statement to issue assembly directives that create a 8456section-relative reference to the given @var{label}, using an integer of the 8457given @var{size}. The label is known to be defined in the given @var{section}. 8458@end defmac 8459 8460@defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label}) 8461A C statement to issue assembly directives that create a self-relative 8462reference to the given @var{label}, using an integer of the given @var{size}. 8463@end defmac 8464 8465@deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x}) 8466If defined, this target hook is a function which outputs a DTP-relative 8467reference to the given TLS symbol of the specified size. 8468@end deftypefn 8469 8470@defmac PUT_SDB_@dots{} 8471Define these macros to override the assembler syntax for the special 8472SDB assembler directives. See @file{sdbout.c} for a list of these 8473macros and their arguments. If the standard syntax is used, you need 8474not define them yourself. 8475@end defmac 8476 8477@defmac SDB_DELIM 8478Some assemblers do not support a semicolon as a delimiter, even between 8479SDB assembler directives. In that case, define this macro to be the 8480delimiter to use (usually @samp{\n}). It is not necessary to define 8481a new set of @code{PUT_SDB_@var{op}} macros if this is the only change 8482required. 8483@end defmac 8484 8485@defmac SDB_ALLOW_UNKNOWN_REFERENCES 8486Define this macro to allow references to unknown structure, 8487union, or enumeration tags to be emitted. Standard COFF does not 8488allow handling of unknown references, MIPS ECOFF has support for 8489it. 8490@end defmac 8491 8492@defmac SDB_ALLOW_FORWARD_REFERENCES 8493Define this macro to allow references to structure, union, or 8494enumeration tags that have not yet been seen to be handled. Some 8495assemblers choke if forward tags are used, while some require it. 8496@end defmac 8497 8498@defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}) 8499A C statement to output SDB debugging information before code for line 8500number @var{line} of the current source file to the stdio stream 8501@var{stream}. The default is to emit an @code{.ln} directive. 8502@end defmac 8503 8504@need 2000 8505@node VMS Debug 8506@subsection Macros for VMS Debug Format 8507 8508@c prevent bad page break with this line 8509Here are macros for VMS debug format. 8510 8511@defmac VMS_DEBUGGING_INFO 8512Define this macro if GCC should produce debugging output for VMS 8513in response to the @option{-g} option. The default behavior for VMS 8514is to generate minimal debug info for a traceback in the absence of 8515@option{-g} unless explicitly overridden with @option{-g0}. This 8516behavior is controlled by @code{OPTIMIZATION_OPTIONS} and 8517@code{OVERRIDE_OPTIONS}. 8518@end defmac 8519 8520@node Floating Point 8521@section Cross Compilation and Floating Point 8522@cindex cross compilation and floating point 8523@cindex floating point and cross compilation 8524 8525While all modern machines use twos-complement representation for integers, 8526there are a variety of representations for floating point numbers. This 8527means that in a cross-compiler the representation of floating point numbers 8528in the compiled program may be different from that used in the machine 8529doing the compilation. 8530 8531Because different representation systems may offer different amounts of 8532range and precision, all floating point constants must be represented in 8533the target machine's format. Therefore, the cross compiler cannot 8534safely use the host machine's floating point arithmetic; it must emulate 8535the target's arithmetic. To ensure consistency, GCC always uses 8536emulation to work with floating point values, even when the host and 8537target floating point formats are identical. 8538 8539The following macros are provided by @file{real.h} for the compiler to 8540use. All parts of the compiler which generate or optimize 8541floating-point calculations must use these macros. They may evaluate 8542their operands more than once, so operands must not have side effects. 8543 8544@defmac REAL_VALUE_TYPE 8545The C data type to be used to hold a floating point value in the target 8546machine's format. Typically this is a @code{struct} containing an 8547array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque 8548quantity. 8549@end defmac 8550 8551@deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 8552Compares for equality the two values, @var{x} and @var{y}. If the target 8553floating point format supports negative zeroes and/or NaNs, 8554@samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and 8555@samp{REAL_VALUES_EQUAL (NaN, NaN)} is false. 8556@end deftypefn 8557 8558@deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) 8559Tests whether @var{x} is less than @var{y}. 8560@end deftypefn 8561 8562@deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x}) 8563Truncates @var{x} to a signed integer, rounding toward zero. 8564@end deftypefn 8565 8566@deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x}) 8567Truncates @var{x} to an unsigned integer, rounding toward zero. If 8568@var{x} is negative, returns zero. 8569@end deftypefn 8570 8571@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode}) 8572Converts @var{string} into a floating point number in the target machine's 8573representation for mode @var{mode}. This routine can handle both 8574decimal and hexadecimal floating point constants, using the syntax 8575defined by the C language for both. 8576@end deftypefn 8577 8578@deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x}) 8579Returns 1 if @var{x} is negative (including negative zero), 0 otherwise. 8580@end deftypefn 8581 8582@deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x}) 8583Determines whether @var{x} represents infinity (positive or negative). 8584@end deftypefn 8585 8586@deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x}) 8587Determines whether @var{x} represents a ``NaN'' (not-a-number). 8588@end deftypefn 8589 8590@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}) 8591Calculates an arithmetic operation on the two floating point values 8592@var{x} and @var{y}, storing the result in @var{output} (which must be a 8593variable). 8594 8595The operation to be performed is specified by @var{code}. Only the 8596following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR}, 8597@code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}. 8598 8599If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the 8600target's floating point format cannot represent infinity, it will call 8601@code{abort}. Callers should check for this situation first, using 8602@code{MODE_HAS_INFINITIES}. @xref{Storage Layout}. 8603@end deftypefn 8604 8605@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x}) 8606Returns the negative of the floating point value @var{x}. 8607@end deftypefn 8608 8609@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x}) 8610Returns the absolute value of @var{x}. 8611@end deftypefn 8612 8613@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x}) 8614Truncates the floating point value @var{x} to fit in @var{mode}. The 8615return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an 8616appropriate bit pattern to be output asa floating constant whose 8617precision accords with mode @var{mode}. 8618@end deftypefn 8619 8620@deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x}) 8621Converts a floating point value @var{x} into a double-precision integer 8622which is then stored into @var{low} and @var{high}. If the value is not 8623integral, it is truncated. 8624@end deftypefn 8625 8626@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}) 8627Converts a double-precision integer found in @var{low} and @var{high}, 8628into a floating point value which is then stored into @var{x}. The 8629value is truncated to fit in mode @var{mode}. 8630@end deftypefn 8631 8632@node Mode Switching 8633@section Mode Switching Instructions 8634@cindex mode switching 8635The following macros control mode switching optimizations: 8636 8637@defmac OPTIMIZE_MODE_SWITCHING (@var{entity}) 8638Define this macro if the port needs extra instructions inserted for mode 8639switching in an optimizing compilation. 8640 8641For an example, the SH4 can perform both single and double precision 8642floating point operations, but to perform a single precision operation, 8643the FPSCR PR bit has to be cleared, while for a double precision 8644operation, this bit has to be set. Changing the PR bit requires a general 8645purpose register as a scratch register, hence these FPSCR sets have to 8646be inserted before reload, i.e.@: you can't put this into instruction emitting 8647or @code{TARGET_MACHINE_DEPENDENT_REORG}. 8648 8649You can have multiple entities that are mode-switched, and select at run time 8650which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should 8651return nonzero for any @var{entity} that needs mode-switching. 8652If you define this macro, you also have to define 8653@code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED}, 8654@code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}. 8655@code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT} 8656are optional. 8657@end defmac 8658 8659@defmac NUM_MODES_FOR_MODE_SWITCHING 8660If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as 8661initializer for an array of integers. Each initializer element 8662N refers to an entity that needs mode switching, and specifies the number 8663of different modes that might need to be set for this entity. 8664The position of the initializer in the initializer---starting counting at 8665zero---determines the integer that is used to refer to the mode-switched 8666entity in question. 8667In macros that take mode arguments / yield a mode result, modes are 8668represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode 8669switch is needed / supplied. 8670@end defmac 8671 8672@defmac MODE_NEEDED (@var{entity}, @var{insn}) 8673@var{entity} is an integer specifying a mode-switched entity. If 8674@code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to 8675return an integer value not larger than the corresponding element in 8676@code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must 8677be switched into prior to the execution of @var{insn}. 8678@end defmac 8679 8680@defmac MODE_AFTER (@var{mode}, @var{insn}) 8681If this macro is defined, it is evaluated for every @var{insn} during 8682mode switching. It determines the mode that an insn results in (if 8683different from the incoming mode). 8684@end defmac 8685 8686@defmac MODE_ENTRY (@var{entity}) 8687If this macro is defined, it is evaluated for every @var{entity} that needs 8688mode switching. It should evaluate to an integer, which is a mode that 8689@var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY} 8690is defined then @code{MODE_EXIT} must be defined. 8691@end defmac 8692 8693@defmac MODE_EXIT (@var{entity}) 8694If this macro is defined, it is evaluated for every @var{entity} that needs 8695mode switching. It should evaluate to an integer, which is a mode that 8696@var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT} 8697is defined then @code{MODE_ENTRY} must be defined. 8698@end defmac 8699 8700@defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n}) 8701This macro specifies the order in which modes for @var{entity} are processed. 87020 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the 8703lowest. The value of the macro should be an integer designating a mode 8704for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode} 8705(@var{entity}, @var{n}) shall be a bijection in 0 @dots{} 8706@code{num_modes_for_mode_switching[@var{entity}] - 1}. 8707@end defmac 8708 8709@defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live}) 8710Generate one or more insns to set @var{entity} to @var{mode}. 8711@var{hard_reg_live} is the set of hard registers live at the point where 8712the insn(s) are to be inserted. 8713@end defmac 8714 8715@node Target Attributes 8716@section Defining target-specific uses of @code{__attribute__} 8717@cindex target attributes 8718@cindex machine attributes 8719@cindex attributes, target-specific 8720 8721Target-specific attributes may be defined for functions, data and types. 8722These are described using the following target hooks; they also need to 8723be documented in @file{extend.texi}. 8724 8725@deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE 8726If defined, this target hook points to an array of @samp{struct 8727attribute_spec} (defined in @file{tree.h}) specifying the machine 8728specific attributes for this target and some of the restrictions on the 8729entities to which these attributes are applied and the arguments they 8730take. 8731@end deftypevr 8732 8733@deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) 8734If defined, this target hook is a function which returns zero if the attributes on 8735@var{type1} and @var{type2} are incompatible, one if they are compatible, 8736and two if they are nearly compatible (which causes a warning to be 8737generated). If this is not defined, machine-specific attributes are 8738supposed always to be compatible. 8739@end deftypefn 8740 8741@deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type}) 8742If defined, this target hook is a function which assigns default attributes to 8743newly defined @var{type}. 8744@end deftypefn 8745 8746@deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) 8747Define this target hook if the merging of type attributes needs special 8748handling. If defined, the result is a list of the combined 8749@code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed 8750that @code{comptypes} has already been called and returned 1. This 8751function may call @code{merge_attributes} to handle machine-independent 8752merging. 8753@end deftypefn 8754 8755@deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl}) 8756Define this target hook if the merging of decl attributes needs special 8757handling. If defined, the result is a list of the combined 8758@code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}. 8759@var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of 8760when this is needed are when one attribute overrides another, or when an 8761attribute is nullified by a subsequent definition. This function may 8762call @code{merge_attributes} to handle machine-independent merging. 8763 8764@findex TARGET_DLLIMPORT_DECL_ATTRIBUTES 8765If the only target-specific handling you require is @samp{dllimport} 8766for Microsoft Windows targets, you should define the macro 8767@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler 8768will then define a function called 8769@code{merge_dllimport_decl_attributes} which can then be defined as 8770the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also 8771add @code{handle_dll_attribute} in the attribute table for your port 8772to perform initial processing of the @samp{dllimport} and 8773@samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and 8774@file{i386/i386.c}, for example. 8775@end deftypefn 8776 8777@deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl}) 8778@var{decl} is a variable or function with @code{__attribute__((dllimport))} 8779specified. Use this hook if the target needs to add extra validation 8780checks to @code{handle_dll_attribute}. 8781@end deftypefn 8782 8783@defmac TARGET_DECLSPEC 8784Define this macro to a nonzero value if you want to treat 8785@code{__declspec(X)} as equivalent to @code{__attribute((X))}. By 8786default, this behavior is enabled only for targets that define 8787@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation 8788of @code{__declspec} is via a built-in macro, but you should not rely 8789on this implementation detail. 8790@end defmac 8791 8792@deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr}) 8793Define this target hook if you want to be able to add attributes to a decl 8794when it is being created. This is normally useful for back ends which 8795wish to implement a pragma by using the attributes which correspond to 8796the pragma's effect. The @var{node} argument is the decl which is being 8797created. The @var{attr_ptr} argument is a pointer to the attribute list 8798for this decl. The list itself should not be modified, since it may be 8799shared with other decls, but attributes may be chained on the head of 8800the list and @code{*@var{attr_ptr}} modified to point to the new 8801attributes, or a copy of the list may be made if further changes are 8802needed. 8803@end deftypefn 8804 8805@deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl}) 8806@cindex inlining 8807This target hook returns @code{true} if it is ok to inline @var{fndecl} 8808into the current function, despite its having target-specific 8809attributes, @code{false} otherwise. By default, if a function has a 8810target specific attribute attached to it, it will not be inlined. 8811@end deftypefn 8812 8813@node MIPS Coprocessors 8814@section Defining coprocessor specifics for MIPS targets. 8815@cindex MIPS coprocessor-definition macros 8816 8817The MIPS specification allows MIPS implementations to have as many as 4 8818coprocessors, each with as many as 32 private registers. GCC supports 8819accessing these registers and transferring values between the registers 8820and memory using asm-ized variables. For example: 8821 8822@smallexample 8823 register unsigned int cp0count asm ("c0r1"); 8824 unsigned int d; 8825 8826 d = cp0count + 3; 8827@end smallexample 8828 8829(``c0r1'' is the default name of register 1 in coprocessor 0; alternate 8830names may be added as described below, or the default names may be 8831overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.) 8832 8833Coprocessor registers are assumed to be epilogue-used; sets to them will 8834be preserved even if it does not appear that the register is used again 8835later in the function. 8836 8837Another note: according to the MIPS spec, coprocessor 1 (if present) is 8838the FPU@. One accesses COP1 registers through standard mips 8839floating-point support; they are not included in this mechanism. 8840 8841There is one macro used in defining the MIPS coprocessor interface which 8842you may want to override in subtargets; it is described below. 8843 8844@defmac ALL_COP_ADDITIONAL_REGISTER_NAMES 8845A comma-separated list (with leading comma) of pairs describing the 8846alternate names of coprocessor registers. The format of each entry should be 8847@smallexample 8848@{ @var{alternatename}, @var{register_number}@} 8849@end smallexample 8850Default: empty. 8851@end defmac 8852 8853@node PCH Target 8854@section Parameters for Precompiled Header Validity Checking 8855@cindex parameters, precompiled headers 8856 8857@deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz}) 8858This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets 8859@samp{*@var{sz}} to the size of the data in bytes. 8860@end deftypefn 8861 8862@deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz}) 8863This hook checks whether the options used to create a PCH file are 8864compatible with the current settings. It returns @code{NULL} 8865if so and a suitable error message if not. Error messages will 8866be presented to the user and must be localized using @samp{_(@var{msg})}. 8867 8868@var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY} 8869when the PCH file was created and @var{sz} is the size of that data in bytes. 8870It's safe to assume that the data was created by the same version of the 8871compiler, so no format checking is needed. 8872 8873The default definition of @code{default_pch_valid_p} should be 8874suitable for most targets. 8875@end deftypefn 8876 8877@deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags}) 8878If this hook is nonnull, the default implementation of 8879@code{TARGET_PCH_VALID_P} will use it to check for compatible values 8880of @code{target_flags}. @var{pch_flags} specifies the value that 8881@code{target_flags} had when the PCH file was created. The return 8882value is the same as for @code{TARGET_PCH_VALID_P}. 8883@end deftypefn 8884 8885@node C++ ABI 8886@section C++ ABI parameters 8887@cindex parameters, c++ abi 8888 8889@deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void) 8890Define this hook to override the integer type used for guard variables. 8891These are used to implement one-time construction of static objects. The 8892default is long_long_integer_type_node. 8893@end deftypefn 8894 8895@deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void) 8896This hook determines how guard variables are used. It should return 8897@code{false} (the default) if first byte should be used. A return value of 8898@code{true} indicates the least significant bit should be used. 8899@end deftypefn 8900 8901@deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type}) 8902This hook returns the size of the cookie to use when allocating an array 8903whose elements have the indicated @var{type}. Assumes that it is already 8904known that a cookie is needed. The default is 8905@code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the 8906IA64/Generic C++ ABI@. 8907@end deftypefn 8908 8909@deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void) 8910This hook should return @code{true} if the element size should be stored in 8911array cookies. The default is to return @code{false}. 8912@end deftypefn 8913 8914@deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export}) 8915If defined by a backend this hook allows the decision made to export 8916class @var{type} to be overruled. Upon entry @var{import_export} 8917will contain 1 if the class is going to be exported, @minus{}1 if it is going 8918to be imported and 0 otherwise. This function should return the 8919modified value and perform any other actions necessary to support the 8920backend's targeted operating system. 8921@end deftypefn 8922 8923@deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void) 8924This hook should return @code{true} if constructors and destructors return 8925the address of the object created/destroyed. The default is to return 8926@code{false}. 8927@end deftypefn 8928 8929@deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void) 8930This hook returns true if the key method for a class (i.e., the method 8931which, if defined in the current translation unit, causes the virtual 8932table to be emitted) may be an inline function. Under the standard 8933Itanium C++ ABI the key method may be an inline function so long as 8934the function is not declared inline in the class definition. Under 8935some variants of the ABI, an inline function can never be the key 8936method. The default is to return @code{true}. 8937@end deftypefn 8938 8939@deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl}) 8940@var{decl} is a virtual table, virtual table table, typeinfo object, 8941or other similar implicit class data object that will be emitted with 8942external linkage in this translation unit. No ELF visibility has been 8943explicitly specified. If the target needs to specify a visibility 8944other than that of the containing class, use this hook to set 8945@code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}. 8946@end deftypefn 8947 8948@deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void) 8949This hook returns true (the default) if virtual tables and other 8950similar implicit class data objects are always COMDAT if they have 8951external linkage. If this hook returns false, then class data for 8952classes whose virtual table will be emitted in only one translation 8953unit will not be COMDAT. 8954@end deftypefn 8955 8956@deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void) 8957This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI) 8958should be used to register static destructors when @option{-fuse-cxa-atexit} 8959is in effect. The default is to return false to use @code{__cxa_atexit}. 8960@end deftypefn 8961 8962@deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type}) 8963@var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been 8964defined. Use this hook to make adjustments to the class (eg, tweak 8965visibility or perform any other required target modifications). 8966@end deftypefn 8967 8968@node Misc 8969@section Miscellaneous Parameters 8970@cindex parameters, miscellaneous 8971 8972@c prevent bad page break with this line 8973Here are several miscellaneous parameters. 8974 8975@defmac HAS_LONG_COND_BRANCH 8976Define this boolean macro to indicate whether or not your architecture 8977has conditional branches that can span all of memory. It is used in 8978conjunction with an optimization that partitions hot and cold basic 8979blocks into separate sections of the executable. If this macro is 8980set to false, gcc will convert any conditional branches that attempt 8981to cross between sections into unconditional branches or indirect jumps. 8982@end defmac 8983 8984@defmac HAS_LONG_UNCOND_BRANCH 8985Define this boolean macro to indicate whether or not your architecture 8986has unconditional branches that can span all of memory. It is used in 8987conjunction with an optimization that partitions hot and cold basic 8988blocks into separate sections of the executable. If this macro is 8989set to false, gcc will convert any unconditional branches that attempt 8990to cross between sections into indirect jumps. 8991@end defmac 8992 8993@defmac CASE_VECTOR_MODE 8994An alias for a machine mode name. This is the machine mode that 8995elements of a jump-table should have. 8996@end defmac 8997 8998@defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body}) 8999Optional: return the preferred mode for an @code{addr_diff_vec} 9000when the minimum and maximum offset are known. If you define this, 9001it enables extra code in branch shortening to deal with @code{addr_diff_vec}. 9002To make this work, you also have to define @code{INSN_ALIGN} and 9003make the alignment for @code{addr_diff_vec} explicit. 9004The @var{body} argument is provided so that the offset_unsigned and scale 9005flags can be updated. 9006@end defmac 9007 9008@defmac CASE_VECTOR_PC_RELATIVE 9009Define this macro to be a C expression to indicate when jump-tables 9010should contain relative addresses. You need not define this macro if 9011jump-tables never contain relative addresses, or jump-tables should 9012contain relative addresses only when @option{-fPIC} or @option{-fPIC} 9013is in effect. 9014@end defmac 9015 9016@defmac CASE_VALUES_THRESHOLD 9017Define this to be the smallest number of different values for which it 9018is best to use a jump-table instead of a tree of conditional branches. 9019The default is four for machines with a @code{casesi} instruction and 9020five otherwise. This is best for most machines. 9021@end defmac 9022 9023@defmac CASE_USE_BIT_TESTS 9024Define this macro to be a C expression to indicate whether C switch 9025statements may be implemented by a sequence of bit tests. This is 9026advantageous on processors that can efficiently implement left shift 9027of 1 by the number of bits held in a register, but inappropriate on 9028targets that would require a loop. By default, this macro returns 9029@code{true} if the target defines an @code{ashlsi3} pattern, and 9030@code{false} otherwise. 9031@end defmac 9032 9033@defmac WORD_REGISTER_OPERATIONS 9034Define this macro if operations between registers with integral mode 9035smaller than a word are always performed on the entire register. 9036Most RISC machines have this property and most CISC machines do not. 9037@end defmac 9038 9039@defmac LOAD_EXTEND_OP (@var{mem_mode}) 9040Define this macro to be a C expression indicating when insns that read 9041memory in @var{mem_mode}, an integral mode narrower than a word, set the 9042bits outside of @var{mem_mode} to be either the sign-extension or the 9043zero-extension of the data read. Return @code{SIGN_EXTEND} for values 9044of @var{mem_mode} for which the 9045insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and 9046@code{UNKNOWN} for other modes. 9047 9048This macro is not called with @var{mem_mode} non-integral or with a width 9049greater than or equal to @code{BITS_PER_WORD}, so you may return any 9050value in this case. Do not define this macro if it would always return 9051@code{UNKNOWN}. On machines where this macro is defined, you will normally 9052define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. 9053 9054You may return a non-@code{UNKNOWN} value even if for some hard registers 9055the sign extension is not performed, if for the @code{REGNO_REG_CLASS} 9056of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero 9057when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any 9058integral mode larger than this but not larger than @code{word_mode}. 9059 9060You must return @code{UNKNOWN} if for some hard registers that allow this 9061mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to 9062@code{word_mode}, but that they can change to another integral mode that 9063is larger then @var{mem_mode} but still smaller than @code{word_mode}. 9064@end defmac 9065 9066@defmac SHORT_IMMEDIATES_SIGN_EXTEND 9067Define this macro if loading short immediate values into registers sign 9068extends. 9069@end defmac 9070 9071@defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC 9072Define this macro if the same instructions that convert a floating 9073point number to a signed fixed point number also convert validly to an 9074unsigned one. 9075@end defmac 9076 9077@deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode}) 9078When @option{-ffast-math} is in effect, GCC tries to optimize 9079divisions by the same divisor, by turning them into multiplications by 9080the reciprocal. This target hook specifies the minimum number of divisions 9081that should be there for GCC to perform the optimization for a variable 9082of mode @var{mode}. The default implementation returns 3 if the machine 9083has an instruction for the division, and 2 if it does not. 9084@end deftypefn 9085 9086@defmac MOVE_MAX 9087The maximum number of bytes that a single instruction can move quickly 9088between memory and registers or between two memory locations. 9089@end defmac 9090 9091@defmac MAX_MOVE_MAX 9092The maximum number of bytes that a single instruction can move quickly 9093between memory and registers or between two memory locations. If this 9094is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the 9095constant value that is the largest value that @code{MOVE_MAX} can have 9096at run-time. 9097@end defmac 9098 9099@defmac SHIFT_COUNT_TRUNCATED 9100A C expression that is nonzero if on this machine the number of bits 9101actually used for the count of a shift operation is equal to the number 9102of bits needed to represent the size of the object being shifted. When 9103this macro is nonzero, the compiler will assume that it is safe to omit 9104a sign-extend, zero-extend, and certain bitwise `and' instructions that 9105truncates the count of a shift operation. On machines that have 9106instructions that act on bit-fields at variable positions, which may 9107include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} 9108also enables deletion of truncations of the values that serve as 9109arguments to bit-field instructions. 9110 9111If both types of instructions truncate the count (for shifts) and 9112position (for bit-field operations), or if no variable-position bit-field 9113instructions exist, you should define this macro. 9114 9115However, on some machines, such as the 80386 and the 680x0, truncation 9116only applies to shift operations and not the (real or pretended) 9117bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on 9118such machines. Instead, add patterns to the @file{md} file that include 9119the implied truncation of the shift instructions. 9120 9121You need not define this macro if it would always have the value of zero. 9122@end defmac 9123 9124@anchor{TARGET_SHIFT_TRUNCATION_MASK} 9125@deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode}) 9126This function describes how the standard shift patterns for @var{mode} 9127deal with shifts by negative amounts or by more than the width of the mode. 9128@xref{shift patterns}. 9129 9130On many machines, the shift patterns will apply a mask @var{m} to the 9131shift count, meaning that a fixed-width shift of @var{x} by @var{y} is 9132equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If 9133this is true for mode @var{mode}, the function should return @var{m}, 9134otherwise it should return 0. A return value of 0 indicates that no 9135particular behavior is guaranteed. 9136 9137Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does 9138@emph{not} apply to general shift rtxes; it applies only to instructions 9139that are generated by the named shift patterns. 9140 9141The default implementation of this function returns 9142@code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED} 9143and 0 otherwise. This definition is always safe, but if 9144@code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns 9145nevertheless truncate the shift count, you may get better code 9146by overriding it. 9147@end deftypefn 9148 9149@defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec}) 9150A C expression which is nonzero if on this machine it is safe to 9151``convert'' an integer of @var{inprec} bits to one of @var{outprec} 9152bits (where @var{outprec} is smaller than @var{inprec}) by merely 9153operating on it as if it had only @var{outprec} bits. 9154 9155On many machines, this expression can be 1. 9156 9157@c rearranged this, removed the phrase "it is reported that". this was 9158@c to fix an overfull hbox. --mew 10feb93 9159When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for 9160modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result. 9161If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in 9162such cases may improve things. 9163@end defmac 9164 9165@deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode}) 9166The representation of an integral mode can be such that the values 9167are always extended to a wider integral mode. Return 9168@code{SIGN_EXTEND} if values of @var{mode} are represented in 9169sign-extended form to @var{rep_mode}. Return @code{UNKNOWN} 9170otherwise. (Currently, none of the targets use zero-extended 9171representation this way so unlike @code{LOAD_EXTEND_OP}, 9172@code{TARGET_MODE_REP_EXTENDED} is expected to return either 9173@code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends 9174@var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next 9175widest integral mode and currently we take advantage of this fact.) 9176 9177Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN} 9178value even if the extension is not performed on certain hard registers 9179as long as for the @code{REGNO_REG_CLASS} of these hard registers 9180@code{CANNOT_CHANGE_MODE_CLASS} returns nonzero. 9181 9182Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP} 9183describe two related properties. If you define 9184@code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want 9185to define @code{LOAD_EXTEND_OP (mode)} to return the same type of 9186extension. 9187 9188In order to enforce the representation of @code{mode}, 9189@code{TRULY_NOOP_TRUNCATION} should return false when truncating to 9190@code{mode}. 9191@end deftypefn 9192 9193@defmac STORE_FLAG_VALUE 9194A C expression describing the value returned by a comparison operator 9195with an integral mode and stored by a store-flag instruction 9196(@samp{s@var{cond}}) when the condition is true. This description must 9197apply to @emph{all} the @samp{s@var{cond}} patterns and all the 9198comparison operators whose results have a @code{MODE_INT} mode. 9199 9200A value of 1 or @minus{}1 means that the instruction implementing the 9201comparison operator returns exactly 1 or @minus{}1 when the comparison is true 9202and 0 when the comparison is false. Otherwise, the value indicates 9203which bits of the result are guaranteed to be 1 when the comparison is 9204true. This value is interpreted in the mode of the comparison 9205operation, which is given by the mode of the first operand in the 9206@samp{s@var{cond}} pattern. Either the low bit or the sign bit of 9207@code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by 9208the compiler. 9209 9210If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will 9211generate code that depends only on the specified bits. It can also 9212replace comparison operators with equivalent operations if they cause 9213the required bits to be set, even if the remaining bits are undefined. 9214For example, on a machine whose comparison operators return an 9215@code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as 9216@samp{0x80000000}, saying that just the sign bit is relevant, the 9217expression 9218 9219@smallexample 9220(ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) 9221@end smallexample 9222 9223@noindent 9224can be converted to 9225 9226@smallexample 9227(ashift:SI @var{x} (const_int @var{n})) 9228@end smallexample 9229 9230@noindent 9231where @var{n} is the appropriate shift count to move the bit being 9232tested into the sign bit. 9233 9234There is no way to describe a machine that always sets the low-order bit 9235for a true value, but does not guarantee the value of any other bits, 9236but we do not know of any machine that has such an instruction. If you 9237are trying to port GCC to such a machine, include an instruction to 9238perform a logical-and of the result with 1 in the pattern for the 9239comparison operators and let us know at @email{gcc@@gcc.gnu.org}. 9240 9241Often, a machine will have multiple instructions that obtain a value 9242from a comparison (or the condition codes). Here are rules to guide the 9243choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions 9244to be used: 9245 9246@itemize @bullet 9247@item 9248Use the shortest sequence that yields a valid definition for 9249@code{STORE_FLAG_VALUE}. It is more efficient for the compiler to 9250``normalize'' the value (convert it to, e.g., 1 or 0) than for the 9251comparison operators to do so because there may be opportunities to 9252combine the normalization with other operations. 9253 9254@item 9255For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being 9256slightly preferred on machines with expensive jumps and 1 preferred on 9257other machines. 9258 9259@item 9260As a second choice, choose a value of @samp{0x80000001} if instructions 9261exist that set both the sign and low-order bits but do not define the 9262others. 9263 9264@item 9265Otherwise, use a value of @samp{0x80000000}. 9266@end itemize 9267 9268Many machines can produce both the value chosen for 9269@code{STORE_FLAG_VALUE} and its negation in the same number of 9270instructions. On those machines, you should also define a pattern for 9271those cases, e.g., one matching 9272 9273@smallexample 9274(set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) 9275@end smallexample 9276 9277Some machines can also perform @code{and} or @code{plus} operations on 9278condition code values with less instructions than the corresponding 9279@samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those 9280machines, define the appropriate patterns. Use the names @code{incscc} 9281and @code{decscc}, respectively, for the patterns which perform 9282@code{plus} or @code{minus} operations on condition code values. See 9283@file{rs6000.md} for some examples. The GNU Superoptizer can be used to 9284find such instruction sequences on other machines. 9285 9286If this macro is not defined, the default value, 1, is used. You need 9287not define @code{STORE_FLAG_VALUE} if the machine has no store-flag 9288instructions, or if the value generated by these instructions is 1. 9289@end defmac 9290 9291@defmac FLOAT_STORE_FLAG_VALUE (@var{mode}) 9292A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is 9293returned when comparison operators with floating-point results are true. 9294Define this macro on machines that have comparison operations that return 9295floating-point values. If there are no such operations, do not define 9296this macro. 9297@end defmac 9298 9299@defmac VECTOR_STORE_FLAG_VALUE (@var{mode}) 9300A C expression that gives a rtx representing the nonzero true element 9301for vector comparisons. The returned rtx should be valid for the inner 9302mode of @var{mode} which is guaranteed to be a vector mode. Define 9303this macro on machines that have vector comparison operations that 9304return a vector result. If there are no such operations, do not define 9305this macro. Typically, this macro is defined as @code{const1_rtx} or 9306@code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent 9307the compiler optimizing such vector comparison operations for the 9308given mode. 9309@end defmac 9310 9311@defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 9312@defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) 9313A C expression that evaluates to true if the architecture defines a value 9314for @code{clz} or @code{ctz} with a zero operand. If so, @var{value} 9315should be set to this value. If this macro is not defined, the value of 9316@code{clz} or @code{ctz} is assumed to be undefined. 9317 9318This macro must be defined if the target's expansion for @code{ffs} 9319relies on a particular value to get correct results. Otherwise it 9320is not necessary, though it may be used to optimize some corner cases. 9321 9322Note that regardless of this macro the ``definedness'' of @code{clz} 9323and @code{ctz} at zero do @emph{not} extend to the builtin functions 9324visible to the user. Thus one may be free to adjust the value at will 9325to match the target expansion of these operations without fear of 9326breaking the API@. 9327@end defmac 9328 9329@defmac Pmode 9330An alias for the machine mode for pointers. On most machines, define 9331this to be the integer mode corresponding to the width of a hardware 9332pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. 9333On some machines you must define this to be one of the partial integer 9334modes, such as @code{PSImode}. 9335 9336The width of @code{Pmode} must be at least as large as the value of 9337@code{POINTER_SIZE}. If it is not equal, you must define the macro 9338@code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended 9339to @code{Pmode}. 9340@end defmac 9341 9342@defmac FUNCTION_MODE 9343An alias for the machine mode used for memory references to functions 9344being called, in @code{call} RTL expressions. On most machines this 9345should be @code{QImode}. 9346@end defmac 9347 9348@defmac STDC_0_IN_SYSTEM_HEADERS 9349In normal operation, the preprocessor expands @code{__STDC__} to the 9350constant 1, to signify that GCC conforms to ISO Standard C@. On some 9351hosts, like Solaris, the system compiler uses a different convention, 9352where @code{__STDC__} is normally 0, but is 1 if the user specifies 9353strict conformance to the C Standard. 9354 9355Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host 9356convention when processing system header files, but when processing user 9357files @code{__STDC__} will always expand to 1. 9358@end defmac 9359 9360@defmac NO_IMPLICIT_EXTERN_C 9361Define this macro if the system header files support C++ as well as C@. 9362This macro inhibits the usual method of using system header files in 9363C++, which is to pretend that the file's contents are enclosed in 9364@samp{extern "C" @{@dots{}@}}. 9365@end defmac 9366 9367@findex #pragma 9368@findex pragma 9369@defmac REGISTER_TARGET_PRAGMAS () 9370Define this macro if you want to implement any target-specific pragmas. 9371If defined, it is a C expression which makes a series of calls to 9372@code{c_register_pragma} or @code{c_register_pragma_with_expansion} 9373for each pragma. The macro may also do any 9374setup required for the pragmas. 9375 9376The primary reason to define this macro is to provide compatibility with 9377other compilers for the same target. In general, we discourage 9378definition of target-specific pragmas for GCC@. 9379 9380If the pragma can be implemented by attributes then you should consider 9381defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well. 9382 9383Preprocessor macros that appear on pragma lines are not expanded. All 9384@samp{#pragma} directives that do not match any registered pragma are 9385silently ignored, unless the user specifies @option{-Wunknown-pragmas}. 9386@end defmac 9387 9388@deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) 9389@deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) 9390 9391Each call to @code{c_register_pragma} or 9392@code{c_register_pragma_with_expansion} establishes one pragma. The 9393@var{callback} routine will be called when the preprocessor encounters a 9394pragma of the form 9395 9396@smallexample 9397#pragma [@var{space}] @var{name} @dots{} 9398@end smallexample 9399 9400@var{space} is the case-sensitive namespace of the pragma, or 9401@code{NULL} to put the pragma in the global namespace. The callback 9402routine receives @var{pfile} as its first argument, which can be passed 9403on to cpplib's functions if necessary. You can lex tokens after the 9404@var{name} by calling @code{pragma_lex}. Tokens that are not read by the 9405callback will be silently ignored. The end of the line is indicated by 9406a token of type @code{CPP_EOF}. Macro expansion occurs on the 9407arguments of pragmas registered with 9408@code{c_register_pragma_with_expansion} but not on the arguments of 9409pragmas registered with @code{c_register_pragma}. 9410 9411For an example use of this routine, see @file{c4x.h} and the callback 9412routines defined in @file{c4x-c.c}. 9413 9414Note that the use of @code{pragma_lex} is specific to the C and C++ 9415compilers. It will not work in the Java or Fortran compilers, or any 9416other language compilers for that matter. Thus if @code{pragma_lex} is going 9417to be called from target-specific code, it must only be done so when 9418building the C and C++ compilers. This can be done by defining the 9419variables @code{c_target_objs} and @code{cxx_target_objs} in the 9420target entry in the @file{config.gcc} file. These variables should name 9421the target-specific, language-specific object file which contains the 9422code that uses @code{pragma_lex}. Note it will also be necessary to add a 9423rule to the makefile fragment pointed to by @code{tmake_file} that shows 9424how to build this object file. 9425@end deftypefun 9426 9427@findex #pragma 9428@findex pragma 9429@defmac HANDLE_SYSV_PRAGMA 9430Define this macro (to a value of 1) if you want the System V style 9431pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name> 9432[=<value>]} to be supported by gcc. 9433 9434The pack pragma specifies the maximum alignment (in bytes) of fields 9435within a structure, in much the same way as the @samp{__aligned__} and 9436@samp{__packed__} @code{__attribute__}s do. A pack value of zero resets 9437the behavior to the default. 9438 9439A subtlety for Microsoft Visual C/C++ style bit-field packing 9440(e.g.@: -mms-bitfields) for targets that support it: 9441When a bit-field is inserted into a packed record, the whole size 9442of the underlying type is used by one or more same-size adjacent 9443bit-fields (that is, if its long:3, 32 bits is used in the record, 9444and any additional adjacent long bit-fields are packed into the same 9445chunk of 32 bits. However, if the size changes, a new field of that 9446size is allocated). 9447 9448If both MS bit-fields and @samp{__attribute__((packed))} are used, 9449the latter will take precedence. If @samp{__attribute__((packed))} is 9450used on a single field when MS bit-fields are in use, it will take 9451precedence for that field, but the alignment of the rest of the structure 9452may affect its placement. 9453 9454The weak pragma only works if @code{SUPPORTS_WEAK} and 9455@code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation 9456of specifically named weak labels, optionally with a value. 9457@end defmac 9458 9459@findex #pragma 9460@findex pragma 9461@defmac HANDLE_PRAGMA_PACK_PUSH_POP 9462Define this macro (to a value of 1) if you want to support the Win32 9463style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma 9464pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum 9465alignment (in bytes) of fields within a structure, in much the same way as 9466the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A 9467pack value of zero resets the behavior to the default. Successive 9468invocations of this pragma cause the previous values to be stacked, so 9469that invocations of @samp{#pragma pack(pop)} will return to the previous 9470value. 9471@end defmac 9472 9473@defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION 9474Define this macro, as well as 9475@code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the 9476arguments of @samp{#pragma pack}. 9477@end defmac 9478 9479@defmac TARGET_DEFAULT_PACK_STRUCT 9480If your target requires a structure packing default other than 0 (meaning 9481the machine default), define this macro to the necessary value (in bytes). 9482This must be a value that would also be valid to use with 9483@samp{#pragma pack()} (that is, a small power of two). 9484@end defmac 9485 9486@defmac DOLLARS_IN_IDENTIFIERS 9487Define this macro to control use of the character @samp{$} in 9488identifier names for the C family of languages. 0 means @samp{$} is 9489not allowed by default; 1 means it is allowed. 1 is the default; 9490there is no need to define this macro in that case. 9491@end defmac 9492 9493@defmac NO_DOLLAR_IN_LABEL 9494Define this macro if the assembler does not accept the character 9495@samp{$} in label names. By default constructors and destructors in 9496G++ have @samp{$} in the identifiers. If this macro is defined, 9497@samp{.} is used instead. 9498@end defmac 9499 9500@defmac NO_DOT_IN_LABEL 9501Define this macro if the assembler does not accept the character 9502@samp{.} in label names. By default constructors and destructors in G++ 9503have names that use @samp{.}. If this macro is defined, these names 9504are rewritten to avoid @samp{.}. 9505@end defmac 9506 9507@defmac INSN_SETS_ARE_DELAYED (@var{insn}) 9508Define this macro as a C expression that is nonzero if it is safe for the 9509delay slot scheduler to place instructions in the delay slot of @var{insn}, 9510even if they appear to use a resource set or clobbered in @var{insn}. 9511@var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that 9512every @code{call_insn} has this behavior. On machines where some @code{insn} 9513or @code{jump_insn} is really a function call and hence has this behavior, 9514you should define this macro. 9515 9516You need not define this macro if it would always return zero. 9517@end defmac 9518 9519@defmac INSN_REFERENCES_ARE_DELAYED (@var{insn}) 9520Define this macro as a C expression that is nonzero if it is safe for the 9521delay slot scheduler to place instructions in the delay slot of @var{insn}, 9522even if they appear to set or clobber a resource referenced in @var{insn}. 9523@var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where 9524some @code{insn} or @code{jump_insn} is really a function call and its operands 9525are registers whose use is actually in the subroutine it calls, you should 9526define this macro. Doing so allows the delay slot scheduler to move 9527instructions which copy arguments into the argument registers into the delay 9528slot of @var{insn}. 9529 9530You need not define this macro if it would always return zero. 9531@end defmac 9532 9533@defmac MULTIPLE_SYMBOL_SPACES 9534Define this macro as a C expression that is nonzero if, in some cases, 9535global symbols from one translation unit may not be bound to undefined 9536symbols in another translation unit without user intervention. For 9537instance, under Microsoft Windows symbols must be explicitly imported 9538from shared libraries (DLLs). 9539 9540You need not define this macro if it would always evaluate to zero. 9541@end defmac 9542 9543@deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers}) 9544This target hook should add to @var{clobbers} @code{STRING_CST} trees for 9545any hard regs the port wishes to automatically clobber for an asm. 9546It should return the result of the last @code{tree_cons} used to add a 9547clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the 9548corresponding parameters to the asm and may be inspected to avoid 9549clobbering a register that is an input or output of the asm. You can use 9550@code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test 9551for overlap with regards to asm-declared registers. 9552@end deftypefn 9553 9554@defmac MATH_LIBRARY 9555Define this macro as a C string constant for the linker argument to link 9556in the system math library, or @samp{""} if the target does not have a 9557separate math library. 9558 9559You need only define this macro if the default of @samp{"-lm"} is wrong. 9560@end defmac 9561 9562@defmac LIBRARY_PATH_ENV 9563Define this macro as a C string constant for the environment variable that 9564specifies where the linker should look for libraries. 9565 9566You need only define this macro if the default of @samp{"LIBRARY_PATH"} 9567is wrong. 9568@end defmac 9569 9570@defmac TARGET_POSIX_IO 9571Define this macro if the target supports the following POSIX@ file 9572functions, access, mkdir and file locking with fcntl / F_SETLKW@. 9573Defining @code{TARGET_POSIX_IO} will enable the test coverage code 9574to use file locking when exiting a program, which avoids race conditions 9575if the program has forked. It will also create directories at run-time 9576for cross-profiling. 9577@end defmac 9578 9579@defmac MAX_CONDITIONAL_EXECUTE 9580 9581A C expression for the maximum number of instructions to execute via 9582conditional execution instructions instead of a branch. A value of 9583@code{BRANCH_COST}+1 is the default if the machine does not use cc0, and 95841 if it does use cc0. 9585@end defmac 9586 9587@defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr}) 9588Used if the target needs to perform machine-dependent modifications on the 9589conditionals used for turning basic blocks into conditionally executed code. 9590@var{ce_info} points to a data structure, @code{struct ce_if_block}, which 9591contains information about the currently processed blocks. @var{true_expr} 9592and @var{false_expr} are the tests that are used for converting the 9593then-block and the else-block, respectively. Set either @var{true_expr} or 9594@var{false_expr} to a null pointer if the tests cannot be converted. 9595@end defmac 9596 9597@defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr}) 9598Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated 9599if-statements into conditions combined by @code{and} and @code{or} operations. 9600@var{bb} contains the basic block that contains the test that is currently 9601being processed and about to be turned into a condition. 9602@end defmac 9603 9604@defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn}) 9605A C expression to modify the @var{PATTERN} of an @var{INSN} that is to 9606be converted to conditional execution format. @var{ce_info} points to 9607a data structure, @code{struct ce_if_block}, which contains information 9608about the currently processed blocks. 9609@end defmac 9610 9611@defmac IFCVT_MODIFY_FINAL (@var{ce_info}) 9612A C expression to perform any final machine dependent modifications in 9613converting code to conditional execution. The involved basic blocks 9614can be found in the @code{struct ce_if_block} structure that is pointed 9615to by @var{ce_info}. 9616@end defmac 9617 9618@defmac IFCVT_MODIFY_CANCEL (@var{ce_info}) 9619A C expression to cancel any machine dependent modifications in 9620converting code to conditional execution. The involved basic blocks 9621can be found in the @code{struct ce_if_block} structure that is pointed 9622to by @var{ce_info}. 9623@end defmac 9624 9625@defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info}) 9626A C expression to initialize any extra fields in a @code{struct ce_if_block} 9627structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro. 9628@end defmac 9629 9630@defmac IFCVT_EXTRA_FIELDS 9631If defined, it should expand to a set of field declarations that will be 9632added to the @code{struct ce_if_block} structure. These should be initialized 9633by the @code{IFCVT_INIT_EXTRA_FIELDS} macro. 9634@end defmac 9635 9636@deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG () 9637If non-null, this hook performs a target-specific pass over the 9638instruction stream. The compiler will run it at all optimization levels, 9639just before the point at which it normally does delayed-branch scheduling. 9640 9641The exact purpose of the hook varies from target to target. Some use 9642it to do transformations that are necessary for correctness, such as 9643laying out in-function constant pools or avoiding hardware hazards. 9644Others use it as an opportunity to do some machine-dependent optimizations. 9645 9646You need not implement the hook if it has nothing to do. The default 9647definition is null. 9648@end deftypefn 9649 9650@deftypefn {Target Hook} void TARGET_INIT_BUILTINS () 9651Define this hook if you have any machine-specific built-in functions 9652that need to be defined. It should be a function that performs the 9653necessary setup. 9654 9655Machine specific built-in functions can be useful to expand special machine 9656instructions that would otherwise not normally be generated because 9657they have no equivalent in the source language (for example, SIMD vector 9658instructions or prefetch instructions). 9659 9660To create a built-in function, call the function 9661@code{lang_hooks.builtin_function} 9662which is defined by the language front end. You can use any type nodes set 9663up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2}; 9664only language front ends that use those two functions will call 9665@samp{TARGET_INIT_BUILTINS}. 9666@end deftypefn 9667 9668@deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore}) 9669 9670Expand a call to a machine specific built-in function that was set up by 9671@samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the 9672function call; the result should go to @var{target} if that is 9673convenient, and have mode @var{mode} if that is convenient. 9674@var{subtarget} may be used as the target for computing one of 9675@var{exp}'s operands. @var{ignore} is nonzero if the value is to be 9676ignored. This function should return the result of the call to the 9677built-in function. 9678@end deftypefn 9679 9680@deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist}) 9681 9682Select a replacement for a machine specific built-in function that 9683was set up by @samp{TARGET_INIT_BUILTINS}. This is done 9684@emph{before} regular type checking, and so allows the target to 9685implement a crude form of function overloading. @var{fndecl} is the 9686declaration of the built-in function. @var{arglist} is the list of 9687arguments passed to the built-in function. The result is a 9688complete expression that implements the operation, usually 9689another @code{CALL_EXPR}. 9690@end deftypefn 9691 9692@deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore}) 9693 9694Fold a call to a machine specific built-in function that was set up by 9695@samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the 9696built-in function. @var{arglist} is the list of arguments passed to 9697the built-in function. The result is another tree containing a 9698simplified expression for the call's result. If @var{ignore} is true 9699the value will be ignored. 9700@end deftypefn 9701 9702@deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn}) 9703 9704Take an instruction in @var{insn} and return NULL if it is valid within a 9705low-overhead loop, otherwise return a string why doloop could not be applied. 9706 9707Many targets use special registers for low-overhead looping. For any 9708instruction that clobbers these this function should return a string indicating 9709the reason why the doloop could not be applied. 9710By default, the RTL loop optimizer does not use a present doloop pattern for 9711loops containing function calls or branch on table instructions. 9712@end deftypefn 9713 9714@defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2}) 9715 9716Take a branch insn in @var{branch1} and another in @var{branch2}. 9717Return true if redirecting @var{branch1} to the destination of 9718@var{branch2} is possible. 9719 9720On some targets, branches may have a limited range. Optimizing the 9721filling of delay slots can result in branches being redirected, and this 9722may in turn cause a branch offset to overflow. 9723@end defmac 9724 9725@deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code}) 9726This target hook returns @code{true} if @var{x} is considered to be commutative. 9727Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider 9728PLUS to be commutative inside a MEM. @var{outer_code} is the rtx code 9729of the enclosing rtl, if known, otherwise it is UNKNOWN. 9730@end deftypefn 9731 9732@deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg}) 9733 9734When the initial value of a hard register has been copied in a pseudo 9735register, it is often not necessary to actually allocate another register 9736to this pseudo register, because the original hard register or a stack slot 9737it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE} 9738is called at the start of register allocation once for each hard register 9739that had its initial value copied by using 9740@code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}. 9741Possible values are @code{NULL_RTX}, if you don't want 9742to do any special allocation, a @code{REG} rtx---that would typically be 9743the hard register itself, if it is known not to be clobbered---or a 9744@code{MEM}. 9745If you are returning a @code{MEM}, this is only a hint for the allocator; 9746it might decide to use another register anyways. 9747You may use @code{current_function_leaf_function} in the hook, functions 9748that use @code{REG_N_SETS}, to determine if the hard 9749register in question will not be clobbered. 9750The default value of this hook is @code{NULL}, which disables any special 9751allocation. 9752@end deftypefn 9753 9754@defmac TARGET_OBJECT_SUFFIX 9755Define this macro to be a C string representing the suffix for object 9756files on your target machine. If you do not define this macro, GCC will 9757use @samp{.o} as the suffix for object files. 9758@end defmac 9759 9760@defmac TARGET_EXECUTABLE_SUFFIX 9761Define this macro to be a C string representing the suffix to be 9762automatically added to executable files on your target machine. If you 9763do not define this macro, GCC will use the null string as the suffix for 9764executable files. 9765@end defmac 9766 9767@defmac COLLECT_EXPORT_LIST 9768If defined, @code{collect2} will scan the individual object files 9769specified on its command line and create an export list for the linker. 9770Define this macro for systems like AIX, where the linker discards 9771object files that are not referenced from @code{main} and uses export 9772lists. 9773@end defmac 9774 9775@defmac MODIFY_JNI_METHOD_CALL (@var{mdecl}) 9776Define this macro to a C expression representing a variant of the 9777method call @var{mdecl}, if Java Native Interface (JNI) methods 9778must be invoked differently from other methods on your target. 9779For example, on 32-bit Microsoft Windows, JNI methods must be invoked using 9780the @code{stdcall} calling convention and this macro is then 9781defined as this expression: 9782 9783@smallexample 9784build_type_attribute_variant (@var{mdecl}, 9785 build_tree_list 9786 (get_identifier ("stdcall"), 9787 NULL)) 9788@end smallexample 9789@end defmac 9790 9791@deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void) 9792This target hook returns @code{true} past the point in which new jump 9793instructions could be created. On machines that require a register for 9794every jump such as the SHmedia ISA of SH5, this point would typically be 9795reload, so this target hook should be defined to a function such as: 9796 9797@smallexample 9798static bool 9799cannot_modify_jumps_past_reload_p () 9800@{ 9801 return (reload_completed || reload_in_progress); 9802@} 9803@end smallexample 9804@end deftypefn 9805 9806@deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void) 9807This target hook returns a register class for which branch target register 9808optimizations should be applied. All registers in this class should be 9809usable interchangeably. After reload, registers in this class will be 9810re-allocated and loads will be hoisted out of loops and be subjected 9811to inter-block scheduling. 9812@end deftypefn 9813 9814@deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen}) 9815Branch target register optimization will by default exclude callee-saved 9816registers 9817that are not already live during the current function; if this target hook 9818returns true, they will be included. The target code must than make sure 9819that all target registers in the class returned by 9820@samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are 9821saved. @var{after_prologue_epilogue_gen} indicates if prologues and 9822epilogues have already been generated. Note, even if you only return 9823true when @var{after_prologue_epilogue_gen} is false, you still are likely 9824to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET} 9825to reserve space for caller-saved target registers. 9826@end deftypefn 9827 9828@defmac POWI_MAX_MULTS 9829If defined, this macro is interpreted as a signed integer C expression 9830that specifies the maximum number of floating point multiplications 9831that should be emitted when expanding exponentiation by an integer 9832constant inline. When this value is defined, exponentiation requiring 9833more than this number of multiplications is implemented by calling the 9834system library's @code{pow}, @code{powf} or @code{powl} routines. 9835The default value places no upper bound on the multiplication count. 9836@end defmac 9837 9838@deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) 9839This target hook should register any extra include files for the 9840target. The parameter @var{stdinc} indicates if normal include files 9841are present. The parameter @var{sysroot} is the system root directory. 9842The parameter @var{iprefix} is the prefix for the gcc directory. 9843@end deftypefn 9844 9845@deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) 9846This target hook should register any extra include files for the 9847target before any standard headers. The parameter @var{stdinc} 9848indicates if normal include files are present. The parameter 9849@var{sysroot} is the system root directory. The parameter 9850@var{iprefix} is the prefix for the gcc directory. 9851@end deftypefn 9852 9853@deftypefn Macro void TARGET_OPTF (char *@var{path}) 9854This target hook should register special include paths for the target. 9855The parameter @var{path} is the include to register. On Darwin 9856systems, this is used for Framework includes, which have semantics 9857that are different from @option{-I}. 9858@end deftypefn 9859 9860@deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl}) 9861This target hook returns @code{true} if it is safe to use a local alias 9862for a virtual function @var{fndecl} when constructing thunks, 9863@code{false} otherwise. By default, the hook returns @code{true} for all 9864functions, if a target supports aliases (i.e.@: defines 9865@code{ASM_OUTPUT_DEF}), @code{false} otherwise, 9866@end deftypefn 9867 9868@defmac TARGET_FORMAT_TYPES 9869If defined, this macro is the name of a global variable containing 9870target-specific format checking information for the @option{-Wformat} 9871option. The default is to have no target-specific format checks. 9872@end defmac 9873 9874@defmac TARGET_N_FORMAT_TYPES 9875If defined, this macro is the number of entries in 9876@code{TARGET_FORMAT_TYPES}. 9877@end defmac 9878 9879@deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING 9880If set to @code{true}, means that the target's memory model does not 9881guarantee that loads which do not depend on one another will access 9882main memory in the order of the instruction stream; if ordering is 9883important, an explicit memory barrier must be used. This is true of 9884many recent processors which implement a policy of ``relaxed,'' 9885``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC, 9886and ia64. The default is @code{false}. 9887@end deftypefn 9888 9889@deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val}) 9890If defined, this macro returns the diagnostic message when it is 9891illegal to pass argument @var{val} to function @var{funcdecl} 9892with prototype @var{typelist}. 9893@end deftypefn 9894 9895@deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype}) 9896If defined, this macro returns the diagnostic message when it is 9897invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL} 9898if validity should be determined by the front end. 9899@end deftypefn 9900 9901@deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type}) 9902If defined, this macro returns the diagnostic message when it is 9903invalid to apply operation @var{op} (where unary plus is denoted by 9904@code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL} 9905if validity should be determined by the front end. 9906@end deftypefn 9907 9908@deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2}) 9909If defined, this macro returns the diagnostic message when it is 9910invalid to apply operation @var{op} to operands of types @var{type1} 9911and @var{type2}, or @code{NULL} if validity should be determined by 9912the front end. 9913@end deftypefn 9914 9915@defmac TARGET_USE_JCR_SECTION 9916This macro determines whether to use the JCR section to register Java 9917classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both 9918SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0. 9919@end defmac 9920