1 This is doc/gcc.info, produced by makeinfo version 4.8 from
2 /scratch/mitchell/gcc-releases/gcc-4.0.1/gcc-4.0.1/gcc/doc/gcc.texi.
4 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
5 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
7 Permission is granted to copy, distribute and/or modify this document
8 under the terms of the GNU Free Documentation License, Version 1.2 or
9 any later version published by the Free Software Foundation; with the
10 Invariant Sections being "GNU General Public License" and "Funding Free
11 Software", the Front-Cover texts being (a) (see below), and with the
12 Back-Cover Texts being (b) (see below). A copy of the license is
13 included in the section entitled "GNU Free Documentation License".
15 (a) The FSF's Front-Cover Text is:
19 (b) The FSF's Back-Cover Text is:
21 You have freedom to copy and modify this GNU Manual, like GNU
22 software. Copies published by the Free Software Foundation raise
23 funds for GNU development.
25 INFO-DIR-SECTION Programming
27 * gcc: (gcc). The GNU Compiler Collection.
29 This file documents the use of the GNU compilers.
31 Copyright (C) 1988, 1989, 1992, 1993, 1994, 1995, 1996, 1997, 1998,
32 1999, 2000, 2001, 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
34 Permission is granted to copy, distribute and/or modify this document
35 under the terms of the GNU Free Documentation License, Version 1.2 or
36 any later version published by the Free Software Foundation; with the
37 Invariant Sections being "GNU General Public License" and "Funding Free
38 Software", the Front-Cover texts being (a) (see below), and with the
39 Back-Cover Texts being (b) (see below). A copy of the license is
40 included in the section entitled "GNU Free Documentation License".
42 (a) The FSF's Front-Cover Text is:
46 (b) The FSF's Back-Cover Text is:
48 You have freedom to copy and modify this GNU Manual, like GNU
49 software. Copies published by the Free Software Foundation raise
50 funds for GNU development.
54 File: gcc.info, Node: Top, Next: G++ and GCC, Up: (DIR)
59 This manual documents how to use the GNU compilers, as well as their
60 features and incompatibilities, and how to report bugs. It corresponds
61 to GCC version 4.0.1. The internals of the GNU compilers, including
62 how to port them to new targets and some information about how to write
63 front ends for new languages, are documented in a separate manual.
64 *Note Introduction: (gccint)Top.
68 * G++ and GCC:: You can compile C or C++ programs.
69 * Standards:: Language standards supported by GCC.
70 * Invoking GCC:: Command options supported by `gcc'.
71 * C Implementation:: How GCC implements the ISO C specification.
72 * C Extensions:: GNU extensions to the C language family.
73 * C++ Extensions:: GNU extensions to the C++ language.
74 * Objective-C:: GNU Objective-C runtime features.
75 * Compatibility:: Binary Compatibility
76 * Gcov:: `gcov'---a test coverage program.
77 * Trouble:: If you have trouble using GCC.
78 * Bugs:: How, why and where to report bugs.
79 * Service:: How to find suppliers of support for GCC.
80 * Contributing:: How to contribute to testing and developing GCC.
82 * Funding:: How to help assure funding for free software.
83 * GNU Project:: The GNU Project and GNU/Linux.
85 * Copying:: GNU General Public License says
86 how you can copy and share GCC.
87 * GNU Free Documentation License:: How you can copy and share this manual.
88 * Contributors:: People who have contributed to GCC.
90 * Option Index:: Index to command line options.
91 * Keyword Index:: Index of concepts and symbol names.
94 File: gcc.info, Node: G++ and GCC, Next: Standards, Prev: Top, Up: Top
96 1 Programming Languages Supported by GCC
97 ****************************************
99 GCC stands for "GNU Compiler Collection". GCC is an integrated
100 distribution of compilers for several major programming languages.
101 These languages currently include C, C++, Objective-C, Objective-C++,
102 Java, Fortran, and Ada.
104 The abbreviation "GCC" has multiple meanings in common use. The
105 current official meaning is "GNU Compiler Collection", which refers
106 generically to the complete suite of tools. The name historically stood
107 for "GNU C Compiler", and this usage is still common when the emphasis
108 is on compiling C programs. Finally, the name is also used when
109 speaking of the "language-independent" component of GCC: code shared
110 among the compilers for all supported languages.
112 The language-independent component of GCC includes the majority of the
113 optimizers, as well as the "back ends" that generate machine code for
116 The part of a compiler that is specific to a particular language is
117 called the "front end". In addition to the front ends that are
118 integrated components of GCC, there are several other front ends that
119 are maintained separately. These support languages such as Pascal,
120 Mercury, and COBOL. To use these, they must be built together with GCC
123 Most of the compilers for languages other than C have their own names.
124 The C++ compiler is G++, the Ada compiler is GNAT, and so on. When we
125 talk about compiling one of those languages, we might refer to that
126 compiler by its own name, or as GCC. Either is correct.
128 Historically, compilers for many languages, including C++ and Fortran,
129 have been implemented as "preprocessors" which emit another high level
130 language such as C. None of the compilers included in GCC are
131 implemented this way; they all generate machine code directly. This
132 sort of preprocessor should not be confused with the "C preprocessor",
133 which is an integral feature of the C, C++, Objective-C and
134 Objective-C++ languages.
137 File: gcc.info, Node: Standards, Next: Invoking GCC, Prev: G++ and GCC, Up: Top
139 2 Language Standards Supported by GCC
140 *************************************
142 For each language compiled by GCC for which there is a standard, GCC
143 attempts to follow one or more versions of that standard, possibly with
144 some exceptions, and possibly with some extensions.
146 GCC supports three versions of the C standard, although support for
147 the most recent version is not yet complete.
149 The original ANSI C standard (X3.159-1989) was ratified in 1989 and
150 published in 1990. This standard was ratified as an ISO standard
151 (ISO/IEC 9899:1990) later in 1990. There were no technical differences
152 between these publications, although the sections of the ANSI standard
153 were renumbered and became clauses in the ISO standard. This standard,
154 in both its forms, is commonly known as "C89", or occasionally as
155 "C90", from the dates of ratification. The ANSI standard, but not the
156 ISO standard, also came with a Rationale document. To select this
157 standard in GCC, use one of the options `-ansi', `-std=c89' or
158 `-std=iso9899:1990'; to obtain all the diagnostics required by the
159 standard, you should also specify `-pedantic' (or `-pedantic-errors' if
160 you want them to be errors rather than warnings). *Note Options
161 Controlling C Dialect: C Dialect Options.
163 Errors in the 1990 ISO C standard were corrected in two Technical
164 Corrigenda published in 1994 and 1996. GCC does not support the
167 An amendment to the 1990 standard was published in 1995. This
168 amendment added digraphs and `__STDC_VERSION__' to the language, but
169 otherwise concerned the library. This amendment is commonly known as
170 "AMD1"; the amended standard is sometimes known as "C94" or "C95". To
171 select this standard in GCC, use the option `-std=iso9899:199409'
172 (with, as for other standard versions, `-pedantic' to receive all
173 required diagnostics).
175 A new edition of the ISO C standard was published in 1999 as ISO/IEC
176 9899:1999, and is commonly known as "C99". GCC has incomplete support
177 for this standard version; see
178 `http://gcc.gnu.org/gcc-4.0/c99status.html' for details. To select this
179 standard, use `-std=c99' or `-std=iso9899:1999'. (While in
180 development, drafts of this standard version were referred to as "C9X".)
182 Errors in the 1999 ISO C standard were corrected in two Technical
183 Corrigenda published in 2001 and 2004. GCC does not support the
186 By default, GCC provides some extensions to the C language that on
187 rare occasions conflict with the C standard. *Note Extensions to the C
188 Language Family: C Extensions. Use of the `-std' options listed above
189 will disable these extensions where they conflict with the C standard
190 version selected. You may also select an extended version of the C
191 language explicitly with `-std=gnu89' (for C89 with GNU extensions) or
192 `-std=gnu99' (for C99 with GNU extensions). The default, if no C
193 language dialect options are given, is `-std=gnu89'; this will change to
194 `-std=gnu99' in some future release when the C99 support is complete.
195 Some features that are part of the C99 standard are accepted as
196 extensions in C89 mode.
198 The ISO C standard defines (in clause 4) two classes of conforming
199 implementation. A "conforming hosted implementation" supports the
200 whole standard including all the library facilities; a "conforming
201 freestanding implementation" is only required to provide certain
202 library facilities: those in `<float.h>', `<limits.h>', `<stdarg.h>',
203 and `<stddef.h>'; since AMD1, also those in `<iso646.h>'; and in C99,
204 also those in `<stdbool.h>' and `<stdint.h>'. In addition, complex
205 types, added in C99, are not required for freestanding implementations.
206 The standard also defines two environments for programs, a
207 "freestanding environment", required of all implementations and which
208 may not have library facilities beyond those required of freestanding
209 implementations, where the handling of program startup and termination
210 are implementation-defined, and a "hosted environment", which is not
211 required, in which all the library facilities are provided and startup
212 is through a function `int main (void)' or `int main (int, char *[])'.
213 An OS kernel would be a freestanding environment; a program using the
214 facilities of an operating system would normally be in a hosted
217 GCC aims towards being usable as a conforming freestanding
218 implementation, or as the compiler for a conforming hosted
219 implementation. By default, it will act as the compiler for a hosted
220 implementation, defining `__STDC_HOSTED__' as `1' and presuming that
221 when the names of ISO C functions are used, they have the semantics
222 defined in the standard. To make it act as a conforming freestanding
223 implementation for a freestanding environment, use the option
224 `-ffreestanding'; it will then define `__STDC_HOSTED__' to `0' and not
225 make assumptions about the meanings of function names from the standard
226 library, with exceptions noted below. To build an OS kernel, you may
227 well still need to make your own arrangements for linking and startup.
228 *Note Options Controlling C Dialect: C Dialect Options.
230 GCC does not provide the library facilities required only of hosted
231 implementations, nor yet all the facilities required by C99 of
232 freestanding implementations; to use the facilities of a hosted
233 environment, you will need to find them elsewhere (for example, in the
234 GNU C library). *Note Standard Libraries: Standard Libraries.
236 Most of the compiler support routines used by GCC are present in
237 `libgcc', but there are a few exceptions. GCC requires the
238 freestanding environment provide `memcpy', `memmove', `memset' and
239 `memcmp'. Finally, if `__builtin_trap' is used, and the target does
240 not implement the `trap' pattern, then GCC will emit a call to `abort'.
242 For references to Technical Corrigenda, Rationale documents and
243 information concerning the history of C that is available online, see
244 `http://gcc.gnu.org/readings.html'
246 There is no formal written standard for Objective-C or Objective-C++.
247 The most authoritative manual is "Object-Oriented Programming and the
248 Objective-C Language", available at a number of web sites:
251 `http://developer.apple.com/documentation/Cocoa/Conceptual/ObjectiveC/'
252 is a recent (and periodically updated) version;
254 * `http://www.toodarkpark.org/computers/objc/' is an older example;
256 * `http://www.gnustep.org' and `http://gcc.gnu.org/readings.html'
257 have additional useful information.
259 There is no standard for treelang, which is a sample language front end
260 for GCC. Its only purpose is as a sample for people wishing to write a
261 new language for GCC. The language is documented in
262 `gcc/treelang/treelang.texi' which can be turned into info or HTML
265 *Note GNAT Reference Manual: (gnat_rm)Top, for information on standard
266 conformance and compatibility of the Ada compiler.
268 *Note Standards: (gfortran)Standards, for details of standards
269 supported by `gfortran'.
271 *Note Compatibility with the Java Platform: (gcj)Compatibility, for
272 details of compatibility between `gcj' and the Java Platform.
275 File: gcc.info, Node: Invoking GCC, Next: C Implementation, Prev: Standards, Up: Top
277 3 GCC Command Options
278 *********************
280 When you invoke GCC, it normally does preprocessing, compilation,
281 assembly and linking. The "overall options" allow you to stop this
282 process at an intermediate stage. For example, the `-c' option says
283 not to run the linker. Then the output consists of object files output
286 Other options are passed on to one stage of processing. Some options
287 control the preprocessor and others the compiler itself. Yet other
288 options control the assembler and linker; most of these are not
289 documented here, since you rarely need to use any of them.
291 Most of the command line options that you can use with GCC are useful
292 for C programs; when an option is only useful with another language
293 (usually C++), the explanation says so explicitly. If the description
294 for a particular option does not mention a source language, you can use
295 that option with all supported languages.
297 *Note Compiling C++ Programs: Invoking G++, for a summary of special
298 options for compiling C++ programs.
300 The `gcc' program accepts options and file names as operands. Many
301 options have multi-letter names; therefore multiple single-letter
302 options may _not_ be grouped: `-dr' is very different from `-d -r'.
304 You can mix options and other arguments. For the most part, the order
305 you use doesn't matter. Order does matter when you use several options
306 of the same kind; for example, if you specify `-L' more than once, the
307 directories are searched in the order specified.
309 Many options have long names starting with `-f' or with `-W'--for
310 example, `-fforce-mem', `-fstrength-reduce', `-Wformat' and so on.
311 Most of these have both positive and negative forms; the negative form
312 of `-ffoo' would be `-fno-foo'. This manual documents only one of
313 these two forms, whichever one is not the default.
315 *Note Option Index::, for an index to GCC's options.
319 * Option Summary:: Brief list of all options, without explanations.
320 * Overall Options:: Controlling the kind of output:
321 an executable, object files, assembler files,
322 or preprocessed source.
323 * Invoking G++:: Compiling C++ programs.
324 * C Dialect Options:: Controlling the variant of C language compiled.
325 * C++ Dialect Options:: Variations on C++.
326 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
328 * Language Independent Options:: Controlling how diagnostics should be
330 * Warning Options:: How picky should the compiler be?
331 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
332 * Optimize Options:: How much optimization?
333 * Preprocessor Options:: Controlling header files and macro definitions.
334 Also, getting dependency information for Make.
335 * Assembler Options:: Passing options to the assembler.
336 * Link Options:: Specifying libraries and so on.
337 * Directory Options:: Where to find header files and libraries.
338 Where to find the compiler executable files.
339 * Spec Files:: How to pass switches to sub-processes.
340 * Target Options:: Running a cross-compiler, or an old version of GCC.
341 * Submodel Options:: Specifying minor hardware or convention variations,
342 such as 68010 vs 68020.
343 * Code Gen Options:: Specifying conventions for function calls, data layout
345 * Environment Variables:: Env vars that affect GCC.
346 * Precompiled Headers:: Compiling a header once, and using it many times.
347 * Running Protoize:: Automatically adding or removing function prototypes.
350 File: gcc.info, Node: Option Summary, Next: Overall Options, Up: Invoking GCC
355 Here is a summary of all the options, grouped by type. Explanations are
356 in the following sections.
359 *Note Options Controlling the Kind of Output: Overall Options.
360 -c -S -E -o FILE -combine -pipe -pass-exit-codes
361 -x LANGUAGE -v -### --help --target-help --version
364 *Note Options Controlling C Dialect: C Dialect Options.
365 -ansi -std=STANDARD -aux-info FILENAME
366 -fno-asm -fno-builtin -fno-builtin-FUNCTION
367 -fhosted -ffreestanding -fms-extensions
368 -trigraphs -no-integrated-cpp -traditional -traditional-cpp
369 -fallow-single-precision -fcond-mismatch
370 -fsigned-bitfields -fsigned-char
371 -funsigned-bitfields -funsigned-char
373 _C++ Language Options_
374 *Note Options Controlling C++ Dialect: C++ Dialect Options.
375 -fabi-version=N -fno-access-control -fcheck-new
376 -fconserve-space -fno-const-strings
377 -fno-elide-constructors
378 -fno-enforce-eh-specs
379 -ffor-scope -fno-for-scope -fno-gnu-keywords
380 -fno-implicit-templates
381 -fno-implicit-inline-templates
382 -fno-implement-inlines -fms-extensions
383 -fno-nonansi-builtins -fno-operator-names
384 -fno-optional-diags -fpermissive
385 -frepo -fno-rtti -fstats -ftemplate-depth-N
386 -fno-threadsafe-statics -fuse-cxa-atexit -fno-weak -nostdinc++
387 -fno-default-inline -fvisibility-inlines-hidden
388 -Wabi -Wctor-dtor-privacy
389 -Wnon-virtual-dtor -Wreorder
390 -Weffc++ -Wno-deprecated -Wstrict-null-sentinel
391 -Wno-non-template-friend -Wold-style-cast
392 -Woverloaded-virtual -Wno-pmf-conversions
395 _Objective-C and Objective-C++ Language Options_
396 *Note Options Controlling Objective-C and Objective-C++ Dialects:
397 Objective-C and Objective-C++ Dialect Options.
398 -fconstant-string-class=CLASS-NAME
399 -fgnu-runtime -fnext-runtime
402 -freplace-objc-classes
405 -Wno-protocol -Wselector -Wundeclared-selector
407 _Language Independent Options_
408 *Note Options to Control Diagnostic Messages Formatting: Language
411 -fdiagnostics-show-location=[once|every-line]
414 *Note Options to Request or Suppress Warnings: Warning Options.
415 -fsyntax-only -pedantic -pedantic-errors
416 -w -Wextra -Wall -Waggregate-return
417 -Wcast-align -Wcast-qual -Wchar-subscripts -Wcomment
418 -Wconversion -Wno-deprecated-declarations
419 -Wdisabled-optimization -Wno-div-by-zero -Wno-endif-labels
420 -Werror -Werror-implicit-function-declaration
421 -Wfatal-errors -Wfloat-equal -Wformat -Wformat=2
422 -Wno-format-extra-args -Wformat-nonliteral
423 -Wformat-security -Wformat-y2k
424 -Wimplicit -Wimplicit-function-declaration -Wimplicit-int
425 -Wimport -Wno-import -Winit-self -Winline
426 -Wno-invalid-offsetof -Winvalid-pch
427 -Wlarger-than-LEN -Wlong-long
428 -Wmain -Wmissing-braces -Wmissing-field-initializers
429 -Wmissing-format-attribute -Wmissing-include-dirs
431 -Wno-multichar -Wnonnull -Wpacked -Wpadded
432 -Wparentheses -Wpointer-arith -Wredundant-decls
433 -Wreturn-type -Wsequence-point -Wshadow
434 -Wsign-compare -Wstrict-aliasing -Wstrict-aliasing=2
435 -Wswitch -Wswitch-default -Wswitch-enum
436 -Wsystem-headers -Wtrigraphs -Wundef -Wuninitialized
437 -Wunknown-pragmas -Wunreachable-code
438 -Wunused -Wunused-function -Wunused-label -Wunused-parameter
439 -Wunused-value -Wunused-variable -Wwrite-strings
442 _C-only Warning Options_
443 -Wbad-function-cast -Wmissing-declarations
444 -Wmissing-prototypes -Wnested-externs -Wold-style-definition
445 -Wstrict-prototypes -Wtraditional
446 -Wdeclaration-after-statement -Wno-pointer-sign
449 *Note Options for Debugging Your Program or GCC: Debugging Options.
450 -dLETTERS -dumpspecs -dumpmachine -dumpversion
451 -fdump-unnumbered -fdump-translation-unit[-N]
452 -fdump-class-hierarchy[-N]
453 -fdump-ipa-all -fdump-ipa-cgraph
455 -fdump-tree-original[-N]
456 -fdump-tree-optimized[-N]
457 -fdump-tree-inlined[-N]
458 -fdump-tree-cfg -fdump-tree-vcg -fdump-tree-alias
460 -fdump-tree-ssa[-N] -fdump-tree-pre[-N]
461 -fdump-tree-ccp[-N] -fdump-tree-dce[-N]
462 -fdump-tree-gimple[-raw] -fdump-tree-mudflap[-N]
465 -fdump-tree-phiopt[-N]
466 -fdump-tree-forwprop[-N]
467 -fdump-tree-copyrename[-N]
468 -fdump-tree-nrv -fdump-tree-vect
471 -ftree-vectorizer-verbose=N
472 -feliminate-dwarf2-dups -feliminate-unused-debug-types
473 -feliminate-unused-debug-symbols -fmem-report -fprofile-arcs -ftree-based-profiling
474 -frandom-seed=STRING -fsched-verbose=N
475 -ftest-coverage -ftime-report -fvar-tracking
476 -g -gLEVEL -gcoff -gdwarf-2
477 -ggdb -gstabs -gstabs+ -gvms -gxcoff -gxcoff+
478 -p -pg -print-file-name=LIBRARY -print-libgcc-file-name
479 -print-multi-directory -print-multi-lib
480 -print-prog-name=PROGRAM -print-search-dirs -Q
483 _Optimization Options_
484 *Note Options that Control Optimization: Optimize Options.
485 -falign-functions=N -falign-jumps=N
486 -falign-labels=N -falign-loops=N
487 -fbounds-check -fmudflap -fmudflapth -fmudflapir
488 -fbranch-probabilities -fprofile-values -fvpt -fbranch-target-load-optimize
489 -fbranch-target-load-optimize2 -fbtr-bb-exclusive
490 -fcaller-saves -fcprop-registers -fcse-follow-jumps
491 -fcse-skip-blocks -fcx-limited-range -fdata-sections
492 -fdelayed-branch -fdelete-null-pointer-checks
493 -fexpensive-optimizations -ffast-math -ffloat-store
494 -fforce-addr -fforce-mem -ffunction-sections
495 -fgcse -fgcse-lm -fgcse-sm -fgcse-las -fgcse-after-reload
496 -floop-optimize -fcrossjumping -fif-conversion -fif-conversion2
497 -finline-functions -finline-limit=N -fkeep-inline-functions
498 -fkeep-static-consts -fmerge-constants -fmerge-all-constants
499 -fmodulo-sched -fno-branch-count-reg
500 -fno-default-inline -fno-defer-pop -floop-optimize2 -fmove-loop-invariants
501 -fno-function-cse -fno-guess-branch-probability
502 -fno-inline -fno-math-errno -fno-peephole -fno-peephole2
503 -funsafe-math-optimizations -ffinite-math-only
504 -fno-trapping-math -fno-zero-initialized-in-bss
505 -fomit-frame-pointer -foptimize-register-move
506 -foptimize-sibling-calls -fprefetch-loop-arrays
507 -fprofile-generate -fprofile-use
508 -fregmove -frename-registers
509 -freorder-blocks -freorder-blocks-and-partition -freorder-functions
510 -frerun-cse-after-loop -frerun-loop-opt
511 -frounding-math -fschedule-insns -fschedule-insns2
512 -fno-sched-interblock -fno-sched-spec -fsched-spec-load
513 -fsched-spec-load-dangerous
514 -fsched-stalled-insns=N -sched-stalled-insns-dep=N
515 -fsched2-use-superblocks
516 -fsched2-use-traces -freschedule-modulo-scheduled-loops
517 -fsignaling-nans -fsingle-precision-constant -fspeculative-prefetching
518 -fstrength-reduce -fstrict-aliasing -ftracer -fthread-jumps
519 -funroll-all-loops -funroll-loops -fpeel-loops
520 -fsplit-ivs-in-unroller -funswitch-loops
521 -fvariable-expansion-in-unroller
522 -ftree-pre -ftree-ccp -ftree-dce -ftree-loop-optimize
523 -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon -fivopts
524 -ftree-dominator-opts -ftree-dse -ftree-copyrename
525 -ftree-ch -ftree-sra -ftree-ter -ftree-lrs -ftree-fre -ftree-vectorize
528 -O -O0 -O1 -O2 -O3 -Os
530 _Preprocessor Options_
531 *Note Options Controlling the Preprocessor: Preprocessor Options.
537 -include FILE -imacros FILE
538 -iprefix FILE -iwithprefix DIR
539 -iwithprefixbefore DIR -isystem DIR
540 -M -MM -MF -MG -MP -MQ -MT -nostdinc
541 -P -fworking-directory -remap
542 -trigraphs -undef -UMACRO -Wp,OPTION
543 -Xpreprocessor OPTION
546 *Note Passing Options to the Assembler: Assembler Options.
547 -Wa,OPTION -Xassembler OPTION
550 *Note Options for Linking: Link Options.
551 OBJECT-FILE-NAME -lLIBRARY
552 -nostartfiles -nodefaultlibs -nostdlib -pie
553 -s -static -static-libgcc -shared -shared-libgcc -symbolic
554 -Wl,OPTION -Xlinker OPTION
558 *Note Options for Directory Search: Directory Options.
559 -BPREFIX -IDIR -iquoteDIR -LDIR -specs=FILE -I-
562 *Note Target Options::.
563 -V VERSION -b MACHINE
565 _Machine Dependent Options_
566 *Note Hardware Models and Configurations: Submodel Options.
570 -mmangle-cpu -mcpu=CPU -mtext=TEXT-SECTION
571 -mdata=DATA-SECTION -mrodata=READONLY-DATA-SECTION
574 -mapcs-frame -mno-apcs-frame
576 -mapcs-stack-check -mno-apcs-stack-check
577 -mapcs-float -mno-apcs-float
578 -mapcs-reentrant -mno-apcs-reentrant
579 -msched-prolog -mno-sched-prolog
580 -mlittle-endian -mbig-endian -mwords-little-endian
581 -mfloat-abi=NAME -msoft-float -mhard-float -mfpe
582 -mthumb-interwork -mno-thumb-interwork
583 -mcpu=NAME -march=NAME -mfpu=NAME
584 -mstructure-size-boundary=N
586 -mlong-calls -mno-long-calls
587 -msingle-pic-base -mno-single-pic-base
590 -mcirrus-fix-invalid-insns -mno-cirrus-fix-invalid-insns
593 -mtpcs-frame -mtpcs-leaf-frame
594 -mcaller-super-interworking -mcallee-super-interworking
597 -mmcu=MCU -msize -minit-stack=N -mno-interrupts
598 -mcall-prologues -mno-tablejump -mtiny-stack -mint8
601 -momit-leaf-frame-pointer -mno-omit-leaf-frame-pointer -mcsync
602 -mno-csync -mlow-64k -mno-low64k -mid-shared-library
603 -mno-id-shared-library -mshared-library-id=N
606 -mcpu=CPU -march=CPU -mtune=CPU
607 -mmax-stack-frame=N -melinux-stacksize=N
608 -metrax4 -metrax100 -mpdebug -mcc-init -mno-side-effects
609 -mstack-align -mdata-align -mconst-align
610 -m32-bit -m16-bit -m8-bit -mno-prologue-epilogue -mno-gotplt
611 -melf -maout -melinux -mlinux -sim -sim2
612 -mmul-bug-workaround -mno-mul-bug-workaround
615 -all_load -allowable_client -arch -arch_errors_fatal
616 -arch_only -bind_at_load -bundle -bundle_loader
617 -client_name -compatibility_version -current_version
619 -dependency-file -dylib_file -dylinker_install_name
620 -dynamic -dynamiclib -exported_symbols_list
621 -filelist -flat_namespace -force_cpusubtype_ALL
622 -force_flat_namespace -headerpad_max_install_names
623 -image_base -init -install_name -keep_private_externs
624 -multi_module -multiply_defined -multiply_defined_unused
625 -noall_load -no_dead_strip_inits_and_terms
626 -nofixprebinding -nomultidefs -noprebind -noseglinkedit
627 -pagezero_size -prebind -prebind_all_twolevel_modules
628 -private_bundle -read_only_relocs -sectalign
629 -sectobjectsymbols -whyload -seg1addr
630 -sectcreate -sectobjectsymbols -sectorder
631 -segaddr -segs_read_only_addr -segs_read_write_addr
632 -seg_addr_table -seg_addr_table_filename -seglinkedit
633 -segprot -segs_read_only_addr -segs_read_write_addr
634 -single_module -static -sub_library -sub_umbrella
635 -twolevel_namespace -umbrella -undefined
636 -unexported_symbols_list -weak_reference_mismatches
637 -whatsloaded -F -gused -gfull -mone-byte-bool
640 -mno-fp-regs -msoft-float -malpha-as -mgas
641 -mieee -mieee-with-inexact -mieee-conformant
642 -mfp-trap-mode=MODE -mfp-rounding-mode=MODE
643 -mtrap-precision=MODE -mbuild-constants
644 -mcpu=CPU-TYPE -mtune=CPU-TYPE
645 -mbwx -mmax -mfix -mcix
646 -mfloat-vax -mfloat-ieee
647 -mexplicit-relocs -msmall-data -mlarge-data
648 -msmall-text -mlarge-text
649 -mmemory-latency=TIME
651 _DEC Alpha/VMS Options_
655 -mgpr-32 -mgpr-64 -mfpr-32 -mfpr-64
656 -mhard-float -msoft-float
657 -malloc-cc -mfixed-cc -mdword -mno-dword
659 -mmedia -mno-media -mmuladd -mno-muladd
660 -mfdpic -minline-plt -mgprel-ro -multilib-library-pic
661 -mlinked-fp -mlong-calls -malign-labels
662 -mlibrary-pic -macc-4 -macc-8
663 -mpack -mno-pack -mno-eflags -mcond-move -mno-cond-move
664 -mscc -mno-scc -mcond-exec -mno-cond-exec
665 -mvliw-branch -mno-vliw-branch
666 -mmulti-cond-exec -mno-multi-cond-exec -mnested-cond-exec
667 -mno-nested-cond-exec -mtomcat-stats
672 -mrelax -mh -ms -mn -mint32 -malign-300
675 -march=ARCHITECTURE-TYPE
676 -mbig-switch -mdisable-fpregs -mdisable-indexing
677 -mfast-indirect-calls -mgas -mgnu-ld -mhp-ld
678 -mfixed-range=REGISTER-RANGE
679 -mjump-in-delay -mlinker-opt -mlong-calls
680 -mlong-load-store -mno-big-switch -mno-disable-fpregs
681 -mno-disable-indexing -mno-fast-indirect-calls -mno-gas
682 -mno-jump-in-delay -mno-long-load-store
683 -mno-portable-runtime -mno-soft-float
684 -mno-space-regs -msoft-float -mpa-risc-1-0
685 -mpa-risc-1-1 -mpa-risc-2-0 -mportable-runtime
686 -mschedule=CPU-TYPE -mspace-regs -msio -mwsio
687 -munix=UNIX-STD -nolibdld -static -threads
689 _i386 and x86-64 Options_
690 -mtune=CPU-TYPE -march=CPU-TYPE
692 -masm=DIALECT -mno-fancy-math-387
693 -mno-fp-ret-in-387 -msoft-float -msvr3-shlib
694 -mno-wide-multiply -mrtd -malign-double
695 -mpreferred-stack-boundary=NUM
696 -mmmx -msse -msse2 -msse3 -m3dnow
697 -mthreads -mno-align-stringops -minline-all-stringops
698 -mpush-args -maccumulate-outgoing-args -m128bit-long-double
699 -m96bit-long-double -mregparm=NUM -momit-leaf-frame-pointer
700 -mno-red-zone -mno-tls-direct-seg-refs
705 -mbig-endian -mlittle-endian -mgnu-as -mgnu-ld -mno-pic
706 -mvolatile-asm-stop -mregister-names -mno-sdata
707 -mconstant-gp -mauto-pic -minline-float-divide-min-latency
708 -minline-float-divide-max-throughput
709 -minline-int-divide-min-latency
710 -minline-int-divide-max-throughput
711 -minline-sqrt-min-latency -minline-sqrt-max-throughput
712 -mno-dwarf2-asm -mearly-stop-bits
713 -mfixed-range=REGISTER-RANGE -mtls-size=TLS-SIZE
714 -mtune=CPU-TYPE -mt -pthread -milp32 -mlp64
719 -malign-loops -mno-align-loops
722 -mmodel=CODE-SIZE-MODEL-TYPE
724 -mno-flush-func -mflush-func=NAME
725 -mno-flush-trap -mflush-trap=NUMBER
729 -m68000 -m68020 -m68020-40 -m68020-60 -m68030 -m68040
730 -m68060 -mcpu32 -m5200 -m68881 -mbitfield -mc68000 -mc68020
731 -mnobitfield -mrtd -mshort -msoft-float -mpcrel
732 -malign-int -mstrict-align -msep-data -mno-sep-data
733 -mshared-library-id=n -mid-shared-library -mno-id-shared-library
736 -m6811 -m6812 -m68hc11 -m68hc12 -m68hcs12
737 -mauto-incdec -minmax -mlong-calls -mshort
738 -msoft-reg-count=COUNT
741 -mhardlit -mno-hardlit -mdiv -mno-div -mrelax-immediates
742 -mno-relax-immediates -mwide-bitfields -mno-wide-bitfields
743 -m4byte-functions -mno-4byte-functions -mcallgraph-data
744 -mno-callgraph-data -mslow-bytes -mno-slow-bytes -mno-lsim
745 -mlittle-endian -mbig-endian -m210 -m340 -mstack-increment
748 -EL -EB -march=ARCH -mtune=ARCH
749 -mips1 -mips2 -mips3 -mips4 -mips32 -mips32r2 -mips64
750 -mips16 -mno-mips16 -mabi=ABI -mabicalls -mno-abicalls
751 -mxgot -mno-xgot -mgp32 -mgp64 -mfp32 -mfp64
752 -mhard-float -msoft-float -msingle-float -mdouble-float
753 -mpaired-single -mips3d
754 -mint64 -mlong64 -mlong32 -msym32 -mno-sym32
755 -GNUM -membedded-data -mno-embedded-data
756 -muninit-const-in-rodata -mno-uninit-const-in-rodata
757 -msplit-addresses -mno-split-addresses
758 -mexplicit-relocs -mno-explicit-relocs
759 -mcheck-zero-division -mno-check-zero-division
760 -mdivide-traps -mdivide-breaks
761 -mmemcpy -mno-memcpy -mlong-calls -mno-long-calls
762 -mmad -mno-mad -mfused-madd -mno-fused-madd -nocpp
763 -mfix-r4000 -mno-fix-r4000 -mfix-r4400 -mno-fix-r4400
764 -mfix-vr4120 -mno-fix-vr4120 -mfix-vr4130
765 -mfix-sb1 -mno-fix-sb1
766 -mflush-func=FUNC -mno-flush-func
767 -mbranch-likely -mno-branch-likely
768 -mfp-exceptions -mno-fp-exceptions
769 -mvr4130-align -mno-vr4130-align
772 -mlibfuncs -mno-libfuncs -mepsilon -mno-epsilon -mabi=gnu
773 -mabi=mmixware -mzero-extend -mknuthdiv -mtoplevel-symbols
774 -melf -mbranch-predict -mno-branch-predict -mbase-addresses
775 -mno-base-addresses -msingle-exit -mno-single-exit
778 -mmult-bug -mno-mult-bug
784 -m32032 -m32332 -m32532 -m32081 -m32381
785 -mmult-add -mnomult-add -msoft-float -mrtd -mnortd
786 -mregparam -mnoregparam -msb -mnosb
787 -mbitfield -mnobitfield -mhimem -mnohimem
790 -mfpu -msoft-float -mac0 -mno-ac0 -m40 -m45 -m10
791 -mbcopy -mbcopy-builtin -mint32 -mno-int16
792 -mint16 -mno-int32 -mfloat32 -mno-float64
793 -mfloat64 -mno-float32 -mabshi -mno-abshi
794 -mbranch-expensive -mbranch-cheap
795 -msplit -mno-split -munix-asm -mdec-asm
797 _PowerPC Options_ See RS/6000 and PowerPC Options.
799 _RS/6000 and PowerPC Options_
802 -mpower -mno-power -mpower2 -mno-power2
803 -mpowerpc -mpowerpc64 -mno-powerpc
804 -maltivec -mno-altivec
805 -mpowerpc-gpopt -mno-powerpc-gpopt
806 -mpowerpc-gfxopt -mno-powerpc-gfxopt
807 -mnew-mnemonics -mold-mnemonics
808 -mfull-toc -mminimal-toc -mno-fp-in-toc -mno-sum-in-toc
809 -m64 -m32 -mxl-compat -mno-xl-compat -mpe
810 -malign-power -malign-natural
811 -msoft-float -mhard-float -mmultiple -mno-multiple
812 -mstring -mno-string -mupdate -mno-update
813 -mfused-madd -mno-fused-madd -mbit-align -mno-bit-align
814 -mstrict-align -mno-strict-align -mrelocatable
815 -mno-relocatable -mrelocatable-lib -mno-relocatable-lib
816 -mtoc -mno-toc -mlittle -mlittle-endian -mbig -mbig-endian
818 -mprioritize-restricted-insns=PRIORITY
819 -msched-costly-dep=DEPENDENCE_TYPE
820 -minsert-sched-nops=SCHEME
821 -mcall-sysv -mcall-netbsd
822 -maix-struct-return -msvr4-struct-return
823 -mabi=altivec -mabi=no-altivec
824 -mabi=spe -mabi=no-spe
827 -mfloat-gprs=yes -mfloat-gprs=no -mfloat-gprs=single -mfloat-gprs=double
828 -mprototype -mno-prototype
829 -msim -mmvme -mads -myellowknife -memb -msdata
830 -msdata=OPT -mvxworks -mwindiss -G NUM -pthread
832 _S/390 and zSeries Options_
833 -mtune=CPU-TYPE -march=CPU-TYPE
834 -mhard-float -msoft-float -mbackchain -mno-backchain
835 -mpacked-stack -mno-packed-stack
836 -msmall-exec -mno-small-exec -mmvcle -mno-mvcle
837 -m64 -m31 -mdebug -mno-debug -mesa -mzarch
838 -mtpf-trace -mno-tpf-trace -mfused-madd -mno-fused-madd
839 -mwarn-framesize -mwarn-dynamicstack -mstack-size -mstack-guard
842 -m1 -m2 -m2e -m3 -m3e
843 -m4-nofpu -m4-single-only -m4-single -m4
844 -m4a-nofpu -m4a-single-only -m4a-single -m4a -m4al
845 -m5-64media -m5-64media-nofpu
846 -m5-32media -m5-32media-nofpu
847 -m5-compact -m5-compact-nofpu
848 -mb -ml -mdalign -mrelax
849 -mbigtable -mfmovd -mhitachi -mrenesas -mno-renesas -mnomacsave
850 -mieee -misize -mpadstruct -mspace
851 -mprefergot -musermode
857 -m32 -m64 -mapp-regs -mno-app-regs
858 -mfaster-structs -mno-faster-structs
859 -mfpu -mno-fpu -mhard-float -msoft-float
860 -mhard-quad-float -msoft-quad-float
861 -mimpure-text -mno-impure-text -mlittle-endian
862 -mstack-bias -mno-stack-bias
863 -munaligned-doubles -mno-unaligned-doubles
864 -mv8plus -mno-v8plus -mvis -mno-vis
868 -Qy -Qn -YP,PATHS -Ym,DIR
870 _TMS320C3x/C4x Options_
871 -mcpu=CPU -mbig -msmall -mregparm -mmemparm
872 -mfast-fix -mmpyi -mbk -mti -mdp-isr-reload
873 -mrpts=COUNT -mrptb -mdb -mloop-unsigned
874 -mparallel-insns -mparallel-mpy -mpreserve-float
877 -mlong-calls -mno-long-calls -mep -mno-ep
878 -mprolog-function -mno-prolog-function -mspace
879 -mtda=N -msda=N -mzda=N
880 -mapp-regs -mno-app-regs
881 -mdisable-callt -mno-disable-callt
889 _x86-64 Options_ See i386 and x86-64 Options.
895 -mconst16 -mno-const16
896 -mfused-madd -mno-fused-madd
897 -mtext-section-literals -mno-text-section-literals
898 -mtarget-align -mno-target-align
899 -mlongcalls -mno-longcalls
901 _zSeries Options_ See S/390 and zSeries Options.
903 _Code Generation Options_
904 *Note Options for Code Generation Conventions: Code Gen Options.
905 -fcall-saved-REG -fcall-used-REG
906 -ffixed-REG -fexceptions
907 -fnon-call-exceptions -funwind-tables
908 -fasynchronous-unwind-tables
909 -finhibit-size-directive -finstrument-functions
910 -fno-common -fno-ident
911 -fpcc-struct-return -fpic -fPIC -fpie -fPIE
912 -freg-struct-return -fshared-data -fshort-enums
913 -fshort-double -fshort-wchar
914 -fverbose-asm -fpack-struct[=N] -fstack-check
915 -fstack-limit-register=REG -fstack-limit-symbol=SYM
916 -fargument-alias -fargument-noalias
917 -fargument-noalias-global -fleading-underscore
919 -ftrapv -fwrapv -fbounds-check
925 * Overall Options:: Controlling the kind of output:
926 an executable, object files, assembler files,
927 or preprocessed source.
928 * C Dialect Options:: Controlling the variant of C language compiled.
929 * C++ Dialect Options:: Variations on C++.
930 * Objective-C and Objective-C++ Dialect Options:: Variations on Objective-C
932 * Language Independent Options:: Controlling how diagnostics should be
934 * Warning Options:: How picky should the compiler be?
935 * Debugging Options:: Symbol tables, measurements, and debugging dumps.
936 * Optimize Options:: How much optimization?
937 * Preprocessor Options:: Controlling header files and macro definitions.
938 Also, getting dependency information for Make.
939 * Assembler Options:: Passing options to the assembler.
940 * Link Options:: Specifying libraries and so on.
941 * Directory Options:: Where to find header files and libraries.
942 Where to find the compiler executable files.
943 * Spec Files:: How to pass switches to sub-processes.
944 * Target Options:: Running a cross-compiler, or an old version of GCC.
947 File: gcc.info, Node: Overall Options, Next: Invoking G++, Prev: Option Summary, Up: Invoking GCC
949 3.2 Options Controlling the Kind of Output
950 ==========================================
952 Compilation can involve up to four stages: preprocessing, compilation
953 proper, assembly and linking, always in that order. GCC is capable of
954 preprocessing and compiling several files either into several assembler
955 input files, or into one assembler input file; then each assembler
956 input file produces an object file, and linking combines all the object
957 files (those newly compiled, and those specified as input) into an
960 For any given input file, the file name suffix determines what kind of
964 C source code which must be preprocessed.
967 C source code which should not be preprocessed.
970 C++ source code which should not be preprocessed.
973 Objective-C source code. Note that you must link with the
974 `libobjc' library to make an Objective-C program work.
977 Objective-C source code which should not be preprocessed.
981 Objective-C++ source code. Note that you must link with the
982 `libobjc' library to make an Objective-C++ program work. Note
983 that `.M' refers to a literal capital M.
986 Objective-C++ source code which should not be preprocessed.
989 C, C++, Objective-C or Objective-C++ header file to be turned into
990 a precompiled header.
999 C++ source code which must be preprocessed. Note that in `.cxx',
1000 the last two letters must both be literally `x'. Likewise, `.C'
1001 refers to a literal capital C.
1005 C++ header file to be turned into a precompiled header.
1010 Fortran source code which should not be preprocessed.
1015 Fortran source code which must be preprocessed (with the
1016 traditional preprocessor).
1019 Fortran source code which must be preprocessed with a RATFOR
1020 preprocessor (not included with GCC).
1024 Fortran 90/95 source code which should not be preprocessed.
1027 Ada source code file which contains a library unit declaration (a
1028 declaration of a package, subprogram, or generic, or a generic
1029 instantiation), or a library unit renaming declaration (a package,
1030 generic, or subprogram renaming declaration). Such files are also
1034 Ada source code file containing a library unit body (a subprogram
1035 or package body). Such files are also called "bodies".
1041 Assembler code which must be preprocessed.
1044 An object file to be fed straight into linking. Any file name
1045 with no recognized suffix is treated this way.
1047 You can specify the input language explicitly with the `-x' option:
1050 Specify explicitly the LANGUAGE for the following input files
1051 (rather than letting the compiler choose a default based on the
1052 file name suffix). This option applies to all following input
1053 files until the next `-x' option. Possible values for LANGUAGE
1055 c c-header c-cpp-output
1056 c++ c++-header c++-cpp-output
1057 objective-c objective-c-header objective-c-cpp-output
1058 objective-c++ objective-c++-header objective-c++-cpp-output
1059 assembler assembler-with-cpp
1061 f77 f77-cpp-input ratfor
1067 Turn off any specification of a language, so that subsequent files
1068 are handled according to their file name suffixes (as they are if
1069 `-x' has not been used at all).
1072 Normally the `gcc' program will exit with the code of 1 if any
1073 phase of the compiler returns a non-success return code. If you
1074 specify `-pass-exit-codes', the `gcc' program will instead return
1075 with numerically highest error produced by any phase that returned
1076 an error indication.
1078 If you only want some of the stages of compilation, you can use `-x'
1079 (or filename suffixes) to tell `gcc' where to start, and one of the
1080 options `-c', `-S', or `-E' to say where `gcc' is to stop. Note that
1081 some combinations (for example, `-x cpp-output -E') instruct `gcc' to
1085 Compile or assemble the source files, but do not link. The linking
1086 stage simply is not done. The ultimate output is in the form of an
1087 object file for each source file.
1089 By default, the object file name for a source file is made by
1090 replacing the suffix `.c', `.i', `.s', etc., with `.o'.
1092 Unrecognized input files, not requiring compilation or assembly,
1096 Stop after the stage of compilation proper; do not assemble. The
1097 output is in the form of an assembler code file for each
1098 non-assembler input file specified.
1100 By default, the assembler file name for a source file is made by
1101 replacing the suffix `.c', `.i', etc., with `.s'.
1103 Input files that don't require compilation are ignored.
1106 Stop after the preprocessing stage; do not run the compiler
1107 proper. The output is in the form of preprocessed source code,
1108 which is sent to the standard output.
1110 Input files which don't require preprocessing are ignored.
1113 Place output in file FILE. This applies regardless to whatever
1114 sort of output is being produced, whether it be an executable file,
1115 an object file, an assembler file or preprocessed C code.
1117 If `-o' is not specified, the default is to put an executable file
1118 in `a.out', the object file for `SOURCE.SUFFIX' in `SOURCE.o', its
1119 assembler file in `SOURCE.s', a precompiled header file in
1120 `SOURCE.SUFFIX.gch', and all preprocessed C source on standard
1124 Print (on standard error output) the commands executed to run the
1125 stages of compilation. Also print the version number of the
1126 compiler driver program and of the preprocessor and the compiler
1130 Like `-v' except the commands are not executed and all command
1131 arguments are quoted. This is useful for shell scripts to capture
1132 the driver-generated command lines.
1135 Use pipes rather than temporary files for communication between the
1136 various stages of compilation. This fails to work on some systems
1137 where the assembler is unable to read from a pipe; but the GNU
1138 assembler has no trouble.
1141 If you are compiling multiple source files, this option tells the
1142 driver to pass all the source files to the compiler at once (for
1143 those languages for which the compiler can handle this). This
1144 will allow intermodule analysis (IMA) to be performed by the
1145 compiler. Currently the only language for which this is supported
1146 is C. If you pass source files for multiple languages to the
1147 driver, using this option, the driver will invoke the compiler(s)
1148 that support IMA once each, passing each compiler all the source
1149 files appropriate for it. For those languages that do not support
1150 IMA this option will be ignored, and the compiler will be invoked
1151 once for each source file in that language. If you use this
1152 option in conjunction with `-save-temps', the compiler will
1153 generate multiple pre-processed files (one for each source file),
1154 but only one (combined) `.o' or `.s' file.
1157 Print (on the standard output) a description of the command line
1158 options understood by `gcc'. If the `-v' option is also specified
1159 then `--help' will also be passed on to the various processes
1160 invoked by `gcc', so that they can display the command line options
1161 they accept. If the `-Wextra' option is also specified then
1162 command line options which have no documentation associated with
1163 them will also be displayed.
1166 Print (on the standard output) a description of target specific
1167 command line options for each tool.
1170 Display the version number and copyrights of the invoked GCC.
1173 File: gcc.info, Node: Invoking G++, Next: C Dialect Options, Prev: Overall Options, Up: Invoking GCC
1175 3.3 Compiling C++ Programs
1176 ==========================
1178 C++ source files conventionally use one of the suffixes `.C', `.cc',
1179 `.cpp', `.CPP', `.c++', `.cp', or `.cxx'; C++ header files often use
1180 `.hh' or `.H'; and preprocessed C++ files use the suffix `.ii'. GCC
1181 recognizes files with these names and compiles them as C++ programs
1182 even if you call the compiler the same way as for compiling C programs
1183 (usually with the name `gcc').
1185 However, C++ programs often require class libraries as well as a
1186 compiler that understands the C++ language--and under some
1187 circumstances, you might want to compile programs or header files from
1188 standard input, or otherwise without a suffix that flags them as C++
1189 programs. You might also like to precompile a C header file with a
1190 `.h' extension to be used in C++ compilations. `g++' is a program that
1191 calls GCC with the default language set to C++, and automatically
1192 specifies linking against the C++ library. On many systems, `g++' is
1193 also installed with the name `c++'.
1195 When you compile C++ programs, you may specify many of the same
1196 command-line options that you use for compiling programs in any
1197 language; or command-line options meaningful for C and related
1198 languages; or options that are meaningful only for C++ programs. *Note
1199 Options Controlling C Dialect: C Dialect Options, for explanations of
1200 options for languages related to C. *Note Options Controlling C++
1201 Dialect: C++ Dialect Options, for explanations of options that are
1202 meaningful only for C++ programs.
1205 File: gcc.info, Node: C Dialect Options, Next: C++ Dialect Options, Prev: Invoking G++, Up: Invoking GCC
1207 3.4 Options Controlling C Dialect
1208 =================================
1210 The following options control the dialect of C (or languages derived
1211 from C, such as C++, Objective-C and Objective-C++) that the compiler
1215 In C mode, support all ISO C90 programs. In C++ mode, remove GNU
1216 extensions that conflict with ISO C++.
1218 This turns off certain features of GCC that are incompatible with
1219 ISO C90 (when compiling C code), or of standard C++ (when
1220 compiling C++ code), such as the `asm' and `typeof' keywords, and
1221 predefined macros such as `unix' and `vax' that identify the type
1222 of system you are using. It also enables the undesirable and
1223 rarely used ISO trigraph feature. For the C compiler, it disables
1224 recognition of C++ style `//' comments as well as the `inline'
1227 The alternate keywords `__asm__', `__extension__', `__inline__'
1228 and `__typeof__' continue to work despite `-ansi'. You would not
1229 want to use them in an ISO C program, of course, but it is useful
1230 to put them in header files that might be included in compilations
1231 done with `-ansi'. Alternate predefined macros such as `__unix__'
1232 and `__vax__' are also available, with or without `-ansi'.
1234 The `-ansi' option does not cause non-ISO programs to be rejected
1235 gratuitously. For that, `-pedantic' is required in addition to
1236 `-ansi'. *Note Warning Options::.
1238 The macro `__STRICT_ANSI__' is predefined when the `-ansi' option
1239 is used. Some header files may notice this macro and refrain from
1240 declaring certain functions or defining certain macros that the
1241 ISO standard doesn't call for; this is to avoid interfering with
1242 any programs that might use these names for other things.
1244 Functions which would normally be built in but do not have
1245 semantics defined by ISO C (such as `alloca' and `ffs') are not
1246 built-in functions with `-ansi' is used. *Note Other built-in
1247 functions provided by GCC: Other Builtins, for details of the
1251 Determine the language standard. This option is currently only
1252 supported when compiling C or C++. A value for this option must be
1253 provided; possible values are
1257 ISO C90 (same as `-ansi').
1260 ISO C90 as modified in amendment 1.
1266 ISO C99. Note that this standard is not yet fully supported;
1267 see `http://gcc.gnu.org/gcc-4.0/c99status.html' for more
1268 information. The names `c9x' and `iso9899:199x' are
1272 Default, ISO C90 plus GNU extensions (including some C99
1277 ISO C99 plus GNU extensions. When ISO C99 is fully
1278 implemented in GCC, this will become the default. The name
1279 `gnu9x' is deprecated.
1282 The 1998 ISO C++ standard plus amendments.
1285 The same as `-std=c++98' plus GNU extensions. This is the
1286 default for C++ code.
1288 Even when this option is not specified, you can still use some of
1289 the features of newer standards in so far as they do not conflict
1290 with previous C standards. For example, you may use
1291 `__restrict__' even when `-std=c99' is not specified.
1293 The `-std' options specifying some version of ISO C have the same
1294 effects as `-ansi', except that features that were not in ISO C90
1295 but are in the specified version (for example, `//' comments and
1296 the `inline' keyword in ISO C99) are not disabled.
1298 *Note Language Standards Supported by GCC: Standards, for details
1299 of these standard versions.
1301 `-aux-info FILENAME'
1302 Output to the given filename prototyped declarations for all
1303 functions declared and/or defined in a translation unit, including
1304 those in header files. This option is silently ignored in any
1305 language other than C.
1307 Besides declarations, the file indicates, in comments, the origin
1308 of each declaration (source file and line), whether the
1309 declaration was implicit, prototyped or unprototyped (`I', `N' for
1310 new or `O' for old, respectively, in the first character after the
1311 line number and the colon), and whether it came from a declaration
1312 or a definition (`C' or `F', respectively, in the following
1313 character). In the case of function definitions, a K&R-style list
1314 of arguments followed by their declarations is also provided,
1315 inside comments, after the declaration.
1318 Do not recognize `asm', `inline' or `typeof' as a keyword, so that
1319 code can use these words as identifiers. You can use the keywords
1320 `__asm__', `__inline__' and `__typeof__' instead. `-ansi' implies
1323 In C++, this switch only affects the `typeof' keyword, since `asm'
1324 and `inline' are standard keywords. You may want to use the
1325 `-fno-gnu-keywords' flag instead, which has the same effect. In
1326 C99 mode (`-std=c99' or `-std=gnu99'), this switch only affects
1327 the `asm' and `typeof' keywords, since `inline' is a standard
1331 `-fno-builtin-FUNCTION'
1332 Don't recognize built-in functions that do not begin with
1333 `__builtin_' as prefix. *Note Other built-in functions provided
1334 by GCC: Other Builtins, for details of the functions affected,
1335 including those which are not built-in functions when `-ansi' or
1336 `-std' options for strict ISO C conformance are used because they
1337 do not have an ISO standard meaning.
1339 GCC normally generates special code to handle certain built-in
1340 functions more efficiently; for instance, calls to `alloca' may
1341 become single instructions that adjust the stack directly, and
1342 calls to `memcpy' may become inline copy loops. The resulting
1343 code is often both smaller and faster, but since the function
1344 calls no longer appear as such, you cannot set a breakpoint on
1345 those calls, nor can you change the behavior of the functions by
1346 linking with a different library. In addition, when a function is
1347 recognized as a built-in function, GCC may use information about
1348 that function to warn about problems with calls to that function,
1349 or to generate more efficient code, even if the resulting code
1350 still contains calls to that function. For example, warnings are
1351 given with `-Wformat' for bad calls to `printf', when `printf' is
1352 built in, and `strlen' is known not to modify global memory.
1354 With the `-fno-builtin-FUNCTION' option only the built-in function
1355 FUNCTION is disabled. FUNCTION must not begin with `__builtin_'.
1356 If a function is named this is not built-in in this version of
1357 GCC, this option is ignored. There is no corresponding
1358 `-fbuiltin-FUNCTION' option; if you wish to enable built-in
1359 functions selectively when using `-fno-builtin' or
1360 `-ffreestanding', you may define macros such as:
1362 #define abs(n) __builtin_abs ((n))
1363 #define strcpy(d, s) __builtin_strcpy ((d), (s))
1366 Assert that compilation takes place in a hosted environment. This
1367 implies `-fbuiltin'. A hosted environment is one in which the
1368 entire standard library is available, and in which `main' has a
1369 return type of `int'. Examples are nearly everything except a
1370 kernel. This is equivalent to `-fno-freestanding'.
1373 Assert that compilation takes place in a freestanding environment.
1374 This implies `-fno-builtin'. A freestanding environment is one
1375 in which the standard library may not exist, and program startup
1376 may not necessarily be at `main'. The most obvious example is an
1377 OS kernel. This is equivalent to `-fno-hosted'.
1379 *Note Language Standards Supported by GCC: Standards, for details
1380 of freestanding and hosted environments.
1383 Accept some non-standard constructs used in Microsoft header files.
1385 Some cases of unnamed fields in structures and unions are only
1386 accepted with this option. *Note Unnamed struct/union fields
1387 within structs/unions: Unnamed Fields, for details.
1390 Support ISO C trigraphs. The `-ansi' option (and `-std' options
1391 for strict ISO C conformance) implies `-trigraphs'.
1393 `-no-integrated-cpp'
1394 Performs a compilation in two passes: preprocessing and compiling.
1395 This option allows a user supplied "cc1", "cc1plus", or "cc1obj"
1396 via the `-B' option. The user supplied compilation step can then
1397 add in an additional preprocessing step after normal preprocessing
1398 but before compiling. The default is to use the integrated cpp
1401 The semantics of this option will change if "cc1", "cc1plus", and
1402 "cc1obj" are merged.
1406 Formerly, these options caused GCC to attempt to emulate a
1407 pre-standard C compiler. They are now only supported with the
1408 `-E' switch. The preprocessor continues to support a pre-standard
1409 mode. See the GNU CPP manual for details.
1412 Allow conditional expressions with mismatched types in the second
1413 and third arguments. The value of such an expression is void.
1414 This option is not supported for C++.
1417 Let the type `char' be unsigned, like `unsigned char'.
1419 Each kind of machine has a default for what `char' should be. It
1420 is either like `unsigned char' by default or like `signed char' by
1423 Ideally, a portable program should always use `signed char' or
1424 `unsigned char' when it depends on the signedness of an object.
1425 But many programs have been written to use plain `char' and expect
1426 it to be signed, or expect it to be unsigned, depending on the
1427 machines they were written for. This option, and its inverse, let
1428 you make such a program work with the opposite default.
1430 The type `char' is always a distinct type from each of `signed
1431 char' or `unsigned char', even though its behavior is always just
1432 like one of those two.
1435 Let the type `char' be signed, like `signed char'.
1437 Note that this is equivalent to `-fno-unsigned-char', which is the
1438 negative form of `-funsigned-char'. Likewise, the option
1439 `-fno-signed-char' is equivalent to `-funsigned-char'.
1441 `-fsigned-bitfields'
1442 `-funsigned-bitfields'
1443 `-fno-signed-bitfields'
1444 `-fno-unsigned-bitfields'
1445 These options control whether a bit-field is signed or unsigned,
1446 when the declaration does not use either `signed' or `unsigned'.
1447 By default, such a bit-field is signed, because this is
1448 consistent: the basic integer types such as `int' are signed types.
1451 File: gcc.info, Node: C++ Dialect Options, Next: Objective-C and Objective-C++ Dialect Options, Prev: C Dialect Options, Up: Invoking GCC
1453 3.5 Options Controlling C++ Dialect
1454 ===================================
1456 This section describes the command-line options that are only meaningful
1457 for C++ programs; but you can also use most of the GNU compiler options
1458 regardless of what language your program is in. For example, you might
1459 compile a file `firstClass.C' like this:
1461 g++ -g -frepo -O -c firstClass.C
1463 In this example, only `-frepo' is an option meant only for C++
1464 programs; you can use the other options with any language supported by
1467 Here is a list of options that are _only_ for compiling C++ programs:
1470 Use version N of the C++ ABI. Version 2 is the version of the C++
1471 ABI that first appeared in G++ 3.4. Version 1 is the version of
1472 the C++ ABI that first appeared in G++ 3.2. Version 0 will always
1473 be the version that conforms most closely to the C++ ABI
1474 specification. Therefore, the ABI obtained using version 0 will
1475 change as ABI bugs are fixed.
1477 The default is version 2.
1479 `-fno-access-control'
1480 Turn off all access checking. This switch is mainly useful for
1481 working around bugs in the access control code.
1484 Check that the pointer returned by `operator new' is non-null
1485 before attempting to modify the storage allocated. This check is
1486 normally unnecessary because the C++ standard specifies that
1487 `operator new' will only return `0' if it is declared `throw()',
1488 in which case the compiler will always check the return value even
1489 without this option. In all other cases, when `operator new' has
1490 a non-empty exception specification, memory exhaustion is
1491 signalled by throwing `std::bad_alloc'. See also `new (nothrow)'.
1494 Put uninitialized or runtime-initialized global variables into the
1495 common segment, as C does. This saves space in the executable at
1496 the cost of not diagnosing duplicate definitions. If you compile
1497 with this flag and your program mysteriously crashes after
1498 `main()' has completed, you may have an object that is being
1499 destroyed twice because two definitions were merged.
1501 This option is no longer useful on most targets, now that support
1502 has been added for putting variables into BSS without making them
1505 `-fno-const-strings'
1506 Give string constants type `char *' instead of type `const char
1507 *'. By default, G++ uses type `const char *' as required by the
1508 standard. Even if you use `-fno-const-strings', you cannot
1509 actually modify the value of a string constant.
1511 This option might be removed in a future release of G++. For
1512 maximum portability, you should structure your code so that it
1513 works with string constants that have type `const char *'.
1515 `-fno-elide-constructors'
1516 The C++ standard allows an implementation to omit creating a
1517 temporary which is only used to initialize another object of the
1518 same type. Specifying this option disables that optimization, and
1519 forces G++ to call the copy constructor in all cases.
1521 `-fno-enforce-eh-specs'
1522 Don't check for violation of exception specifications at runtime.
1523 This option violates the C++ standard, but may be useful for
1524 reducing code size in production builds, much like defining
1525 `NDEBUG'. The compiler will still optimize based on the exception
1530 If `-ffor-scope' is specified, the scope of variables declared in
1531 a for-init-statement is limited to the `for' loop itself, as
1532 specified by the C++ standard. If `-fno-for-scope' is specified,
1533 the scope of variables declared in a for-init-statement extends to
1534 the end of the enclosing scope, as was the case in old versions of
1535 G++, and other (traditional) implementations of C++.
1537 The default if neither flag is given to follow the standard, but
1538 to allow and give a warning for old-style code that would
1539 otherwise be invalid, or have different behavior.
1542 Do not recognize `typeof' as a keyword, so that code can use this
1543 word as an identifier. You can use the keyword `__typeof__'
1544 instead. `-ansi' implies `-fno-gnu-keywords'.
1546 `-fno-implicit-templates'
1547 Never emit code for non-inline templates which are instantiated
1548 implicitly (i.e. by use); only emit code for explicit
1549 instantiations. *Note Template Instantiation::, for more
1552 `-fno-implicit-inline-templates'
1553 Don't emit code for implicit instantiations of inline templates,
1554 either. The default is to handle inlines differently so that
1555 compiles with and without optimization will need the same set of
1556 explicit instantiations.
1558 `-fno-implement-inlines'
1559 To save space, do not emit out-of-line copies of inline functions
1560 controlled by `#pragma implementation'. This will cause linker
1561 errors if these functions are not inlined everywhere they are
1565 Disable pedantic warnings about constructs used in MFC, such as
1566 implicit int and getting a pointer to member function via
1567 non-standard syntax.
1569 `-fno-nonansi-builtins'
1570 Disable built-in declarations of functions that are not mandated by
1571 ANSI/ISO C. These include `ffs', `alloca', `_exit', `index',
1572 `bzero', `conjf', and other related functions.
1574 `-fno-operator-names'
1575 Do not treat the operator name keywords `and', `bitand', `bitor',
1576 `compl', `not', `or' and `xor' as synonyms as keywords.
1578 `-fno-optional-diags'
1579 Disable diagnostics that the standard says a compiler does not
1580 need to issue. Currently, the only such diagnostic issued by G++
1581 is the one for a name having multiple meanings within a class.
1584 Downgrade some diagnostics about nonconformant code from errors to
1585 warnings. Thus, using `-fpermissive' will allow some
1586 nonconforming code to compile.
1589 Enable automatic template instantiation at link time. This option
1590 also implies `-fno-implicit-templates'. *Note Template
1591 Instantiation::, for more information.
1594 Disable generation of information about every class with virtual
1595 functions for use by the C++ runtime type identification features
1596 (`dynamic_cast' and `typeid'). If you don't use those parts of
1597 the language, you can save some space by using this flag. Note
1598 that exception handling uses the same information, but it will
1599 generate it as needed.
1602 Emit statistics about front-end processing at the end of the
1603 compilation. This information is generally only useful to the G++
1606 `-ftemplate-depth-N'
1607 Set the maximum instantiation depth for template classes to N. A
1608 limit on the template instantiation depth is needed to detect
1609 endless recursions during template class instantiation. ANSI/ISO
1610 C++ conforming programs must not rely on a maximum depth greater
1613 `-fno-threadsafe-statics'
1614 Do not emit the extra code to use the routines specified in the C++
1615 ABI for thread-safe initialization of local statics. You can use
1616 this option to reduce code size slightly in code that doesn't need
1620 Register destructors for objects with static storage duration with
1621 the `__cxa_atexit' function rather than the `atexit' function.
1622 This option is required for fully standards-compliant handling of
1623 static destructors, but will only work if your C library supports
1626 `-fvisibility-inlines-hidden'
1627 Causes all inlined methods to be marked with `__attribute__
1628 ((visibility ("hidden")))' so that they do not appear in the
1629 export table of a DSO and do not require a PLT indirection when
1630 used within the DSO. Enabling this option can have a dramatic
1631 effect on load and link times of a DSO as it massively reduces the
1632 size of the dynamic export table when the library makes heavy use
1633 of templates. While it can cause bloating through duplication of
1634 code within each DSO where it is used, often the wastage is less
1635 than the considerable space occupied by a long symbol name in the
1636 export table which is typical when using templates and namespaces.
1637 For even more savings, combine with the `-fvisibility=hidden'
1641 Do not use weak symbol support, even if it is provided by the
1642 linker. By default, G++ will use weak symbols if they are
1643 available. This option exists only for testing, and should not be
1644 used by end-users; it will result in inferior code and has no
1645 benefits. This option may be removed in a future release of G++.
1648 Do not search for header files in the standard directories
1649 specific to C++, but do still search the other standard
1650 directories. (This option is used when building the C++ library.)
1652 In addition, these optimization, warning, and code generation options
1653 have meanings only for C++ programs:
1655 `-fno-default-inline'
1656 Do not assume `inline' for functions defined inside a class scope.
1657 *Note Options That Control Optimization: Optimize Options. Note
1658 that these functions will have linkage like inline functions; they
1659 just won't be inlined by default.
1662 Warn when G++ generates code that is probably not compatible with
1663 the vendor-neutral C++ ABI. Although an effort has been made to
1664 warn about all such cases, there are probably some cases that are
1665 not warned about, even though G++ is generating incompatible code.
1666 There may also be cases where warnings are emitted even though
1667 the code that is generated will be compatible.
1669 You should rewrite your code to avoid these warnings if you are
1670 concerned about the fact that code generated by G++ may not be
1671 binary compatible with code generated by other compilers.
1673 The known incompatibilities at this point include:
1675 * Incorrect handling of tail-padding for bit-fields. G++ may
1676 attempt to pack data into the same byte as a base class. For
1679 struct A { virtual void f(); int f1 : 1; };
1680 struct B : public A { int f2 : 1; };
1682 In this case, G++ will place `B::f2' into the same byte
1683 as`A::f1'; other compilers will not. You can avoid this
1684 problem by explicitly padding `A' so that its size is a
1685 multiple of the byte size on your platform; that will cause
1686 G++ and other compilers to layout `B' identically.
1688 * Incorrect handling of tail-padding for virtual bases. G++
1689 does not use tail padding when laying out virtual bases. For
1692 struct A { virtual void f(); char c1; };
1693 struct B { B(); char c2; };
1694 struct C : public A, public virtual B {};
1696 In this case, G++ will not place `B' into the tail-padding for
1697 `A'; other compilers will. You can avoid this problem by
1698 explicitly padding `A' so that its size is a multiple of its
1699 alignment (ignoring virtual base classes); that will cause
1700 G++ and other compilers to layout `C' identically.
1702 * Incorrect handling of bit-fields with declared widths greater
1703 than that of their underlying types, when the bit-fields
1704 appear in a union. For example:
1706 union U { int i : 4096; };
1708 Assuming that an `int' does not have 4096 bits, G++ will make
1709 the union too small by the number of bits in an `int'.
1711 * Empty classes can be placed at incorrect offsets. For
1721 struct C : public B, public A {};
1723 G++ will place the `A' base class of `C' at a nonzero offset;
1724 it should be placed at offset zero. G++ mistakenly believes
1725 that the `A' data member of `B' is already at offset zero.
1727 * Names of template functions whose types involve `typename' or
1728 template template parameters can be mangled incorrectly.
1730 template <typename Q>
1731 void f(typename Q::X) {}
1733 template <template <typename> class Q>
1734 void f(typename Q<int>::X) {}
1736 Instantiations of these templates may be mangled incorrectly.
1739 `-Wctor-dtor-privacy (C++ only)'
1740 Warn when a class seems unusable because all the constructors or
1741 destructors in that class are private, and it has neither friends
1742 nor public static member functions.
1744 `-Wnon-virtual-dtor (C++ only)'
1745 Warn when a class appears to be polymorphic, thereby requiring a
1746 virtual destructor, yet it declares a non-virtual one. This
1747 warning is enabled by `-Wall'.
1749 `-Wreorder (C++ only)'
1750 Warn when the order of member initializers given in the code does
1751 not match the order in which they must be executed. For instance:
1756 A(): j (0), i (1) { }
1759 The compiler will rearrange the member initializers for `i' and
1760 `j' to match the declaration order of the members, emitting a
1761 warning to that effect. This warning is enabled by `-Wall'.
1763 The following `-W...' options are not affected by `-Wall'.
1765 `-Weffc++ (C++ only)'
1766 Warn about violations of the following style guidelines from Scott
1767 Meyers' `Effective C++' book:
1769 * Item 11: Define a copy constructor and an assignment
1770 operator for classes with dynamically allocated memory.
1772 * Item 12: Prefer initialization to assignment in constructors.
1774 * Item 14: Make destructors virtual in base classes.
1776 * Item 15: Have `operator=' return a reference to `*this'.
1778 * Item 23: Don't try to return a reference when you must
1782 Also warn about violations of the following style guidelines from
1783 Scott Meyers' `More Effective C++' book:
1785 * Item 6: Distinguish between prefix and postfix forms of
1786 increment and decrement operators.
1788 * Item 7: Never overload `&&', `||', or `,'.
1791 When selecting this option, be aware that the standard library
1792 headers do not obey all of these guidelines; use `grep -v' to
1793 filter out those warnings.
1795 `-Wno-deprecated (C++ only)'
1796 Do not warn about usage of deprecated features. *Note Deprecated
1799 `-Wstrict-null-sentinel (C++ only)'
1800 Warn also about the use of an uncasted `NULL' as sentinel. When
1801 compiling only with GCC this is a valid sentinel, as `NULL' is
1802 defined to `__null'. Although it is a null pointer constant not a
1803 null pointer, it is guaranteed to of the same size as a pointer.
1804 But this use is not portable across different compilers.
1806 `-Wno-non-template-friend (C++ only)'
1807 Disable warnings when non-templatized friend functions are declared
1808 within a template. Since the advent of explicit template
1809 specification support in G++, if the name of the friend is an
1810 unqualified-id (i.e., `friend foo(int)'), the C++ language
1811 specification demands that the friend declare or define an
1812 ordinary, nontemplate function. (Section 14.5.3). Before G++
1813 implemented explicit specification, unqualified-ids could be
1814 interpreted as a particular specialization of a templatized
1815 function. Because this non-conforming behavior is no longer the
1816 default behavior for G++, `-Wnon-template-friend' allows the
1817 compiler to check existing code for potential trouble spots and is
1818 on by default. This new compiler behavior can be turned off with
1819 `-Wno-non-template-friend' which keeps the conformant compiler code
1820 but disables the helpful warning.
1822 `-Wold-style-cast (C++ only)'
1823 Warn if an old-style (C-style) cast to a non-void type is used
1824 within a C++ program. The new-style casts (`static_cast',
1825 `reinterpret_cast', and `const_cast') are less vulnerable to
1826 unintended effects and much easier to search for.
1828 `-Woverloaded-virtual (C++ only)'
1829 Warn when a function declaration hides virtual functions from a
1830 base class. For example, in:
1836 struct B: public A {
1840 the `A' class version of `f' is hidden in `B', and code like:
1845 will fail to compile.
1847 `-Wno-pmf-conversions (C++ only)'
1848 Disable the diagnostic for converting a bound pointer to member
1849 function to a plain pointer.
1851 `-Wsign-promo (C++ only)'
1852 Warn when overload resolution chooses a promotion from unsigned or
1853 enumerated type to a signed type, over a conversion to an unsigned
1854 type of the same size. Previous versions of G++ would try to
1855 preserve unsignedness, but the standard mandates the current
1860 A& operator = (int);
1869 In this example, G++ will synthesize a default `A& operator =
1870 (const A&);', while cfront will use the user-defined `operator ='.
1873 File: gcc.info, Node: Objective-C and Objective-C++ Dialect Options, Next: Language Independent Options, Prev: C++ Dialect Options, Up: Invoking GCC
1875 3.6 Options Controlling Objective-C and Objective-C++ Dialects
1876 ==============================================================
1878 (NOTE: This manual does not describe the Objective-C and Objective-C++
1879 languages themselves. See *Note Language Standards Supported by GCC:
1880 Standards, for references.)
1882 This section describes the command-line options that are only
1883 meaningful for Objective-C and Objective-C++ programs, but you can also
1884 use most of the language-independent GNU compiler options. For
1885 example, you might compile a file `some_class.m' like this:
1887 gcc -g -fgnu-runtime -O -c some_class.m
1889 In this example, `-fgnu-runtime' is an option meant only for
1890 Objective-C and Objective-C++ programs; you can use the other options
1891 with any language supported by GCC.
1893 Note that since Objective-C is an extension of the C language,
1894 Objective-C compilations may also use options specific to the C
1895 front-end (e.g., `-Wtraditional'). Similarly, Objective-C++
1896 compilations may use C++-specific options (e.g., `-Wabi').
1898 Here is a list of options that are _only_ for compiling Objective-C
1899 and Objective-C++ programs:
1901 `-fconstant-string-class=CLASS-NAME'
1902 Use CLASS-NAME as the name of the class to instantiate for each
1903 literal string specified with the syntax `@"..."'. The default
1904 class name is `NXConstantString' if the GNU runtime is being used,
1905 and `NSConstantString' if the NeXT runtime is being used (see
1906 below). The `-fconstant-cfstrings' option, if also present, will
1907 override the `-fconstant-string-class' setting and cause `@"..."'
1908 literals to be laid out as constant CoreFoundation strings.
1911 Generate object code compatible with the standard GNU Objective-C
1912 runtime. This is the default for most types of systems.
1915 Generate output compatible with the NeXT runtime. This is the
1916 default for NeXT-based systems, including Darwin and Mac OS X.
1917 The macro `__NEXT_RUNTIME__' is predefined if (and only if) this
1920 `-fno-nil-receivers'
1921 Assume that all Objective-C message dispatches (e.g., `[receiver
1922 message:arg]') in this translation unit ensure that the receiver
1923 is not `nil'. This allows for more efficient entry points in the
1924 runtime to be used. Currently, this option is only available in
1925 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
1928 Enable syntactic support for structured exception handling in
1929 Objective-C, similar to what is offered by C++ and Java.
1930 Currently, this option is only available in conjunction with the
1931 NeXT runtime on Mac OS X 10.3 and later.
1938 @catch (AnObjCClass *exc) {
1945 @catch (AnotherClass *exc) {
1948 @catch (id allOthers) {
1957 The `@throw' statement may appear anywhere in an Objective-C or
1958 Objective-C++ program; when used inside of a `@catch' block, the
1959 `@throw' may appear without an argument (as shown above), in which
1960 case the object caught by the `@catch' will be rethrown.
1962 Note that only (pointers to) Objective-C objects may be thrown and
1963 caught using this scheme. When an object is thrown, it will be
1964 caught by the nearest `@catch' clause capable of handling objects
1965 of that type, analogously to how `catch' blocks work in C++ and
1966 Java. A `@catch(id ...)' clause (as shown above) may also be
1967 provided to catch any and all Objective-C exceptions not caught by
1968 previous `@catch' clauses (if any).
1970 The `@finally' clause, if present, will be executed upon exit from
1971 the immediately preceding `@try ... @catch' section. This will
1972 happen regardless of whether any exceptions are thrown, caught or
1973 rethrown inside the `@try ... @catch' section, analogously to the
1974 behavior of the `finally' clause in Java.
1976 There are several caveats to using the new exception mechanism:
1978 * Although currently designed to be binary compatible with
1979 `NS_HANDLER'-style idioms provided by the `NSException'
1980 class, the new exceptions can only be used on Mac OS X 10.3
1981 (Panther) and later systems, due to additional functionality
1982 needed in the (NeXT) Objective-C runtime.
1984 * As mentioned above, the new exceptions do not support handling
1985 types other than Objective-C objects. Furthermore, when
1986 used from Objective-C++, the Objective-C exception model does
1987 not interoperate with C++ exceptions at this time. This
1988 means you cannot `@throw' an exception from Objective-C and
1989 `catch' it in C++, or vice versa (i.e., `throw ... @catch').
1991 The `-fobjc-exceptions' switch also enables the use of
1992 synchronization blocks for thread-safe execution:
1994 @synchronized (ObjCClass *guard) {
1998 Upon entering the `@synchronized' block, a thread of execution
1999 shall first check whether a lock has been placed on the
2000 corresponding `guard' object by another thread. If it has, the
2001 current thread shall wait until the other thread relinquishes its
2002 lock. Once `guard' becomes available, the current thread will
2003 place its own lock on it, execute the code contained in the
2004 `@synchronized' block, and finally relinquish the lock (thereby
2005 making `guard' available to other threads).
2007 Unlike Java, Objective-C does not allow for entire methods to be
2008 marked `@synchronized'. Note that throwing exceptions out of
2009 `@synchronized' blocks is allowed, and will cause the guarding
2010 object to be unlocked properly.
2012 `-freplace-objc-classes'
2013 Emit a special marker instructing `ld(1)' not to statically link in
2014 the resulting object file, and allow `dyld(1)' to load it in at
2015 run time instead. This is used in conjunction with the
2016 Fix-and-Continue debugging mode, where the object file in question
2017 may be recompiled and dynamically reloaded in the course of
2018 program execution, without the need to restart the program itself.
2019 Currently, Fix-and-Continue functionality is only available in
2020 conjunction with the NeXT runtime on Mac OS X 10.3 and later.
2023 When compiling for the NeXT runtime, the compiler ordinarily
2024 replaces calls to `objc_getClass("...")' (when the name of the
2025 class is known at compile time) with static class references that
2026 get initialized at load time, which improves run-time performance.
2027 Specifying the `-fzero-link' flag suppresses this behavior and
2028 causes calls to `objc_getClass("...")' to be retained. This is
2029 useful in Zero-Link debugging mode, since it allows for individual
2030 class implementations to be modified during program execution.
2033 Dump interface declarations for all classes seen in the source
2034 file to a file named `SOURCENAME.decl'.
2037 If a class is declared to implement a protocol, a warning is
2038 issued for every method in the protocol that is not implemented by
2039 the class. The default behavior is to issue a warning for every
2040 method not explicitly implemented in the class, even if a method
2041 implementation is inherited from the superclass. If you use the
2042 `-Wno-protocol' option, then methods inherited from the superclass
2043 are considered to be implemented, and no warning is issued for
2047 Warn if multiple methods of different types for the same selector
2048 are found during compilation. The check is performed on the list
2049 of methods in the final stage of compilation. Additionally, a
2050 check is performed for each selector appearing in a
2051 `@selector(...)' expression, and a corresponding method for that
2052 selector has been found during compilation. Because these checks
2053 scan the method table only at the end of compilation, these
2054 warnings are not produced if the final stage of compilation is not
2055 reached, for example because an error is found during compilation,
2056 or because the `-fsyntax-only' option is being used.
2058 `-Wundeclared-selector'
2059 Warn if a `@selector(...)' expression referring to an undeclared
2060 selector is found. A selector is considered undeclared if no
2061 method with that name has been declared before the
2062 `@selector(...)' expression, either explicitly in an `@interface'
2063 or `@protocol' declaration, or implicitly in an `@implementation'
2064 section. This option always performs its checks as soon as a
2065 `@selector(...)' expression is found, while `-Wselector' only
2066 performs its checks in the final stage of compilation. This also
2067 enforces the coding style convention that methods and selectors
2068 must be declared before being used.
2070 `-print-objc-runtime-info'
2071 Generate C header describing the largest structure that is passed
2076 File: gcc.info, Node: Language Independent Options, Next: Warning Options, Prev: Objective-C and Objective-C++ Dialect Options, Up: Invoking GCC
2078 3.7 Options to Control Diagnostic Messages Formatting
2079 =====================================================
2081 Traditionally, diagnostic messages have been formatted irrespective of
2082 the output device's aspect (e.g. its width, ...). The options described
2083 below can be used to control the diagnostic messages formatting
2084 algorithm, e.g. how many characters per line, how often source location
2085 information should be reported. Right now, only the C++ front end can
2086 honor these options. However it is expected, in the near future, that
2087 the remaining front ends would be able to digest them correctly.
2089 `-fmessage-length=N'
2090 Try to format error messages so that they fit on lines of about N
2091 characters. The default is 72 characters for `g++' and 0 for the
2092 rest of the front ends supported by GCC. If N is zero, then no
2093 line-wrapping will be done; each error message will appear on a
2096 `-fdiagnostics-show-location=once'
2097 Only meaningful in line-wrapping mode. Instructs the diagnostic
2098 messages reporter to emit _once_ source location information; that
2099 is, in case the message is too long to fit on a single physical
2100 line and has to be wrapped, the source location won't be emitted
2101 (as prefix) again, over and over, in subsequent continuation
2102 lines. This is the default behavior.
2104 `-fdiagnostics-show-location=every-line'
2105 Only meaningful in line-wrapping mode. Instructs the diagnostic
2106 messages reporter to emit the same source location information (as
2107 prefix) for physical lines that result from the process of breaking
2108 a message which is too long to fit on a single line.
2112 File: gcc.info, Node: Warning Options, Next: Debugging Options, Prev: Language Independent Options, Up: Invoking GCC
2114 3.8 Options to Request or Suppress Warnings
2115 ===========================================
2117 Warnings are diagnostic messages that report constructions which are
2118 not inherently erroneous but which are risky or suggest there may have
2121 You can request many specific warnings with options beginning `-W',
2122 for example `-Wimplicit' to request warnings on implicit declarations.
2123 Each of these specific warning options also has a negative form
2124 beginning `-Wno-' to turn off warnings; for example, `-Wno-implicit'.
2125 This manual lists only one of the two forms, whichever is not the
2128 The following options control the amount and kinds of warnings produced
2129 by GCC; for further, language-specific options also refer to *Note C++
2130 Dialect Options:: and *Note Objective-C and Objective-C++ Dialect
2134 Check the code for syntax errors, but don't do anything beyond
2138 Issue all the warnings demanded by strict ISO C and ISO C++;
2139 reject all programs that use forbidden extensions, and some other
2140 programs that do not follow ISO C and ISO C++. For ISO C, follows
2141 the version of the ISO C standard specified by any `-std' option
2144 Valid ISO C and ISO C++ programs should compile properly with or
2145 without this option (though a rare few will require `-ansi' or a
2146 `-std' option specifying the required version of ISO C). However,
2147 without this option, certain GNU extensions and traditional C and
2148 C++ features are supported as well. With this option, they are
2151 `-pedantic' does not cause warning messages for use of the
2152 alternate keywords whose names begin and end with `__'. Pedantic
2153 warnings are also disabled in the expression that follows
2154 `__extension__'. However, only system header files should use
2155 these escape routes; application programs should avoid them.
2156 *Note Alternate Keywords::.
2158 Some users try to use `-pedantic' to check programs for strict ISO
2159 C conformance. They soon find that it does not do quite what they
2160 want: it finds some non-ISO practices, but not all--only those for
2161 which ISO C _requires_ a diagnostic, and some others for which
2162 diagnostics have been added.
2164 A feature to report any failure to conform to ISO C might be
2165 useful in some instances, but would require considerable
2166 additional work and would be quite different from `-pedantic'. We
2167 don't have plans to support such a feature in the near future.
2169 Where the standard specified with `-std' represents a GNU extended
2170 dialect of C, such as `gnu89' or `gnu99', there is a corresponding
2171 "base standard", the version of ISO C on which the GNU extended
2172 dialect is based. Warnings from `-pedantic' are given where they
2173 are required by the base standard. (It would not make sense for
2174 such warnings to be given only for features not in the specified
2175 GNU C dialect, since by definition the GNU dialects of C include
2176 all features the compiler supports with the given option, and
2177 there would be nothing to warn about.)
2180 Like `-pedantic', except that errors are produced rather than
2184 Inhibit all warning messages.
2187 Inhibit warning messages about the use of `#import'.
2190 Warn if an array subscript has type `char'. This is a common cause
2191 of error, as programmers often forget that this type is signed on
2192 some machines. This warning is enabled by `-Wall'.
2195 Warn whenever a comment-start sequence `/*' appears in a `/*'
2196 comment, or whenever a Backslash-Newline appears in a `//' comment.
2197 This warning is enabled by `-Wall'.
2200 This option causes the compiler to abort compilation on the first
2201 error occurred rather than trying to keep going and printing
2202 further error messages.
2205 Check calls to `printf' and `scanf', etc., to make sure that the
2206 arguments supplied have types appropriate to the format string
2207 specified, and that the conversions specified in the format string
2208 make sense. This includes standard functions, and others
2209 specified by format attributes (*note Function Attributes::), in
2210 the `printf', `scanf', `strftime' and `strfmon' (an X/Open
2211 extension, not in the C standard) families (or other
2212 target-specific families). Which functions are checked without
2213 format attributes having been specified depends on the standard
2214 version selected, and such checks of functions without the
2215 attribute specified are disabled by `-ffreestanding' or
2218 The formats are checked against the format features supported by
2219 GNU libc version 2.2. These include all ISO C90 and C99 features,
2220 as well as features from the Single Unix Specification and some
2221 BSD and GNU extensions. Other library implementations may not
2222 support all these features; GCC does not support warning about
2223 features that go beyond a particular library's limitations.
2224 However, if `-pedantic' is used with `-Wformat', warnings will be
2225 given about format features not in the selected standard version
2226 (but not for `strfmon' formats, since those are not in any version
2227 of the C standard). *Note Options Controlling C Dialect: C
2230 Since `-Wformat' also checks for null format arguments for several
2231 functions, `-Wformat' also implies `-Wnonnull'.
2233 `-Wformat' is included in `-Wall'. For more control over some
2234 aspects of format checking, the options `-Wformat-y2k',
2235 `-Wno-format-extra-args', `-Wno-format-zero-length',
2236 `-Wformat-nonliteral', `-Wformat-security', and `-Wformat=2' are
2237 available, but are not included in `-Wall'.
2240 If `-Wformat' is specified, also warn about `strftime' formats
2241 which may yield only a two-digit year.
2243 `-Wno-format-extra-args'
2244 If `-Wformat' is specified, do not warn about excess arguments to a
2245 `printf' or `scanf' format function. The C standard specifies
2246 that such arguments are ignored.
2248 Where the unused arguments lie between used arguments that are
2249 specified with `$' operand number specifications, normally
2250 warnings are still given, since the implementation could not know
2251 what type to pass to `va_arg' to skip the unused arguments.
2252 However, in the case of `scanf' formats, this option will suppress
2253 the warning if the unused arguments are all pointers, since the
2254 Single Unix Specification says that such unused arguments are
2257 `-Wno-format-zero-length'
2258 If `-Wformat' is specified, do not warn about zero-length formats.
2259 The C standard specifies that zero-length formats are allowed.
2261 `-Wformat-nonliteral'
2262 If `-Wformat' is specified, also warn if the format string is not a
2263 string literal and so cannot be checked, unless the format function
2264 takes its format arguments as a `va_list'.
2267 If `-Wformat' is specified, also warn about uses of format
2268 functions that represent possible security problems. At present,
2269 this warns about calls to `printf' and `scanf' functions where the
2270 format string is not a string literal and there are no format
2271 arguments, as in `printf (foo);'. This may be a security hole if
2272 the format string came from untrusted input and contains `%n'.
2273 (This is currently a subset of what `-Wformat-nonliteral' warns
2274 about, but in future warnings may be added to `-Wformat-security'
2275 that are not included in `-Wformat-nonliteral'.)
2278 Enable `-Wformat' plus format checks not included in `-Wformat'.
2279 Currently equivalent to `-Wformat -Wformat-nonliteral
2280 -Wformat-security -Wformat-y2k'.
2283 Warn about passing a null pointer for arguments marked as
2284 requiring a non-null value by the `nonnull' function attribute.
2286 `-Wnonnull' is included in `-Wall' and `-Wformat'. It can be
2287 disabled with the `-Wno-nonnull' option.
2289 `-Winit-self (C, C++, Objective-C and Objective-C++ only)'
2290 Warn about uninitialized variables which are initialized with
2291 themselves. Note this option can only be used with the
2292 `-Wuninitialized' option, which in turn only works with `-O1' and
2295 For example, GCC will warn about `i' being uninitialized in the
2296 following snippet only when `-Winit-self' has been specified:
2304 Warn when a declaration does not specify a type. This warning is
2307 `-Wimplicit-function-declaration'
2308 `-Werror-implicit-function-declaration'
2309 Give a warning (or error) whenever a function is used before being
2310 declared. The form `-Wno-error-implicit-function-declaration' is
2311 not supported. This warning is enabled by `-Wall' (as a warning,
2315 Same as `-Wimplicit-int' and `-Wimplicit-function-declaration'.
2316 This warning is enabled by `-Wall'.
2319 Warn if the type of `main' is suspicious. `main' should be a
2320 function with external linkage, returning int, taking either zero
2321 arguments, two, or three arguments of appropriate types. This
2322 warning is enabled by `-Wall'.
2325 Warn if an aggregate or union initializer is not fully bracketed.
2326 In the following example, the initializer for `a' is not fully
2327 bracketed, but that for `b' is fully bracketed.
2329 int a[2][2] = { 0, 1, 2, 3 };
2330 int b[2][2] = { { 0, 1 }, { 2, 3 } };
2332 This warning is enabled by `-Wall'.
2334 `-Wmissing-include-dirs (C, C++, Objective-C and Objective-C++ only)'
2335 Warn if a user-supplied include directory does not exist.
2338 Warn if parentheses are omitted in certain contexts, such as when
2339 there is an assignment in a context where a truth value is
2340 expected, or when operators are nested whose precedence people
2341 often get confused about. Only the warning for an assignment used
2342 as a truth value is supported when compiling C++; the other
2343 warnings are only supported when compiling C.
2345 Also warn if a comparison like `x<=y<=z' appears; this is
2346 equivalent to `(x<=y ? 1 : 0) <= z', which is a different
2347 interpretation from that of ordinary mathematical notation.
2349 Also warn about constructions where there may be confusion to which
2350 `if' statement an `else' branch belongs. Here is an example of
2361 In C, every `else' branch belongs to the innermost possible `if'
2362 statement, which in this example is `if (b)'. This is often not
2363 what the programmer expected, as illustrated in the above example
2364 by indentation the programmer chose. When there is the potential
2365 for this confusion, GCC will issue a warning when this flag is
2366 specified. To eliminate the warning, add explicit braces around
2367 the innermost `if' statement so there is no way the `else' could
2368 belong to the enclosing `if'. The resulting code would look like
2381 This warning is enabled by `-Wall'.
2384 Warn about code that may have undefined semantics because of
2385 violations of sequence point rules in the C standard.
2387 The C standard defines the order in which expressions in a C
2388 program are evaluated in terms of "sequence points", which
2389 represent a partial ordering between the execution of parts of the
2390 program: those executed before the sequence point, and those
2391 executed after it. These occur after the evaluation of a full
2392 expression (one which is not part of a larger expression), after
2393 the evaluation of the first operand of a `&&', `||', `? :' or `,'
2394 (comma) operator, before a function is called (but after the
2395 evaluation of its arguments and the expression denoting the called
2396 function), and in certain other places. Other than as expressed
2397 by the sequence point rules, the order of evaluation of
2398 subexpressions of an expression is not specified. All these rules
2399 describe only a partial order rather than a total order, since,
2400 for example, if two functions are called within one expression
2401 with no sequence point between them, the order in which the
2402 functions are called is not specified. However, the standards
2403 committee have ruled that function calls do not overlap.
2405 It is not specified when between sequence points modifications to
2406 the values of objects take effect. Programs whose behavior
2407 depends on this have undefined behavior; the C standard specifies
2408 that "Between the previous and next sequence point an object shall
2409 have its stored value modified at most once by the evaluation of
2410 an expression. Furthermore, the prior value shall be read only to
2411 determine the value to be stored.". If a program breaks these
2412 rules, the results on any particular implementation are entirely
2415 Examples of code with undefined behavior are `a = a++;', `a[n] =
2416 b[n++]' and `a[i++] = i;'. Some more complicated cases are not
2417 diagnosed by this option, and it may give an occasional false
2418 positive result, but in general it has been found fairly effective
2419 at detecting this sort of problem in programs.
2421 The present implementation of this option only works for C
2422 programs. A future implementation may also work for C++ programs.
2424 The C standard is worded confusingly, therefore there is some
2425 debate over the precise meaning of the sequence point rules in
2426 subtle cases. Links to discussions of the problem, including
2427 proposed formal definitions, may be found on the GCC readings
2428 page, at `http://gcc.gnu.org/readings.html'.
2430 This warning is enabled by `-Wall'.
2433 Warn whenever a function is defined with a return-type that
2434 defaults to `int'. Also warn about any `return' statement with no
2435 return-value in a function whose return-type is not `void'.
2437 For C, also warn if the return type of a function has a type
2438 qualifier such as `const'. Such a type qualifier has no effect,
2439 since the value returned by a function is not an lvalue. ISO C
2440 prohibits qualified `void' return types on function definitions,
2441 so such return types always receive a warning even without this
2444 For C++, a function without return type always produces a
2445 diagnostic message, even when `-Wno-return-type' is specified.
2446 The only exceptions are `main' and functions defined in system
2449 This warning is enabled by `-Wall'.
2452 Warn whenever a `switch' statement has an index of enumerated type
2453 and lacks a `case' for one or more of the named codes of that
2454 enumeration. (The presence of a `default' label prevents this
2455 warning.) `case' labels outside the enumeration range also
2456 provoke warnings when this option is used. This warning is
2460 Warn whenever a `switch' statement does not have a `default' case.
2463 Warn whenever a `switch' statement has an index of enumerated type
2464 and lacks a `case' for one or more of the named codes of that
2465 enumeration. `case' labels outside the enumeration range also
2466 provoke warnings when this option is used.
2469 Warn if any trigraphs are encountered that might change the
2470 meaning of the program (trigraphs within comments are not warned
2471 about). This warning is enabled by `-Wall'.
2474 Warn whenever a static function is declared but not defined or a
2475 non\-inline static function is unused. This warning is enabled by
2479 Warn whenever a label is declared but not used. This warning is
2482 To suppress this warning use the `unused' attribute (*note
2483 Variable Attributes::).
2485 `-Wunused-parameter'
2486 Warn whenever a function parameter is unused aside from its
2489 To suppress this warning use the `unused' attribute (*note
2490 Variable Attributes::).
2493 Warn whenever a local variable or non-constant static variable is
2494 unused aside from its declaration This warning is enabled by
2497 To suppress this warning use the `unused' attribute (*note
2498 Variable Attributes::).
2501 Warn whenever a statement computes a result that is explicitly not
2502 used. This warning is enabled by `-Wall'.
2504 To suppress this warning cast the expression to `void'.
2507 All the above `-Wunused' options combined.
2509 In order to get a warning about an unused function parameter, you
2510 must either specify `-Wextra -Wunused' (note that `-Wall' implies
2511 `-Wunused'), or separately specify `-Wunused-parameter'.
2514 Warn if an automatic variable is used without first being
2515 initialized or if a variable may be clobbered by a `setjmp' call.
2517 These warnings are possible only in optimizing compilation,
2518 because they require data flow information that is computed only
2519 when optimizing. If you don't specify `-O', you simply won't get
2522 If you want to warn about code which uses the uninitialized value
2523 of the variable in its own initializer, use the `-Winit-self'
2526 These warnings occur for individual uninitialized or clobbered
2527 elements of structure, union or array variables as well as for
2528 variables which are uninitialized or clobbered as a whole. They do
2529 not occur for variables or elements declared `volatile'. Because
2530 these warnings depend on optimization, the exact variables or
2531 elements for which there are warnings will depend on the precise
2532 optimization options and version of GCC used.
2534 Note that there may be no warning about a variable that is used
2535 only to compute a value that itself is never used, because such
2536 computations may be deleted by data flow analysis before the
2537 warnings are printed.
2539 These warnings are made optional because GCC is not smart enough
2540 to see all the reasons why the code might be correct despite
2541 appearing to have an error. Here is one example of how this can
2557 If the value of `y' is always 1, 2 or 3, then `x' is always
2558 initialized, but GCC doesn't know this. Here is another common
2563 if (change_y) save_y = y, y = new_y;
2565 if (change_y) y = save_y;
2568 This has no bug because `save_y' is used only if it is set.
2570 This option also warns when a non-volatile automatic variable
2571 might be changed by a call to `longjmp'. These warnings as well
2572 are possible only in optimizing compilation.
2574 The compiler sees only the calls to `setjmp'. It cannot know
2575 where `longjmp' will be called; in fact, a signal handler could
2576 call it at any point in the code. As a result, you may get a
2577 warning even when there is in fact no problem because `longjmp'
2578 cannot in fact be called at the place which would cause a problem.
2580 Some spurious warnings can be avoided if you declare all the
2581 functions you use that never return as `noreturn'. *Note Function
2584 This warning is enabled by `-Wall'.
2587 Warn when a #pragma directive is encountered which is not
2588 understood by GCC. If this command line option is used, warnings
2589 will even be issued for unknown pragmas in system header files.
2590 This is not the case if the warnings were only enabled by the
2591 `-Wall' command line option.
2594 This option is only active when `-fstrict-aliasing' is active. It
2595 warns about code which might break the strict aliasing rules that
2596 the compiler is using for optimization. The warning does not
2597 catch all cases, but does attempt to catch the more common
2598 pitfalls. It is included in `-Wall'.
2600 `-Wstrict-aliasing=2'
2601 This option is only active when `-fstrict-aliasing' is active. It
2602 warns about all code which might break the strict aliasing rules
2603 that the compiler is using for optimization. This warning catches
2604 all cases, but it will also give a warning for some ambiguous
2605 cases that are safe.
2608 All of the above `-W' options combined. This enables all the
2609 warnings about constructions that some users consider
2610 questionable, and that are easy to avoid (or modify to prevent the
2611 warning), even in conjunction with macros. This also enables some
2612 language-specific warnings described in *Note C++ Dialect
2613 Options:: and *Note Objective-C and Objective-C++ Dialect
2616 The following `-W...' options are not implied by `-Wall'. Some of
2617 them warn about constructions that users generally do not consider
2618 questionable, but which occasionally you might wish to check for;
2619 others warn about constructions that are necessary or hard to avoid in
2620 some cases, and there is no simple way to modify the code to suppress
2624 (This option used to be called `-W'. The older name is still
2625 supported, but the newer name is more descriptive.) Print extra
2626 warning messages for these events:
2628 * A function can return either with or without a value.
2629 (Falling off the end of the function body is considered
2630 returning without a value.) For example, this function would
2631 evoke such a warning:
2639 * An expression-statement or the left-hand side of a comma
2640 expression contains no side effects. To suppress the
2641 warning, cast the unused expression to void. For example, an
2642 expression such as `x[i,j]' will cause a warning, but
2643 `x[(void)i,j]' will not.
2645 * An unsigned value is compared against zero with `<' or `>='.
2647 * Storage-class specifiers like `static' are not the first
2648 things in a declaration. According to the C Standard, this
2649 usage is obsolescent.
2651 * If `-Wall' or `-Wunused' is also specified, warn about unused
2654 * A comparison between signed and unsigned values could produce
2655 an incorrect result when the signed value is converted to
2656 unsigned. (But don't warn if `-Wno-sign-compare' is also
2659 * An aggregate has an initializer which does not initialize all
2660 members. This warning can be independently controlled by
2661 `-Wmissing-field-initializers'.
2663 * A function parameter is declared without a type specifier in
2664 K&R-style functions:
2668 * An empty body occurs in an `if' or `else' statement.
2670 * A pointer is compared against integer zero with `<', `<=',
2673 * A variable might be changed by `longjmp' or `vfork'.
2675 * Any of several floating-point events that often indicate
2676 errors, such as overflow, underflow, loss of precision, etc.
2678 * (C++ only) An enumerator and a non-enumerator both appear in
2679 a conditional expression.
2681 * (C++ only) A non-static reference or non-static `const'
2682 member appears in a class without constructors.
2684 * (C++ only) Ambiguous virtual bases.
2686 * (C++ only) Subscripting an array which has been declared
2689 * (C++ only) Taking the address of a variable which has been
2690 declared `register'.
2692 * (C++ only) A base class is not initialized in a derived
2693 class' copy constructor.
2696 Do not warn about compile-time integer division by zero. Floating
2697 point division by zero is not warned about, as it can be a
2698 legitimate way of obtaining infinities and NaNs.
2701 Print warning messages for constructs found in system header files.
2702 Warnings from system headers are normally suppressed, on the
2703 assumption that they usually do not indicate real problems and
2704 would only make the compiler output harder to read. Using this
2705 command line option tells GCC to emit warnings from system headers
2706 as if they occurred in user code. However, note that using
2707 `-Wall' in conjunction with this option will _not_ warn about
2708 unknown pragmas in system headers--for that, `-Wunknown-pragmas'
2712 Warn if floating point values are used in equality comparisons.
2714 The idea behind this is that sometimes it is convenient (for the
2715 programmer) to consider floating-point values as approximations to
2716 infinitely precise real numbers. If you are doing this, then you
2717 need to compute (by analyzing the code, or in some other way) the
2718 maximum or likely maximum error that the computation introduces,
2719 and allow for it when performing comparisons (and when producing
2720 output, but that's a different problem). In particular, instead
2721 of testing for equality, you would check to see whether the two
2722 values have ranges that overlap; and this is done with the
2723 relational operators, so equality comparisons are probably
2726 `-Wtraditional (C only)'
2727 Warn about certain constructs that behave differently in
2728 traditional and ISO C. Also warn about ISO C constructs that have
2729 no traditional C equivalent, and/or problematic constructs which
2732 * Macro parameters that appear within string literals in the
2733 macro body. In traditional C macro replacement takes place
2734 within string literals, but does not in ISO C.
2736 * In traditional C, some preprocessor directives did not exist.
2737 Traditional preprocessors would only consider a line to be a
2738 directive if the `#' appeared in column 1 on the line.
2739 Therefore `-Wtraditional' warns about directives that
2740 traditional C understands but would ignore because the `#'
2741 does not appear as the first character on the line. It also
2742 suggests you hide directives like `#pragma' not understood by
2743 traditional C by indenting them. Some traditional
2744 implementations would not recognize `#elif', so it suggests
2745 avoiding it altogether.
2747 * A function-like macro that appears without arguments.
2749 * The unary plus operator.
2751 * The `U' integer constant suffix, or the `F' or `L' floating
2752 point constant suffixes. (Traditional C does support the `L'
2753 suffix on integer constants.) Note, these suffixes appear in
2754 macros defined in the system headers of most modern systems,
2755 e.g. the `_MIN'/`_MAX' macros in `<limits.h>'. Use of these
2756 macros in user code might normally lead to spurious warnings,
2757 however GCC's integrated preprocessor has enough context to
2758 avoid warning in these cases.
2760 * A function declared external in one block and then used after
2761 the end of the block.
2763 * A `switch' statement has an operand of type `long'.
2765 * A non-`static' function declaration follows a `static' one.
2766 This construct is not accepted by some traditional C
2769 * The ISO type of an integer constant has a different width or
2770 signedness from its traditional type. This warning is only
2771 issued if the base of the constant is ten. I.e. hexadecimal
2772 or octal values, which typically represent bit patterns, are
2775 * Usage of ISO string concatenation is detected.
2777 * Initialization of automatic aggregates.
2779 * Identifier conflicts with labels. Traditional C lacks a
2780 separate namespace for labels.
2782 * Initialization of unions. If the initializer is zero, the
2783 warning is omitted. This is done under the assumption that
2784 the zero initializer in user code appears conditioned on e.g.
2785 `__STDC__' to avoid missing initializer warnings and relies
2786 on default initialization to zero in the traditional C case.
2788 * Conversions by prototypes between fixed/floating point values
2789 and vice versa. The absence of these prototypes when
2790 compiling with traditional C would cause serious problems.
2791 This is a subset of the possible conversion warnings, for the
2792 full set use `-Wconversion'.
2794 * Use of ISO C style function definitions. This warning
2795 intentionally is _not_ issued for prototype declarations or
2796 variadic functions because these ISO C features will appear
2797 in your code when using libiberty's traditional C
2798 compatibility macros, `PARAMS' and `VPARAMS'. This warning
2799 is also bypassed for nested functions because that feature is
2800 already a GCC extension and thus not relevant to traditional
2803 `-Wdeclaration-after-statement (C only)'
2804 Warn when a declaration is found after a statement in a block.
2805 This construct, known from C++, was introduced with ISO C99 and is
2806 by default allowed in GCC. It is not supported by ISO C90 and was
2807 not supported by GCC versions before GCC 3.0. *Note Mixed
2811 Warn if an undefined identifier is evaluated in an `#if' directive.
2814 Do not warn whenever an `#else' or an `#endif' are followed by
2818 Warn whenever a local variable shadows another local variable,
2819 parameter or global variable or whenever a built-in function is
2823 Warn whenever an object of larger than LEN bytes is defined.
2826 Warn about anything that depends on the "size of" a function type
2827 or of `void'. GNU C assigns these types a size of 1, for
2828 convenience in calculations with `void *' pointers and pointers to
2831 `-Wbad-function-cast (C only)'
2832 Warn whenever a function call is cast to a non-matching type. For
2833 example, warn if `int malloc()' is cast to `anything *'.
2836 Warn whenever a pointer is cast so as to remove a type qualifier
2837 from the target type. For example, warn if a `const char *' is
2838 cast to an ordinary `char *'.
2841 Warn whenever a pointer is cast such that the required alignment
2842 of the target is increased. For example, warn if a `char *' is
2843 cast to an `int *' on machines where integers can only be accessed
2844 at two- or four-byte boundaries.
2847 When compiling C, give string constants the type `const
2848 char[LENGTH]' so that copying the address of one into a
2849 non-`const' `char *' pointer will get a warning; when compiling
2850 C++, warn about the deprecated conversion from string constants to
2851 `char *'. These warnings will help you find at compile time code
2852 that can try to write into a string constant, but only if you have
2853 been very careful about using `const' in declarations and
2854 prototypes. Otherwise, it will just be a nuisance; this is why we
2855 did not make `-Wall' request these warnings.
2858 Warn if a prototype causes a type conversion that is different
2859 from what would happen to the same argument in the absence of a
2860 prototype. This includes conversions of fixed point to floating
2861 and vice versa, and conversions changing the width or signedness
2862 of a fixed point argument except when the same as the default
2865 Also, warn if a negative integer constant expression is implicitly
2866 converted to an unsigned type. For example, warn about the
2867 assignment `x = -1' if `x' is unsigned. But do not warn about
2868 explicit casts like `(unsigned) -1'.
2871 Warn when a comparison between signed and unsigned values could
2872 produce an incorrect result when the signed value is converted to
2873 unsigned. This warning is also enabled by `-Wextra'; to get the
2874 other warnings of `-Wextra' without this warning, use `-Wextra
2877 `-Waggregate-return'
2878 Warn if any functions that return structures or unions are defined
2879 or called. (In languages where you can return an array, this also
2882 `-Wstrict-prototypes (C only)'
2883 Warn if a function is declared or defined without specifying the
2884 argument types. (An old-style function definition is permitted
2885 without a warning if preceded by a declaration which specifies the
2888 `-Wold-style-definition (C only)'
2889 Warn if an old-style function definition is used. A warning is
2890 given even if there is a previous prototype.
2892 `-Wmissing-prototypes (C only)'
2893 Warn if a global function is defined without a previous prototype
2894 declaration. This warning is issued even if the definition itself
2895 provides a prototype. The aim is to detect global functions that
2896 fail to be declared in header files.
2898 `-Wmissing-declarations (C only)'
2899 Warn if a global function is defined without a previous
2900 declaration. Do so even if the definition itself provides a
2901 prototype. Use this option to detect global functions that are
2902 not declared in header files.
2904 `-Wmissing-field-initializers'
2905 Warn if a structure's initializer has some fields missing. For
2906 example, the following code would cause such a warning, because
2907 `x.h' is implicitly zero:
2909 struct s { int f, g, h; };
2910 struct s x = { 3, 4 };
2912 This option does not warn about designated initializers, so the
2913 following modification would not trigger a warning:
2915 struct s { int f, g, h; };
2916 struct s x = { .f = 3, .g = 4 };
2918 This warning is included in `-Wextra'. To get other `-Wextra'
2919 warnings without this one, use `-Wextra
2920 -Wno-missing-field-initializers'.
2922 `-Wmissing-noreturn'
2923 Warn about functions which might be candidates for attribute
2924 `noreturn'. Note these are only possible candidates, not absolute
2925 ones. Care should be taken to manually verify functions actually
2926 do not ever return before adding the `noreturn' attribute,
2927 otherwise subtle code generation bugs could be introduced. You
2928 will not get a warning for `main' in hosted C environments.
2930 `-Wmissing-format-attribute'
2931 If `-Wformat' is enabled, also warn about functions which might be
2932 candidates for `format' attributes. Note these are only possible
2933 candidates, not absolute ones. GCC will guess that `format'
2934 attributes might be appropriate for any function that calls a
2935 function like `vprintf' or `vscanf', but this might not always be
2936 the case, and some functions for which `format' attributes are
2937 appropriate may not be detected. This option has no effect unless
2938 `-Wformat' is enabled (possibly by `-Wall').
2941 Do not warn if a multicharacter constant (`'FOOF'') is used.
2942 Usually they indicate a typo in the user's code, as they have
2943 implementation-defined values, and should not be used in portable
2946 `-Wno-deprecated-declarations'
2947 Do not warn about uses of functions, variables, and types marked as
2948 deprecated by using the `deprecated' attribute. (*note Function
2949 Attributes::, *note Variable Attributes::, *note Type
2953 Warn if a structure is given the packed attribute, but the packed
2954 attribute has no effect on the layout or size of the structure.
2955 Such structures may be mis-aligned for little benefit. For
2956 instance, in this code, the variable `f.x' in `struct bar' will be
2957 misaligned even though `struct bar' does not itself have the
2963 } __attribute__((packed));
2970 Warn if padding is included in a structure, either to align an
2971 element of the structure or to align the whole structure.
2972 Sometimes when this happens it is possible to rearrange the fields
2973 of the structure to reduce the padding and so make the structure
2977 Warn if anything is declared more than once in the same scope,
2978 even in cases where multiple declaration is valid and changes
2981 `-Wnested-externs (C only)'
2982 Warn if an `extern' declaration is encountered within a function.
2984 `-Wunreachable-code'
2985 Warn if the compiler detects that code will never be executed.
2987 This option is intended to warn when the compiler detects that at
2988 least a whole line of source code will never be executed, because
2989 some condition is never satisfied or because it is after a
2990 procedure that never returns.
2992 It is possible for this option to produce a warning even though
2993 there are circumstances under which part of the affected line can
2994 be executed, so care should be taken when removing
2995 apparently-unreachable code.
2997 For instance, when a function is inlined, a warning may mean that
2998 the line is unreachable in only one inlined copy of the function.
3000 This option is not made part of `-Wall' because in a debugging
3001 version of a program there is often substantial code which checks
3002 correct functioning of the program and is, hopefully, unreachable
3003 because the program does work. Another common use of unreachable
3004 code is to provide behavior which is selectable at compile-time.
3007 Warn if a function can not be inlined and it was declared as
3008 inline. Even with this option, the compiler will not warn about
3009 failures to inline functions declared in system headers.
3011 The compiler uses a variety of heuristics to determine whether or
3012 not to inline a function. For example, the compiler takes into
3013 account the size of the function being inlined and the amount of
3014 inlining that has already been done in the current function.
3015 Therefore, seemingly insignificant changes in the source program
3016 can cause the warnings produced by `-Winline' to appear or
3019 `-Wno-invalid-offsetof (C++ only)'
3020 Suppress warnings from applying the `offsetof' macro to a non-POD
3021 type. According to the 1998 ISO C++ standard, applying `offsetof'
3022 to a non-POD type is undefined. In existing C++ implementations,
3023 however, `offsetof' typically gives meaningful results even when
3024 applied to certain kinds of non-POD types. (Such as a simple
3025 `struct' that fails to be a POD type only by virtue of having a
3026 constructor.) This flag is for users who are aware that they are
3027 writing nonportable code and who have deliberately chosen to
3028 ignore the warning about it.
3030 The restrictions on `offsetof' may be relaxed in a future version
3031 of the C++ standard.
3034 Warn if a precompiled header (*note Precompiled Headers::) is
3035 found in the search path but can't be used.
3038 Warn if `long long' type is used. This is default. To inhibit
3039 the warning messages, use `-Wno-long-long'. Flags `-Wlong-long'
3040 and `-Wno-long-long' are taken into account only when `-pedantic'
3044 Warn if variadic macros are used in pedantic ISO C90 mode, or the
3045 GNU alternate syntax when in pedantic ISO C99 mode. This is
3046 default. To inhibit the warning messages, use
3047 `-Wno-variadic-macros'.
3049 `-Wdisabled-optimization'
3050 Warn if a requested optimization pass is disabled. This warning
3051 does not generally indicate that there is anything wrong with your
3052 code; it merely indicates that GCC's optimizers were unable to
3053 handle the code effectively. Often, the problem is that your code
3054 is too big or too complex; GCC will refuse to optimize programs
3055 when the optimization itself is likely to take inordinate amounts
3059 Don't warn for pointer argument passing or assignment with
3060 different signedness. Only useful in the negative form since this
3061 warning is enabled by default. This option is only supported for
3065 Make all warnings into errors.
3068 File: gcc.info, Node: Debugging Options, Next: Optimize Options, Prev: Warning Options, Up: Invoking GCC
3070 3.9 Options for Debugging Your Program or GCC
3071 =============================================
3073 GCC has various special options that are used for debugging either your
3077 Produce debugging information in the operating system's native
3078 format (stabs, COFF, XCOFF, or DWARF 2). GDB can work with this
3079 debugging information.
3081 On most systems that use stabs format, `-g' enables use of extra
3082 debugging information that only GDB can use; this extra information
3083 makes debugging work better in GDB but will probably make other
3084 debuggers crash or refuse to read the program. If you want to
3085 control for certain whether to generate the extra information, use
3086 `-gstabs+', `-gstabs', `-gxcoff+', `-gxcoff', or `-gvms' (see
3089 GCC allows you to use `-g' with `-O'. The shortcuts taken by
3090 optimized code may occasionally produce surprising results: some
3091 variables you declared may not exist at all; flow of control may
3092 briefly move where you did not expect it; some statements may not
3093 be executed because they compute constant results or their values
3094 were already at hand; some statements may execute in different
3095 places because they were moved out of loops.
3097 Nevertheless it proves possible to debug optimized output. This
3098 makes it reasonable to use the optimizer for programs that might
3101 The following options are useful when GCC is generated with the
3102 capability for more than one debugging format.
3105 Produce debugging information for use by GDB. This means to use
3106 the most expressive format available (DWARF 2, stabs, or the
3107 native format if neither of those are supported), including GDB
3108 extensions if at all possible.
3111 Produce debugging information in stabs format (if that is
3112 supported), without GDB extensions. This is the format used by
3113 DBX on most BSD systems. On MIPS, Alpha and System V Release 4
3114 systems this option produces stabs debugging output which is not
3115 understood by DBX or SDB. On System V Release 4 systems this
3116 option requires the GNU assembler.
3118 `-feliminate-unused-debug-symbols'
3119 Produce debugging information in stabs format (if that is
3120 supported), for only symbols that are actually used.
3123 Produce debugging information in stabs format (if that is
3124 supported), using GNU extensions understood only by the GNU
3125 debugger (GDB). The use of these extensions is likely to make
3126 other debuggers crash or refuse to read the program.
3129 Produce debugging information in COFF format (if that is
3130 supported). This is the format used by SDB on most System V
3131 systems prior to System V Release 4.
3134 Produce debugging information in XCOFF format (if that is
3135 supported). This is the format used by the DBX debugger on IBM
3139 Produce debugging information in XCOFF format (if that is
3140 supported), using GNU extensions understood only by the GNU
3141 debugger (GDB). The use of these extensions is likely to make
3142 other debuggers crash or refuse to read the program, and may cause
3143 assemblers other than the GNU assembler (GAS) to fail with an
3147 Produce debugging information in DWARF version 2 format (if that is
3148 supported). This is the format used by DBX on IRIX 6. With this
3149 option, GCC uses features of DWARF version 3 when they are useful;
3150 version 3 is upward compatible with version 2, but may still cause
3151 problems for older debuggers.
3154 Produce debugging information in VMS debug format (if that is
3155 supported). This is the format used by DEBUG on VMS systems.
3163 Request debugging information and also use LEVEL to specify how
3164 much information. The default level is 2.
3166 Level 1 produces minimal information, enough for making backtraces
3167 in parts of the program that you don't plan to debug. This
3168 includes descriptions of functions and external variables, but no
3169 information about local variables and no line numbers.
3171 Level 3 includes extra information, such as all the macro
3172 definitions present in the program. Some debuggers support macro
3173 expansion when you use `-g3'.
3175 `-gdwarf-2' does not accept a concatenated debug level, because
3176 GCC used to support an option `-gdwarf' that meant to generate
3177 debug information in version 1 of the DWARF format (which is very
3178 different from version 2), and it would have been too confusing.
3179 That debug format is long obsolete, but the option cannot be
3180 changed now. Instead use an additional `-gLEVEL' option to change
3181 the debug level for DWARF2.
3183 `-feliminate-dwarf2-dups'
3184 Compress DWARF2 debugging information by eliminating duplicated
3185 information about each symbol. This option only makes sense when
3186 generating DWARF2 debugging information with `-gdwarf-2'.
3189 Generate extra code to write profile information suitable for the
3190 analysis program `prof'. You must use this option when compiling
3191 the source files you want data about, and you must also use it when
3195 Generate extra code to write profile information suitable for the
3196 analysis program `gprof'. You must use this option when compiling
3197 the source files you want data about, and you must also use it when
3201 Makes the compiler print out each function name as it is compiled,
3202 and print some statistics about each pass when it finishes.
3205 Makes the compiler print some statistics about the time consumed
3206 by each pass when it finishes.
3209 Makes the compiler print some statistics about permanent memory
3210 allocation when it finishes.
3213 Add code so that program flow "arcs" are instrumented. During
3214 execution the program records how many times each branch and call
3215 is executed and how many times it is taken or returns. When the
3216 compiled program exits it saves this data to a file called
3217 `AUXNAME.gcda' for each source file. The data may be used for
3218 profile-directed optimizations (`-fbranch-probabilities'), or for
3219 test coverage analysis (`-ftest-coverage'). Each object file's
3220 AUXNAME is generated from the name of the output file, if
3221 explicitly specified and it is not the final executable, otherwise
3222 it is the basename of the source file. In both cases any suffix
3223 is removed (e.g. `foo.gcda' for input file `dir/foo.c', or
3224 `dir/foo.gcda' for output file specified as `-o dir/foo.o').
3226 * Compile the source files with `-fprofile-arcs' plus
3227 optimization and code generation options. For test coverage
3228 analysis, use the additional `-ftest-coverage' option. You
3229 do not need to profile every source file in a program.
3231 * Link your object files with `-lgcov' or `-fprofile-arcs' (the
3232 latter implies the former).
3234 * Run the program on a representative workload to generate the
3235 arc profile information. This may be repeated any number of
3236 times. You can run concurrent instances of your program, and
3237 provided that the file system supports locking, the data
3238 files will be correctly updated. Also `fork' calls are
3239 detected and correctly handled (double counting will not
3242 * For profile-directed optimizations, compile the source files
3243 again with the same optimization and code generation options
3244 plus `-fbranch-probabilities' (*note Options that Control
3245 Optimization: Optimize Options.).
3247 * For test coverage analysis, use `gcov' to produce human
3248 readable information from the `.gcno' and `.gcda' files.
3249 Refer to the `gcov' documentation for further information.
3252 With `-fprofile-arcs', for each function of your program GCC
3253 creates a program flow graph, then finds a spanning tree for the
3254 graph. Only arcs that are not on the spanning tree have to be
3255 instrumented: the compiler adds code to count the number of times
3256 that these arcs are executed. When an arc is the only exit or
3257 only entrance to a block, the instrumentation code can be added to
3258 the block; otherwise, a new basic block must be created to hold
3259 the instrumentation code.
3261 `-ftree-based-profiling'
3262 This option is used in addition to `-fprofile-arcs' or
3263 `-fbranch-probabilities' to control whether those optimizations
3264 are performed on a tree-based or rtl-based internal representation.
3265 If you use this option when compiling with `-fprofile-arcs', you
3266 must also use it when compiling later with
3267 `-fbranch-probabilities'. Currently the tree-based optimization
3268 is in an early stage of development, and this option is
3269 recommended only for those people working on improving it.
3272 Produce a notes file that the `gcov' code-coverage utility (*note
3273 `gcov'--a Test Coverage Program: Gcov.) can use to show program
3274 coverage. Each source file's note file is called `AUXNAME.gcno'.
3275 Refer to the `-fprofile-arcs' option above for a description of
3276 AUXNAME and instructions on how to generate test coverage data.
3277 Coverage data will match the source files more closely, if you do
3283 Says to make debugging dumps during compilation at times specified
3284 by LETTERS. This is used for debugging the RTL-based passes of
3285 the compiler. The file names for most of the dumps are made by
3286 appending a pass number and a word to the DUMPNAME. DUMPNAME is
3287 generated from the name of the output file, if explicitly
3288 specified and it is not an executable, otherwise it is the
3289 basename of the source file.
3291 Most debug dumps can be enabled either passing a letter to the `-d'
3292 option, or with a long `-fdump-rtl' switch; here are the possible
3293 letters for use in LETTERS and PASS, and their meanings:
3296 Annotate the assembler output with miscellaneous debugging
3301 Dump after computing branch probabilities, to `FILE.09.bp'.
3305 Dump after block reordering, to `FILE.30.bbro'.
3308 `-fdump-rtl-combine'
3309 Dump after instruction combination, to the file
3315 `-dC' and `-fdump-rtl-ce1' enable dumping after the first if
3316 conversion, to the file `FILE.11.ce1'. `-dC' and
3317 `-fdump-rtl-ce2' enable dumping after the second if
3318 conversion, to the file `FILE.18.ce2'.
3323 `-dd' and `-fdump-rtl-btl' enable dumping after branch target
3324 load optimization, to `FILE.31.btl'. `-dd' and
3325 `-fdump-rtl-dbr' enable dumping after delayed branch
3326 scheduling, to `FILE.36.dbr'.
3329 Dump all macro definitions, at the end of preprocessing, in
3330 addition to normal output.
3334 Dump after the third if conversion, to `FILE.28.ce3'.
3339 `-df' and `-fdump-rtl-cfg' enable dumping after control and
3340 data flow analysis, to `FILE.08.cfg'. `-df' and
3341 `-fdump-rtl-cfg' enable dumping dump after life analysis, to
3346 Dump after global register allocation, to `FILE.23.greg'.
3351 `-dG' and `-fdump-rtl-gcse' enable dumping after GCSE, to
3352 `FILE.05.gcse'. `-dG' and `-fdump-rtl-bypass' enable dumping
3353 after jump bypassing and control flow optimizations, to
3358 Dump after finalization of EH handling code, to `FILE.02.eh'.
3361 `-fdump-rtl-sibling'
3362 Dump after sibling call optimizations, to `FILE.01.sibling'.
3366 Dump after the first jump optimization, to `FILE.03.jump'.
3370 Dump after conversion from registers to stack, to
3375 Dump after local register allocation, to `FILE.22.lreg'.
3380 `-dL' and `-fdump-rtl-loop' enable dumping after the first
3381 loop optimization pass, to `FILE.06.loop'. `-dL' and
3382 `-fdump-rtl-loop2' enable dumping after the second pass, to
3387 Dump after modulo scheduling, to `FILE.20.sms'.
3391 Dump after performing the machine dependent reorganization
3392 pass, to `FILE.35.mach'.
3396 Dump after register renumbering, to `FILE.29.rnreg'.
3399 `-fdump-rtl-regmove'
3400 Dump after the register move pass, to `FILE.19.regmove'.
3403 `-fdump-rtl-postreload'
3404 Dump after post-reload optimizations, to `FILE.24.postreload'.
3408 Dump after RTL generation, to `FILE.00.expand'.
3412 Dump after the second scheduling pass, to `FILE.32.sched2'.
3416 Dump after CSE (including the jump optimization that
3417 sometimes follows CSE), to `FILE.04.cse'.
3421 Dump after the first scheduling pass, to `FILE.21.sched'.
3425 Dump after the second CSE pass (including the jump
3426 optimization that sometimes follows CSE), to `FILE.15.cse2'.
3430 Dump after running tracer, to `FILE.12.tracer'.
3434 `-fdump-rtl-vartrack'
3435 `-dV' and `-fdump-rtl-vpt' enable dumping after the value
3436 profile transformations, to `FILE.10.vpt'. `-dV' and
3437 `-fdump-rtl-vartrack' enable dumping after variable tracking,
3438 to `FILE.34.vartrack'.
3442 Dump after the second flow pass, to `FILE.26.flow2'.
3445 `-fdump-rtl-peephole2'
3446 Dump after the peephole pass, to `FILE.27.peephole2'.
3450 Dump after live range splitting, to `FILE.14.web'.
3454 Produce all the dumps listed above.
3457 Produce a core dump whenever an error occurs.
3460 Print statistics on memory usage, at the end of the run, to
3464 Annotate the assembler output with a comment indicating which
3465 pattern and alternative was used. The length of each
3466 instruction is also printed.
3469 Dump the RTL in the assembler output as a comment before each
3470 instruction. Also turns on `-dp' annotation.
3473 For each of the other indicated dump files (either with `-d'
3474 or `-fdump-rtl-PASS'), dump a representation of the control
3475 flow graph suitable for viewing with VCG to `FILE.PASS.vcg'.
3478 Just generate RTL for a function instead of compiling it.
3479 Usually used with `r' (`-fdump-rtl-expand').
3482 Dump debugging information during parsing, to standard error.
3485 When doing debugging dumps (see `-d' option above), suppress
3486 instruction numbers and line number note output. This makes it
3487 more feasible to use diff on debugging dumps for compiler
3488 invocations with different options, in particular with and without
3491 `-fdump-translation-unit (C and C++ only)'
3492 `-fdump-translation-unit-OPTIONS (C and C++ only)'
3493 Dump a representation of the tree structure for the entire
3494 translation unit to a file. The file name is made by appending
3495 `.tu' to the source file name. If the `-OPTIONS' form is used,
3496 OPTIONS controls the details of the dump as described for the
3497 `-fdump-tree' options.
3499 `-fdump-class-hierarchy (C++ only)'
3500 `-fdump-class-hierarchy-OPTIONS (C++ only)'
3501 Dump a representation of each class's hierarchy and virtual
3502 function table layout to a file. The file name is made by
3503 appending `.class' to the source file name. If the `-OPTIONS'
3504 form is used, OPTIONS controls the details of the dump as
3505 described for the `-fdump-tree' options.
3508 Control the dumping at various stages of inter-procedural analysis
3509 language tree to a file. The file name is generated by appending
3510 a switch specific suffix to the source file name. The following
3514 Enables all inter-procedural analysis dumps; currently the
3515 only produced dump is the `cgraph' dump.
3518 Dumps information about call-graph optimization, unused
3519 function removal, and inlining decisions.
3521 `-fdump-tree-SWITCH (C and C++ only)'
3522 `-fdump-tree-SWITCH-OPTIONS (C and C++ only)'
3523 Control the dumping at various stages of processing the
3524 intermediate language tree to a file. The file name is generated
3525 by appending a switch specific suffix to the source file name. If
3526 the `-OPTIONS' form is used, OPTIONS is a list of `-' separated
3527 options that control the details of the dump. Not all options are
3528 applicable to all dumps, those which are not meaningful will be
3529 ignored. The following options are available
3532 Print the address of each node. Usually this is not
3533 meaningful as it changes according to the environment and
3534 source file. Its primary use is for tying up a dump file
3535 with a debug environment.
3538 Inhibit dumping of members of a scope or body of a function
3539 merely because that scope has been reached. Only dump such
3540 items when they are directly reachable by some other path.
3541 When dumping pretty-printed trees, this option inhibits
3542 dumping the bodies of control structures.
3545 Print a raw representation of the tree. By default, trees are
3546 pretty-printed into a C-like representation.
3549 Enable more detailed dumps (not honored by every dump option).
3552 Enable dumping various statistics about the pass (not honored
3553 by every dump option).
3556 Enable showing basic block boundaries (disabled in raw dumps).
3559 Enable showing virtual operands for every statement.
3562 Enable showing line numbers for statements.
3565 Enable showing the unique ID (`DECL_UID') for each variable.
3568 Turn on all options, except `raw', `slim' and `lineno'.
3570 The following tree dumps are possible:
3572 Dump before any tree based optimization, to `FILE.original'.
3575 Dump after all tree based optimization, to `FILE.optimized'.
3578 Dump after function inlining, to `FILE.inlined'.
3581 Dump each function before and after the gimplification pass
3582 to a file. The file name is made by appending `.gimple' to
3583 the source file name.
3586 Dump the control flow graph of each function to a file. The
3587 file name is made by appending `.cfg' to the source file name.
3590 Dump the control flow graph of each function to a file in VCG
3591 format. The file name is made by appending `.vcg' to the
3592 source file name. Note that if the file contains more than
3593 one function, the generated file cannot be used directly by
3594 VCG. You will need to cut and paste each function's graph
3595 into its own separate file first.
3598 Dump each function after copying loop headers. The file name
3599 is made by appending `.ch' to the source file name.
3602 Dump SSA related information to a file. The file name is
3603 made by appending `.ssa' to the source file name.
3606 Dump aliasing information for each function. The file name
3607 is made by appending `.alias' to the source file name.
3610 Dump each function after CCP. The file name is made by
3611 appending `.ccp' to the source file name.
3614 Dump trees after partial redundancy elimination. The file
3615 name is made by appending `.pre' to the source file name.
3618 Dump trees after full redundancy elimination. The file name
3619 is made by appending `.fre' to the source file name.
3622 Dump each function after dead code elimination. The file
3623 name is made by appending `.dce' to the source file name.
3626 Dump each function after adding mudflap instrumentation. The
3627 file name is made by appending `.mudflap' to the source file
3631 Dump each function after performing scalar replacement of
3632 aggregates. The file name is made by appending `.sra' to the
3636 Dump each function after applying dominator tree
3637 optimizations. The file name is made by appending `.dom' to
3638 the source file name.
3641 Dump each function after applying dead store elimination.
3642 The file name is made by appending `.dse' to the source file
3646 Dump each function after optimizing PHI nodes into
3647 straightline code. The file name is made by appending
3648 `.phiopt' to the source file name.
3651 Dump each function after forward propagating single use
3652 variables. The file name is made by appending `.forwprop' to
3653 the source file name.
3656 Dump each function after applying the copy rename
3657 optimization. The file name is made by appending
3658 `.copyrename' to the source file name.
3661 Dump each function after applying the named return value
3662 optimization on generic trees. The file name is made by
3663 appending `.nrv' to the source file name.
3666 Dump each function after applying vectorization of loops.
3667 The file name is made by appending `.vect' to the source file
3671 Enable all the available tree dumps with the flags provided
3674 `-ftree-vectorizer-verbose=N'
3675 This option controls the amount of debugging output the vectorizer
3676 prints. This information is written to standard error, unless
3677 `-fdump-tree-all' or `-fdump-tree-vect' is specified, in which
3678 case it is output to the usual dump listing file, `.vect'.
3680 `-frandom-seed=STRING'
3681 This option provides a seed that GCC uses when it would otherwise
3682 use random numbers. It is used to generate certain symbol names
3683 that have to be different in every compiled file. It is also used
3684 to place unique stamps in coverage data files and the object files
3685 that produce them. You can use the `-frandom-seed' option to
3686 produce reproducibly identical object files.
3688 The STRING should be different for every file you compile.
3691 On targets that use instruction scheduling, this option controls
3692 the amount of debugging output the scheduler prints. This
3693 information is written to standard error, unless `-dS' or `-dR' is
3694 specified, in which case it is output to the usual dump listing
3695 file, `.sched' or `.sched2' respectively. However for N greater
3696 than nine, the output is always printed to standard error.
3698 For N greater than zero, `-fsched-verbose' outputs the same
3699 information as `-dRS'. For N greater than one, it also output
3700 basic block probabilities, detailed ready list information and
3701 unit/insn info. For N greater than two, it includes RTL at abort
3702 point, control-flow and regions info. And for N over four,
3703 `-fsched-verbose' also includes dependence info.
3706 Store the usual "temporary" intermediate files permanently; place
3707 them in the current directory and name them based on the source
3708 file. Thus, compiling `foo.c' with `-c -save-temps' would produce
3709 files `foo.i' and `foo.s', as well as `foo.o'. This creates a
3710 preprocessed `foo.i' output file even though the compiler now
3711 normally uses an integrated preprocessor.
3713 When used in combination with the `-x' command line option,
3714 `-save-temps' is sensible enough to avoid over writing an input
3715 source file with the same extension as an intermediate file. The
3716 corresponding intermediate file may be obtained by renaming the
3717 source file before using `-save-temps'.
3720 Report the CPU time taken by each subprocess in the compilation
3721 sequence. For C source files, this is the compiler proper and
3722 assembler (plus the linker if linking is done). The output looks
3728 The first number on each line is the "user time", that is time
3729 spent executing the program itself. The second number is "system
3730 time", time spent executing operating system routines on behalf of
3731 the program. Both numbers are in seconds.
3734 Run variable tracking pass. It computes where variables are
3735 stored at each position in code. Better debugging information is
3736 then generated (if the debugging information format supports this
3739 It is enabled by default when compiling with optimization (`-Os',
3740 `-O', `-O2', ...), debugging information (`-g') and the debug info
3743 `-print-file-name=LIBRARY'
3744 Print the full absolute name of the library file LIBRARY that
3745 would be used when linking--and don't do anything else. With this
3746 option, GCC does not compile or link anything; it just prints the
3749 `-print-multi-directory'
3750 Print the directory name corresponding to the multilib selected by
3751 any other switches present in the command line. This directory is
3752 supposed to exist in `GCC_EXEC_PREFIX'.
3755 Print the mapping from multilib directory names to compiler
3756 switches that enable them. The directory name is separated from
3757 the switches by `;', and each switch starts with an `@' instead of
3758 the `-', without spaces between multiple switches. This is
3759 supposed to ease shell-processing.
3761 `-print-prog-name=PROGRAM'
3762 Like `-print-file-name', but searches for a program such as `cpp'.
3764 `-print-libgcc-file-name'
3765 Same as `-print-file-name=libgcc.a'.
3767 This is useful when you use `-nostdlib' or `-nodefaultlibs' but
3768 you do want to link with `libgcc.a'. You can do
3770 gcc -nostdlib FILES... `gcc -print-libgcc-file-name`
3772 `-print-search-dirs'
3773 Print the name of the configured installation directory and a list
3774 of program and library directories `gcc' will search--and don't do
3777 This is useful when `gcc' prints the error message `installation
3778 problem, cannot exec cpp0: No such file or directory'. To resolve
3779 this you either need to put `cpp0' and the other compiler
3780 components where `gcc' expects to find them, or you can set the
3781 environment variable `GCC_EXEC_PREFIX' to the directory where you
3782 installed them. Don't forget the trailing `/'. *Note Environment
3786 Print the compiler's target machine (for example,
3787 `i686-pc-linux-gnu')--and don't do anything else.
3790 Print the compiler version (for example, `3.0')--and don't do
3794 Print the compiler's built-in specs--and don't do anything else.
3795 (This is used when GCC itself is being built.) *Note Spec Files::.
3797 `-feliminate-unused-debug-types'
3798 Normally, when producing DWARF2 output, GCC will emit debugging
3799 information for all types declared in a compilation unit,
3800 regardless of whether or not they are actually used in that
3801 compilation unit. Sometimes this is useful, such as if, in the
3802 debugger, you want to cast a value to a type that is not actually
3803 used in your program (but is declared). More often, however, this
3804 results in a significant amount of wasted space. With this
3805 option, GCC will avoid producing debug symbol output for types
3806 that are nowhere used in the source file being compiled.
3809 File: gcc.info, Node: Optimize Options, Next: Preprocessor Options, Prev: Debugging Options, Up: Invoking GCC
3811 3.10 Options That Control Optimization
3812 ======================================
3814 These options control various sorts of optimizations.
3816 Without any optimization option, the compiler's goal is to reduce the
3817 cost of compilation and to make debugging produce the expected results.
3818 Statements are independent: if you stop the program with a breakpoint
3819 between statements, you can then assign a new value to any variable or
3820 change the program counter to any other statement in the function and
3821 get exactly the results you would expect from the source code.
3823 Turning on optimization flags makes the compiler attempt to improve
3824 the performance and/or code size at the expense of compilation time and
3825 possibly the ability to debug the program.
3827 The compiler performs optimization based on the knowledge it has of
3828 the program. Optimization levels `-O2' and above, in particular,
3829 enable _unit-at-a-time_ mode, which allows the compiler to consider
3830 information gained from later functions in the file when compiling a
3831 function. Compiling multiple files at once to a single output file in
3832 _unit-at-a-time_ mode allows the compiler to use information gained
3833 from all of the files when compiling each of them.
3835 Not all optimizations are controlled directly by a flag. Only
3836 optimizations that have a flag are listed.
3840 Optimize. Optimizing compilation takes somewhat more time, and a
3841 lot more memory for a large function.
3843 With `-O', the compiler tries to reduce code size and execution
3844 time, without performing any optimizations that take a great deal
3845 of compilation time.
3847 `-O' turns on the following optimization flags:
3850 -fguess-branch-probability
3857 -ftree-dominator-opts
3867 `-O' also turns on `-fomit-frame-pointer' on machines where doing
3868 so does not interfere with debugging.
3870 `-O' doesn't turn on `-ftree-sra' for the Ada compiler. This
3871 option must be explicitly specified on the command line to be
3872 enabled for the Ada compiler.
3875 Optimize even more. GCC performs nearly all supported
3876 optimizations that do not involve a space-speed tradeoff. The
3877 compiler does not perform loop unrolling or function inlining when
3878 you specify `-O2'. As compared to `-O', this option increases
3879 both compilation time and the performance of the generated code.
3881 `-O2' turns on all optimization flags specified by `-O'. It also
3882 turns on the following optimization flags:
3885 -foptimize-sibling-calls
3886 -fcse-follow-jumps -fcse-skip-blocks
3888 -fexpensive-optimizations
3890 -frerun-cse-after-loop -frerun-loop-opt
3894 -fschedule-insns -fschedule-insns2
3895 -fsched-interblock -fsched-spec
3898 -fdelete-null-pointer-checks
3899 -freorder-blocks -freorder-functions
3901 -falign-functions -falign-jumps
3902 -falign-loops -falign-labels
3905 Please note the warning under `-fgcse' about invoking `-O2' on
3906 programs that use computed gotos.
3909 Optimize yet more. `-O3' turns on all optimizations specified by
3910 `-O2' and also turns on the `-finline-functions',
3911 `-funswitch-loops' and `-fgcse-after-reload' options.
3914 Do not optimize. This is the default.
3917 Optimize for size. `-Os' enables all `-O2' optimizations that do
3918 not typically increase code size. It also performs further
3919 optimizations designed to reduce code size.
3921 `-Os' disables the following optimization flags:
3922 -falign-functions -falign-jumps -falign-loops
3923 -falign-labels -freorder-blocks -freorder-blocks-and-partition -fprefetch-loop-arrays
3925 If you use multiple `-O' options, with or without level numbers,
3926 the last such option is the one that is effective.
3928 Options of the form `-fFLAG' specify machine-independent flags. Most
3929 flags have both positive and negative forms; the negative form of
3930 `-ffoo' would be `-fno-foo'. In the table below, only one of the forms
3931 is listed--the one you typically will use. You can figure out the
3932 other form by either removing `no-' or adding it.
3934 The following options control specific optimizations. They are either
3935 activated by `-O' options or are related to ones that are. You can use
3936 the following flags in the rare cases when "fine-tuning" of
3937 optimizations to be performed is desired.
3939 `-fno-default-inline'
3940 Do not make member functions inline by default merely because they
3941 are defined inside the class scope (C++ only). Otherwise, when
3942 you specify `-O', member functions defined inside class scope are
3943 compiled inline by default; i.e., you don't need to add `inline'
3944 in front of the member function name.
3947 Always pop the arguments to each function call as soon as that
3948 function returns. For machines which must pop arguments after a
3949 function call, the compiler normally lets arguments accumulate on
3950 the stack for several function calls and pops them all at once.
3952 Disabled at levels `-O', `-O2', `-O3', `-Os'.
3955 Force memory operands to be copied into registers before doing
3956 arithmetic on them. This produces better code by making all memory
3957 references potential common subexpressions. When they are not
3958 common subexpressions, instruction combination should eliminate
3959 the separate register-load.
3961 Enabled at levels `-O2', `-O3', `-Os'.
3964 Force memory address constants to be copied into registers before
3965 doing arithmetic on them. This may produce better code just as
3968 `-fomit-frame-pointer'
3969 Don't keep the frame pointer in a register for functions that
3970 don't need one. This avoids the instructions to save, set up and
3971 restore frame pointers; it also makes an extra register available
3972 in many functions. *It also makes debugging impossible on some
3975 On some machines, such as the VAX, this flag has no effect, because
3976 the standard calling sequence automatically handles the frame
3977 pointer and nothing is saved by pretending it doesn't exist. The
3978 machine-description macro `FRAME_POINTER_REQUIRED' controls
3979 whether a target machine supports this flag. *Note Register
3980 Usage: (gccint)Registers.
3982 Enabled at levels `-O', `-O2', `-O3', `-Os'.
3984 `-foptimize-sibling-calls'
3985 Optimize sibling and tail recursive calls.
3987 Enabled at levels `-O2', `-O3', `-Os'.
3990 Don't pay attention to the `inline' keyword. Normally this option
3991 is used to keep the compiler from expanding any functions inline.
3992 Note that if you are not optimizing, no functions can be expanded
3995 `-finline-functions'
3996 Integrate all simple functions into their callers. The compiler
3997 heuristically decides which functions are simple enough to be worth
3998 integrating in this way.
4000 If all calls to a given function are integrated, and the function
4001 is declared `static', then the function is normally not output as
4002 assembler code in its own right.
4004 Enabled at level `-O3'.
4007 By default, GCC limits the size of functions that can be inlined.
4008 This flag allows the control of this limit for functions that are
4009 explicitly marked as inline (i.e., marked with the inline keyword
4010 or defined within the class definition in c++). N is the size of
4011 functions that can be inlined in number of pseudo instructions
4012 (not counting parameter handling). The default value of N is 600.
4013 Increasing this value can result in more inlined code at the cost
4014 of compilation time and memory consumption. Decreasing usually
4015 makes the compilation faster and less code will be inlined (which
4016 presumably means slower programs). This option is particularly
4017 useful for programs that use inlining heavily such as those based
4018 on recursive templates with C++.
4020 Inlining is actually controlled by a number of parameters, which
4021 may be specified individually by using `--param NAME=VALUE'. The
4022 `-finline-limit=N' option sets some of these parameters as follows:
4024 `max-inline-insns-single'
4027 `max-inline-insns-auto'
4031 is set to 130 or N/4, whichever is smaller.
4033 `max-inline-insns-rtl'
4036 See below for a documentation of the individual parameters
4037 controlling inlining.
4039 _Note:_ pseudo instruction represents, in this particular context,
4040 an abstract measurement of function's size. In no way, it
4041 represents a count of assembly instructions and as such its exact
4042 meaning might change from one release to an another.
4044 `-fkeep-inline-functions'
4045 In C, emit `static' functions that are declared `inline' into the
4046 object file, even if the function has been inlined into all of its
4047 callers. This switch does not affect functions using the `extern
4048 inline' extension in GNU C. In C++, emit any and all inline
4049 functions into the object file.
4051 `-fkeep-static-consts'
4052 Emit variables declared `static const' when optimization isn't
4053 turned on, even if the variables aren't referenced.
4055 GCC enables this option by default. If you want to force the
4056 compiler to check if the variable was referenced, regardless of
4057 whether or not optimization is turned on, use the
4058 `-fno-keep-static-consts' option.
4061 Attempt to merge identical constants (string constants and
4062 floating point constants) across compilation units.
4064 This option is the default for optimized compilation if the
4065 assembler and linker support it. Use `-fno-merge-constants' to
4066 inhibit this behavior.
4068 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4070 `-fmerge-all-constants'
4071 Attempt to merge identical constants and identical variables.
4073 This option implies `-fmerge-constants'. In addition to
4074 `-fmerge-constants' this considers e.g. even constant initialized
4075 arrays or initialized constant variables with integral or floating
4076 point types. Languages like C or C++ require each non-automatic
4077 variable to have distinct location, so using this option will
4078 result in non-conforming behavior.
4081 Perform swing modulo scheduling immediately before the first
4082 scheduling pass. This pass looks at innermost loops and reorders
4083 their instructions by overlapping different iterations.
4085 `-fno-branch-count-reg'
4086 Do not use "decrement and branch" instructions on a count register,
4087 but instead generate a sequence of instructions that decrement a
4088 register, compare it against zero, then branch based upon the
4089 result. This option is only meaningful on architectures that
4090 support such instructions, which include x86, PowerPC, IA-64 and
4093 The default is `-fbranch-count-reg', enabled when
4094 `-fstrength-reduce' is enabled.
4097 Do not put function addresses in registers; make each instruction
4098 that calls a constant function contain the function's address
4101 This option results in less efficient code, but some strange hacks
4102 that alter the assembler output may be confused by the
4103 optimizations performed when this option is not used.
4105 The default is `-ffunction-cse'
4107 `-fno-zero-initialized-in-bss'
4108 If the target supports a BSS section, GCC by default puts
4109 variables that are initialized to zero into BSS. This can save
4110 space in the resulting code.
4112 This option turns off this behavior because some programs
4113 explicitly rely on variables going to the data section. E.g., so
4114 that the resulting executable can find the beginning of that
4115 section and/or make assumptions based on that.
4117 The default is `-fzero-initialized-in-bss'.
4120 For front-ends that support it, generate additional code to check
4121 that indices used to access arrays are within the declared range.
4122 This is currently only supported by the Java and Fortran
4123 front-ends, where this option defaults to true and false
4126 `-fmudflap -fmudflapth -fmudflapir'
4127 For front-ends that support it (C and C++), instrument all risky
4128 pointer/array dereferencing operations, some standard library
4129 string/heap functions, and some other associated constructs with
4130 range/validity tests. Modules so instrumented should be immune to
4131 buffer overflows, invalid heap use, and some other classes of C/C++
4132 programming errors. The instrumentation relies on a separate
4133 runtime library (`libmudflap'), which will be linked into a
4134 program if `-fmudflap' is given at link time. Run-time behavior
4135 of the instrumented program is controlled by the `MUDFLAP_OPTIONS'
4136 environment variable. See `env MUDFLAP_OPTIONS=-help a.out' for
4139 Use `-fmudflapth' instead of `-fmudflap' to compile and to link if
4140 your program is multi-threaded. Use `-fmudflapir', in addition to
4141 `-fmudflap' or `-fmudflapth', if instrumentation should ignore
4142 pointer reads. This produces less instrumentation (and therefore
4143 faster execution) and still provides some protection against
4144 outright memory corrupting writes, but allows erroneously read
4145 data to propagate within a program.
4148 Perform the optimizations of loop strength reduction and
4149 elimination of iteration variables.
4151 Enabled at levels `-O2', `-O3', `-Os'.
4154 Perform optimizations where we check to see if a jump branches to a
4155 location where another comparison subsumed by the first is found.
4156 If so, the first branch is redirected to either the destination of
4157 the second branch or a point immediately following it, depending
4158 on whether the condition is known to be true or false.
4160 Enabled at levels `-O2', `-O3', `-Os'.
4162 `-fcse-follow-jumps'
4163 In common subexpression elimination, scan through jump instructions
4164 when the target of the jump is not reached by any other path. For
4165 example, when CSE encounters an `if' statement with an `else'
4166 clause, CSE will follow the jump when the condition tested is
4169 Enabled at levels `-O2', `-O3', `-Os'.
4172 This is similar to `-fcse-follow-jumps', but causes CSE to follow
4173 jumps which conditionally skip over blocks. When CSE encounters a
4174 simple `if' statement with no else clause, `-fcse-skip-blocks'
4175 causes CSE to follow the jump around the body of the `if'.
4177 Enabled at levels `-O2', `-O3', `-Os'.
4179 `-frerun-cse-after-loop'
4180 Re-run common subexpression elimination after loop optimizations
4183 Enabled at levels `-O2', `-O3', `-Os'.
4186 Run the loop optimizer twice.
4188 Enabled at levels `-O2', `-O3', `-Os'.
4191 Perform a global common subexpression elimination pass. This pass
4192 also performs global constant and copy propagation.
4194 _Note:_ When compiling a program using computed gotos, a GCC
4195 extension, you may get better runtime performance if you disable
4196 the global common subexpression elimination pass by adding
4197 `-fno-gcse' to the command line.
4199 Enabled at levels `-O2', `-O3', `-Os'.
4202 When `-fgcse-lm' is enabled, global common subexpression
4203 elimination will attempt to move loads which are only killed by
4204 stores into themselves. This allows a loop containing a
4205 load/store sequence to be changed to a load outside the loop, and
4206 a copy/store within the loop.
4208 Enabled by default when gcse is enabled.
4211 When `-fgcse-sm' is enabled, a store motion pass is run after
4212 global common subexpression elimination. This pass will attempt
4213 to move stores out of loops. When used in conjunction with
4214 `-fgcse-lm', loops containing a load/store sequence can be changed
4215 to a load before the loop and a store after the loop.
4217 Not enabled at any optimization level.
4220 When `-fgcse-las' is enabled, the global common subexpression
4221 elimination pass eliminates redundant loads that come after stores
4222 to the same memory location (both partial and full redundancies).
4224 Not enabled at any optimization level.
4226 `-fgcse-after-reload'
4227 When `-fgcse-after-reload' is enabled, a redundant load elimination
4228 pass is performed after reload. The purpose of this pass is to
4229 cleanup redundant spilling.
4232 Perform loop optimizations: move constant expressions out of
4233 loops, simplify exit test conditions and optionally do
4234 strength-reduction as well.
4236 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4239 Perform loop optimizations using the new loop optimizer. The
4240 optimizations (loop unrolling, peeling and unswitching, loop
4241 invariant motion) are enabled by separate flags.
4244 Perform cross-jumping transformation. This transformation unifies
4245 equivalent code and save code size. The resulting code may or may
4246 not perform better than without cross-jumping.
4248 Enabled at levels `-O2', `-O3', `-Os'.
4251 Attempt to transform conditional jumps into branch-less
4252 equivalents. This include use of conditional moves, min, max, set
4253 flags and abs instructions, and some tricks doable by standard
4254 arithmetics. The use of conditional execution on chips where it
4255 is available is controlled by `if-conversion2'.
4257 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4260 Use conditional execution (where available) to transform
4261 conditional jumps into branch-less equivalents.
4263 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4265 `-fdelete-null-pointer-checks'
4266 Use global dataflow analysis to identify and eliminate useless
4267 checks for null pointers. The compiler assumes that dereferencing
4268 a null pointer would have halted the program. If a pointer is
4269 checked after it has already been dereferenced, it cannot be null.
4271 In some environments, this assumption is not true, and programs can
4272 safely dereference null pointers. Use
4273 `-fno-delete-null-pointer-checks' to disable this optimization for
4274 programs which depend on that behavior.
4276 Enabled at levels `-O2', `-O3', `-Os'.
4278 `-fexpensive-optimizations'
4279 Perform a number of minor optimizations that are relatively
4282 Enabled at levels `-O2', `-O3', `-Os'.
4284 `-foptimize-register-move'
4286 Attempt to reassign register numbers in move instructions and as
4287 operands of other simple instructions in order to maximize the
4288 amount of register tying. This is especially helpful on machines
4289 with two-operand instructions.
4291 Note `-fregmove' and `-foptimize-register-move' are the same
4294 Enabled at levels `-O2', `-O3', `-Os'.
4297 If supported for the target machine, attempt to reorder
4298 instructions to exploit instruction slots available after delayed
4299 branch instructions.
4301 Enabled at levels `-O', `-O2', `-O3', `-Os'.
4304 If supported for the target machine, attempt to reorder
4305 instructions to eliminate execution stalls due to required data
4306 being unavailable. This helps machines that have slow floating
4307 point or memory load instructions by allowing other instructions
4308 to be issued until the result of the load or floating point
4309 instruction is required.
4311 Enabled at levels `-O2', `-O3', `-Os'.
4314 Similar to `-fschedule-insns', but requests an additional pass of
4315 instruction scheduling after register allocation has been done.
4316 This is especially useful on machines with a relatively small
4317 number of registers and where memory load instructions take more
4320 Enabled at levels `-O2', `-O3', `-Os'.
4322 `-fno-sched-interblock'
4323 Don't schedule instructions across basic blocks. This is normally
4324 enabled by default when scheduling before register allocation, i.e.
4325 with `-fschedule-insns' or at `-O2' or higher.
4328 Don't allow speculative motion of non-load instructions. This is
4329 normally enabled by default when scheduling before register
4330 allocation, i.e. with `-fschedule-insns' or at `-O2' or higher.
4333 Allow speculative motion of some load instructions. This only
4334 makes sense when scheduling before register allocation, i.e. with
4335 `-fschedule-insns' or at `-O2' or higher.
4337 `-fsched-spec-load-dangerous'
4338 Allow speculative motion of more load instructions. This only
4339 makes sense when scheduling before register allocation, i.e. with
4340 `-fschedule-insns' or at `-O2' or higher.
4342 `-fsched-stalled-insns=N'
4343 Define how many insns (if any) can be moved prematurely from the
4344 queue of stalled insns into the ready list, during the second
4347 `-fsched-stalled-insns-dep=N'
4348 Define how many insn groups (cycles) will be examined for a
4349 dependency on a stalled insn that is candidate for premature
4350 removal from the queue of stalled insns. Has an effect only
4351 during the second scheduling pass, and only if
4352 `-fsched-stalled-insns' is used and its value is not zero.
4354 `-fsched2-use-superblocks'
4355 When scheduling after register allocation, do use superblock
4356 scheduling algorithm. Superblock scheduling allows motion across
4357 basic block boundaries resulting on faster schedules. This option
4358 is experimental, as not all machine descriptions used by GCC model
4359 the CPU closely enough to avoid unreliable results from the
4362 This only makes sense when scheduling after register allocation,
4363 i.e. with `-fschedule-insns2' or at `-O2' or higher.
4365 `-fsched2-use-traces'
4366 Use `-fsched2-use-superblocks' algorithm when scheduling after
4367 register allocation and additionally perform code duplication in
4368 order to increase the size of superblocks using tracer pass. See
4369 `-ftracer' for details on trace formation.
4371 This mode should produce faster but significantly longer programs.
4372 Also without `-fbranch-probabilities' the traces constructed may
4373 not match the reality and hurt the performance. This only makes
4374 sense when scheduling after register allocation, i.e. with
4375 `-fschedule-insns2' or at `-O2' or higher.
4377 `-freschedule-modulo-scheduled-loops'
4378 The modulo scheduling comes before the traditional scheduling, if
4379 a loop was modulo scheduled we may want to prevent the later
4380 scheduling passes from changing its schedule, we use this option
4384 Enable values to be allocated in registers that will be clobbered
4385 by function calls, by emitting extra instructions to save and
4386 restore the registers around such calls. Such allocation is done
4387 only when it seems to result in better code than would otherwise
4390 This option is always enabled by default on certain machines,
4391 usually those which have no call-preserved registers to use
4394 Enabled at levels `-O2', `-O3', `-Os'.
4397 Perform Partial Redundancy Elimination (PRE) on trees. This flag
4398 is enabled by default at `-O2' and `-O3'.
4401 Perform Full Redundancy Elimination (FRE) on trees. The difference
4402 between FRE and PRE is that FRE only considers expressions that
4403 are computed on all paths leading to the redundant computation.
4404 This analysis faster than PRE, though it exposes fewer
4405 redundancies. This flag is enabled by default at `-O' and higher.
4408 Perform sparse conditional constant propagation (CCP) on trees.
4409 This flag is enabled by default at `-O' and higher.
4412 Perform dead code elimination (DCE) on trees. This flag is
4413 enabled by default at `-O' and higher.
4415 `-ftree-dominator-opts'
4416 Perform dead code elimination (DCE) on trees. This flag is
4417 enabled by default at `-O' and higher.
4420 Perform loop header copying on trees. This is beneficial since it
4421 increases effectiveness of code motion optimizations. It also
4422 saves one jump. This flag is enabled by default at `-O' and
4423 higher. It is not enabled for `-Os', since it usually increases
4426 `-ftree-loop-optimize'
4427 Perform loop optimizations on trees. This flag is enabled by
4428 default at `-O' and higher.
4430 `-ftree-loop-linear'
4431 Perform linear loop transformations on tree. This flag can
4432 improve cache performance and allow further loop optimizations to
4436 Perform loop invariant motion on trees. This pass moves only
4437 invariants that would be hard to handle at RTL level (function
4438 calls, operations that expand to nontrivial sequences of insns).
4439 With `-funswitch-loops' it also moves operands of conditions that
4440 are invariant out of the loop, so that we can use just trivial
4441 invariantness analysis in loop unswitching. The pass also includes
4444 `-ftree-loop-ivcanon'
4445 Create a canonical counter for number of iterations in the loop
4446 for that determining number of iterations requires complicated
4447 analysis. Later optimizations then may determine the number
4448 easily. Useful especially in connection with unrolling.
4451 Perform induction variable optimizations (strength reduction,
4452 induction variable merging and induction variable elimination) on
4456 Perform scalar replacement of aggregates. This pass replaces
4457 structure references with scalars to prevent committing structures
4458 to memory too early. This flag is enabled by default at `-O' and
4462 Perform copy renaming on trees. This pass attempts to rename
4463 compiler temporaries to other variables at copy locations, usually
4464 resulting in variable names which more closely resemble the
4465 original variables. This flag is enabled by default at `-O' and
4469 Perform temporary expression replacement during the SSA->normal
4470 phase. Single use/single def temporaries are replaced at their
4471 use location with their defining expression. This results in
4472 non-GIMPLE code, but gives the expanders much more complex trees
4473 to work on resulting in better RTL generation. This is enabled by
4474 default at `-O' and higher.
4477 Perform live range splitting during the SSA->normal phase.
4478 Distinct live ranges of a variable are split into unique
4479 variables, allowing for better optimization later. This is
4480 enabled by default at `-O' and higher.
4483 Perform loop vectorization on trees.
4486 Perform tail duplication to enlarge superblock size. This
4487 transformation simplifies the control flow of the function
4488 allowing other optimizations to do better job.
4491 Unroll loops whose number of iterations can be determined at
4492 compile time or upon entry to the loop. `-funroll-loops' implies
4493 both `-fstrength-reduce' and `-frerun-cse-after-loop'. This
4494 option makes code larger, and may or may not make it run faster.
4496 `-funroll-all-loops'
4497 Unroll all loops, even if their number of iterations is uncertain
4498 when the loop is entered. This usually makes programs run more
4499 slowly. `-funroll-all-loops' implies the same options as
4502 `-fsplit-ivs-in-unroller'
4503 Enables expressing of values of induction variables in later
4504 iterations of the unrolled loop using the value in the first
4505 iteration. This breaks long dependency chains, thus improving
4506 efficiency of the scheduling passes (for best results, `-fweb'
4507 should be used as well).
4509 Combination of `-fweb' and CSE is often sufficient to obtain the
4510 same effect. However in cases the loop body is more complicated
4511 than a single basic block, this is not reliable. It also does not
4512 work at all on some of the architectures due to restrictions in
4515 This optimization is enabled by default.
4517 `-fvariable-expansion-in-unroller'
4518 With this option, the compiler will create multiple copies of some
4519 local variables when unrolling a loop which can result in superior
4522 `-fprefetch-loop-arrays'
4523 If supported by the target machine, generate instructions to
4524 prefetch memory to improve the performance of loops that access
4527 These options may generate better or worse code; results are highly
4528 dependent on the structure of loops within the source code.
4532 Disable any machine-specific peephole optimizations. The
4533 difference between `-fno-peephole' and `-fno-peephole2' is in how
4534 they are implemented in the compiler; some targets use one, some
4535 use the other, a few use both.
4537 `-fpeephole' is enabled by default. `-fpeephole2' enabled at
4538 levels `-O2', `-O3', `-Os'.
4540 `-fno-guess-branch-probability'
4541 Do not guess branch probabilities using heuristics.
4543 GCC will use heuristics to guess branch probabilities if they are
4544 not provided by profiling feedback (`-fprofile-arcs'). These
4545 heuristics are based on the control flow graph. If some branch
4546 probabilities are specified by `__builtin_expect', then the
4547 heuristics will be used to guess branch probabilities for the rest
4548 of the control flow graph, taking the `__builtin_expect' info into
4549 account. The interactions between the heuristics and
4550 `__builtin_expect' can be complex, and in some cases, it may be
4551 useful to disable the heuristics so that the effects of
4552 `__builtin_expect' are easier to understand.
4554 The default is `-fguess-branch-probability' at levels `-O', `-O2',
4558 Reorder basic blocks in the compiled function in order to reduce
4559 number of taken branches and improve code locality.
4561 Enabled at levels `-O2', `-O3'.
4563 `-freorder-blocks-and-partition'
4564 In addition to reordering basic blocks in the compiled function,
4565 in order to reduce number of taken branches, partitions hot and
4566 cold basic blocks into separate sections of the assembly and .o
4567 files, to improve paging and cache locality performance.
4569 This optimization is automatically turned off in the presence of
4570 exception handling, for linkonce sections, for functions with a
4571 user-defined section attribute and on any architecture that does
4572 not support named sections.
4574 `-freorder-functions'
4575 Reorder functions in the object file in order to improve code
4576 locality. This is implemented by using special subsections
4577 `.text.hot' for most frequently executed functions and
4578 `.text.unlikely' for unlikely executed functions. Reordering is
4579 done by the linker so object file format must support named
4580 sections and linker must place them in a reasonable way.
4582 Also profile feedback must be available in to make this option
4583 effective. See `-fprofile-arcs' for details.
4585 Enabled at levels `-O2', `-O3', `-Os'.
4588 Allows the compiler to assume the strictest aliasing rules
4589 applicable to the language being compiled. For C (and C++), this
4590 activates optimizations based on the type of expressions. In
4591 particular, an object of one type is assumed never to reside at
4592 the same address as an object of a different type, unless the
4593 types are almost the same. For example, an `unsigned int' can
4594 alias an `int', but not a `void*' or a `double'. A character type
4595 may alias any other type.
4597 Pay special attention to code like this:
4608 The practice of reading from a different union member than the one
4609 most recently written to (called "type-punning") is common. Even
4610 with `-fstrict-aliasing', type-punning is allowed, provided the
4611 memory is accessed through the union type. So, the code above
4612 will work as expected. However, this code might not:
4621 Every language that wishes to perform language-specific alias
4622 analysis should define a function that computes, given an `tree'
4623 node, an alias set for the node. Nodes in different alias sets
4624 are not allowed to alias. For an example, see the C front-end
4625 function `c_get_alias_set'.
4627 Enabled at levels `-O2', `-O3', `-Os'.
4630 `-falign-functions=N'
4631 Align the start of functions to the next power-of-two greater than
4632 N, skipping up to N bytes. For instance, `-falign-functions=32'
4633 aligns functions to the next 32-byte boundary, but
4634 `-falign-functions=24' would align to the next 32-byte boundary
4635 only if this can be done by skipping 23 bytes or less.
4637 `-fno-align-functions' and `-falign-functions=1' are equivalent
4638 and mean that functions will not be aligned.
4640 Some assemblers only support this flag when N is a power of two;
4641 in that case, it is rounded up.
4643 If N is not specified or is zero, use a machine-dependent default.
4645 Enabled at levels `-O2', `-O3'.
4649 Align all branch targets to a power-of-two boundary, skipping up to
4650 N bytes like `-falign-functions'. This option can easily make
4651 code slower, because it must insert dummy operations for when the
4652 branch target is reached in the usual flow of the code.
4654 `-fno-align-labels' and `-falign-labels=1' are equivalent and mean
4655 that labels will not be aligned.
4657 If `-falign-loops' or `-falign-jumps' are applicable and are
4658 greater than this value, then their values are used instead.
4660 If N is not specified or is zero, use a machine-dependent default
4661 which is very likely to be `1', meaning no alignment.
4663 Enabled at levels `-O2', `-O3'.
4667 Align loops to a power-of-two boundary, skipping up to N bytes
4668 like `-falign-functions'. The hope is that the loop will be
4669 executed many times, which will make up for any execution of the
4672 `-fno-align-loops' and `-falign-loops=1' are equivalent and mean
4673 that loops will not be aligned.
4675 If N is not specified or is zero, use a machine-dependent default.
4677 Enabled at levels `-O2', `-O3'.
4681 Align branch targets to a power-of-two boundary, for branch targets
4682 where the targets can only be reached by jumping, skipping up to N
4683 bytes like `-falign-functions'. In this case, no dummy operations
4686 `-fno-align-jumps' and `-falign-jumps=1' are equivalent and mean
4687 that loops will not be aligned.
4689 If N is not specified or is zero, use a machine-dependent default.
4691 Enabled at levels `-O2', `-O3'.
4694 Parse the whole compilation unit before starting to produce code.
4695 This allows some extra optimizations to take place but consumes
4696 more memory (in general). There are some compatibility issues
4697 with _unit-at-at-time_ mode:
4698 * enabling _unit-at-a-time_ mode may change the order in which
4699 functions, variables, and top-level `asm' statements are
4700 emitted, and will likely break code relying on some particular
4701 ordering. The majority of such top-level `asm' statements,
4702 though, can be replaced by `section' attributes.
4704 * _unit-at-a-time_ mode removes unreferenced static variables
4705 and functions are removed. This may result in undefined
4706 references when an `asm' statement refers directly to
4707 variables or functions that are otherwise unused. In that
4708 case either the variable/function shall be listed as an
4709 operand of the `asm' statement operand or, in the case of
4710 top-level `asm' statements the attribute `used' shall be used
4713 * Static functions now can use non-standard passing conventions
4714 that may break `asm' statements calling functions directly.
4715 Again, attribute `used' will prevent this behavior.
4717 As a temporary workaround, `-fno-unit-at-a-time' can be used, but
4718 this scheme may not be supported by future releases of GCC.
4720 Enabled at levels `-O2', `-O3'.
4723 Constructs webs as commonly used for register allocation purposes
4724 and assign each web individual pseudo register. This allows the
4725 register allocation pass to operate on pseudos directly, but also
4726 strengthens several other optimization passes, such as CSE, loop
4727 optimizer and trivial dead code remover. It can, however, make
4728 debugging impossible, since variables will no longer stay in a
4731 Enabled at levels `-O2', `-O3', `-Os', on targets where the
4732 default format for debugging information supports variable
4735 `-fno-cprop-registers'
4736 After register allocation and post-register allocation instruction
4737 splitting, we perform a copy-propagation pass to try to reduce
4738 scheduling dependencies and occasionally eliminate the copy.
4740 Disabled at levels `-O', `-O2', `-O3', `-Os'.
4742 `-fprofile-generate'
4743 Enable options usually used for instrumenting application to
4744 produce profile useful for later recompilation with profile
4745 feedback based optimization. You must use `-fprofile-generate'
4746 both when compiling and when linking your program.
4748 The following options are enabled: `-fprofile-arcs',
4749 `-fprofile-values', `-fvpt'.
4752 Enable profile feedback directed optimizations, and optimizations
4753 generally profitable only with profile feedback available.
4755 The following options are enabled: `-fbranch-probabilities',
4756 `-fvpt', `-funroll-loops', `-fpeel-loops', `-ftracer'.
4759 The following options control compiler behavior regarding floating
4760 point arithmetic. These options trade off between speed and
4761 correctness. All must be specifically enabled.
4764 Do not store floating point variables in registers, and inhibit
4765 other options that might change whether a floating point value is
4766 taken from a register or memory.
4768 This option prevents undesirable excess precision on machines such
4769 as the 68000 where the floating registers (of the 68881) keep more
4770 precision than a `double' is supposed to have. Similarly for the
4771 x86 architecture. For most programs, the excess precision does
4772 only good, but a few programs rely on the precise definition of
4773 IEEE floating point. Use `-ffloat-store' for such programs, after
4774 modifying them to store all pertinent intermediate computations
4778 Sets `-fno-math-errno', `-funsafe-math-optimizations',
4779 `-fno-trapping-math', `-ffinite-math-only', `-fno-rounding-math',
4780 `-fno-signaling-nans' and `fcx-limited-range'.
4782 This option causes the preprocessor macro `__FAST_MATH__' to be
4785 This option should never be turned on by any `-O' option since it
4786 can result in incorrect output for programs which depend on an
4787 exact implementation of IEEE or ISO rules/specifications for math
4791 Do not set ERRNO after calling math functions that are executed
4792 with a single instruction, e.g., sqrt. A program that relies on
4793 IEEE exceptions for math error handling may want to use this flag
4794 for speed while maintaining IEEE arithmetic compatibility.
4796 This option should never be turned on by any `-O' option since it
4797 can result in incorrect output for programs which depend on an
4798 exact implementation of IEEE or ISO rules/specifications for math
4801 The default is `-fmath-errno'.
4803 `-funsafe-math-optimizations'
4804 Allow optimizations for floating-point arithmetic that (a) assume
4805 that arguments and results are valid and (b) may violate IEEE or
4806 ANSI standards. When used at link-time, it may include libraries
4807 or startup files that change the default FPU control word or other
4808 similar optimizations.
4810 This option should never be turned on by any `-O' option since it
4811 can result in incorrect output for programs which depend on an
4812 exact implementation of IEEE or ISO rules/specifications for math
4815 The default is `-fno-unsafe-math-optimizations'.
4817 `-ffinite-math-only'
4818 Allow optimizations for floating-point arithmetic that assume that
4819 arguments and results are not NaNs or +-Infs.
4821 This option should never be turned on by any `-O' option since it
4822 can result in incorrect output for programs which depend on an
4823 exact implementation of IEEE or ISO rules/specifications.
4825 The default is `-fno-finite-math-only'.
4827 `-fno-trapping-math'
4828 Compile code assuming that floating-point operations cannot
4829 generate user-visible traps. These traps include division by
4830 zero, overflow, underflow, inexact result and invalid operation.
4831 This option implies `-fno-signaling-nans'. Setting this option
4832 may allow faster code if one relies on "non-stop" IEEE arithmetic,
4835 This option should never be turned on by any `-O' option since it
4836 can result in incorrect output for programs which depend on an
4837 exact implementation of IEEE or ISO rules/specifications for math
4840 The default is `-ftrapping-math'.
4843 Disable transformations and optimizations that assume default
4844 floating point rounding behavior. This is round-to-zero for all
4845 floating point to integer conversions, and round-to-nearest for
4846 all other arithmetic truncations. This option should be specified
4847 for programs that change the FP rounding mode dynamically, or that
4848 may be executed with a non-default rounding mode. This option
4849 disables constant folding of floating point expressions at
4850 compile-time (which may be affected by rounding mode) and
4851 arithmetic transformations that are unsafe in the presence of
4852 sign-dependent rounding modes.
4854 The default is `-fno-rounding-math'.
4856 This option is experimental and does not currently guarantee to
4857 disable all GCC optimizations that are affected by rounding mode.
4858 Future versions of GCC may provide finer control of this setting
4859 using C99's `FENV_ACCESS' pragma. This command line option will
4860 be used to specify the default state for `FENV_ACCESS'.
4863 Compile code assuming that IEEE signaling NaNs may generate
4864 user-visible traps during floating-point operations. Setting this
4865 option disables optimizations that may change the number of
4866 exceptions visible with signaling NaNs. This option implies
4869 This option causes the preprocessor macro `__SUPPORT_SNAN__' to be
4872 The default is `-fno-signaling-nans'.
4874 This option is experimental and does not currently guarantee to
4875 disable all GCC optimizations that affect signaling NaN behavior.
4877 `-fsingle-precision-constant'
4878 Treat floating point constant as single precision constant instead
4879 of implicitly converting it to double precision constant.
4881 `-fcx-limited-range'
4882 `-fno-cx-limited-range'
4883 When enabled, this option states that a range reduction step is not
4884 needed when performing complex division. The default is
4885 `-fno-cx-limited-range', but is enabled by `-ffast-math'.
4887 This option controls the default setting of the ISO C99
4888 `CX_LIMITED_RANGE' pragma. Nevertheless, the option applies to
4892 The following options control optimizations that may improve
4893 performance, but are not enabled by any `-O' options. This section
4894 includes experimental options that may produce broken code.
4896 `-fbranch-probabilities'
4897 After running a program compiled with `-fprofile-arcs' (*note
4898 Options for Debugging Your Program or `gcc': Debugging Options.),
4899 you can compile it a second time using `-fbranch-probabilities',
4900 to improve optimizations based on the number of times each branch
4901 was taken. When the program compiled with `-fprofile-arcs' exits
4902 it saves arc execution counts to a file called `SOURCENAME.gcda'
4903 for each source file The information in this data file is very
4904 dependent on the structure of the generated code, so you must use
4905 the same source code and the same optimization options for both
4908 With `-fbranch-probabilities', GCC puts a `REG_BR_PROB' note on
4909 each `JUMP_INSN' and `CALL_INSN'. These can be used to improve
4910 optimization. Currently, they are only used in one place: in
4911 `reorg.c', instead of guessing which path a branch is mostly to
4912 take, the `REG_BR_PROB' values are used to exactly determine which
4913 path is taken more often.
4916 If combined with `-fprofile-arcs', it adds code so that some data
4917 about values of expressions in the program is gathered.
4919 With `-fbranch-probabilities', it reads back the data gathered
4920 from profiling values of expressions and adds `REG_VALUE_PROFILE'
4921 notes to instructions for their later usage in optimizations.
4923 Enabled with `-fprofile-generate' and `-fprofile-use'.
4926 If combined with `-fprofile-arcs', it instructs the compiler to add
4927 a code to gather information about values of expressions.
4929 With `-fbranch-probabilities', it reads back the data gathered and
4930 actually performs the optimizations based on them. Currently the
4931 optimizations include specialization of division operation using
4932 the knowledge about the value of the denominator.
4934 `-fspeculative-prefetching'
4935 If combined with `-fprofile-arcs', it instructs the compiler to add
4936 a code to gather information about addresses of memory references
4939 With `-fbranch-probabilities', it reads back the data gathered and
4940 issues prefetch instructions according to them. In addition to
4941 the opportunities noticed by `-fprefetch-loop-arrays', it also
4942 notices more complicated memory access patterns--for example
4943 accesses to the data stored in linked list whose elements are
4944 usually allocated sequentially.
4946 In order to prevent issuing double prefetches, usage of
4947 `-fspeculative-prefetching' implies `-fno-prefetch-loop-arrays'.
4949 Enabled with `-fprofile-generate' and `-fprofile-use'.
4951 `-frename-registers'
4952 Attempt to avoid false dependencies in scheduled code by making use
4953 of registers left over after register allocation. This
4954 optimization will most benefit processors with lots of registers.
4955 Depending on the debug information format adopted by the target,
4956 however, it can make debugging impossible, since variables will no
4957 longer stay in a "home register".
4959 Not enabled by default at any level because it has known bugs.
4962 Perform tail duplication to enlarge superblock size. This
4963 transformation simplifies the control flow of the function
4964 allowing other optimizations to do better job.
4966 Enabled with `-fprofile-use'.
4969 Unroll loops whose number of iterations can be determined at
4970 compile time or upon entry to the loop. `-funroll-loops' implies
4971 `-frerun-cse-after-loop'. It also turns on complete loop peeling
4972 (i.e. complete removal of loops with small constant number of
4973 iterations). This option makes code larger, and may or may not
4976 Enabled with `-fprofile-use'.
4978 `-funroll-all-loops'
4979 Unroll all loops, even if their number of iterations is uncertain
4980 when the loop is entered. This usually makes programs run more
4981 slowly. `-funroll-all-loops' implies the same options as
4985 Peels the loops for that there is enough information that they do
4986 not roll much (from profile feedback). It also turns on complete
4987 loop peeling (i.e. complete removal of loops with small constant
4988 number of iterations).
4990 Enabled with `-fprofile-use'.
4992 `-fmove-loop-invariants'
4993 Enables the loop invariant motion pass in the new loop optimizer.
4994 Enabled at level `-O1'
4997 Move branches with loop invariant conditions out of the loop, with
4998 duplicates of the loop on both branches (modified according to
4999 result of the condition).
5001 `-fprefetch-loop-arrays'
5002 If supported by the target machine, generate instructions to
5003 prefetch memory to improve the performance of loops that access
5006 Disabled at level `-Os'.
5008 `-ffunction-sections'
5010 Place each function or data item into its own section in the output
5011 file if the target supports arbitrary sections. The name of the
5012 function or the name of the data item determines the section's name
5015 Use these options on systems where the linker can perform
5016 optimizations to improve locality of reference in the instruction
5017 space. Most systems using the ELF object format and SPARC
5018 processors running Solaris 2 have linkers with such optimizations.
5019 AIX may have these optimizations in the future.
5021 Only use these options when there are significant benefits from
5022 doing so. When you specify these options, the assembler and
5023 linker will create larger object and executable files and will
5024 also be slower. You will not be able to use `gprof' on all
5025 systems if you specify this option and you may have problems with
5026 debugging if you specify both this option and `-g'.
5028 `-fbranch-target-load-optimize'
5029 Perform branch target register load optimization before prologue /
5030 epilogue threading. The use of target registers can typically be
5031 exposed only during reload, thus hoisting loads out of loops and
5032 doing inter-block scheduling needs a separate optimization pass.
5034 `-fbranch-target-load-optimize2'
5035 Perform branch target register load optimization after prologue /
5038 `-fbtr-bb-exclusive'
5039 When performing branch target register load optimization, don't
5040 reuse branch target registers in within any basic block.
5042 `--param NAME=VALUE'
5043 In some places, GCC uses various constants to control the amount of
5044 optimization that is done. For example, GCC will not inline
5045 functions that contain more that a certain number of instructions.
5046 You can control some of these constants on the command-line using
5047 the `--param' option.
5049 The names of specific parameters, and the meaning of the values,
5050 are tied to the internals of the compiler, and are subject to
5051 change without notice in future releases.
5053 In each case, the VALUE is an integer. The allowable choices for
5054 NAME are given in the following table:
5056 `sra-max-structure-size'
5057 The maximum structure size, in bytes, at which the scalar
5058 replacement of aggregates (SRA) optimization will perform
5059 block copies. The default value, 0, implies that GCC will
5060 select the most appropriate size itself.
5062 `sra-field-structure-ratio'
5063 The threshold ratio (as a percentage) between instantiated
5064 fields and the complete structure size. We say that if the
5065 ratio of the number of bytes in instantiated fields to the
5066 number of bytes in the complete structure exceeds this
5067 parameter, then block copies are not used. The default is 75.
5069 `max-crossjump-edges'
5070 The maximum number of incoming edges to consider for
5071 crossjumping. The algorithm used by `-fcrossjumping' is
5072 O(N^2) in the number of edges incoming to each block.
5073 Increasing values mean more aggressive optimization, making
5074 the compile time increase with probably small improvement in
5077 `min-crossjump-insns'
5078 The minimum number of instructions which must be matched at
5079 the end of two blocks before crossjumping will be performed
5080 on them. This value is ignored in the case where all
5081 instructions in the block being crossjumped from are matched.
5082 The default value is 5.
5084 `max-goto-duplication-insns'
5085 The maximum number of instructions to duplicate to a block
5086 that jumps to a computed goto. To avoid O(N^2) behavior in a
5087 number of passes, GCC factors computed gotos early in the
5088 compilation process, and unfactors them as late as possible.
5089 Only computed jumps at the end of a basic blocks with no more
5090 than max-goto-duplication-insns are unfactored. The default
5093 `max-delay-slot-insn-search'
5094 The maximum number of instructions to consider when looking
5095 for an instruction to fill a delay slot. If more than this
5096 arbitrary number of instructions is searched, the time
5097 savings from filling the delay slot will be minimal so stop
5098 searching. Increasing values mean more aggressive
5099 optimization, making the compile time increase with probably
5100 small improvement in executable run time.
5102 `max-delay-slot-live-search'
5103 When trying to fill delay slots, the maximum number of
5104 instructions to consider when searching for a block with
5105 valid live register information. Increasing this arbitrarily
5106 chosen value means more aggressive optimization, increasing
5107 the compile time. This parameter should be removed when the
5108 delay slot code is rewritten to maintain the control-flow
5112 The approximate maximum amount of memory that will be
5113 allocated in order to perform the global common subexpression
5114 elimination optimization. If more memory than specified is
5115 required, the optimization will not be done.
5118 The maximum number of passes of GCSE to run. The default is
5121 `max-pending-list-length'
5122 The maximum number of pending dependencies scheduling will
5123 allow before flushing the current state and starting over.
5124 Large functions with few branches or calls can create
5125 excessively large lists which needlessly consume memory and
5128 `max-inline-insns-single'
5129 Several parameters control the tree inliner used in gcc.
5130 This number sets the maximum number of instructions (counted
5131 in GCC's internal representation) in a single function that
5132 the tree inliner will consider for inlining. This only
5133 affects functions declared inline and methods implemented in
5134 a class declaration (C++). The default value is 450.
5136 `max-inline-insns-auto'
5137 When you use `-finline-functions' (included in `-O3'), a lot
5138 of functions that would otherwise not be considered for
5139 inlining by the compiler will be investigated. To those
5140 functions, a different (more restrictive) limit compared to
5141 functions declared inline can be applied. The default value
5144 `large-function-insns'
5145 The limit specifying really large functions. For functions
5146 larger than this limit after inlining inlining is constrained
5147 by `--param large-function-growth'. This parameter is useful
5148 primarily to avoid extreme compilation time caused by
5149 non-linear algorithms used by the backend. This parameter is
5150 ignored when `-funit-at-a-time' is not used. The default
5153 `large-function-growth'
5154 Specifies maximal growth of large function caused by inlining
5155 in percents. This parameter is ignored when
5156 `-funit-at-a-time' is not used. The default value is 100
5157 which limits large function growth to 2.0 times the original
5160 `inline-unit-growth'
5161 Specifies maximal overall growth of the compilation unit
5162 caused by inlining. This parameter is ignored when
5163 `-funit-at-a-time' is not used. The default value is 50
5164 which limits unit growth to 1.5 times the original size.
5166 `max-inline-insns-recursive'
5167 `max-inline-insns-recursive-auto'
5168 Specifies maximum number of instructions out-of-line copy of
5169 self recursive inline function can grow into by performing
5172 For functions declared inline `--param
5173 max-inline-insns-recursive' is taken into acount. For
5174 function not declared inline, recursive inlining happens only
5175 when `-finline-functions' (included in `-O3') is enabled and
5176 `--param max-inline-insns-recursive-auto' is used. The
5177 default value is 450.
5179 `max-inline-recursive-depth'
5180 `max-inline-recursive-depth-auto'
5181 Specifies maximum recursion depth used by the recursive
5184 For functions declared inline `--param
5185 max-inline-recursive-depth' is taken into acount. For
5186 function not declared inline, recursive inlining happens only
5187 when `-finline-functions' (included in `-O3') is enabled and
5188 `--param max-inline-recursive-depth-auto' is used. The
5189 default value is 450.
5192 Specify cost of call instruction relative to simple
5193 arithmetics operations (having cost of 1). Increasing this
5194 cost disqualify inlinining of non-leaf functions and at same
5195 time increase size of leaf function that is believed to
5196 reduce function size by being inlined. In effect it increase
5197 amount of inlining for code having large abstraction penalty
5198 (many functions that just pass the argumetns to other
5199 functions) and decrease inlining for code with low
5200 abstraction penalty. Default value is 16.
5202 `max-unrolled-insns'
5203 The maximum number of instructions that a loop should have if
5204 that loop is unrolled, and if the loop is unrolled, it
5205 determines how many times the loop code is unrolled.
5207 `max-average-unrolled-insns'
5208 The maximum number of instructions biased by probabilities of
5209 their execution that a loop should have if that loop is
5210 unrolled, and if the loop is unrolled, it determines how many
5211 times the loop code is unrolled.
5214 The maximum number of unrollings of a single loop.
5217 The maximum number of instructions that a loop should have if
5218 that loop is peeled, and if the loop is peeled, it determines
5219 how many times the loop code is peeled.
5222 The maximum number of peelings of a single loop.
5224 `max-completely-peeled-insns'
5225 The maximum number of insns of a completely peeled loop.
5227 `max-completely-peel-times'
5228 The maximum number of iterations of a loop to be suitable for
5231 `max-unswitch-insns'
5232 The maximum number of insns of an unswitched loop.
5234 `max-unswitch-level'
5235 The maximum number of branches unswitched in a single loop.
5238 The minimum cost of an expensive expression in the loop
5241 `iv-consider-all-candidates-bound'
5242 Bound on number of candidates for induction variables below
5243 that all candidates are considered for each use in induction
5244 variable optimizations. Only the most relevant candidates
5245 are considered if there are more candidates, to avoid
5246 quadratic time complexity.
5248 `iv-max-considered-uses'
5249 The induction variable optimizations give up on loops that
5250 contain more induction variable uses.
5252 `iv-always-prune-cand-set-bound'
5253 If number of candidates in the set is smaller than this value,
5254 we always try to remove unnecessary ivs from the set during
5255 its optimization when a new iv is added to the set.
5257 `scev-max-expr-size'
5258 Bound on size of expressions used in the scalar evolutions
5259 analyzer. Large expressions slow the analyzer.
5261 `max-iterations-to-track'
5262 The maximum number of iterations of a loop the brute force
5263 algorithm for analysis of # of iterations of the loop tries
5266 `hot-bb-count-fraction'
5267 Select fraction of the maximal count of repetitions of basic
5268 block in program given basic block needs to have to be
5271 `hot-bb-frequency-fraction'
5272 Select fraction of the maximal frequency of executions of
5273 basic block in function given basic block needs to have to be
5276 `tracer-dynamic-coverage'
5277 `tracer-dynamic-coverage-feedback'
5278 This value is used to limit superblock formation once the
5279 given percentage of executed instructions is covered. This
5280 limits unnecessary code size expansion.
5282 The `tracer-dynamic-coverage-feedback' is used only when
5283 profile feedback is available. The real profiles (as opposed
5284 to statically estimated ones) are much less balanced allowing
5285 the threshold to be larger value.
5287 `tracer-max-code-growth'
5288 Stop tail duplication once code growth has reached given
5289 percentage. This is rather hokey argument, as most of the
5290 duplicates will be eliminated later in cross jumping, so it
5291 may be set to much higher values than is the desired code
5294 `tracer-min-branch-ratio'
5295 Stop reverse growth when the reverse probability of best edge
5296 is less than this threshold (in percent).
5298 `tracer-min-branch-ratio'
5299 `tracer-min-branch-ratio-feedback'
5300 Stop forward growth if the best edge do have probability
5301 lower than this threshold.
5303 Similarly to `tracer-dynamic-coverage' two values are
5304 present, one for compilation for profile feedback and one for
5305 compilation without. The value for compilation with profile
5306 feedback needs to be more conservative (higher) in order to
5307 make tracer effective.
5309 `max-cse-path-length'
5310 Maximum number of basic blocks on path that cse considers.
5313 `global-var-threshold'
5314 Counts the number of function calls (N) and the number of
5315 call-clobbered variables (V). If NxV is larger than this
5316 limit, a single artificial variable will be created to
5317 represent all the call-clobbered variables at function call
5318 sites. This artificial variable will then be made to alias
5319 every call-clobbered variable. (done as `int * size_t' on
5320 the host machine; beware overflow).
5323 Maximum number of virtual operands allowed to represent
5324 aliases before triggering the alias grouping heuristic.
5325 Alias grouping reduces compile times and memory consumption
5326 needed for aliasing at the expense of precision loss in alias
5330 GCC uses a garbage collector to manage its own memory
5331 allocation. This parameter specifies the minimum percentage
5332 by which the garbage collector's heap should be allowed to
5333 expand between collections. Tuning this may improve
5334 compilation speed; it has no effect on code generation.
5336 The default is 30% + 70% * (RAM/1GB) with an upper bound of
5337 100% when RAM >= 1GB. If `getrlimit' is available, the
5338 notion of "RAM" is the smallest of actual RAM and
5339 `RLIMIT_DATA' or `RLIMIT_AS'. If GCC is not able to
5340 calculate RAM on a particular platform, the lower bound of
5341 30% is used. Setting this parameter and `ggc-min-heapsize'
5342 to zero causes a full collection to occur at every
5343 opportunity. This is extremely slow, but can be useful for
5347 Minimum size of the garbage collector's heap before it begins
5348 bothering to collect garbage. The first collection occurs
5349 after the heap expands by `ggc-min-expand'% beyond
5350 `ggc-min-heapsize'. Again, tuning this may improve
5351 compilation speed, and has no effect on code generation.
5353 The default is the smaller of RAM/8, RLIMIT_RSS, or a limit
5354 which tries to ensure that RLIMIT_DATA or RLIMIT_AS are not
5355 exceeded, but with a lower bound of 4096 (four megabytes) and
5356 an upper bound of 131072 (128 megabytes). If GCC is not able
5357 to calculate RAM on a particular platform, the lower bound is
5358 used. Setting this parameter very large effectively disables
5359 garbage collection. Setting this parameter and
5360 `ggc-min-expand' to zero causes a full collection to occur at
5363 `max-reload-search-insns'
5364 The maximum number of instruction reload should look backward
5365 for equivalent register. Increasing values mean more
5366 aggressive optimization, making the compile time increase
5367 with probably slightly better performance. The default value
5370 `max-cselib-memory-location'
5371 The maximum number of memory locations cselib should take
5372 into acount. Increasing values mean more aggressive
5373 optimization, making the compile time increase with probably
5374 slightly better performance. The default value is 500.
5376 `reorder-blocks-duplicate'
5377 `reorder-blocks-duplicate-feedback'
5378 Used by basic block reordering pass to decide whether to use
5379 unconditional branch or duplicate the code on its
5380 destination. Code is duplicated when its estimated size is
5381 smaller than this value multiplied by the estimated size of
5382 unconditional jump in the hot spots of the program.
5384 The `reorder-block-duplicate-feedback' is used only when
5385 profile feedback is available and may be set to higher values
5386 than `reorder-block-duplicate' since information about the
5387 hot spots is more accurate.
5389 `max-sched-region-blocks'
5390 The maximum number of blocks in a region to be considered for
5391 interblock scheduling. The default value is 10.
5393 `max-sched-region-insns'
5394 The maximum number of insns in a region to be considered for
5395 interblock scheduling. The default value is 100.
5397 `max-last-value-rtl'
5398 The maximum size measured as number of RTLs that can be
5399 recorded in an expression in combiner for a pseudo register
5400 as last known value of that register. The default is 10000.
5402 `integer-share-limit'
5403 Small integer constants can use a shared data structure,
5404 reducing the compiler's memory usage and increasing its
5405 speed. This sets the maximum value of a shared integer
5406 constant's. The default value is 256.
5410 File: gcc.info, Node: Preprocessor Options, Next: Assembler Options, Prev: Optimize Options, Up: Invoking GCC
5412 3.11 Options Controlling the Preprocessor
5413 =========================================
5415 These options control the C preprocessor, which is run on each C source
5416 file before actual compilation.
5418 If you use the `-E' option, nothing is done except preprocessing.
5419 Some of these options make sense only together with `-E' because they
5420 cause the preprocessor output to be unsuitable for actual compilation.
5422 You can use `-Wp,OPTION' to bypass the compiler driver and pass
5423 OPTION directly through to the preprocessor. If OPTION contains
5424 commas, it is split into multiple options at the commas. However,
5425 many options are modified, translated or interpreted by the
5426 compiler driver before being passed to the preprocessor, and `-Wp'
5427 forcibly bypasses this phase. The preprocessor's direct interface
5428 is undocumented and subject to change, so whenever possible you
5429 should avoid using `-Wp' and let the driver handle the options
5432 `-Xpreprocessor OPTION'
5433 Pass OPTION as an option to the preprocessor. You can use this to
5434 supply system-specific preprocessor options which GCC does not
5435 know how to recognize.
5437 If you want to pass an option that takes an argument, you must use
5438 `-Xpreprocessor' twice, once for the option and once for the
5442 Predefine NAME as a macro, with definition `1'.
5444 `-D NAME=DEFINITION'
5445 The contents of DEFINITION are tokenized and processed as if they
5446 appeared during translation phase three in a `#define' directive.
5447 In particular, the definition will be truncated by embedded
5450 If you are invoking the preprocessor from a shell or shell-like
5451 program you may need to use the shell's quoting syntax to protect
5452 characters such as spaces that have a meaning in the shell syntax.
5454 If you wish to define a function-like macro on the command line,
5455 write its argument list with surrounding parentheses before the
5456 equals sign (if any). Parentheses are meaningful to most shells,
5457 so you will need to quote the option. With `sh' and `csh',
5458 `-D'NAME(ARGS...)=DEFINITION'' works.
5460 `-D' and `-U' options are processed in the order they are given on
5461 the command line. All `-imacros FILE' and `-include FILE' options
5462 are processed after all `-D' and `-U' options.
5465 Cancel any previous definition of NAME, either built in or
5466 provided with a `-D' option.
5469 Do not predefine any system-specific or GCC-specific macros. The
5470 standard predefined macros remain defined.
5473 Add the directory DIR to the list of directories to be searched
5474 for header files. Directories named by `-I' are searched before
5475 the standard system include directories. If the directory DIR is
5476 a standard system include directory, the option is ignored to
5477 ensure that the default search order for system directories and
5478 the special treatment of system headers are not defeated .
5481 Write output to FILE. This is the same as specifying FILE as the
5482 second non-option argument to `cpp'. `gcc' has a different
5483 interpretation of a second non-option argument, so you must use
5484 `-o' to specify the output file.
5487 Turns on all optional warnings which are desirable for normal code.
5488 At present this is `-Wcomment', `-Wtrigraphs', `-Wmultichar' and a
5489 warning about integer promotion causing a change of sign in `#if'
5490 expressions. Note that many of the preprocessor's warnings are on
5491 by default and have no options to control them.
5495 Warn whenever a comment-start sequence `/*' appears in a `/*'
5496 comment, or whenever a backslash-newline appears in a `//' comment.
5497 (Both forms have the same effect.)
5500 Most trigraphs in comments cannot affect the meaning of the
5501 program. However, a trigraph that would form an escaped newline
5502 (`??/' at the end of a line) can, by changing where the comment
5503 begins or ends. Therefore, only trigraphs that would form escaped
5504 newlines produce warnings inside a comment.
5506 This option is implied by `-Wall'. If `-Wall' is not given, this
5507 option is still enabled unless trigraphs are enabled. To get
5508 trigraph conversion without warnings, but get the other `-Wall'
5509 warnings, use `-trigraphs -Wall -Wno-trigraphs'.
5512 Warn about certain constructs that behave differently in
5513 traditional and ISO C. Also warn about ISO C constructs that have
5514 no traditional C equivalent, and problematic constructs which
5518 Warn the first time `#import' is used.
5521 Warn whenever an identifier which is not a macro is encountered in
5522 an `#if' directive, outside of `defined'. Such identifiers are
5526 Warn about macros defined in the main file that are unused. A
5527 macro is "used" if it is expanded or tested for existence at least
5528 once. The preprocessor will also warn if the macro has not been
5529 used at the time it is redefined or undefined.
5531 Built-in macros, macros defined on the command line, and macros
5532 defined in include files are not warned about.
5534 _Note:_ If a macro is actually used, but only used in skipped
5535 conditional blocks, then CPP will report it as unused. To avoid
5536 the warning in such a case, you might improve the scope of the
5537 macro's definition by, for example, moving it into the first
5538 skipped block. Alternatively, you could provide a dummy use with
5541 #if defined the_macro_causing_the_warning
5545 Warn whenever an `#else' or an `#endif' are followed by text.
5546 This usually happens in code of the form
5554 The second and third `FOO' should be in comments, but often are not
5555 in older programs. This warning is on by default.
5558 Make all warnings into hard errors. Source code which triggers
5559 warnings will be rejected.
5562 Issue warnings for code in system headers. These are normally
5563 unhelpful in finding bugs in your own code, therefore suppressed.
5564 If you are responsible for the system library, you may want to see
5568 Suppress all warnings, including those which GNU CPP issues by
5572 Issue all the mandatory diagnostics listed in the C standard.
5573 Some of them are left out by default, since they trigger
5574 frequently on harmless code.
5577 Issue all the mandatory diagnostics, and make all mandatory
5578 diagnostics into errors. This includes mandatory diagnostics that
5579 GCC issues without `-pedantic' but treats as warnings.
5582 Instead of outputting the result of preprocessing, output a rule
5583 suitable for `make' describing the dependencies of the main source
5584 file. The preprocessor outputs one `make' rule containing the
5585 object file name for that source file, a colon, and the names of
5586 all the included files, including those coming from `-include' or
5587 `-imacros' command line options.
5589 Unless specified explicitly (with `-MT' or `-MQ'), the object file
5590 name consists of the basename of the source file with any suffix
5591 replaced with object file suffix. If there are many included
5592 files then the rule is split into several lines using `\'-newline.
5593 The rule has no commands.
5595 This option does not suppress the preprocessor's debug output,
5596 such as `-dM'. To avoid mixing such debug output with the
5597 dependency rules you should explicitly specify the dependency
5598 output file with `-MF', or use an environment variable like
5599 `DEPENDENCIES_OUTPUT' (*note Environment Variables::). Debug
5600 output will still be sent to the regular output stream as normal.
5602 Passing `-M' to the driver implies `-E', and suppresses warnings
5603 with an implicit `-w'.
5606 Like `-M' but do not mention header files that are found in system
5607 header directories, nor header files that are included, directly
5608 or indirectly, from such a header.
5610 This implies that the choice of angle brackets or double quotes in
5611 an `#include' directive does not in itself determine whether that
5612 header will appear in `-MM' dependency output. This is a slight
5613 change in semantics from GCC versions 3.0 and earlier.
5616 When used with `-M' or `-MM', specifies a file to write the
5617 dependencies to. If no `-MF' switch is given the preprocessor
5618 sends the rules to the same place it would have sent preprocessed
5621 When used with the driver options `-MD' or `-MMD', `-MF' overrides
5622 the default dependency output file.
5625 In conjunction with an option such as `-M' requesting dependency
5626 generation, `-MG' assumes missing header files are generated files
5627 and adds them to the dependency list without raising an error.
5628 The dependency filename is taken directly from the `#include'
5629 directive without prepending any path. `-MG' also suppresses
5630 preprocessed output, as a missing header file renders this useless.
5632 This feature is used in automatic updating of makefiles.
5635 This option instructs CPP to add a phony target for each dependency
5636 other than the main file, causing each to depend on nothing. These
5637 dummy rules work around errors `make' gives if you remove header
5638 files without updating the `Makefile' to match.
5640 This is typical output:
5642 test.o: test.c test.h
5647 Change the target of the rule emitted by dependency generation. By
5648 default CPP takes the name of the main input file, including any
5649 path, deletes any file suffix such as `.c', and appends the
5650 platform's usual object suffix. The result is the target.
5652 An `-MT' option will set the target to be exactly the string you
5653 specify. If you want multiple targets, you can specify them as a
5654 single argument to `-MT', or use multiple `-MT' options.
5656 For example, `-MT '$(objpfx)foo.o'' might give
5658 $(objpfx)foo.o: foo.c
5661 Same as `-MT', but it quotes any characters which are special to
5662 Make. `-MQ '$(objpfx)foo.o'' gives
5664 $$(objpfx)foo.o: foo.c
5666 The default target is automatically quoted, as if it were given
5670 `-MD' is equivalent to `-M -MF FILE', except that `-E' is not
5671 implied. The driver determines FILE based on whether an `-o'
5672 option is given. If it is, the driver uses its argument but with
5673 a suffix of `.d', otherwise it take the basename of the input file
5674 and applies a `.d' suffix.
5676 If `-MD' is used in conjunction with `-E', any `-o' switch is
5677 understood to specify the dependency output file (but *note -MF:
5678 dashMF.), but if used without `-E', each `-o' is understood to
5679 specify a target object file.
5681 Since `-E' is not implied, `-MD' can be used to generate a
5682 dependency output file as a side-effect of the compilation process.
5685 Like `-MD' except mention only user header files, not system
5689 When using precompiled headers (*note Precompiled Headers::), this
5690 flag will cause the dependency-output flags to also list the files
5691 from the precompiled header's dependencies. If not specified only
5692 the precompiled header would be listed and not the files that were
5693 used to create it because those files are not consulted when a
5694 precompiled header is used.
5697 This option allows use of a precompiled header (*note Precompiled
5698 Headers::) together with `-E'. It inserts a special `#pragma',
5699 `#pragma GCC pch_preprocess "<filename>"' in the output to mark
5700 the place where the precompiled header was found, and its
5701 filename. When `-fpreprocessed' is in use, GCC recognizes this
5702 `#pragma' and loads the PCH.
5704 This option is off by default, because the resulting preprocessed
5705 output is only really suitable as input to GCC. It is switched on
5708 You should not write this `#pragma' in your own code, but it is
5709 safe to edit the filename if the PCH file is available in a
5710 different location. The filename may be absolute or it may be
5711 relative to GCC's current directory.
5716 `-x assembler-with-cpp'
5717 Specify the source language: C, C++, Objective-C, or assembly.
5718 This has nothing to do with standards conformance or extensions;
5719 it merely selects which base syntax to expect. If you give none
5720 of these options, cpp will deduce the language from the extension
5721 of the source file: `.c', `.cc', `.m', or `.S'. Some other common
5722 extensions for C++ and assembly are also recognized. If cpp does
5723 not recognize the extension, it will treat the file as C; this is
5724 the most generic mode.
5726 _Note:_ Previous versions of cpp accepted a `-lang' option which
5727 selected both the language and the standards conformance level.
5728 This option has been removed, because it conflicts with the `-l'
5733 Specify the standard to which the code should conform. Currently
5734 CPP knows about C and C++ standards; others may be added in the
5737 STANDARD may be one of:
5740 The ISO C standard from 1990. `c89' is the customary
5741 shorthand for this version of the standard.
5743 The `-ansi' option is equivalent to `-std=c89'.
5746 The 1990 C standard, as amended in 1994.
5752 The revised ISO C standard, published in December 1999.
5753 Before publication, this was known as C9X.
5756 The 1990 C standard plus GNU extensions. This is the default.
5760 The 1999 C standard plus GNU extensions.
5763 The 1998 ISO C++ standard plus amendments.
5766 The same as `-std=c++98' plus GNU extensions. This is the
5767 default for C++ code.
5770 Split the include path. Any directories specified with `-I'
5771 options before `-I-' are searched only for headers requested with
5772 `#include "FILE"'; they are not searched for `#include <FILE>'.
5773 If additional directories are specified with `-I' options after
5774 the `-I-', those directories are searched for all `#include'
5777 In addition, `-I-' inhibits the use of the directory of the current
5778 file directory as the first search directory for `#include "FILE"'.
5779 This option has been deprecated.
5782 Do not search the standard system directories for header files.
5783 Only the directories you have specified with `-I' options (and the
5784 directory of the current file, if appropriate) are searched.
5787 Do not search for header files in the C++-specific standard
5788 directories, but do still search the other standard directories.
5789 (This option is used when building the C++ library.)
5792 Process FILE as if `#include "file"' appeared as the first line of
5793 the primary source file. However, the first directory searched
5794 for FILE is the preprocessor's working directory _instead of_ the
5795 directory containing the main source file. If not found there, it
5796 is searched for in the remainder of the `#include "..."' search
5799 If multiple `-include' options are given, the files are included
5800 in the order they appear on the command line.
5803 Exactly like `-include', except that any output produced by
5804 scanning FILE is thrown away. Macros it defines remain defined.
5805 This allows you to acquire all the macros from a header without
5806 also processing its declarations.
5808 All files specified by `-imacros' are processed before all files
5809 specified by `-include'.
5812 Search DIR for header files, but do it _after_ all directories
5813 specified with `-I' and the standard system directories have been
5814 exhausted. DIR is treated as a system include directory.
5817 Specify PREFIX as the prefix for subsequent `-iwithprefix'
5818 options. If the prefix represents a directory, you should include
5822 `-iwithprefixbefore DIR'
5823 Append DIR to the prefix specified previously with `-iprefix', and
5824 add the resulting directory to the include search path.
5825 `-iwithprefixbefore' puts it in the same place `-I' would;
5826 `-iwithprefix' puts it where `-idirafter' would.
5829 Search DIR for header files, after all directories specified by
5830 `-I' but before the standard system directories. Mark it as a
5831 system directory, so that it gets the same special treatment as is
5832 applied to the standard system directories.
5835 Search DIR only for header files requested with `#include "FILE"';
5836 they are not searched for `#include <FILE>', before all
5837 directories specified by `-I' and before the standard system
5840 `-fdollars-in-identifiers'
5841 Accept `$' in identifiers.
5844 Indicate to the preprocessor that the input file has already been
5845 preprocessed. This suppresses things like macro expansion,
5846 trigraph conversion, escaped newline splicing, and processing of
5847 most directives. The preprocessor still recognizes and removes
5848 comments, so that you can pass a file preprocessed with `-C' to
5849 the compiler without problems. In this mode the integrated
5850 preprocessor is little more than a tokenizer for the front ends.
5852 `-fpreprocessed' is implicit if the input file has one of the
5853 extensions `.i', `.ii' or `.mi'. These are the extensions that
5854 GCC uses for preprocessed files created by `-save-temps'.
5857 Set the distance between tab stops. This helps the preprocessor
5858 report correct column numbers in warnings or errors, even if tabs
5859 appear on the line. If the value is less than 1 or greater than
5860 100, the option is ignored. The default is 8.
5862 `-fexec-charset=CHARSET'
5863 Set the execution character set, used for string and character
5864 constants. The default is UTF-8. CHARSET can be any encoding
5865 supported by the system's `iconv' library routine.
5867 `-fwide-exec-charset=CHARSET'
5868 Set the wide execution character set, used for wide string and
5869 character constants. The default is UTF-32 or UTF-16, whichever
5870 corresponds to the width of `wchar_t'. As with `-fexec-charset',
5871 CHARSET can be any encoding supported by the system's `iconv'
5872 library routine; however, you will have problems with encodings
5873 that do not fit exactly in `wchar_t'.
5875 `-finput-charset=CHARSET'
5876 Set the input character set, used for translation from the
5877 character set of the input file to the source character set used
5878 by GCC. If the locale does not specify, or GCC cannot get this
5879 information from the locale, the default is UTF-8. This can be
5880 overridden by either the locale or this command line option.
5881 Currently the command line option takes precedence if there's a
5882 conflict. CHARSET can be any encoding supported by the system's
5883 `iconv' library routine.
5885 `-fworking-directory'
5886 Enable generation of linemarkers in the preprocessor output that
5887 will let the compiler know the current working directory at the
5888 time of preprocessing. When this option is enabled, the
5889 preprocessor will emit, after the initial linemarker, a second
5890 linemarker with the current working directory followed by two
5891 slashes. GCC will use this directory, when it's present in the
5892 preprocessed input, as the directory emitted as the current
5893 working directory in some debugging information formats. This
5894 option is implicitly enabled if debugging information is enabled,
5895 but this can be inhibited with the negated form
5896 `-fno-working-directory'. If the `-P' flag is present in the
5897 command line, this option has no effect, since no `#line'
5898 directives are emitted whatsoever.
5901 Do not print column numbers in diagnostics. This may be necessary
5902 if diagnostics are being scanned by a program that does not
5903 understand the column numbers, such as `dejagnu'.
5905 `-A PREDICATE=ANSWER'
5906 Make an assertion with the predicate PREDICATE and answer ANSWER.
5907 This form is preferred to the older form `-A PREDICATE(ANSWER)',
5908 which is still supported, because it does not use shell special
5911 `-A -PREDICATE=ANSWER'
5912 Cancel an assertion with the predicate PREDICATE and answer ANSWER.
5915 CHARS is a sequence of one or more of the following characters,
5916 and must not be preceded by a space. Other characters are
5917 interpreted by the compiler proper, or reserved for future
5918 versions of GCC, and so are silently ignored. If you specify
5919 characters whose behavior conflicts, the result is undefined.
5922 Instead of the normal output, generate a list of `#define'
5923 directives for all the macros defined during the execution of
5924 the preprocessor, including predefined macros. This gives
5925 you a way of finding out what is predefined in your version
5926 of the preprocessor. Assuming you have no file `foo.h', the
5929 touch foo.h; cpp -dM foo.h
5931 will show all the predefined macros.
5934 Like `M' except in two respects: it does _not_ include the
5935 predefined macros, and it outputs _both_ the `#define'
5936 directives and the result of preprocessing. Both kinds of
5937 output go to the standard output file.
5940 Like `D', but emit only the macro names, not their expansions.
5943 Output `#include' directives in addition to the result of
5947 Inhibit generation of linemarkers in the output from the
5948 preprocessor. This might be useful when running the preprocessor
5949 on something that is not C code, and will be sent to a program
5950 which might be confused by the linemarkers.
5953 Do not discard comments. All comments are passed through to the
5954 output file, except for comments in processed directives, which
5955 are deleted along with the directive.
5957 You should be prepared for side effects when using `-C'; it causes
5958 the preprocessor to treat comments as tokens in their own right.
5959 For example, comments appearing at the start of what would be a
5960 directive line have the effect of turning that line into an
5961 ordinary source line, since the first token on the line is no
5965 Do not discard comments, including during macro expansion. This is
5966 like `-C', except that comments contained within macros are also
5967 passed through to the output file where the macro is expanded.
5969 In addition to the side-effects of the `-C' option, the `-CC'
5970 option causes all C++-style comments inside a macro to be
5971 converted to C-style comments. This is to prevent later use of
5972 that macro from inadvertently commenting out the remainder of the
5975 The `-CC' option is generally used to support lint comments.
5978 Try to imitate the behavior of old-fashioned C preprocessors, as
5979 opposed to ISO C preprocessors.
5982 Process trigraph sequences. These are three-character sequences,
5983 all starting with `??', that are defined by ISO C to stand for
5984 single characters. For example, `??/' stands for `\', so `'??/n''
5985 is a character constant for a newline. By default, GCC ignores
5986 trigraphs, but in standard-conforming modes it converts them. See
5987 the `-std' and `-ansi' options.
5989 The nine trigraphs and their replacements are
5991 Trigraph: ??( ??) ??< ??> ??= ??/ ??' ??! ??-
5992 Replacement: [ ] { } # \ ^ | ~
5995 Enable special code to work around file systems which only permit
5996 very short file names, such as MS-DOS.
6000 Print text describing all the command line options instead of
6001 preprocessing anything.
6004 Verbose mode. Print out GNU CPP's version number at the beginning
6005 of execution, and report the final form of the include path.
6008 Print the name of each header file used, in addition to other
6009 normal activities. Each name is indented to show how deep in the
6010 `#include' stack it is. Precompiled header files are also
6011 printed, even if they are found to be invalid; an invalid
6012 precompiled header file is printed with `...x' and a valid one
6017 Print out GNU CPP's version number. With one dash, proceed to
6018 preprocess as normal. With two dashes, exit immediately.
6021 File: gcc.info, Node: Assembler Options, Next: Link Options, Prev: Preprocessor Options, Up: Invoking GCC
6023 3.12 Passing Options to the Assembler
6024 =====================================
6026 You can pass options to the assembler.
6029 Pass OPTION as an option to the assembler. If OPTION contains
6030 commas, it is split into multiple options at the commas.
6032 `-Xassembler OPTION'
6033 Pass OPTION as an option to the assembler. You can use this to
6034 supply system-specific assembler options which GCC does not know
6037 If you want to pass an option that takes an argument, you must use
6038 `-Xassembler' twice, once for the option and once for the argument.
6042 File: gcc.info, Node: Link Options, Next: Directory Options, Prev: Assembler Options, Up: Invoking GCC
6044 3.13 Options for Linking
6045 ========================
6047 These options come into play when the compiler links object files into
6048 an executable output file. They are meaningless if the compiler is not
6052 A file name that does not end in a special recognized suffix is
6053 considered to name an object file or library. (Object files are
6054 distinguished from libraries by the linker according to the file
6055 contents.) If linking is done, these object files are used as
6056 input to the linker.
6061 If any of these options is used, then the linker is not run, and
6062 object file names should not be used as arguments. *Note Overall
6067 Search the library named LIBRARY when linking. (The second
6068 alternative with the library as a separate argument is only for
6069 POSIX compliance and is not recommended.)
6071 It makes a difference where in the command you write this option;
6072 the linker searches and processes libraries and object files in
6073 the order they are specified. Thus, `foo.o -lz bar.o' searches
6074 library `z' after file `foo.o' but before `bar.o'. If `bar.o'
6075 refers to functions in `z', those functions may not be loaded.
6077 The linker searches a standard list of directories for the library,
6078 which is actually a file named `libLIBRARY.a'. The linker then
6079 uses this file as if it had been specified precisely by name.
6081 The directories searched include several standard system
6082 directories plus any that you specify with `-L'.
6084 Normally the files found this way are library files--archive files
6085 whose members are object files. The linker handles an archive
6086 file by scanning through it for members which define symbols that
6087 have so far been referenced but not defined. But if the file that
6088 is found is an ordinary object file, it is linked in the usual
6089 fashion. The only difference between using an `-l' option and
6090 specifying a file name is that `-l' surrounds LIBRARY with `lib'
6091 and `.a' and searches several directories.
6094 You need this special case of the `-l' option in order to link an
6095 Objective-C or Objective-C++ program.
6098 Do not use the standard system startup files when linking. The
6099 standard system libraries are used normally, unless `-nostdlib' or
6100 `-nodefaultlibs' is used.
6103 Do not use the standard system libraries when linking. Only the
6104 libraries you specify will be passed to the linker. The standard
6105 startup files are used normally, unless `-nostartfiles' is used.
6106 The compiler may generate calls to `memcmp', `memset', `memcpy'
6107 and `memmove'. These entries are usually resolved by entries in
6108 libc. These entry points should be supplied through some other
6109 mechanism when this option is specified.
6112 Do not use the standard system startup files or libraries when
6113 linking. No startup files and only the libraries you specify will
6114 be passed to the linker. The compiler may generate calls to
6115 `memcmp', `memset', `memcpy' and `memmove'. These entries are
6116 usually resolved by entries in libc. These entry points should be
6117 supplied through some other mechanism when this option is
6120 One of the standard libraries bypassed by `-nostdlib' and
6121 `-nodefaultlibs' is `libgcc.a', a library of internal subroutines
6122 that GCC uses to overcome shortcomings of particular machines, or
6123 special needs for some languages. (*Note Interfacing to GCC
6124 Output: (gccint)Interface, for more discussion of `libgcc.a'.) In
6125 most cases, you need `libgcc.a' even when you want to avoid other
6126 standard libraries. In other words, when you specify `-nostdlib'
6127 or `-nodefaultlibs' you should usually specify `-lgcc' as well.
6128 This ensures that you have no unresolved references to internal GCC
6129 library subroutines. (For example, `__main', used to ensure C++
6130 constructors will be called; *note `collect2': (gccint)Collect2.)
6133 Produce a position independent executable on targets which support
6134 it. For predictable results, you must also specify the same set
6135 of options that were used to generate code (`-fpie', `-fPIE', or
6136 model suboptions) when you specify this option.
6139 Remove all symbol table and relocation information from the
6143 On systems that support dynamic linking, this prevents linking
6144 with the shared libraries. On other systems, this option has no
6148 Produce a shared object which can then be linked with other
6149 objects to form an executable. Not all systems support this
6150 option. For predictable results, you must also specify the same
6151 set of options that were used to generate code (`-fpic', `-fPIC',
6152 or model suboptions) when you specify this option.(1)
6156 On systems that provide `libgcc' as a shared library, these options
6157 force the use of either the shared or static version respectively.
6158 If no shared version of `libgcc' was built when the compiler was
6159 configured, these options have no effect.
6161 There are several situations in which an application should use the
6162 shared `libgcc' instead of the static version. The most common of
6163 these is when the application wishes to throw and catch exceptions
6164 across different shared libraries. In that case, each of the
6165 libraries as well as the application itself should use the shared
6168 Therefore, the G++ and GCJ drivers automatically add
6169 `-shared-libgcc' whenever you build a shared library or a main
6170 executable, because C++ and Java programs typically use
6171 exceptions, so this is the right thing to do.
6173 If, instead, you use the GCC driver to create shared libraries,
6174 you may find that they will not always be linked with the shared
6175 `libgcc'. If GCC finds, at its configuration time, that you have
6176 a non-GNU linker or a GNU linker that does not support option
6177 `--eh-frame-hdr', it will link the shared version of `libgcc' into
6178 shared libraries by default. Otherwise, it will take advantage of
6179 the linker and optimize away the linking with the shared version
6180 of `libgcc', linking with the static version of libgcc by default.
6181 This allows exceptions to propagate through such shared
6182 libraries, without incurring relocation costs at library load time.
6184 However, if a library or main executable is supposed to throw or
6185 catch exceptions, you must link it using the G++ or GCJ driver, as
6186 appropriate for the languages used in the program, or using the
6187 option `-shared-libgcc', such that it is linked with the shared
6191 Bind references to global symbols when building a shared object.
6192 Warn about any unresolved references (unless overridden by the
6193 link editor option `-Xlinker -z -Xlinker defs'). Only a few
6194 systems support this option.
6197 Pass OPTION as an option to the linker. You can use this to
6198 supply system-specific linker options which GCC does not know how
6201 If you want to pass an option that takes an argument, you must use
6202 `-Xlinker' twice, once for the option and once for the argument.
6203 For example, to pass `-assert definitions', you must write
6204 `-Xlinker -assert -Xlinker definitions'. It does not work to write
6205 `-Xlinker "-assert definitions"', because this passes the entire
6206 string as a single argument, which is not what the linker expects.
6209 Pass OPTION as an option to the linker. If OPTION contains
6210 commas, it is split into multiple options at the commas.
6213 Pretend the symbol SYMBOL is undefined, to force linking of
6214 library modules to define it. You can use `-u' multiple times with
6215 different symbols to force loading of additional library modules.
6217 ---------- Footnotes ----------
6219 (1) On some systems, `gcc -shared' needs to build supplementary stub
6220 code for constructors to work. On multi-libbed systems, `gcc -shared'
6221 must select the correct support libraries to link against. Failing to
6222 supply the correct flags may lead to subtle defects. Supplying them in
6223 cases where they are not necessary is innocuous.
6226 File: gcc.info, Node: Directory Options, Next: Spec Files, Prev: Link Options, Up: Invoking GCC
6228 3.14 Options for Directory Search
6229 =================================
6231 These options specify directories to search for header files, for
6232 libraries and for parts of the compiler:
6235 Add the directory DIR to the head of the list of directories to be
6236 searched for header files. This can be used to override a system
6237 header file, substituting your own version, since these
6238 directories are searched before the system header file
6239 directories. However, you should not use this option to add
6240 directories that contain vendor-supplied system header files (use
6241 `-isystem' for that). If you use more than one `-I' option, the
6242 directories are scanned in left-to-right order; the standard
6243 system directories come after.
6245 If a standard system include directory, or a directory specified
6246 with `-isystem', is also specified with `-I', the `-I' option will
6247 be ignored. The directory will still be searched but as a system
6248 directory at its normal position in the system include chain.
6249 This is to ensure that GCC's procedure to fix buggy system headers
6250 and the ordering for the include_next directive are not
6251 inadvertently changed. If you really need to change the search
6252 order for system directories, use the `-nostdinc' and/or
6256 Add the directory DIR to the head of the list of directories to be
6257 searched for header files only for the case of `#include "FILE"';
6258 they are not searched for `#include <FILE>', otherwise just like
6262 Add directory DIR to the list of directories to be searched for
6266 This option specifies where to find the executables, libraries,
6267 include files, and data files of the compiler itself.
6269 The compiler driver program runs one or more of the subprograms
6270 `cpp', `cc1', `as' and `ld'. It tries PREFIX as a prefix for each
6271 program it tries to run, both with and without `MACHINE/VERSION/'
6272 (*note Target Options::).
6274 For each subprogram to be run, the compiler driver first tries the
6275 `-B' prefix, if any. If that name is not found, or if `-B' was
6276 not specified, the driver tries two standard prefixes, which are
6277 `/usr/lib/gcc/' and `/usr/local/lib/gcc/'. If neither of those
6278 results in a file name that is found, the unmodified program name
6279 is searched for using the directories specified in your `PATH'
6280 environment variable.
6282 The compiler will check to see if the path provided by the `-B'
6283 refers to a directory, and if necessary it will add a directory
6284 separator character at the end of the path.
6286 `-B' prefixes that effectively specify directory names also apply
6287 to libraries in the linker, because the compiler translates these
6288 options into `-L' options for the linker. They also apply to
6289 includes files in the preprocessor, because the compiler
6290 translates these options into `-isystem' options for the
6291 preprocessor. In this case, the compiler appends `include' to the
6294 The run-time support file `libgcc.a' can also be searched for using
6295 the `-B' prefix, if needed. If it is not found there, the two
6296 standard prefixes above are tried, and that is all. The file is
6297 left out of the link if it is not found by those means.
6299 Another way to specify a prefix much like the `-B' prefix is to use
6300 the environment variable `GCC_EXEC_PREFIX'. *Note Environment
6303 As a special kludge, if the path provided by `-B' is
6304 `[dir/]stageN/', where N is a number in the range 0 to 9, then it
6305 will be replaced by `[dir/]include'. This is to help with
6306 boot-strapping the compiler.
6309 Process FILE after the compiler reads in the standard `specs'
6310 file, in order to override the defaults that the `gcc' driver
6311 program uses when determining what switches to pass to `cc1',
6312 `cc1plus', `as', `ld', etc. More than one `-specs=FILE' can be
6313 specified on the command line, and they are processed in order,
6317 This option has been deprecated. Please use `-iquote' instead for
6318 `-I' directories before the `-I-' and remove the `-I-'. Any
6319 directories you specify with `-I' options before the `-I-' option
6320 are searched only for the case of `#include "FILE"'; they are not
6321 searched for `#include <FILE>'.
6323 If additional directories are specified with `-I' options after
6324 the `-I-', these directories are searched for all `#include'
6325 directives. (Ordinarily _all_ `-I' directories are used this way.)
6327 In addition, the `-I-' option inhibits the use of the current
6328 directory (where the current input file came from) as the first
6329 search directory for `#include "FILE"'. There is no way to
6330 override this effect of `-I-'. With `-I.' you can specify
6331 searching the directory which was current when the compiler was
6332 invoked. That is not exactly the same as what the preprocessor
6333 does by default, but it is often satisfactory.
6335 `-I-' does not inhibit the use of the standard system directories
6336 for header files. Thus, `-I-' and `-nostdinc' are independent.
6339 File: gcc.info, Node: Spec Files, Next: Target Options, Prev: Directory Options, Up: Invoking GCC
6341 3.15 Specifying subprocesses and the switches to pass to them
6342 =============================================================
6344 `gcc' is a driver program. It performs its job by invoking a sequence
6345 of other programs to do the work of compiling, assembling and linking.
6346 GCC interprets its command-line parameters and uses these to deduce
6347 which programs it should invoke, and which command-line options it
6348 ought to place on their command lines. This behavior is controlled by
6349 "spec strings". In most cases there is one spec string for each
6350 program that GCC can invoke, but a few programs have multiple spec
6351 strings to control their behavior. The spec strings built into GCC can
6352 be overridden by using the `-specs=' command-line switch to specify a
6355 "Spec files" are plaintext files that are used to construct spec
6356 strings. They consist of a sequence of directives separated by blank
6357 lines. The type of directive is determined by the first non-whitespace
6358 character on the line and it can be one of the following:
6361 Issues a COMMAND to the spec file processor. The commands that can
6365 Search for FILE and insert its text at the current point in
6368 `%include_noerr <FILE>'
6369 Just like `%include', but do not generate an error message if
6370 the include file cannot be found.
6372 `%rename OLD_NAME NEW_NAME'
6373 Rename the spec string OLD_NAME to NEW_NAME.
6377 This tells the compiler to create, override or delete the named
6378 spec string. All lines after this directive up to the next
6379 directive or blank line are considered to be the text for the spec
6380 string. If this results in an empty string then the spec will be
6381 deleted. (Or, if the spec did not exist, then nothing will
6382 happened.) Otherwise, if the spec does not currently exist a new
6383 spec will be created. If the spec does exist then its contents
6384 will be overridden by the text of this directive, unless the first
6385 character of that text is the `+' character, in which case the
6386 text will be appended to the spec.
6389 Creates a new `[SUFFIX] spec' pair. All lines after this directive
6390 and up to the next directive or blank line are considered to make
6391 up the spec string for the indicated suffix. When the compiler
6392 encounters an input file with the named suffix, it will processes
6393 the spec string in order to work out how to compile that file.
6399 This says that any input file whose name ends in `.ZZ' should be
6400 passed to the program `z-compile', which should be invoked with the
6401 command-line switch `-input' and with the result of performing the
6402 `%i' substitution. (See below.)
6404 As an alternative to providing a spec string, the text that
6405 follows a suffix directive can be one of the following:
6408 This says that the suffix is an alias for a known LANGUAGE.
6409 This is similar to using the `-x' command-line switch to GCC
6410 to specify a language explicitly. For example:
6415 Says that .ZZ files are, in fact, C++ source files.
6418 This causes an error messages saying:
6420 NAME compiler not installed on this system.
6422 GCC already has an extensive list of suffixes built into it. This
6423 directive will add an entry to the end of the list of suffixes, but
6424 since the list is searched from the end backwards, it is
6425 effectively possible to override earlier entries using this
6429 GCC has the following spec strings built into it. Spec files can
6430 override these strings or create their own. Note that individual
6431 targets can also add their own spec strings to this list.
6433 asm Options to pass to the assembler
6434 asm_final Options to pass to the assembler post-processor
6435 cpp Options to pass to the C preprocessor
6436 cc1 Options to pass to the C compiler
6437 cc1plus Options to pass to the C++ compiler
6438 endfile Object files to include at the end of the link
6439 link Options to pass to the linker
6440 lib Libraries to include on the command line to the linker
6441 libgcc Decides which GCC support library to pass to the linker
6442 linker Sets the name of the linker
6443 predefines Defines to be passed to the C preprocessor
6444 signed_char Defines to pass to CPP to say whether `char' is signed
6446 startfile Object files to include at the start of the link
6448 Here is a small example of a spec file:
6453 --start-group -lgcc -lc -leval1 --end-group %(old_lib)
6455 This example renames the spec called `lib' to `old_lib' and then
6456 overrides the previous definition of `lib' with a new one. The new
6457 definition adds in some extra command-line options before including the
6458 text of the old definition.
6460 "Spec strings" are a list of command-line options to be passed to their
6461 corresponding program. In addition, the spec strings can contain
6462 `%'-prefixed sequences to substitute variable text or to conditionally
6463 insert text into the command line. Using these constructs it is
6464 possible to generate quite complex command lines.
6466 Here is a table of all defined `%'-sequences for spec strings. Note
6467 that spaces are not generated automatically around the results of
6468 expanding these sequences. Therefore you can concatenate them together
6469 or combine them with constant text in a single argument.
6472 Substitute one `%' into the program name or argument.
6475 Substitute the name of the input file being processed.
6478 Substitute the basename of the input file being processed. This
6479 is the substring up to (and not including) the last period and not
6480 including the directory.
6483 This is the same as `%b', but include the file suffix (text after
6487 Marks the argument containing or following the `%d' as a temporary
6488 file name, so that that file will be deleted if GCC exits
6489 successfully. Unlike `%g', this contributes no text to the
6493 Substitute a file name that has suffix SUFFIX and is chosen once
6494 per compilation, and mark the argument in the same way as `%d'.
6495 To reduce exposure to denial-of-service attacks, the file name is
6496 now chosen in a way that is hard to predict even when previously
6497 chosen file names are known. For example, `%g.s ... %g.o ... %g.s'
6498 might turn into `ccUVUUAU.s ccXYAXZ12.o ccUVUUAU.s'. SUFFIX
6499 matches the regexp `[.A-Za-z]*' or the special string `%O', which
6500 is treated exactly as if `%O' had been preprocessed. Previously,
6501 `%g' was simply substituted with a file name chosen once per
6502 compilation, without regard to any appended suffix (which was
6503 therefore treated just like ordinary text), making such attacks
6504 more likely to succeed.
6507 Like `%g', but generates a new temporary file name even if
6508 `%uSUFFIX' was already seen.
6511 Substitutes the last file name generated with `%uSUFFIX',
6512 generating a new one if there is no such last file name. In the
6513 absence of any `%uSUFFIX', this is just like `%gSUFFIX', except
6514 they don't share the same suffix _space_, so `%g.s ... %U.s ...
6515 %g.s ... %U.s' would involve the generation of two distinct file
6516 names, one for each `%g.s' and another for each `%U.s'.
6517 Previously, `%U' was simply substituted with a file name chosen
6518 for the previous `%u', without regard to any appended suffix.
6521 Substitutes the name of the `HOST_BIT_BUCKET', if any, and if it is
6522 writable, and if save-temps is off; otherwise, substitute the name
6523 of a temporary file, just like `%u'. This temporary file is not
6524 meant for communication between processes, but rather as a junk
6529 Like `%g', except if `-pipe' is in effect. In that case `%|'
6530 substitutes a single dash and `%m' substitutes nothing at all.
6531 These are the two most common ways to instruct a program that it
6532 should read from standard input or write to standard output. If
6533 you need something more elaborate you can use an `%{pipe:`X'}'
6534 construct: see for example `f/lang-specs.h'.
6537 Substitutes .SUFFIX for the suffixes of a matched switch's args
6538 when it is subsequently output with `%*'. SUFFIX is terminated by
6539 the next space or %.
6542 Marks the argument containing or following the `%w' as the
6543 designated output file of this compilation. This puts the argument
6544 into the sequence of arguments that `%o' will substitute later.
6547 Substitutes the names of all the output files, with spaces
6548 automatically placed around them. You should write spaces around
6549 the `%o' as well or the results are undefined. `%o' is for use in
6550 the specs for running the linker. Input files whose names have no
6551 recognized suffix are not compiled at all, but they are included
6552 among the output files, so they will be linked.
6555 Substitutes the suffix for object files. Note that this is
6556 handled specially when it immediately follows `%g, %u, or %U',
6557 because of the need for those to form complete file names. The
6558 handling is such that `%O' is treated exactly as if it had already
6559 been substituted, except that `%g, %u, and %U' do not currently
6560 support additional SUFFIX characters following `%O' as they would
6561 following, for example, `.o'.
6564 Substitutes the standard macro predefinitions for the current
6565 target machine. Use this when running `cpp'.
6568 Like `%p', but puts `__' before and after the name of each
6569 predefined macro, except for macros that start with `__' or with
6570 `_L', where L is an uppercase letter. This is for ISO C.
6573 Substitute any of `-iprefix' (made from `GCC_EXEC_PREFIX'),
6574 `-isysroot' (made from `TARGET_SYSTEM_ROOT'), and `-isystem' (made
6575 from `COMPILER_PATH' and `-B' options) as necessary.
6578 Current argument is the name of a library or startup file of some
6579 sort. Search for that file in a standard list of directories and
6580 substitute the full name found.
6583 Print STR as an error message. STR is terminated by a newline.
6584 Use this when inconsistent options are detected.
6587 Substitute the contents of spec string NAME at this point.
6590 Like `%(...)' but put `__' around `-D' arguments.
6593 Accumulate an option for `%X'.
6596 Output the accumulated linker options specified by `-Wl' or a `%x'
6600 Output the accumulated assembler options specified by `-Wa'.
6603 Output the accumulated preprocessor options specified by `-Wp'.
6606 Process the `asm' spec. This is used to compute the switches to
6607 be passed to the assembler.
6610 Process the `asm_final' spec. This is a spec string for passing
6611 switches to an assembler post-processor, if such a program is
6615 Process the `link' spec. This is the spec for computing the
6616 command line passed to the linker. Typically it will make use of
6617 the `%L %G %S %D and %E' sequences.
6620 Dump out a `-L' option for each directory that GCC believes might
6621 contain startup files. If the target supports multilibs then the
6622 current multilib directory will be prepended to each of these
6626 Process the `lib' spec. This is a spec string for deciding which
6627 libraries should be included on the command line to the linker.
6630 Process the `libgcc' spec. This is a spec string for deciding
6631 which GCC support library should be included on the command line
6635 Process the `startfile' spec. This is a spec for deciding which
6636 object files should be the first ones passed to the linker.
6637 Typically this might be a file named `crt0.o'.
6640 Process the `endfile' spec. This is a spec string that specifies
6641 the last object files that will be passed to the linker.
6644 Process the `cpp' spec. This is used to construct the arguments
6645 to be passed to the C preprocessor.
6648 Process the `cc1' spec. This is used to construct the options to
6649 be passed to the actual C compiler (`cc1').
6652 Process the `cc1plus' spec. This is used to construct the options
6653 to be passed to the actual C++ compiler (`cc1plus').
6656 Substitute the variable part of a matched option. See below.
6657 Note that each comma in the substituted string is replaced by a
6661 Remove all occurrences of `-S' from the command line. Note--this
6662 command is position dependent. `%' commands in the spec string
6663 before this one will see `-S', `%' commands in the spec string
6664 after this one will not.
6667 Call the named function FUNCTION, passing it ARGS. ARGS is first
6668 processed as a nested spec string, then split into an argument
6669 vector in the usual fashion. The function returns a string which
6670 is processed as if it had appeared literally as part of the
6673 The following built-in spec functions are provided:
6676 The `if-exists' spec function takes one argument, an absolute
6677 pathname to a file. If the file exists, `if-exists' returns
6678 the pathname. Here is a small example of its usage:
6681 crt0%O%s %:if-exists(crti%O%s) crtbegin%O%s
6684 The `if-exists-else' spec function is similar to the
6685 `if-exists' spec function, except that it takes two
6686 arguments. The first argument is an absolute pathname to a
6687 file. If the file exists, `if-exists-else' returns the
6688 pathname. If it does not exist, it returns the second
6689 argument. This way, `if-exists-else' can be used to select
6690 one file or another, based on the existence of the first.
6691 Here is a small example of its usage:
6694 crt0%O%s %:if-exists(crti%O%s) \
6695 %:if-exists-else(crtbeginT%O%s crtbegin%O%s)
6698 The `replace-outfile' spec function takes two arguments. It
6699 looks for the first argument in the outfiles array and
6700 replaces it with the second argument. Here is a small
6701 example of its usage:
6703 %{fgnu-runtime:%:replace-outfile(-lobjc -lobjc-gnu)}
6707 Substitutes the `-S' switch, if that switch was given to GCC. If
6708 that switch was not specified, this substitutes nothing. Note that
6709 the leading dash is omitted when specifying this option, and it is
6710 automatically inserted if the substitution is performed. Thus the
6711 spec string `%{foo}' would match the command-line option `-foo'
6712 and would output the command line option `-foo'.
6715 Like %{`S'} but mark last argument supplied within as a file to be
6719 Substitutes all the switches specified to GCC whose names start
6720 with `-S', but which also take an argument. This is used for
6721 switches like `-o', `-D', `-I', etc. GCC considers `-o foo' as
6722 being one switch whose names starts with `o'. %{o*} would
6723 substitute this text, including the space. Thus two arguments
6727 Like %{`S'*}, but preserve order of `S' and `T' options (the order
6728 of `S' and `T' in the spec is not significant). There can be any
6729 number of ampersand-separated variables; for each the wild card is
6730 optional. Useful for CPP as `%{D*&U*&A*}'.
6733 Substitutes `X', if the `-S' switch was given to GCC.
6736 Substitutes `X', if the `-S' switch was _not_ given to GCC.
6739 Substitutes `X' if one or more switches whose names start with
6740 `-S' are specified to GCC. Normally `X' is substituted only once,
6741 no matter how many such switches appeared. However, if `%*'
6742 appears somewhere in `X', then `X' will be substituted once for
6743 each matching switch, with the `%*' replaced by the part of that
6744 switch that matched the `*'.
6747 Substitutes `X', if processing a file with suffix `S'.
6750 Substitutes `X', if _not_ processing a file with suffix `S'.
6753 Substitutes `X' if either `-S' or `-P' was given to GCC. This may
6754 be combined with `!', `.', and `*' sequences as well, although
6755 they have a stronger binding than the `|'. If `%*' appears in
6756 `X', all of the alternatives must be starred, and only the first
6757 matching alternative is substituted.
6759 For example, a spec string like this:
6761 %{.c:-foo} %{!.c:-bar} %{.c|d:-baz} %{!.c|d:-boggle}
6763 will output the following command-line options from the following
6764 input command-line options:
6768 -d fred.c -foo -baz -boggle
6769 -d jim.d -bar -baz -boggle
6772 If `S' was given to GCC, substitutes `X'; else if `T' was given to
6773 GCC, substitutes `Y'; else substitutes `D'. There can be as many
6774 clauses as you need. This may be combined with `.', `!', `|', and
6778 The conditional text `X' in a %{`S':`X'} or similar construct may
6779 contain other nested `%' constructs or spaces, or even newlines. They
6780 are processed as usual, as described above. Trailing white space in
6781 `X' is ignored. White space may also appear anywhere on the left side
6782 of the colon in these constructs, except between `.' or `*' and the
6785 The `-O', `-f', `-m', and `-W' switches are handled specifically in
6786 these constructs. If another value of `-O' or the negated form of a
6787 `-f', `-m', or `-W' switch is found later in the command line, the
6788 earlier switch value is ignored, except with {`S'*} where `S' is just
6789 one letter, which passes all matching options.
6791 The character `|' at the beginning of the predicate text is used to
6792 indicate that a command should be piped to the following command, but
6793 only if `-pipe' is specified.
6795 It is built into GCC which switches take arguments and which do not.
6796 (You might think it would be useful to generalize this to allow each
6797 compiler's spec to say which switches take arguments. But this cannot
6798 be done in a consistent fashion. GCC cannot even decide which input
6799 files have been specified without knowing which switches take arguments,
6800 and it must know which input files to compile in order to tell which
6803 GCC also knows implicitly that arguments starting in `-l' are to be
6804 treated as compiler output files, and passed to the linker in their
6805 proper position among the other output files.
6808 File: gcc.info, Node: Target Options, Next: Submodel Options, Prev: Spec Files, Up: Invoking GCC
6810 3.16 Specifying Target Machine and Compiler Version
6811 ===================================================
6813 The usual way to run GCC is to run the executable called `gcc', or
6814 `<machine>-gcc' when cross-compiling, or `<machine>-gcc-<version>' to
6815 run a version other than the one that was installed last. Sometimes
6816 this is inconvenient, so GCC provides options that will switch to
6817 another cross-compiler or version.
6820 The argument MACHINE specifies the target machine for compilation.
6822 The value to use for MACHINE is the same as was specified as the
6823 machine type when configuring GCC as a cross-compiler. For
6824 example, if a cross-compiler was configured with `configure
6825 i386v', meaning to compile for an 80386 running System V, then you
6826 would specify `-b i386v' to run that cross compiler.
6829 The argument VERSION specifies which version of GCC to run. This
6830 is useful when multiple versions are installed. For example,
6831 VERSION might be `2.0', meaning to run GCC version 2.0.
6833 The `-V' and `-b' options work by running the
6834 `<machine>-gcc-<version>' executable, so there's no real reason to use
6835 them if you can just run that directly.
6838 File: gcc.info, Node: Submodel Options, Next: Code Gen Options, Prev: Target Options, Up: Invoking GCC
6840 3.17 Hardware Models and Configurations
6841 =======================================
6843 Earlier we discussed the standard option `-b' which chooses among
6844 different installed compilers for completely different target machines,
6845 such as VAX vs. 68000 vs. 80386.
6847 In addition, each of these target machine types can have its own
6848 special options, starting with `-m', to choose among various hardware
6849 models or configurations--for example, 68010 vs 68020, floating
6850 coprocessor or none. A single installed version of the compiler can
6851 compile for any model or configuration, according to the options
6854 Some configurations of the compiler also support additional special
6855 options, usually for compatibility with other compilers on the same
6858 These options are defined by the macro `TARGET_SWITCHES' in the
6859 machine description. The default for the options is also defined by
6860 that macro, which enables you to change the defaults.
6867 * Blackfin Options::
6870 * DEC Alpha Options::
6871 * DEC Alpha/VMS Options::
6875 * i386 and x86-64 Options::
6887 * RS/6000 and PowerPC Options::
6888 * S/390 and zSeries Options::
6891 * System V Options::
6892 * TMS320C3x/C4x Options::
6896 * Xstormy16 Options::
6901 File: gcc.info, Node: ARC Options, Next: ARM Options, Up: Submodel Options
6906 These options are defined for ARC implementations:
6909 Compile code for little endian mode. This is the default.
6912 Compile code for big endian mode.
6915 Prepend the name of the cpu to all public symbol names. In
6916 multiple-processor systems, there are many ARC variants with
6917 different instruction and register set characteristics. This flag
6918 prevents code compiled for one cpu to be linked with code compiled
6919 for another. No facility exists for handling variants that are
6920 "almost identical". This is an all or nothing option.
6923 Compile code for ARC variant CPU. Which variants are supported
6924 depend on the configuration. All variants support `-mcpu=base',
6925 this is the default.
6927 `-mtext=TEXT-SECTION'
6928 `-mdata=DATA-SECTION'
6929 `-mrodata=READONLY-DATA-SECTION'
6930 Put functions, data, and readonly data in TEXT-SECTION,
6931 DATA-SECTION, and READONLY-DATA-SECTION respectively by default.
6932 This can be overridden with the `section' attribute. *Note
6933 Variable Attributes::.
6937 File: gcc.info, Node: ARM Options, Next: AVR Options, Prev: ARC Options, Up: Submodel Options
6942 These `-m' options are defined for Advanced RISC Machines (ARM)
6946 Generate code for the specified ABI. Permissible values are:
6947 `apcs-gnu', `atpcs', `aapcs' and `iwmmxt'.
6950 Generate a stack frame that is compliant with the ARM Procedure
6951 Call Standard for all functions, even if this is not strictly
6952 necessary for correct execution of the code. Specifying
6953 `-fomit-frame-pointer' with this option will cause the stack
6954 frames not to be generated for leaf functions. The default is
6958 This is a synonym for `-mapcs-frame'.
6961 Generate code which supports calling between the ARM and Thumb
6962 instruction sets. Without this option the two instruction sets
6963 cannot be reliably used inside one program. The default is
6964 `-mno-thumb-interwork', since slightly larger code is generated
6965 when `-mthumb-interwork' is specified.
6968 Prevent the reordering of instructions in the function prolog, or
6969 the merging of those instruction with the instructions in the
6970 function's body. This means that all functions will start with a
6971 recognizable set of instructions (or in fact one of a choice from
6972 a small set of different function prologues), and this information
6973 can be used to locate the start if functions inside an executable
6974 piece of code. The default is `-msched-prolog'.
6977 Generate output containing floating point instructions. This is
6981 Generate output containing library calls for floating point.
6982 *Warning:* the requisite libraries are not available for all ARM
6983 targets. Normally the facilities of the machine's usual C
6984 compiler are used, but this cannot be done directly in
6985 cross-compilation. You must make your own arrangements to provide
6986 suitable library functions for cross-compilation.
6988 `-msoft-float' changes the calling convention in the output file;
6989 therefore, it is only useful if you compile _all_ of a program with
6990 this option. In particular, you need to compile `libgcc.a', the
6991 library that comes with GCC, with `-msoft-float' in order for this
6995 Specifies which ABI to use for floating point values. Permissible
6996 values are: `soft', `softfp' and `hard'.
6998 `soft' and `hard' are equivalent to `-msoft-float' and
6999 `-mhard-float' respectively. `softfp' allows the generation of
7000 floating point instructions, but still uses the soft-float calling
7004 Generate code for a processor running in little-endian mode. This
7005 is the default for all standard configurations.
7008 Generate code for a processor running in big-endian mode; the
7009 default is to compile code for a little-endian processor.
7011 `-mwords-little-endian'
7012 This option only applies when generating code for big-endian
7013 processors. Generate code for a little-endian word order but a
7014 big-endian byte order. That is, a byte order of the form
7015 `32107654'. Note: this option should only be used if you require
7016 compatibility with code for big-endian ARM processors generated by
7017 versions of the compiler prior to 2.8.
7020 This specifies the name of the target ARM processor. GCC uses
7021 this name to determine what kind of instructions it can emit when
7022 generating assembly code. Permissible names are: `arm2', `arm250',
7023 `arm3', `arm6', `arm60', `arm600', `arm610', `arm620', `arm7',
7024 `arm7m', `arm7d', `arm7dm', `arm7di', `arm7dmi', `arm70', `arm700',
7025 `arm700i', `arm710', `arm710c', `arm7100', `arm7500', `arm7500fe',
7026 `arm7tdmi', `arm7tdmi-s', `arm8', `strongarm', `strongarm110',
7027 `strongarm1100', `arm8', `arm810', `arm9', `arm9e', `arm920',
7028 `arm920t', `arm922t', `arm946e-s', `arm966e-s', `arm968e-s',
7029 `arm926ej-s', `arm940t', `arm9tdmi', `arm10tdmi', `arm1020t',
7030 `arm1026ej-s', `arm10e', `arm1020e', `arm1022e', `arm1136j-s',
7031 `arm1136jf-s', `mpcore', `mpcorenovfp', `arm1176jz-s',
7032 `arm1176jzf-s', `xscale', `iwmmxt', `ep9312'.
7035 This option is very similar to the `-mcpu=' option, except that
7036 instead of specifying the actual target processor type, and hence
7037 restricting which instructions can be used, it specifies that GCC
7038 should tune the performance of the code as if the target were of
7039 the type specified in this option, but still choosing the
7040 instructions that it will generate based on the cpu specified by a
7041 `-mcpu=' option. For some ARM implementations better performance
7042 can be obtained by using this option.
7045 This specifies the name of the target ARM architecture. GCC uses
7046 this name to determine what kind of instructions it can emit when
7047 generating assembly code. This option can be used in conjunction
7048 with or instead of the `-mcpu=' option. Permissible names are:
7049 `armv2', `armv2a', `armv3', `armv3m', `armv4', `armv4t', `armv5',
7050 `armv5t', `armv5te', `armv6', `armv6j', `iwmmxt', `ep9312'.
7055 This specifies what floating point hardware (or hardware
7056 emulation) is available on the target. Permissible names are:
7057 `fpa', `fpe2', `fpe3', `maverick', `vfp'. `-mfp' and `-mfpe' are
7058 synonyms for `-mfpu'=`fpe'NUMBER, for compatibility with older
7061 If `-msoft-float' is specified this specifies the format of
7062 floating point values.
7064 `-mstructure-size-boundary=N'
7065 The size of all structures and unions will be rounded up to a
7066 multiple of the number of bits set by this option. Permissible
7067 values are 8, 32 and 64. The default value varies for different
7068 toolchains. For the COFF targeted toolchain the default value is
7069 8. A value of 64 is only allowed if the underlying ABI supports
7072 Specifying the larger number can produce faster, more efficient
7073 code, but can also increase the size of the program. Different
7074 values are potentially incompatible. Code compiled with one value
7075 cannot necessarily expect to work with code or libraries compiled
7076 with another value, if they exchange information using structures
7079 `-mabort-on-noreturn'
7080 Generate a call to the function `abort' at the end of a `noreturn'
7081 function. It will be executed if the function tries to return.
7085 Tells the compiler to perform function calls by first loading the
7086 address of the function into a register and then performing a
7087 subroutine call on this register. This switch is needed if the
7088 target function will lie outside of the 64 megabyte addressing
7089 range of the offset based version of subroutine call instruction.
7091 Even if this switch is enabled, not all function calls will be
7092 turned into long calls. The heuristic is that static functions,
7093 functions which have the `short-call' attribute, functions that
7094 are inside the scope of a `#pragma no_long_calls' directive and
7095 functions whose definitions have already been compiled within the
7096 current compilation unit, will not be turned into long calls. The
7097 exception to this rule is that weak function definitions,
7098 functions with the `long-call' attribute or the `section'
7099 attribute, and functions that are within the scope of a `#pragma
7100 long_calls' directive, will always be turned into long calls.
7102 This feature is not enabled by default. Specifying
7103 `-mno-long-calls' will restore the default behavior, as will
7104 placing the function calls within the scope of a `#pragma
7105 long_calls_off' directive. Note these switches have no effect on
7106 how the compiler generates code to handle function calls via
7109 `-mnop-fun-dllimport'
7110 Disable support for the `dllimport' attribute.
7113 Treat the register used for PIC addressing as read-only, rather
7114 than loading it in the prologue for each function. The run-time
7115 system is responsible for initializing this register with an
7116 appropriate value before execution begins.
7118 `-mpic-register=REG'
7119 Specify the register to be used for PIC addressing. The default
7120 is R10 unless stack-checking is enabled, when R9 is used.
7122 `-mcirrus-fix-invalid-insns'
7123 Insert NOPs into the instruction stream to in order to work around
7124 problems with invalid Maverick instruction combinations. This
7125 option is only valid if the `-mcpu=ep9312' option has been used to
7126 enable generation of instructions for the Cirrus Maverick floating
7127 point co-processor. This option is not enabled by default, since
7128 the problem is only present in older Maverick implementations.
7129 The default can be re-enabled by use of the
7130 `-mno-cirrus-fix-invalid-insns' switch.
7132 `-mpoke-function-name'
7133 Write the name of each function into the text section, directly
7134 preceding the function prologue. The generated code is similar to
7138 .ascii "arm_poke_function_name", 0
7141 .word 0xff000000 + (t1 - t0)
7142 arm_poke_function_name
7144 stmfd sp!, {fp, ip, lr, pc}
7147 When performing a stack backtrace, code can inspect the value of
7148 `pc' stored at `fp + 0'. If the trace function then looks at
7149 location `pc - 12' and the top 8 bits are set, then we know that
7150 there is a function name embedded immediately preceding this
7151 location and has length `((pc[-3]) & 0xff000000)'.
7154 Generate code for the 16-bit Thumb instruction set. The default
7155 is to use the 32-bit ARM instruction set.
7158 Generate a stack frame that is compliant with the Thumb Procedure
7159 Call Standard for all non-leaf functions. (A leaf function is one
7160 that does not call any other functions.) The default is
7164 Generate a stack frame that is compliant with the Thumb Procedure
7165 Call Standard for all leaf functions. (A leaf function is one
7166 that does not call any other functions.) The default is
7167 `-mno-apcs-leaf-frame'.
7169 `-mcallee-super-interworking'
7170 Gives all externally visible functions in the file being compiled
7171 an ARM instruction set header which switches to Thumb mode before
7172 executing the rest of the function. This allows these functions
7173 to be called from non-interworking code.
7175 `-mcaller-super-interworking'
7176 Allows calls via function pointers (including virtual functions) to
7177 execute correctly regardless of whether the target code has been
7178 compiled for interworking or not. There is a small overhead in
7179 the cost of executing a function pointer if this option is enabled.
7183 File: gcc.info, Node: AVR Options, Next: Blackfin Options, Prev: ARM Options, Up: Submodel Options
7188 These options are defined for AVR implementations:
7191 Specify ATMEL AVR instruction set or MCU type.
7193 Instruction set avr1 is for the minimal AVR core, not supported by
7194 the C compiler, only for assembler programs (MCU types: at90s1200,
7195 attiny10, attiny11, attiny12, attiny15, attiny28).
7197 Instruction set avr2 (default) is for the classic AVR core with up
7198 to 8K program memory space (MCU types: at90s2313, at90s2323,
7199 attiny22, at90s2333, at90s2343, at90s4414, at90s4433, at90s4434,
7200 at90s8515, at90c8534, at90s8535).
7202 Instruction set avr3 is for the classic AVR core with up to 128K
7203 program memory space (MCU types: atmega103, atmega603, at43usb320,
7206 Instruction set avr4 is for the enhanced AVR core with up to 8K
7207 program memory space (MCU types: atmega8, atmega83, atmega85).
7209 Instruction set avr5 is for the enhanced AVR core with up to 128K
7210 program memory space (MCU types: atmega16, atmega161, atmega163,
7211 atmega32, atmega323, atmega64, atmega128, at43usb355, at94k).
7214 Output instruction sizes to the asm file.
7217 Specify the initial stack address, which may be a symbol or
7218 numeric value, `__stack' is the default.
7221 Generated code is not compatible with hardware interrupts. Code
7222 size will be smaller.
7225 Functions prologues/epilogues expanded as call to appropriate
7226 subroutines. Code size will be smaller.
7229 Do not generate tablejump insns which sometimes increase code size.
7232 Change only the low 8 bits of the stack pointer.
7235 Assume int to be 8 bit integer. This affects the sizes of all
7236 types: A char will be 1 byte, an int will be 1 byte, an long will
7237 be 2 bytes and long long will be 4 bytes. Please note that this
7238 option does not comply to the C standards, but it will provide you
7239 with smaller code size.
7242 File: gcc.info, Node: Blackfin Options, Next: CRIS Options, Prev: AVR Options, Up: Submodel Options
7244 3.17.4 Blackfin Options
7245 -----------------------
7247 `-momit-leaf-frame-pointer'
7248 Don't keep the frame pointer in a register for leaf functions.
7249 This avoids the instructions to save, set up and restore frame
7250 pointers and makes an extra register available in leaf functions.
7251 The option `-fomit-frame-pointer' removes the frame pointer for
7252 all functions which might make debugging harder.
7255 When enabled, the compiler will ensure that the generated code
7256 does not contain speculative loads after jump instructions. This
7257 option is enabled by default.
7260 Don't generate extra code to prevent speculative loads from
7264 When enabled, the compiler is free to take advantage of the
7265 knowledge that the entire program fits into the low 64k of memory.
7268 Assume that the program is arbitrarily large. This is the default.
7270 `-mid-shared-library'
7271 Generate code that supports shared libraries via the library ID
7272 method. This allows for execute in place and shared libraries in
7273 an environment without virtual memory management. This option
7276 `-mno-id-shared-library'
7277 Generate code that doesn't assume ID based shared libraries are
7278 being used. This is the default.
7280 `-mshared-library-id=n'
7281 Specified the identification number of the ID based shared library
7282 being compiled. Specifying a value of 0 will generate more
7283 compact code, specifying other values will force the allocation of
7284 that number to the current library but is no more space or time
7285 efficient than omitting this option.
7288 File: gcc.info, Node: CRIS Options, Next: Darwin Options, Prev: Blackfin Options, Up: Submodel Options
7293 These options are defined specifically for the CRIS ports.
7295 `-march=ARCHITECTURE-TYPE'
7296 `-mcpu=ARCHITECTURE-TYPE'
7297 Generate code for the specified architecture. The choices for
7298 ARCHITECTURE-TYPE are `v3', `v8' and `v10' for respectively
7299 ETRAX 4, ETRAX 100, and ETRAX 100 LX. Default is `v0' except for
7300 cris-axis-linux-gnu, where the default is `v10'.
7302 `-mtune=ARCHITECTURE-TYPE'
7303 Tune to ARCHITECTURE-TYPE everything applicable about the generated
7304 code, except for the ABI and the set of available instructions.
7305 The choices for ARCHITECTURE-TYPE are the same as for
7306 `-march=ARCHITECTURE-TYPE'.
7308 `-mmax-stack-frame=N'
7309 Warn when the stack frame of a function exceeds N bytes.
7311 `-melinux-stacksize=N'
7312 Only available with the `cris-axis-aout' target. Arranges for
7313 indications in the program to the kernel loader that the stack of
7314 the program should be set to N bytes.
7318 The options `-metrax4' and `-metrax100' are synonyms for
7319 `-march=v3' and `-march=v8' respectively.
7321 `-mmul-bug-workaround'
7322 `-mno-mul-bug-workaround'
7323 Work around a bug in the `muls' and `mulu' instructions for CPU
7324 models where it applies. This option is active by default.
7327 Enable CRIS-specific verbose debug-related information in the
7328 assembly code. This option also has the effect to turn off the
7329 `#NO_APP' formatted-code indicator to the assembler at the
7330 beginning of the assembly file.
7333 Do not use condition-code results from previous instruction;
7334 always emit compare and test instructions before use of condition
7338 Do not emit instructions with side-effects in addressing modes
7339 other than post-increment.
7347 These options (no-options) arranges (eliminate arrangements) for
7348 the stack-frame, individual data and constants to be aligned for
7349 the maximum single data access size for the chosen CPU model. The
7350 default is to arrange for 32-bit alignment. ABI details such as
7351 structure layout are not affected by these options.
7356 Similar to the stack- data- and const-align options above, these
7357 options arrange for stack-frame, writable data and constants to
7358 all be 32-bit, 16-bit or 8-bit aligned. The default is 32-bit
7361 `-mno-prologue-epilogue'
7362 `-mprologue-epilogue'
7363 With `-mno-prologue-epilogue', the normal function prologue and
7364 epilogue that sets up the stack-frame are omitted and no return
7365 instructions or return sequences are generated in the code. Use
7366 this option only together with visual inspection of the compiled
7367 code: no warnings or errors are generated when call-saved
7368 registers must be saved, or storage for local variable needs to be
7373 With `-fpic' and `-fPIC', don't generate (do generate) instruction
7374 sequences that load addresses for functions from the PLT part of
7375 the GOT rather than (traditional on other architectures) calls to
7376 the PLT. The default is `-mgotplt'.
7379 Legacy no-op option only recognized with the cris-axis-aout target.
7382 Legacy no-op option only recognized with the cris-axis-elf and
7383 cris-axis-linux-gnu targets.
7386 Only recognized with the cris-axis-aout target, where it selects a
7387 GNU/linux-like multilib, include files and instruction set for
7391 Legacy no-op option only recognized with the cris-axis-linux-gnu
7395 This option, recognized for the cris-axis-aout and cris-axis-elf
7396 arranges to link with input-output functions from a simulator
7397 library. Code, initialized data and zero-initialized data are
7398 allocated consecutively.
7401 Like `-sim', but pass linker options to locate initialized data at
7402 0x40000000 and zero-initialized data at 0x80000000.
7405 File: gcc.info, Node: Darwin Options, Next: DEC Alpha Options, Prev: CRIS Options, Up: Submodel Options
7407 3.17.6 Darwin Options
7408 ---------------------
7410 These options are defined for all architectures running the Darwin
7413 FSF GCC on Darwin does not create "fat" object files; it will create
7414 an object file for the single architecture that it was built to target.
7415 Apple's GCC on Darwin does create "fat" files if multiple `-arch'
7416 options are used; it does so by running the compiler or linker multiple
7417 times and joining the results together with `lipo'.
7419 The subtype of the file created (like `ppc7400' or `ppc970' or `i686')
7420 is determined by the flags that specify the ISA that GCC is targetting,
7421 like `-mcpu' or `-march'. The `-force_cpusubtype_ALL' option can be
7422 used to override this.
7424 The Darwin tools vary in their behavior when presented with an ISA
7425 mismatch. The assembler, `as', will only permit instructions to be
7426 used that are valid for the subtype of the file it is generating, so
7427 you cannot put 64-bit instructions in an `ppc750' object file. The
7428 linker for shared libraries, `/usr/bin/libtool', will fail and print an
7429 error if asked to create a shared library with a less restrictive
7430 subtype than its input files (for instance, trying to put a `ppc970'
7431 object file in a `ppc7400' library). The linker for executables, `ld',
7432 will quietly give the executable the most restrictive subtype of any of
7436 Add the framework directory DIR to the head of the list of
7437 directories to be searched for header files. These directories are
7438 interleaved with those specified by `-I' options and are scanned
7439 in a left-to-right order.
7441 A framework directory is a directory with frameworks in it. A
7442 framework is a directory with a `"Headers"' and/or
7443 `"PrivateHeaders"' directory contained directly in it that ends in
7444 `".framework"'. The name of a framework is the name of this
7445 directory excluding the `".framework"'. Headers associated with
7446 the framework are found in one of those two directories, with
7447 `"Headers"' being searched first. A subframework is a framework
7448 directory that is in a framework's `"Frameworks"' directory.
7449 Includes of subframework headers can only appear in a header of a
7450 framework that contains the subframework, or in a sibling
7451 subframework header. Two subframeworks are siblings if they occur
7452 in the same framework. A subframework should not have the same
7453 name as a framework, a warning will be issued if this is violated.
7454 Currently a subframework cannot have subframeworks, in the
7455 future, the mechanism may be extended to support this. The
7456 standard frameworks can be found in `"/System/Library/Frameworks"'
7457 and `"/Library/Frameworks"'. An example include looks like
7458 `#include <Framework/header.h>', where `Framework' denotes the
7459 name of the framework and header.h is found in the
7460 `"PrivateHeaders"' or `"Headers"' directory.
7463 Emit debugging information for symbols that are used. For STABS
7464 debugging format, this enables `-feliminate-unused-debug-symbols'.
7465 This is by default ON.
7468 Emit debugging information for all symbols and types.
7471 Override the defaults for `bool' so that `sizeof(bool)==1'. By
7472 default `sizeof(bool)' is `4' when compiling for Darwin/PowerPC
7473 and `1' when compiling for Darwin/x86, so this option has no
7476 *Warning:* The `-mone-byte-bool' switch causes GCC to generate
7477 code that is not binary compatible with code generated without
7478 that switch. Using this switch may require recompiling all other
7479 modules in a program, including system libraries. Use this switch
7480 to conform to a non-default data model.
7482 `-mfix-and-continue'
7483 `-ffix-and-continue'
7485 Generate code suitable for fast turn around development. Needed to
7486 enable gdb to dynamically load `.o' files into already running
7487 programs. `-findirect-data' and `-ffix-and-continue' are provided
7488 for backwards compatibility.
7491 Loads all members of static archive libraries. See man ld(1) for
7494 `-arch_errors_fatal'
7495 Cause the errors having to do with files that have the wrong
7496 architecture to be fatal.
7499 Causes the output file to be marked such that the dynamic linker
7500 will bind all undefined references when the file is loaded or
7504 Produce a Mach-o bundle format file. See man ld(1) for more
7507 `-bundle_loader EXECUTABLE'
7508 This option specifies the EXECUTABLE that will be loading the build
7509 output file being linked. See man ld(1) for more information.
7512 When passed this option, GCC will produce a dynamic library
7513 instead of an executable when linking, using the Darwin `libtool'
7516 `-force_cpusubtype_ALL'
7517 This causes GCC's output file to have the ALL subtype, instead of
7518 one controlled by the `-mcpu' or `-march' option.
7520 `-allowable_client CLIENT_NAME'
7522 `-compatibility_version'
7527 `-dylinker_install_name'
7529 `-exported_symbols_list'
7532 `-force_flat_namespace'
7533 `-headerpad_max_install_names'
7537 `-keep_private_externs'
7540 `-multiply_defined_unused'
7542 `-no_dead_strip_inits_and_terms'
7549 `-prebind_all_twolevel_modules'
7553 `-sectobjectsymbols'
7557 `-sectobjectsymbols'
7560 `-segs_read_only_addr'
7561 `-segs_read_write_addr'
7563 `-seg_addr_table_filename'
7566 `-segs_read_only_addr'
7567 `-segs_read_write_addr'
7572 `-twolevel_namespace'
7575 `-unexported_symbols_list'
7576 `-weak_reference_mismatches'
7578 These options are passed to the Darwin linker. The Darwin linker
7579 man page describes them in detail.
7582 File: gcc.info, Node: DEC Alpha Options, Next: DEC Alpha/VMS Options, Prev: Darwin Options, Up: Submodel Options
7584 3.17.7 DEC Alpha Options
7585 ------------------------
7587 These `-m' options are defined for the DEC Alpha implementations:
7591 Use (do not use) the hardware floating-point instructions for
7592 floating-point operations. When `-msoft-float' is specified,
7593 functions in `libgcc.a' will be used to perform floating-point
7594 operations. Unless they are replaced by routines that emulate the
7595 floating-point operations, or compiled in such a way as to call
7596 such emulations routines, these routines will issue floating-point
7597 operations. If you are compiling for an Alpha without
7598 floating-point operations, you must ensure that the library is
7599 built so as not to call them.
7601 Note that Alpha implementations without floating-point operations
7602 are required to have floating-point registers.
7606 Generate code that uses (does not use) the floating-point register
7607 set. `-mno-fp-regs' implies `-msoft-float'. If the floating-point
7608 register set is not used, floating point operands are passed in
7609 integer registers as if they were integers and floating-point
7610 results are passed in `$0' instead of `$f0'. This is a
7611 non-standard calling sequence, so any function with a
7612 floating-point argument or return value called by code compiled
7613 with `-mno-fp-regs' must also be compiled with that option.
7615 A typical use of this option is building a kernel that does not
7616 use, and hence need not save and restore, any floating-point
7620 The Alpha architecture implements floating-point hardware
7621 optimized for maximum performance. It is mostly compliant with
7622 the IEEE floating point standard. However, for full compliance,
7623 software assistance is required. This option generates code fully
7624 IEEE compliant code _except_ that the INEXACT-FLAG is not
7625 maintained (see below). If this option is turned on, the
7626 preprocessor macro `_IEEE_FP' is defined during compilation. The
7627 resulting code is less efficient but is able to correctly support
7628 denormalized numbers and exceptional IEEE values such as
7629 not-a-number and plus/minus infinity. Other Alpha compilers call
7630 this option `-ieee_with_no_inexact'.
7632 `-mieee-with-inexact'
7633 This is like `-mieee' except the generated code also maintains the
7634 IEEE INEXACT-FLAG. Turning on this option causes the generated
7635 code to implement fully-compliant IEEE math. In addition to
7636 `_IEEE_FP', `_IEEE_FP_EXACT' is defined as a preprocessor macro.
7637 On some Alpha implementations the resulting code may execute
7638 significantly slower than the code generated by default. Since
7639 there is very little code that depends on the INEXACT-FLAG, you
7640 should normally not specify this option. Other Alpha compilers
7641 call this option `-ieee_with_inexact'.
7643 `-mfp-trap-mode=TRAP-MODE'
7644 This option controls what floating-point related traps are enabled.
7645 Other Alpha compilers call this option `-fptm TRAP-MODE'. The
7646 trap mode can be set to one of four values:
7649 This is the default (normal) setting. The only traps that
7650 are enabled are the ones that cannot be disabled in software
7651 (e.g., division by zero trap).
7654 In addition to the traps enabled by `n', underflow traps are
7658 Like `su', but the instructions are marked to be safe for
7659 software completion (see Alpha architecture manual for
7663 Like `su', but inexact traps are enabled as well.
7665 `-mfp-rounding-mode=ROUNDING-MODE'
7666 Selects the IEEE rounding mode. Other Alpha compilers call this
7667 option `-fprm ROUNDING-MODE'. The ROUNDING-MODE can be one of:
7670 Normal IEEE rounding mode. Floating point numbers are
7671 rounded towards the nearest machine number or towards the
7672 even machine number in case of a tie.
7675 Round towards minus infinity.
7678 Chopped rounding mode. Floating point numbers are rounded
7682 Dynamic rounding mode. A field in the floating point control
7683 register (FPCR, see Alpha architecture reference manual)
7684 controls the rounding mode in effect. The C library
7685 initializes this register for rounding towards plus infinity.
7686 Thus, unless your program modifies the FPCR, `d' corresponds
7687 to round towards plus infinity.
7689 `-mtrap-precision=TRAP-PRECISION'
7690 In the Alpha architecture, floating point traps are imprecise.
7691 This means without software assistance it is impossible to recover
7692 from a floating trap and program execution normally needs to be
7693 terminated. GCC can generate code that can assist operating
7694 system trap handlers in determining the exact location that caused
7695 a floating point trap. Depending on the requirements of an
7696 application, different levels of precisions can be selected:
7699 Program precision. This option is the default and means a
7700 trap handler can only identify which program caused a
7701 floating point exception.
7704 Function precision. The trap handler can determine the
7705 function that caused a floating point exception.
7708 Instruction precision. The trap handler can determine the
7709 exact instruction that caused a floating point exception.
7711 Other Alpha compilers provide the equivalent options called
7712 `-scope_safe' and `-resumption_safe'.
7715 This option marks the generated code as IEEE conformant. You must
7716 not use this option unless you also specify `-mtrap-precision=i'
7717 and either `-mfp-trap-mode=su' or `-mfp-trap-mode=sui'. Its only
7718 effect is to emit the line `.eflag 48' in the function prologue of
7719 the generated assembly file. Under DEC Unix, this has the effect
7720 that IEEE-conformant math library routines will be linked in.
7723 Normally GCC examines a 32- or 64-bit integer constant to see if
7724 it can construct it from smaller constants in two or three
7725 instructions. If it cannot, it will output the constant as a
7726 literal and generate code to load it from the data segment at
7729 Use this option to require GCC to construct _all_ integer constants
7730 using code, even if it takes more instructions (the maximum is
7733 You would typically use this option to build a shared library
7734 dynamic loader. Itself a shared library, it must relocate itself
7735 in memory before it can find the variables and constants in its
7740 Select whether to generate code to be assembled by the
7741 vendor-supplied assembler (`-malpha-as') or by the GNU assembler
7752 Indicate whether GCC should generate code to use the optional BWX,
7753 CIX, FIX and MAX instruction sets. The default is to use the
7754 instruction sets supported by the CPU type specified via `-mcpu='
7755 option or that of the CPU on which GCC was built if none was
7760 Generate code that uses (does not use) VAX F and G floating point
7761 arithmetic instead of IEEE single and double precision.
7764 `-mno-explicit-relocs'
7765 Older Alpha assemblers provided no way to generate symbol
7766 relocations except via assembler macros. Use of these macros does
7767 not allow optimal instruction scheduling. GNU binutils as of
7768 version 2.12 supports a new syntax that allows the compiler to
7769 explicitly mark which relocations should apply to which
7770 instructions. This option is mostly useful for debugging, as GCC
7771 detects the capabilities of the assembler when it is built and
7772 sets the default accordingly.
7776 When `-mexplicit-relocs' is in effect, static data is accessed via
7777 "gp-relative" relocations. When `-msmall-data' is used, objects 8
7778 bytes long or smaller are placed in a "small data area" (the
7779 `.sdata' and `.sbss' sections) and are accessed via 16-bit
7780 relocations off of the `$gp' register. This limits the size of
7781 the small data area to 64KB, but allows the variables to be
7782 directly accessed via a single instruction.
7784 The default is `-mlarge-data'. With this option the data area is
7785 limited to just below 2GB. Programs that require more than 2GB of
7786 data must use `malloc' or `mmap' to allocate the data in the heap
7787 instead of in the program's data segment.
7789 When generating code for shared libraries, `-fpic' implies
7790 `-msmall-data' and `-fPIC' implies `-mlarge-data'.
7794 When `-msmall-text' is used, the compiler assumes that the code of
7795 the entire program (or shared library) fits in 4MB, and is thus
7796 reachable with a branch instruction. When `-msmall-data' is used,
7797 the compiler can assume that all local symbols share the same
7798 `$gp' value, and thus reduce the number of instructions required
7799 for a function call from 4 to 1.
7801 The default is `-mlarge-text'.
7804 Set the instruction set and instruction scheduling parameters for
7805 machine type CPU_TYPE. You can specify either the `EV' style name
7806 or the corresponding chip number. GCC supports scheduling
7807 parameters for the EV4, EV5 and EV6 family of processors and will
7808 choose the default values for the instruction set from the
7809 processor you specify. If you do not specify a processor type,
7810 GCC will default to the processor on which the compiler was built.
7812 Supported values for CPU_TYPE are
7817 Schedules as an EV4 and has no instruction set extensions.
7821 Schedules as an EV5 and has no instruction set extensions.
7825 Schedules as an EV5 and supports the BWX extension.
7830 Schedules as an EV5 and supports the BWX and MAX extensions.
7834 Schedules as an EV6 and supports the BWX, FIX, and MAX
7839 Schedules as an EV6 and supports the BWX, CIX, FIX, and MAX
7843 Set only the instruction scheduling parameters for machine type
7844 CPU_TYPE. The instruction set is not changed.
7846 `-mmemory-latency=TIME'
7847 Sets the latency the scheduler should assume for typical memory
7848 references as seen by the application. This number is highly
7849 dependent on the memory access patterns used by the application
7850 and the size of the external cache on the machine.
7852 Valid options for TIME are
7855 A decimal number representing clock cycles.
7861 The compiler contains estimates of the number of clock cycles
7862 for "typical" EV4 & EV5 hardware for the Level 1, 2 & 3 caches
7863 (also called Dcache, Scache, and Bcache), as well as to main
7864 memory. Note that L3 is only valid for EV5.
7868 File: gcc.info, Node: DEC Alpha/VMS Options, Next: FRV Options, Prev: DEC Alpha Options, Up: Submodel Options
7870 3.17.8 DEC Alpha/VMS Options
7871 ----------------------------
7873 These `-m' options are defined for the DEC Alpha/VMS implementations:
7875 `-mvms-return-codes'
7876 Return VMS condition codes from main. The default is to return
7877 POSIX style condition (e.g. error) codes.
7880 File: gcc.info, Node: FRV Options, Next: H8/300 Options, Prev: DEC Alpha/VMS Options, Up: Submodel Options
7886 Only use the first 32 general purpose registers.
7889 Use all 64 general purpose registers.
7892 Use only the first 32 floating point registers.
7895 Use all 64 floating point registers
7898 Use hardware instructions for floating point operations.
7901 Use library routines for floating point operations.
7904 Dynamically allocate condition code registers.
7907 Do not try to dynamically allocate condition code registers, only
7908 use `icc0' and `fcc0'.
7911 Change ABI to use double word insns.
7914 Do not use double word instructions.
7917 Use floating point double instructions.
7920 Do not use floating point double instructions.
7923 Use media instructions.
7926 Do not use media instructions.
7929 Use multiply and add/subtract instructions.
7932 Do not use multiply and add/subtract instructions.
7935 Select the FDPIC ABI, that uses function descriptors to represent
7936 pointers to functions. Without any PIC/PIE-related options, it
7937 implies `-fPIE'. With `-fpic' or `-fpie', it assumes GOT entries
7938 and small data are within a 12-bit range from the GOT base
7939 address; with `-fPIC' or `-fPIE', GOT offsets are computed with 32
7943 Enable inlining of PLT entries in function calls to functions that
7944 are not known to bind locally. It has no effect without `-mfdpic'.
7945 It's enabled by default if optimizing for speed and compiling for
7946 shared libraries (i.e., `-fPIC' or `-fpic'), or when an
7947 optimization option such as `-O3' or above is present in the
7951 Assume a large TLS segment when generating thread-local code.
7954 Do not assume a large TLS segment when generating thread-local
7958 Enable the use of `GPREL' relocations in the FDPIC ABI for data
7959 that is known to be in read-only sections. It's enabled by
7960 default, except for `-fpic' or `-fpie': even though it may help
7961 make the global offset table smaller, it trades 1 instruction for
7962 4. With `-fPIC' or `-fPIE', it trades 3 instructions for 4, one
7963 of which may be shared by multiple symbols, and it avoids the need
7964 for a GOT entry for the referenced symbol, so it's more likely to
7965 be a win. If it is not, `-mno-gprel-ro' can be used to disable it.
7967 `-multilib-library-pic'
7968 Link with the (library, not FD) pic libraries. It's implied by
7969 `-mlibrary-pic', as well as by `-fPIC' and `-fpic' without
7970 `-mfdpic'. You should never have to use it explicitly.
7973 Follow the EABI requirement of always creating a frame pointer
7974 whenever a stack frame is allocated. This option is enabled by
7975 default and can be disabled with `-mno-linked-fp'.
7978 Use indirect addressing to call functions outside the current
7979 compilation unit. This allows the functions to be placed anywhere
7980 within the 32-bit address space.
7983 Try to align labels to an 8-byte boundary by inserting nops into
7984 the previous packet. This option only has an effect when VLIW
7985 packing is enabled. It doesn't create new packets; it merely adds
7986 nops to existing ones.
7989 Generate position-independent EABI code.
7992 Use only the first four media accumulator registers.
7995 Use all eight media accumulator registers.
7998 Pack VLIW instructions.
8001 Do not pack VLIW instructions.
8004 Do not mark ABI switches in e_flags.
8007 Enable the use of conditional-move instructions (default).
8009 This switch is mainly for debugging the compiler and will likely
8010 be removed in a future version.
8013 Disable the use of conditional-move instructions.
8015 This switch is mainly for debugging the compiler and will likely
8016 be removed in a future version.
8019 Enable the use of conditional set instructions (default).
8021 This switch is mainly for debugging the compiler and will likely
8022 be removed in a future version.
8025 Disable the use of conditional set instructions.
8027 This switch is mainly for debugging the compiler and will likely
8028 be removed in a future version.
8031 Enable the use of conditional execution (default).
8033 This switch is mainly for debugging the compiler and will likely
8034 be removed in a future version.
8037 Disable the use of conditional execution.
8039 This switch is mainly for debugging the compiler and will likely
8040 be removed in a future version.
8043 Run a pass to pack branches into VLIW instructions (default).
8045 This switch is mainly for debugging the compiler and will likely
8046 be removed in a future version.
8049 Do not run a pass to pack branches into VLIW instructions.
8051 This switch is mainly for debugging the compiler and will likely
8052 be removed in a future version.
8055 Enable optimization of `&&' and `||' in conditional execution
8058 This switch is mainly for debugging the compiler and will likely
8059 be removed in a future version.
8061 `-mno-multi-cond-exec'
8062 Disable optimization of `&&' and `||' in conditional execution.
8064 This switch is mainly for debugging the compiler and will likely
8065 be removed in a future version.
8067 `-mnested-cond-exec'
8068 Enable nested conditional execution optimizations (default).
8070 This switch is mainly for debugging the compiler and will likely
8071 be removed in a future version.
8073 `-mno-nested-cond-exec'
8074 Disable nested conditional execution optimizations.
8076 This switch is mainly for debugging the compiler and will likely
8077 be removed in a future version.
8080 Cause gas to print out tomcat statistics.
8083 Select the processor type for which to generate code. Possible
8084 values are `frv', `fr550', `tomcat', `fr500', `fr450', `fr405',
8085 `fr400', `fr300' and `simple'.
8089 File: gcc.info, Node: H8/300 Options, Next: HPPA Options, Prev: FRV Options, Up: Submodel Options
8091 3.17.10 H8/300 Options
8092 ----------------------
8094 These `-m' options are defined for the H8/300 implementations:
8097 Shorten some address references at link time, when possible; uses
8098 the linker option `-relax'. *Note `ld' and the H8/300:
8099 (ld)H8/300, for a fuller description.
8102 Generate code for the H8/300H.
8105 Generate code for the H8S.
8108 Generate code for the H8S and H8/300H in the normal mode. This
8109 switch must be used either with `-mh' or `-ms'.
8112 Generate code for the H8S/2600. This switch must be used with
8116 Make `int' data 32 bits by default.
8119 On the H8/300H and H8S, use the same alignment rules as for the
8120 H8/300. The default for the H8/300H and H8S is to align longs and
8121 floats on 4 byte boundaries. `-malign-300' causes them to be
8122 aligned on 2 byte boundaries. This option has no effect on the
8126 File: gcc.info, Node: HPPA Options, Next: i386 and x86-64 Options, Prev: H8/300 Options, Up: Submodel Options
8128 3.17.11 HPPA Options
8129 --------------------
8131 These `-m' options are defined for the HPPA family of computers:
8133 `-march=ARCHITECTURE-TYPE'
8134 Generate code for the specified architecture. The choices for
8135 ARCHITECTURE-TYPE are `1.0' for PA 1.0, `1.1' for PA 1.1, and
8136 `2.0' for PA 2.0 processors. Refer to `/usr/lib/sched.models' on
8137 an HP-UX system to determine the proper architecture option for
8138 your machine. Code compiled for lower numbered architectures will
8139 run on higher numbered architectures, but not the other way around.
8144 Synonyms for `-march=1.0', `-march=1.1', and `-march=2.0'
8148 Generate code suitable for big switch tables. Use this option
8149 only if the assembler/linker complain about out of range branches
8150 within a switch table.
8153 Fill delay slots of function calls with unconditional jump
8154 instructions by modifying the return pointer for the function call
8155 to be the target of the conditional jump.
8158 Prevent floating point registers from being used in any manner.
8159 This is necessary for compiling kernels which perform lazy context
8160 switching of floating point registers. If you use this option and
8161 attempt to perform floating point operations, the compiler will
8164 `-mdisable-indexing'
8165 Prevent the compiler from using indexing address modes. This
8166 avoids some rather obscure problems when compiling MIG generated
8170 Generate code that assumes the target has no space registers.
8171 This allows GCC to generate faster indirect calls and use unscaled
8172 index address modes.
8174 Such code is suitable for level 0 PA systems and kernels.
8176 `-mfast-indirect-calls'
8177 Generate code that assumes calls never cross space boundaries.
8178 This allows GCC to emit code which performs faster indirect calls.
8180 This option will not work in the presence of shared libraries or
8183 `-mfixed-range=REGISTER-RANGE'
8184 Generate code treating the given register range as fixed registers.
8185 A fixed register is one that the register allocator can not use.
8186 This is useful when compiling kernel code. A register range is
8187 specified as two registers separated by a dash. Multiple register
8188 ranges can be specified separated by a comma.
8191 Generate 3-instruction load and store sequences as sometimes
8192 required by the HP-UX 10 linker. This is equivalent to the `+k'
8193 option to the HP compilers.
8195 `-mportable-runtime'
8196 Use the portable calling conventions proposed by HP for ELF
8200 Enable the use of assembler directives only GAS understands.
8202 `-mschedule=CPU-TYPE'
8203 Schedule code according to the constraints for the machine type
8204 CPU-TYPE. The choices for CPU-TYPE are `700' `7100', `7100LC',
8205 `7200', `7300' and `8000'. Refer to `/usr/lib/sched.models' on an
8206 HP-UX system to determine the proper scheduling option for your
8207 machine. The default scheduling is `8000'.
8210 Enable the optimization pass in the HP-UX linker. Note this makes
8211 symbolic debugging impossible. It also triggers a bug in the
8212 HP-UX 8 and HP-UX 9 linkers in which they give bogus error
8213 messages when linking some programs.
8216 Generate output containing library calls for floating point.
8217 *Warning:* the requisite libraries are not available for all HPPA
8218 targets. Normally the facilities of the machine's usual C
8219 compiler are used, but this cannot be done directly in
8220 cross-compilation. You must make your own arrangements to provide
8221 suitable library functions for cross-compilation. The embedded
8222 target `hppa1.1-*-pro' does provide software floating point
8225 `-msoft-float' changes the calling convention in the output file;
8226 therefore, it is only useful if you compile _all_ of a program with
8227 this option. In particular, you need to compile `libgcc.a', the
8228 library that comes with GCC, with `-msoft-float' in order for this
8232 Generate the predefine, `_SIO', for server IO. The default is
8233 `-mwsio'. This generates the predefines, `__hp9000s700',
8234 `__hp9000s700__' and `_WSIO', for workstation IO. These options
8235 are available under HP-UX and HI-UX.
8238 Use GNU ld specific options. This passes `-shared' to ld when
8239 building a shared library. It is the default when GCC is
8240 configured, explicitly or implicitly, with the GNU linker. This
8241 option does not have any affect on which ld is called, it only
8242 changes what parameters are passed to that ld. The ld that is
8243 called is determined by the `--with-ld' configure option, GCC's
8244 program search path, and finally by the user's `PATH'. The linker
8245 used by GCC can be printed using `which `gcc -print-prog-name=ld`'.
8248 Use HP ld specific options. This passes `-b' to ld when building
8249 a shared library and passes `+Accept TypeMismatch' to ld on all
8250 links. It is the default when GCC is configured, explicitly or
8251 implicitly, with the HP linker. This option does not have any
8252 affect on which ld is called, it only changes what parameters are
8253 passed to that ld. The ld that is called is determined by the
8254 `--with-ld' configure option, GCC's program search path, and
8255 finally by the user's `PATH'. The linker used by GCC can be
8256 printed using `which `gcc -print-prog-name=ld`'.
8259 Generate code that uses long call sequences. This ensures that a
8260 call is always able to reach linker generated stubs. The default
8261 is to generate long calls only when the distance from the call
8262 site to the beginning of the function or translation unit, as the
8263 case may be, exceeds a predefined limit set by the branch type
8264 being used. The limits for normal calls are 7,600,000 and 240,000
8265 bytes, respectively for the PA 2.0 and PA 1.X architectures.
8266 Sibcalls are always limited at 240,000 bytes.
8268 Distances are measured from the beginning of functions when using
8269 the `-ffunction-sections' option, or when using the `-mgas' and
8270 `-mno-portable-runtime' options together under HP-UX with the SOM
8273 It is normally not desirable to use this option as it will degrade
8274 performance. However, it may be useful in large applications,
8275 particularly when partial linking is used to build the application.
8277 The types of long calls used depends on the capabilities of the
8278 assembler and linker, and the type of code being generated. The
8279 impact on systems that support long absolute calls, and long pic
8280 symbol-difference or pc-relative calls should be relatively small.
8281 However, an indirect call is used on 32-bit ELF systems in pic code
8282 and it is quite long.
8285 Generate compiler predefines and select a startfile for the
8286 specified UNIX standard. The choices for UNIX-STD are `93', `95'
8287 and `98'. `93' is supported on all HP-UX versions. `95' is
8288 available on HP-UX 10.10 and later. `98' is available on HP-UX
8289 11.11 and later. The default values are `93' for HP-UX 10.00,
8290 `95' for HP-UX 10.10 though to 11.00, and `98' for HP-UX 11.11 and
8293 `-munix=93' provides the same predefines as GCC 3.3 and 3.4.
8294 `-munix=95' provides additional predefines for `XOPEN_UNIX' and
8295 `_XOPEN_SOURCE_EXTENDED', and the startfile `unix95.o'.
8296 `-munix=98' provides additional predefines for `_XOPEN_UNIX',
8297 `_XOPEN_SOURCE_EXTENDED', `_INCLUDE__STDC_A1_SOURCE' and
8298 `_INCLUDE_XOPEN_SOURCE_500', and the startfile `unix98.o'.
8300 It is _important_ to note that this option changes the interfaces
8301 for various library routines. It also affects the operational
8302 behavior of the C library. Thus, _extreme_ care is needed in
8305 Library code that is intended to operate with more than one UNIX
8306 standard must test, set and restore the variable
8307 __XPG4_EXTENDED_MASK as appropriate. Most GNU software doesn't
8308 provide this capability.
8311 Suppress the generation of link options to search libdld.sl when
8312 the `-static' option is specified on HP-UX 10 and later.
8315 The HP-UX implementation of setlocale in libc has a dependency on
8316 libdld.sl. There isn't an archive version of libdld.sl. Thus,
8317 when the `-static' option is specified, special link options are
8318 needed to resolve this dependency.
8320 On HP-UX 10 and later, the GCC driver adds the necessary options to
8321 link with libdld.sl when the `-static' option is specified. This
8322 causes the resulting binary to be dynamic. On the 64-bit port,
8323 the linkers generate dynamic binaries by default in any case. The
8324 `-nolibdld' option can be used to prevent the GCC driver from
8325 adding these link options.
8328 Add support for multithreading with the "dce thread" library under
8329 HP-UX. This option sets flags for both the preprocessor and
8333 File: gcc.info, Node: i386 and x86-64 Options, Next: IA-64 Options, Prev: HPPA Options, Up: Submodel Options
8335 3.17.12 Intel 386 and AMD x86-64 Options
8336 ----------------------------------------
8338 These `-m' options are defined for the i386 and x86-64 family of
8342 Tune to CPU-TYPE everything applicable about the generated code,
8343 except for the ABI and the set of available instructions. The
8344 choices for CPU-TYPE are:
8346 Original Intel's i386 CPU.
8349 Intel's i486 CPU. (No scheduling is implemented for this
8353 Intel Pentium CPU with no MMX support.
8356 Intel PentiumMMX CPU based on Pentium core with MMX
8357 instruction set support.
8360 Intel PentiumPro CPU.
8363 Intel Pentium2 CPU based on PentiumPro core with MMX
8364 instruction set support.
8366 _pentium3, pentium3m_
8367 Intel Pentium3 CPU based on PentiumPro core with MMX and SSE
8368 instruction set support.
8371 Low power version of Intel Pentium3 CPU with MMX, SSE and
8372 SSE2 instruction set support. Used by Centrino notebooks.
8374 _pentium4, pentium4m_
8375 Intel Pentium4 CPU with MMX, SSE and SSE2 instruction set
8379 Improved version of Intel Pentium4 CPU with MMX, SSE, SSE2
8380 and SSE3 instruction set support.
8383 Improved version of Intel Pentium4 CPU with 64-bit
8384 extensions, MMX, SSE, SSE2 and SSE3 instruction set support.
8387 AMD K6 CPU with MMX instruction set support.
8390 Improved versions of AMD K6 CPU with MMX and 3dNOW!
8391 instruction set support.
8393 _athlon, athlon-tbird_
8394 AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and SSE
8395 prefetch instructions support.
8397 _athlon-4, athlon-xp, athlon-mp_
8398 Improved AMD Athlon CPU with MMX, 3dNOW!, enhanced 3dNOW! and
8399 full SSE instruction set support.
8401 _k8, opteron, athlon64, athlon-fx_
8402 AMD K8 core based CPUs with x86-64 instruction set support.
8403 (This supersets MMX, SSE, SSE2, 3dNOW!, enhanced 3dNOW! and
8404 64-bit instruction set extensions.)
8407 IDT Winchip C6 CPU, dealt in same way as i486 with additional
8408 MMX instruction set support.
8411 IDT Winchip2 CPU, dealt in same way as i486 with additional
8412 MMX and 3dNOW! instruction set support.
8415 Via C3 CPU with MMX and 3dNOW! instruction set support. (No
8416 scheduling is implemented for this chip.)
8419 Via C3-2 CPU with MMX and SSE instruction set support. (No
8420 scheduling is implemented for this chip.)
8422 While picking a specific CPU-TYPE will schedule things
8423 appropriately for that particular chip, the compiler will not
8424 generate any code that does not run on the i386 without the
8425 `-march=CPU-TYPE' option being used.
8428 Generate instructions for the machine type CPU-TYPE. The choices
8429 for CPU-TYPE are the same as for `-mtune'. Moreover, specifying
8430 `-march=CPU-TYPE' implies `-mtune=CPU-TYPE'.
8433 A deprecated synonym for `-mtune'.
8439 These options are synonyms for `-mtune=i386', `-mtune=i486',
8440 `-mtune=pentium', and `-mtune=pentiumpro' respectively. These
8441 synonyms are deprecated.
8444 Generate floating point arithmetics for selected unit UNIT. The
8445 choices for UNIT are:
8448 Use the standard 387 floating point coprocessor present
8449 majority of chips and emulated otherwise. Code compiled with
8450 this option will run almost everywhere. The temporary
8451 results are computed in 80bit precision instead of precision
8452 specified by the type resulting in slightly different results
8453 compared to most of other chips. See `-ffloat-store' for
8454 more detailed description.
8456 This is the default choice for i386 compiler.
8459 Use scalar floating point instructions present in the SSE
8460 instruction set. This instruction set is supported by
8461 Pentium3 and newer chips, in the AMD line by Athlon-4,
8462 Athlon-xp and Athlon-mp chips. The earlier version of SSE
8463 instruction set supports only single precision arithmetics,
8464 thus the double and extended precision arithmetics is still
8465 done using 387. Later version, present only in Pentium4 and
8466 the future AMD x86-64 chips supports double precision
8469 For the i386 compiler, you need to use `-march=CPU-TYPE',
8470 `-msse' or `-msse2' switches to enable SSE extensions and
8471 make this option effective. For the x86-64 compiler, these
8472 extensions are enabled by default.
8474 The resulting code should be considerably faster in the
8475 majority of cases and avoid the numerical instability
8476 problems of 387 code, but may break some existing code that
8477 expects temporaries to be 80bit.
8479 This is the default choice for the x86-64 compiler.
8482 Attempt to utilize both instruction sets at once. This
8483 effectively double the amount of available registers and on
8484 chips with separate execution units for 387 and SSE the
8485 execution resources too. Use this option with care, as it is
8486 still experimental, because the GCC register allocator does
8487 not model separate functional units well resulting in
8488 instable performance.
8491 Output asm instructions using selected DIALECT. Supported choices
8492 are `intel' or `att' (the default one).
8496 Control whether or not the compiler uses IEEE floating point
8497 comparisons. These handle correctly the case where the result of a
8498 comparison is unordered.
8501 Generate output containing library calls for floating point.
8502 *Warning:* the requisite libraries are not part of GCC. Normally
8503 the facilities of the machine's usual C compiler are used, but
8504 this can't be done directly in cross-compilation. You must make
8505 your own arrangements to provide suitable library functions for
8508 On machines where a function returns floating point results in the
8509 80387 register stack, some floating point opcodes may be emitted
8510 even if `-msoft-float' is used.
8512 `-mno-fp-ret-in-387'
8513 Do not use the FPU registers for return values of functions.
8515 The usual calling convention has functions return values of types
8516 `float' and `double' in an FPU register, even if there is no FPU.
8517 The idea is that the operating system should emulate an FPU.
8519 The option `-mno-fp-ret-in-387' causes such values to be returned
8520 in ordinary CPU registers instead.
8522 `-mno-fancy-math-387'
8523 Some 387 emulators do not support the `sin', `cos' and `sqrt'
8524 instructions for the 387. Specify this option to avoid generating
8525 those instructions. This option is the default on FreeBSD,
8526 OpenBSD and NetBSD. This option is overridden when `-march'
8527 indicates that the target cpu will always have an FPU and so the
8528 instruction will not need emulation. As of revision 2.6.1, these
8529 instructions are not generated unless you also use the
8530 `-funsafe-math-optimizations' switch.
8534 Control whether GCC aligns `double', `long double', and `long
8535 long' variables on a two word boundary or a one word boundary.
8536 Aligning `double' variables on a two word boundary will produce
8537 code that runs somewhat faster on a `Pentium' at the expense of
8540 *Warning:* if you use the `-malign-double' switch, structures
8541 containing the above types will be aligned differently than the
8542 published application binary interface specifications for the 386
8543 and will not be binary compatible with structures in code compiled
8544 without that switch.
8546 `-m96bit-long-double'
8547 `-m128bit-long-double'
8548 These switches control the size of `long double' type. The i386
8549 application binary interface specifies the size to be 96 bits, so
8550 `-m96bit-long-double' is the default in 32 bit mode.
8552 Modern architectures (Pentium and newer) would prefer `long double'
8553 to be aligned to an 8 or 16 byte boundary. In arrays or structures
8554 conforming to the ABI, this would not be possible. So specifying a
8555 `-m128bit-long-double' will align `long double' to a 16 byte
8556 boundary by padding the `long double' with an additional 32 bit
8559 In the x86-64 compiler, `-m128bit-long-double' is the default
8560 choice as its ABI specifies that `long double' is to be aligned on
8563 Notice that neither of these options enable any extra precision
8564 over the x87 standard of 80 bits for a `long double'.
8566 *Warning:* if you override the default value for your target ABI,
8567 the structures and arrays containing `long double' variables will
8568 change their size as well as function calling convention for
8569 function taking `long double' will be modified. Hence they will
8570 not be binary compatible with arrays or structures in code
8571 compiled without that switch.
8575 Control whether GCC places uninitialized local variables into the
8576 `bss' or `data' segments. `-msvr3-shlib' places them into `bss'.
8577 These options are meaningful only on System V Release 3.
8580 Use a different function-calling convention, in which functions
8581 that take a fixed number of arguments return with the `ret' NUM
8582 instruction, which pops their arguments while returning. This
8583 saves one instruction in the caller since there is no need to pop
8584 the arguments there.
8586 You can specify that an individual function is called with this
8587 calling sequence with the function attribute `stdcall'. You can
8588 also override the `-mrtd' option by using the function attribute
8589 `cdecl'. *Note Function Attributes::.
8591 *Warning:* this calling convention is incompatible with the one
8592 normally used on Unix, so you cannot use it if you need to call
8593 libraries compiled with the Unix compiler.
8595 Also, you must provide function prototypes for all functions that
8596 take variable numbers of arguments (including `printf'); otherwise
8597 incorrect code will be generated for calls to those functions.
8599 In addition, seriously incorrect code will result if you call a
8600 function with too many arguments. (Normally, extra arguments are
8601 harmlessly ignored.)
8604 Control how many registers are used to pass integer arguments. By
8605 default, no registers are used to pass arguments, and at most 3
8606 registers can be used. You can control this behavior for a
8607 specific function by using the function attribute `regparm'.
8608 *Note Function Attributes::.
8610 *Warning:* if you use this switch, and NUM is nonzero, then you
8611 must build all modules with the same value, including any
8612 libraries. This includes the system libraries and startup modules.
8614 `-mpreferred-stack-boundary=NUM'
8615 Attempt to keep the stack boundary aligned to a 2 raised to NUM
8616 byte boundary. If `-mpreferred-stack-boundary' is not specified,
8617 the default is 4 (16 bytes or 128 bits), except when optimizing
8618 for code size (`-Os'), in which case the default is the minimum
8619 correct alignment (4 bytes for x86, and 8 bytes for x86-64).
8621 On Pentium and PentiumPro, `double' and `long double' values
8622 should be aligned to an 8 byte boundary (see `-malign-double') or
8623 suffer significant run time performance penalties. On Pentium
8624 III, the Streaming SIMD Extension (SSE) data type `__m128' suffers
8625 similar penalties if it is not 16 byte aligned.
8627 To ensure proper alignment of this values on the stack, the stack
8628 boundary must be as aligned as that required by any value stored
8629 on the stack. Further, every function must be generated such that
8630 it keeps the stack aligned. Thus calling a function compiled with
8631 a higher preferred stack boundary from a function compiled with a
8632 lower preferred stack boundary will most likely misalign the
8633 stack. It is recommended that libraries that use callbacks always
8634 use the default setting.
8636 This extra alignment does consume extra stack space, and generally
8637 increases code size. Code that is sensitive to stack space usage,
8638 such as embedded systems and operating system kernels, may want to
8639 reduce the preferred alignment to `-mpreferred-stack-boundary=2'.
8655 These switches enable or disable the use of built-in functions
8656 that allow direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow
8657 extensions of the instruction set.
8659 *Note X86 Built-in Functions::, for details of the functions
8660 enabled and disabled by these switches.
8662 To have SSE/SSE2 instructions generated automatically from
8663 floating-point code, see `-mfpmath=sse'.
8667 Use PUSH operations to store outgoing parameters. This method is
8668 shorter and usually equally fast as method using SUB/MOV
8669 operations and is enabled by default. In some cases disabling it
8670 may improve performance because of improved scheduling and reduced
8673 `-maccumulate-outgoing-args'
8674 If enabled, the maximum amount of space required for outgoing
8675 arguments will be computed in the function prologue. This is
8676 faster on most modern CPUs because of reduced dependencies,
8677 improved scheduling and reduced stack usage when preferred stack
8678 boundary is not equal to 2. The drawback is a notable increase in
8679 code size. This switch implies `-mno-push-args'.
8682 Support thread-safe exception handling on `Mingw32'. Code that
8683 relies on thread-safe exception handling must compile and link all
8684 code with the `-mthreads' option. When compiling, `-mthreads'
8685 defines `-D_MT'; when linking, it links in a special thread helper
8686 library `-lmingwthrd' which cleans up per thread exception
8689 `-mno-align-stringops'
8690 Do not align destination of inlined string operations. This
8691 switch reduces code size and improves performance in case the
8692 destination is already aligned, but GCC doesn't know about it.
8694 `-minline-all-stringops'
8695 By default GCC inlines string operations only when destination is
8696 known to be aligned at least to 4 byte boundary. This enables
8697 more inlining, increase code size, but may improve performance of
8698 code that depends on fast memcpy, strlen and memset for short
8701 `-momit-leaf-frame-pointer'
8702 Don't keep the frame pointer in a register for leaf functions.
8703 This avoids the instructions to save, set up and restore frame
8704 pointers and makes an extra register available in leaf functions.
8705 The option `-fomit-frame-pointer' removes the frame pointer for
8706 all functions which might make debugging harder.
8708 `-mtls-direct-seg-refs'
8709 `-mno-tls-direct-seg-refs'
8710 Controls whether TLS variables may be accessed with offsets from
8711 the TLS segment register (`%gs' for 32-bit, `%fs' for 64-bit), or
8712 whether the thread base pointer must be added. Whether or not this
8713 is legal depends on the operating system, and whether it maps the
8714 segment to cover the entire TLS area.
8716 For systems that use GNU libc, the default is on.
8718 These `-m' switches are supported in addition to the above on AMD
8719 x86-64 processors in 64-bit environments.
8723 Generate code for a 32-bit or 64-bit environment. The 32-bit
8724 environment sets int, long and pointer to 32 bits and generates
8725 code that runs on any i386 system. The 64-bit environment sets
8726 int to 32 bits and long and pointer to 64 bits and generates code
8727 for AMD's x86-64 architecture.
8730 Do not use a so called red zone for x86-64 code. The red zone is
8731 mandated by the x86-64 ABI, it is a 128-byte area beyond the
8732 location of the stack pointer that will not be modified by signal
8733 or interrupt handlers and therefore can be used for temporary data
8734 without adjusting the stack pointer. The flag `-mno-red-zone'
8735 disables this red zone.
8738 Generate code for the small code model: the program and its
8739 symbols must be linked in the lower 2 GB of the address space.
8740 Pointers are 64 bits. Programs can be statically or dynamically
8741 linked. This is the default code model.
8744 Generate code for the kernel code model. The kernel runs in the
8745 negative 2 GB of the address space. This model has to be used for
8749 Generate code for the medium model: The program is linked in the
8750 lower 2 GB of the address space but symbols can be located
8751 anywhere in the address space. Programs can be statically or
8752 dynamically linked, but building of shared libraries are not
8753 supported with the medium model.
8756 Generate code for the large model: This model makes no assumptions
8757 about addresses and sizes of sections. Currently GCC does not
8758 implement this model.
8761 File: gcc.info, Node: IA-64 Options, Next: M32R/D Options, Prev: i386 and x86-64 Options, Up: Submodel Options
8763 3.17.13 IA-64 Options
8764 ---------------------
8766 These are the `-m' options defined for the Intel IA-64 architecture.
8769 Generate code for a big endian target. This is the default for
8773 Generate code for a little endian target. This is the default for
8778 Generate (or don't) code for the GNU assembler. This is the
8783 Generate (or don't) code for the GNU linker. This is the default.
8786 Generate code that does not use a global pointer register. The
8787 result is not position independent code, and violates the IA-64
8790 `-mvolatile-asm-stop'
8791 `-mno-volatile-asm-stop'
8792 Generate (or don't) a stop bit immediately before and after
8793 volatile asm statements.
8796 `-mno-register-names'
8797 Generate (or don't) `in', `loc', and `out' register names for the
8798 stacked registers. This may make assembler output more readable.
8802 Disable (or enable) optimizations that use the small data section.
8803 This may be useful for working around optimizer bugs.
8806 Generate code that uses a single constant global pointer value.
8807 This is useful when compiling kernel code.
8810 Generate code that is self-relocatable. This implies
8811 `-mconstant-gp'. This is useful when compiling firmware code.
8813 `-minline-float-divide-min-latency'
8814 Generate code for inline divides of floating point values using
8815 the minimum latency algorithm.
8817 `-minline-float-divide-max-throughput'
8818 Generate code for inline divides of floating point values using
8819 the maximum throughput algorithm.
8821 `-minline-int-divide-min-latency'
8822 Generate code for inline divides of integer values using the
8823 minimum latency algorithm.
8825 `-minline-int-divide-max-throughput'
8826 Generate code for inline divides of integer values using the
8827 maximum throughput algorithm.
8829 `-minline-sqrt-min-latency'
8830 Generate code for inline square roots using the minimum latency
8833 `-minline-sqrt-max-throughput'
8834 Generate code for inline square roots using the maximum throughput
8839 Don't (or do) generate assembler code for the DWARF2 line number
8840 debugging info. This may be useful when not using the GNU
8844 `-mno-early-stop-bits'
8845 Allow stop bits to be placed earlier than immediately preceding the
8846 instruction that triggered the stop bit. This can improve
8847 instruction scheduling, but does not always do so.
8849 `-mfixed-range=REGISTER-RANGE'
8850 Generate code treating the given register range as fixed registers.
8851 A fixed register is one that the register allocator can not use.
8852 This is useful when compiling kernel code. A register range is
8853 specified as two registers separated by a dash. Multiple register
8854 ranges can be specified separated by a comma.
8856 `-mtls-size=TLS-SIZE'
8857 Specify bit size of immediate TLS offsets. Valid values are 14,
8860 `-mtune-arch=CPU-TYPE'
8861 Tune the instruction scheduling for a particular CPU, Valid values
8862 are itanium, itanium1, merced, itanium2, and mckinley.
8866 Add support for multithreading using the POSIX threads library.
8867 This option sets flags for both the preprocessor and linker. It
8868 does not affect the thread safety of object code produced by the
8869 compiler or that of libraries supplied with it. These are HP-UX
8874 Generate code for a 32-bit or 64-bit environment. The 32-bit
8875 environment sets int, long and pointer to 32 bits. The 64-bit
8876 environment sets int to 32 bits and long and pointer to 64 bits.
8877 These are HP-UX specific flags.
8881 File: gcc.info, Node: M32R/D Options, Next: M680x0 Options, Prev: IA-64 Options, Up: Submodel Options
8883 3.17.14 M32R/D Options
8884 ----------------------
8886 These `-m' options are defined for Renesas M32R/D architectures:
8889 Generate code for the M32R/2.
8892 Generate code for the M32R/X.
8895 Generate code for the M32R. This is the default.
8898 Assume all objects live in the lower 16MB of memory (so that their
8899 addresses can be loaded with the `ld24' instruction), and assume
8900 all subroutines are reachable with the `bl' instruction. This is
8903 The addressability of a particular object can be set with the
8907 Assume objects may be anywhere in the 32-bit address space (the
8908 compiler will generate `seth/add3' instructions to load their
8909 addresses), and assume all subroutines are reachable with the `bl'
8913 Assume objects may be anywhere in the 32-bit address space (the
8914 compiler will generate `seth/add3' instructions to load their
8915 addresses), and assume subroutines may not be reachable with the
8916 `bl' instruction (the compiler will generate the much slower
8917 `seth/add3/jl' instruction sequence).
8920 Disable use of the small data area. Variables will be put into
8921 one of `.data', `bss', or `.rodata' (unless the `section'
8922 attribute has been specified). This is the default.
8924 The small data area consists of sections `.sdata' and `.sbss'.
8925 Objects may be explicitly put in the small data area with the
8926 `section' attribute using one of these sections.
8929 Put small global and static data in the small data area, but do not
8930 generate special code to reference them.
8933 Put small global and static data in the small data area, and
8934 generate special instructions to reference them.
8937 Put global and static objects less than or equal to NUM bytes into
8938 the small data or bss sections instead of the normal data or bss
8939 sections. The default value of NUM is 8. The `-msdata' option
8940 must be set to one of `sdata' or `use' for this option to have any
8943 All modules should be compiled with the same `-G NUM' value.
8944 Compiling with different values of NUM may or may not work; if it
8945 doesn't the linker will give an error message--incorrect code will
8949 Makes the M32R specific code in the compiler display some
8950 statistics that might help in debugging programs.
8953 Align all loops to a 32-byte boundary.
8956 Do not enforce a 32-byte alignment for loops. This is the default.
8958 `-missue-rate=NUMBER'
8959 Issue NUMBER instructions per cycle. NUMBER can only be 1 or 2.
8961 `-mbranch-cost=NUMBER'
8962 NUMBER can only be 1 or 2. If it is 1 then branches will be
8963 preferred over conditional code, if it is 2, then the opposite will
8966 `-mflush-trap=NUMBER'
8967 Specifies the trap number to use to flush the cache. The default
8968 is 12. Valid numbers are between 0 and 15 inclusive.
8971 Specifies that the cache cannot be flushed by using a trap.
8974 Specifies the name of the operating system function to call to
8975 flush the cache. The default is __flush_cache_, but a function
8976 call will only be used if a trap is not available.
8979 Indicates that there is no OS function for flushing the cache.
8983 File: gcc.info, Node: M680x0 Options, Next: M68hc1x Options, Prev: M32R/D Options, Up: Submodel Options
8985 3.17.15 M680x0 Options
8986 ----------------------
8988 These are the `-m' options defined for the 68000 series. The default
8989 values for these options depends on which style of 68000 was selected
8990 when the compiler was configured; the defaults for the most common
8991 choices are given below.
8995 Generate output for a 68000. This is the default when the
8996 compiler is configured for 68000-based systems.
8998 Use this option for microcontrollers with a 68000 or EC000 core,
8999 including the 68008, 68302, 68306, 68307, 68322, 68328 and 68356.
9003 Generate output for a 68020. This is the default when the
9004 compiler is configured for 68020-based systems.
9007 Generate output containing 68881 instructions for floating point.
9008 This is the default for most 68020 systems unless `--nfp' was
9009 specified when the compiler was configured.
9012 Generate output for a 68030. This is the default when the
9013 compiler is configured for 68030-based systems.
9016 Generate output for a 68040. This is the default when the
9017 compiler is configured for 68040-based systems.
9019 This option inhibits the use of 68881/68882 instructions that have
9020 to be emulated by software on the 68040. Use this option if your
9021 68040 does not have code to emulate those instructions.
9024 Generate output for a 68060. This is the default when the
9025 compiler is configured for 68060-based systems.
9027 This option inhibits the use of 68020 and 68881/68882 instructions
9028 that have to be emulated by software on the 68060. Use this
9029 option if your 68060 does not have code to emulate those
9033 Generate output for a CPU32. This is the default when the
9034 compiler is configured for CPU32-based systems.
9036 Use this option for microcontrollers with a CPU32 or CPU32+ core,
9037 including the 68330, 68331, 68332, 68333, 68334, 68336, 68340,
9038 68341, 68349 and 68360.
9041 Generate output for a 520X "coldfire" family cpu. This is the
9042 default when the compiler is configured for 520X-based systems.
9044 Use this option for microcontroller with a 5200 core, including
9045 the MCF5202, MCF5203, MCF5204 and MCF5202.
9048 Generate output for a 68040, without using any of the new
9049 instructions. This results in code which can run relatively
9050 efficiently on either a 68020/68881 or a 68030 or a 68040. The
9051 generated code does use the 68881 instructions that are emulated
9055 Generate output for a 68060, without using any of the new
9056 instructions. This results in code which can run relatively
9057 efficiently on either a 68020/68881 or a 68030 or a 68040. The
9058 generated code does use the 68881 instructions that are emulated
9062 Generate output containing library calls for floating point.
9063 *Warning:* the requisite libraries are not available for all m68k
9064 targets. Normally the facilities of the machine's usual C
9065 compiler are used, but this can't be done directly in
9066 cross-compilation. You must make your own arrangements to provide
9067 suitable library functions for cross-compilation. The embedded
9068 targets `m68k-*-aout' and `m68k-*-coff' do provide software
9069 floating point support.
9072 Consider type `int' to be 16 bits wide, like `short int'.
9073 Additionally, parameters passed on the stack are also aligned to a
9074 16-bit boundary even on targets whose API mandates promotion to
9078 Do not use the bit-field instructions. The `-m68000', `-mcpu32'
9079 and `-m5200' options imply `-mnobitfield'.
9082 Do use the bit-field instructions. The `-m68020' option implies
9083 `-mbitfield'. This is the default if you use a configuration
9084 designed for a 68020.
9087 Use a different function-calling convention, in which functions
9088 that take a fixed number of arguments return with the `rtd'
9089 instruction, which pops their arguments while returning. This
9090 saves one instruction in the caller since there is no need to pop
9091 the arguments there.
9093 This calling convention is incompatible with the one normally used
9094 on Unix, so you cannot use it if you need to call libraries
9095 compiled with the Unix compiler.
9097 Also, you must provide function prototypes for all functions that
9098 take variable numbers of arguments (including `printf'); otherwise
9099 incorrect code will be generated for calls to those functions.
9101 In addition, seriously incorrect code will result if you call a
9102 function with too many arguments. (Normally, extra arguments are
9103 harmlessly ignored.)
9105 The `rtd' instruction is supported by the 68010, 68020, 68030,
9106 68040, 68060 and CPU32 processors, but not by the 68000 or 5200.
9110 Control whether GCC aligns `int', `long', `long long', `float',
9111 `double', and `long double' variables on a 32-bit boundary
9112 (`-malign-int') or a 16-bit boundary (`-mno-align-int'). Aligning
9113 variables on 32-bit boundaries produces code that runs somewhat
9114 faster on processors with 32-bit busses at the expense of more
9117 *Warning:* if you use the `-malign-int' switch, GCC will align
9118 structures containing the above types differently than most
9119 published application binary interface specifications for the m68k.
9122 Use the pc-relative addressing mode of the 68000 directly, instead
9123 of using a global offset table. At present, this option implies
9124 `-fpic', allowing at most a 16-bit offset for pc-relative
9125 addressing. `-fPIC' is not presently supported with `-mpcrel',
9126 though this could be supported for 68020 and higher processors.
9130 Do not (do) assume that unaligned memory references will be
9131 handled by the system.
9134 Generate code that allows the data segment to be located in a
9135 different area of memory from the text segment. This allows for
9136 execute in place in an environment without virtual memory
9137 management. This option implies `-fPIC'.
9140 Generate code that assumes that the data segment follows the text
9141 segment. This is the default.
9143 `-mid-shared-library'
9144 Generate code that supports shared libraries via the library ID
9145 method. This allows for execute in place and shared libraries in
9146 an environment without virtual memory management. This option
9149 `-mno-id-shared-library'
9150 Generate code that doesn't assume ID based shared libraries are
9151 being used. This is the default.
9153 `-mshared-library-id=n'
9154 Specified the identification number of the ID based shared library
9155 being compiled. Specifying a value of 0 will generate more
9156 compact code, specifying other values will force the allocation of
9157 that number to the current library but is no more space or time
9158 efficient than omitting this option.
9162 File: gcc.info, Node: M68hc1x Options, Next: MCore Options, Prev: M680x0 Options, Up: Submodel Options
9164 3.17.16 M68hc1x Options
9165 -----------------------
9167 These are the `-m' options defined for the 68hc11 and 68hc12
9168 microcontrollers. The default values for these options depends on
9169 which style of microcontroller was selected when the compiler was
9170 configured; the defaults for the most common choices are given below.
9174 Generate output for a 68HC11. This is the default when the
9175 compiler is configured for 68HC11-based systems.
9179 Generate output for a 68HC12. This is the default when the
9180 compiler is configured for 68HC12-based systems.
9184 Generate output for a 68HCS12.
9187 Enable the use of 68HC12 pre and post auto-increment and
9188 auto-decrement addressing modes.
9192 Enable the use of 68HC12 min and max instructions.
9196 Treat all calls as being far away (near). If calls are assumed to
9197 be far away, the compiler will use the `call' instruction to call
9198 a function and the `rtc' instruction for returning.
9201 Consider type `int' to be 16 bits wide, like `short int'.
9203 `-msoft-reg-count=COUNT'
9204 Specify the number of pseudo-soft registers which are used for the
9205 code generation. The maximum number is 32. Using more pseudo-soft
9206 register may or may not result in better code depending on the
9207 program. The default is 4 for 68HC11 and 2 for 68HC12.
9211 File: gcc.info, Node: MCore Options, Next: MIPS Options, Prev: M68hc1x Options, Up: Submodel Options
9213 3.17.17 MCore Options
9214 ---------------------
9216 These are the `-m' options defined for the Motorola M*Core processors.
9220 Inline constants into the code stream if it can be done in two
9221 instructions or less.
9225 Use the divide instruction. (Enabled by default).
9228 `-mno-relax-immediate'
9229 Allow arbitrary sized immediates in bit operations.
9232 `-mno-wide-bitfields'
9233 Always treat bit-fields as int-sized.
9236 `-mno-4byte-functions'
9237 Force all functions to be aligned to a four byte boundary.
9240 `-mno-callgraph-data'
9241 Emit callgraph information.
9245 Prefer word access when reading byte quantities.
9249 Generate code for a little endian target.
9253 Generate code for the 210 processor.
9256 File: gcc.info, Node: MIPS Options, Next: MMIX Options, Prev: MCore Options, Up: Submodel Options
9258 3.17.18 MIPS Options
9259 --------------------
9262 Generate big-endian code.
9265 Generate little-endian code. This is the default for `mips*el-*-*'
9269 Generate code that will run on ARCH, which can be the name of a
9270 generic MIPS ISA, or the name of a particular processor. The ISA
9271 names are: `mips1', `mips2', `mips3', `mips4', `mips32',
9272 `mips32r2', and `mips64'. The processor names are: `4kc', `4kp',
9273 `5kc', `20kc', `m4k', `r2000', `r3000', `r3900', `r4000', `r4400',
9274 `r4600', `r4650', `r6000', `r8000', `rm7000', `rm9000', `orion',
9275 `sb1', `vr4100', `vr4111', `vr4120', `vr4130', `vr4300', `vr5000',
9276 `vr5400' and `vr5500'. The special value `from-abi' selects the
9277 most compatible architecture for the selected ABI (that is,
9278 `mips1' for 32-bit ABIs and `mips3' for 64-bit ABIs).
9280 In processor names, a final `000' can be abbreviated as `k' (for
9281 example, `-march=r2k'). Prefixes are optional, and `vr' may be
9284 GCC defines two macros based on the value of this option. The
9285 first is `_MIPS_ARCH', which gives the name of target
9286 architecture, as a string. The second has the form
9287 `_MIPS_ARCH_FOO', where FOO is the capitalized value of
9288 `_MIPS_ARCH'. For example, `-march=r2000' will set `_MIPS_ARCH'
9289 to `"r2000"' and define the macro `_MIPS_ARCH_R2000'.
9291 Note that the `_MIPS_ARCH' macro uses the processor names given
9292 above. In other words, it will have the full prefix and will not
9293 abbreviate `000' as `k'. In the case of `from-abi', the macro
9294 names the resolved architecture (either `"mips1"' or `"mips3"').
9295 It names the default architecture when no `-march' option is given.
9298 Optimize for ARCH. Among other things, this option controls the
9299 way instructions are scheduled, and the perceived cost of
9300 arithmetic operations. The list of ARCH values is the same as for
9303 When this option is not used, GCC will optimize for the processor
9304 specified by `-march'. By using `-march' and `-mtune' together,
9305 it is possible to generate code that will run on a family of
9306 processors, but optimize the code for one particular member of
9309 `-mtune' defines the macros `_MIPS_TUNE' and `_MIPS_TUNE_FOO',
9310 which work in the same way as the `-march' ones described above.
9313 Equivalent to `-march=mips1'.
9316 Equivalent to `-march=mips2'.
9319 Equivalent to `-march=mips3'.
9322 Equivalent to `-march=mips4'.
9325 Equivalent to `-march=mips32'.
9328 Equivalent to `-march=mips32r2'.
9331 Equivalent to `-march=mips64'.
9335 Use (do not use) the MIPS16 ISA.
9342 Generate code for the given ABI.
9344 Note that the EABI has a 32-bit and a 64-bit variant. GCC normally
9345 generates 64-bit code when you select a 64-bit architecture, but
9346 you can use `-mgp32' to get 32-bit code instead.
9348 For information about the O64 ABI, see
9349 `http://gcc.gnu.org/projects/mipso64-abi.html'.
9353 Generate (do not generate) SVR4-style position-independent code.
9354 `-mabicalls' is the default for SVR4-based systems.
9358 Lift (do not lift) the usual restrictions on the size of the global
9361 GCC normally uses a single instruction to load values from the GOT.
9362 While this is relatively efficient, it will only work if the GOT
9363 is smaller than about 64k. Anything larger will cause the linker
9364 to report an error such as:
9366 relocation truncated to fit: R_MIPS_GOT16 foobar
9368 If this happens, you should recompile your code with `-mxgot'. It
9369 should then work with very large GOTs, although it will also be
9370 less efficient, since it will take three instructions to fetch the
9371 value of a global symbol.
9373 Note that some linkers can create multiple GOTs. If you have such
9374 a linker, you should only need to use `-mxgot' when a single object
9375 file accesses more than 64k's worth of GOT entries. Very few do.
9377 These options have no effect unless GCC is generating position
9381 Assume that general-purpose registers are 32 bits wide.
9384 Assume that general-purpose registers are 64 bits wide.
9387 Assume that floating-point registers are 32 bits wide.
9390 Assume that floating-point registers are 64 bits wide.
9393 Use floating-point coprocessor instructions.
9396 Do not use floating-point coprocessor instructions. Implement
9397 floating-point calculations using library calls instead.
9400 Assume that the floating-point coprocessor only supports
9401 single-precision operations.
9404 Assume that the floating-point coprocessor supports
9405 double-precision operations. This is the default.
9408 `-mno-paired-single'
9409 Use (do not use) paired-single floating-point instructions. *Note
9410 MIPS Paired-Single Support::. This option can only be used when
9411 generating 64-bit code and requires hardware floating-point
9412 support to be enabled.
9416 Use (do not use) the MIPS-3D ASE. *Note MIPS-3D Built-in
9417 Functions::. The option `-mips3d' implies `-mpaired-single'.
9420 Force `int' and `long' types to be 64 bits wide. See `-mlong32'
9421 for an explanation of the default and the way that the pointer
9424 This option has been deprecated and will be removed in a future
9428 Force `long' types to be 64 bits wide. See `-mlong32' for an
9429 explanation of the default and the way that the pointer size is
9433 Force `long', `int', and pointer types to be 32 bits wide.
9435 The default size of `int's, `long's and pointers depends on the
9436 ABI. All the supported ABIs use 32-bit `int's. The n64 ABI uses
9437 64-bit `long's, as does the 64-bit EABI; the others use 32-bit
9438 `long's. Pointers are the same size as `long's, or the same size
9439 as integer registers, whichever is smaller.
9443 Assume (do not assume) that all symbols have 32-bit values,
9444 regardless of the selected ABI. This option is useful in
9445 combination with `-mabi=64' and `-mno-abicalls' because it allows
9446 GCC to generate shorter and faster references to symbolic
9450 Put global and static items less than or equal to NUM bytes into
9451 the small data or bss section instead of the normal data or bss
9452 section. This allows the data to be accessed using a single
9455 All modules should be compiled with the same `-G NUM' value.
9458 `-mno-embedded-data'
9459 Allocate variables to the read-only data section first if
9460 possible, then next in the small data section if possible,
9461 otherwise in data. This gives slightly slower code than the
9462 default, but reduces the amount of RAM required when executing,
9463 and thus may be preferred for some embedded systems.
9465 `-muninit-const-in-rodata'
9466 `-mno-uninit-const-in-rodata'
9467 Put uninitialized `const' variables in the read-only data section.
9468 This option is only meaningful in conjunction with
9472 `-mno-split-addresses'
9473 Enable (disable) use of the `%hi()' and `%lo()' assembler
9474 relocation operators. This option has been superseded by
9475 `-mexplicit-relocs' but is retained for backwards compatibility.
9478 `-mno-explicit-relocs'
9479 Use (do not use) assembler relocation operators when dealing with
9480 symbolic addresses. The alternative, selected by
9481 `-mno-explicit-relocs', is to use assembler macros instead.
9483 `-mexplicit-relocs' is the default if GCC was configured to use an
9484 assembler that supports relocation operators.
9486 `-mcheck-zero-division'
9487 `-mno-check-zero-division'
9488 Trap (do not trap) on integer division by zero. The default is
9489 `-mcheck-zero-division'.
9493 MIPS systems check for division by zero by generating either a
9494 conditional trap or a break instruction. Using traps results in
9495 smaller code, but is only supported on MIPS II and later. Also,
9496 some versions of the Linux kernel have a bug that prevents trap
9497 from generating the proper signal (`SIGFPE'). Use
9498 `-mdivide-traps' to allow conditional traps on architectures that
9499 support them and `-mdivide-breaks' to force the use of breaks.
9501 The default is usually `-mdivide-traps', but this can be
9502 overridden at configure time using `--with-divide=breaks'.
9503 Divide-by-zero checks can be completely disabled using
9504 `-mno-check-zero-division'.
9508 Force (do not force) the use of `memcpy()' for non-trivial block
9509 moves. The default is `-mno-memcpy', which allows GCC to inline
9510 most constant-sized copies.
9514 Disable (do not disable) use of the `jal' instruction. Calling
9515 functions using `jal' is more efficient but requires the caller
9516 and callee to be in the same 256 megabyte segment.
9518 This option has no effect on abicalls code. The default is
9523 Enable (disable) use of the `mad', `madu' and `mul' instructions,
9524 as provided by the R4650 ISA.
9528 Enable (disable) use of the floating point multiply-accumulate
9529 instructions, when they are available. The default is
9532 When multiply-accumulate instructions are used, the intermediate
9533 product is calculated to infinite precision and is not subject to
9534 the FCSR Flush to Zero bit. This may be undesirable in some
9538 Tell the MIPS assembler to not run its preprocessor over user
9539 assembler files (with a `.s' suffix) when assembling them.
9543 Work around certain R4000 CPU errata:
9544 - A double-word or a variable shift may give an incorrect
9545 result if executed immediately after starting an integer
9548 - A double-word or a variable shift may give an incorrect
9549 result if executed while an integer multiplication is in
9552 - An integer division may give an incorrect result if started
9553 in a delay slot of a taken branch or a jump.
9557 Work around certain R4400 CPU errata:
9558 - A double-word or a variable shift may give an incorrect
9559 result if executed immediately after starting an integer
9564 Work around certain VR4120 errata:
9565 - `dmultu' does not always produce the correct result.
9567 - `div' and `ddiv' do not always produce the correct result if
9568 one of the operands is negative.
9569 The workarounds for the division errata rely on special functions
9570 in `libgcc.a'. At present, these functions are only provided by
9571 the `mips64vr*-elf' configurations.
9573 Other VR4120 errata require a nop to be inserted between certain
9574 pairs of instructions. These errata are handled by the assembler,
9578 Work around the VR4130 `mflo'/`mfhi' errata. The workarounds are
9579 implemented by the assembler rather than by GCC, although GCC will
9580 avoid using `mflo' and `mfhi' if the VR4130 `macc', `macchi',
9581 `dmacc' and `dmacchi' instructions are available instead.
9585 Work around certain SB-1 CPU core errata. (This flag currently
9586 works around the SB-1 revision 2 "F1" and "F2" floating point
9591 Specifies the function to call to flush the I and D caches, or to
9592 not call any such function. If called, the function must take the
9593 same arguments as the common `_flush_func()', that is, the address
9594 of the memory range for which the cache is being flushed, the size
9595 of the memory range, and the number 3 (to flush both caches). The
9596 default depends on the target GCC was configured for, but commonly
9597 is either `_flush_func' or `__cpu_flush'.
9600 `-mno-branch-likely'
9601 Enable or disable use of Branch Likely instructions, regardless of
9602 the default for the selected architecture. By default, Branch
9603 Likely instructions may be generated if they are supported by the
9604 selected architecture. An exception is for the MIPS32 and MIPS64
9605 architectures and processors which implement those architectures;
9606 for those, Branch Likely instructions will not be generated by
9607 default because the MIPS32 and MIPS64 architectures specifically
9608 deprecate their use.
9611 `-mno-fp-exceptions'
9612 Specifies whether FP exceptions are enabled. This affects how we
9613 schedule FP instructions for some processors. The default is that
9614 FP exceptions are enabled.
9616 For instance, on the SB-1, if FP exceptions are disabled, and we
9617 are emitting 64-bit code, then we can use both FP pipes.
9618 Otherwise, we can only use one FP pipe.
9622 The VR4130 pipeline is two-way superscalar, but can only issue two
9623 instructions together if the first one is 8-byte aligned. When
9624 this option is enabled, GCC will align pairs of instructions that
9625 it thinks should execute in parallel.
9627 This option only has an effect when optimizing for the VR4130. It
9628 normally makes code faster, but at the expense of making it bigger.
9629 It is enabled by default at optimization level `-O3'.
9632 File: gcc.info, Node: MMIX Options, Next: MN10300 Options, Prev: MIPS Options, Up: Submodel Options
9634 3.17.19 MMIX Options
9635 --------------------
9637 These options are defined for the MMIX:
9641 Specify that intrinsic library functions are being compiled,
9642 passing all values in registers, no matter the size.
9646 Generate floating-point comparison instructions that compare with
9647 respect to the `rE' epsilon register.
9651 Generate code that passes function parameters and return values
9652 that (in the called function) are seen as registers `$0' and up,
9653 as opposed to the GNU ABI which uses global registers `$231' and
9658 When reading data from memory in sizes shorter than 64 bits, use
9659 (do not use) zero-extending load instructions by default, rather
9660 than sign-extending ones.
9664 Make the result of a division yielding a remainder have the same
9665 sign as the divisor. With the default, `-mno-knuthdiv', the sign
9666 of the remainder follows the sign of the dividend. Both methods
9667 are arithmetically valid, the latter being almost exclusively used.
9669 `-mtoplevel-symbols'
9670 `-mno-toplevel-symbols'
9671 Prepend (do not prepend) a `:' to all global symbols, so the
9672 assembly code can be used with the `PREFIX' assembly directive.
9675 Generate an executable in the ELF format, rather than the default
9676 `mmo' format used by the `mmix' simulator.
9679 `-mno-branch-predict'
9680 Use (do not use) the probable-branch instructions, when static
9681 branch prediction indicates a probable branch.
9684 `-mno-base-addresses'
9685 Generate (do not generate) code that uses _base addresses_. Using
9686 a base address automatically generates a request (handled by the
9687 assembler and the linker) for a constant to be set up in a global
9688 register. The register is used for one or more base address
9689 requests within the range 0 to 255 from the value held in the
9690 register. The generally leads to short and fast code, but the
9691 number of different data items that can be addressed is limited.
9692 This means that a program that uses lots of static data may
9693 require `-mno-base-addresses'.
9697 Force (do not force) generated code to have a single exit point in
9701 File: gcc.info, Node: MN10300 Options, Next: NS32K Options, Prev: MMIX Options, Up: Submodel Options
9703 3.17.20 MN10300 Options
9704 -----------------------
9706 These `-m' options are defined for Matsushita MN10300 architectures:
9709 Generate code to avoid bugs in the multiply instructions for the
9710 MN10300 processors. This is the default.
9713 Do not generate code to avoid bugs in the multiply instructions
9714 for the MN10300 processors.
9717 Generate code which uses features specific to the AM33 processor.
9720 Do not generate code which uses features specific to the AM33
9721 processor. This is the default.
9724 Do not link in the C run-time initialization object file.
9727 Indicate to the linker that it should perform a relaxation
9728 optimization pass to shorten branches, calls and absolute memory
9729 addresses. This option only has an effect when used on the
9730 command line for the final link step.
9732 This option makes symbolic debugging impossible.
9735 File: gcc.info, Node: NS32K Options, Next: PDP-11 Options, Prev: MN10300 Options, Up: Submodel Options
9737 3.17.21 NS32K Options
9738 ---------------------
9740 These are the `-m' options defined for the 32000 series. The default
9741 values for these options depends on which style of 32000 was selected
9742 when the compiler was configured; the defaults for the most common
9743 choices are given below.
9747 Generate output for a 32032. This is the default when the
9748 compiler is configured for 32032 and 32016 based systems.
9752 Generate output for a 32332. This is the default when the
9753 compiler is configured for 32332-based systems.
9757 Generate output for a 32532. This is the default when the
9758 compiler is configured for 32532-based systems.
9761 Generate output containing 32081 instructions for floating point.
9762 This is the default for all systems.
9765 Generate output containing 32381 instructions for floating point.
9766 This also implies `-m32081'. The 32381 is only compatible with
9767 the 32332 and 32532 cpus. This is the default for the
9768 pc532-netbsd configuration.
9771 Try and generate multiply-add floating point instructions `polyF'
9772 and `dotF'. This option is only available if the `-m32381' option
9773 is in effect. Using these instructions requires changes to
9774 register allocation which generally has a negative impact on
9775 performance. This option should only be enabled when compiling
9776 code particularly likely to make heavy use of multiply-add
9780 Do not try and generate multiply-add floating point instructions
9781 `polyF' and `dotF'. This is the default on all platforms.
9784 Generate output containing library calls for floating point.
9785 *Warning:* the requisite libraries may not be available.
9789 Control whether or not the compiler uses IEEE floating point
9790 comparisons. These handle correctly the case where the result of a
9791 comparison is unordered. *Warning:* the requisite kernel support
9792 may not be available.
9795 Do not use the bit-field instructions. On some machines it is
9796 faster to use shifting and masking operations. This is the
9797 default for the pc532.
9800 Do use the bit-field instructions. This is the default for all
9801 platforms except the pc532.
9804 Use a different function-calling convention, in which functions
9805 that take a fixed number of arguments return pop their arguments
9806 on return with the `ret' instruction.
9808 This calling convention is incompatible with the one normally used
9809 on Unix, so you cannot use it if you need to call libraries
9810 compiled with the Unix compiler.
9812 Also, you must provide function prototypes for all functions that
9813 take variable numbers of arguments (including `printf'); otherwise
9814 incorrect code will be generated for calls to those functions.
9816 In addition, seriously incorrect code will result if you call a
9817 function with too many arguments. (Normally, extra arguments are
9818 harmlessly ignored.)
9820 This option takes its name from the 680x0 `rtd' instruction.
9823 Use a different function-calling convention where the first two
9824 arguments are passed in registers.
9826 This calling convention is incompatible with the one normally used
9827 on Unix, so you cannot use it if you need to call libraries
9828 compiled with the Unix compiler.
9831 Do not pass any arguments in registers. This is the default for
9835 It is OK to use the sb as an index register which is always loaded
9836 with zero. This is the default for the pc532-netbsd target.
9839 The sb register is not available for use or has not been
9840 initialized to zero by the run time system. This is the default
9841 for all targets except the pc532-netbsd. It is also implied
9842 whenever `-mhimem' or `-fpic' is set.
9845 Many ns32000 series addressing modes use displacements of up to
9846 512MB. If an address is above 512MB then displacements from zero
9847 can not be used. This option causes code to be generated which
9848 can be loaded above 512MB. This may be useful for operating
9849 systems or ROM code.
9852 Assume code will be loaded in the first 512MB of virtual address
9853 space. This is the default for all platforms.
9857 File: gcc.info, Node: PDP-11 Options, Next: PowerPC Options, Prev: NS32K Options, Up: Submodel Options
9859 3.17.22 PDP-11 Options
9860 ----------------------
9862 These options are defined for the PDP-11:
9865 Use hardware FPP floating point. This is the default. (FIS
9866 floating point on the PDP-11/40 is not supported.)
9869 Do not use hardware floating point.
9872 Return floating-point results in ac0 (fr0 in Unix assembler
9876 Return floating-point results in memory. This is the default.
9879 Generate code for a PDP-11/40.
9882 Generate code for a PDP-11/45. This is the default.
9885 Generate code for a PDP-11/10.
9888 Use inline `movmemhi' patterns for copying memory. This is the
9892 Do not use inline `movmemhi' patterns for copying memory.
9896 Use 16-bit `int'. This is the default.
9904 Use 64-bit `float'. This is the default.
9911 Use `abshi2' pattern. This is the default.
9914 Do not use `abshi2' pattern.
9916 `-mbranch-expensive'
9917 Pretend that branches are expensive. This is for experimenting
9918 with code generation only.
9921 Do not pretend that branches are expensive. This is the default.
9924 Generate code for a system with split I&D.
9927 Generate code for a system without split I&D. This is the default.
9930 Use Unix assembler syntax. This is the default when configured for
9934 Use DEC assembler syntax. This is the default when configured for
9935 any PDP-11 target other than `pdp11-*-bsd'.
9938 File: gcc.info, Node: PowerPC Options, Next: RS/6000 and PowerPC Options, Prev: PDP-11 Options, Up: Submodel Options
9940 3.17.23 PowerPC Options
9941 -----------------------
9943 These are listed under *Note RS/6000 and PowerPC Options::.
9946 File: gcc.info, Node: RS/6000 and PowerPC Options, Next: S/390 and zSeries Options, Prev: PowerPC Options, Up: Submodel Options
9948 3.17.24 IBM RS/6000 and PowerPC Options
9949 ---------------------------------------
9951 These `-m' options are defined for the IBM RS/6000 and PowerPC:
9959 `-mno-powerpc-gpopt'
9961 `-mno-powerpc-gfxopt'
9964 GCC supports two related instruction set architectures for the
9965 RS/6000 and PowerPC. The "POWER" instruction set are those
9966 instructions supported by the `rios' chip set used in the original
9967 RS/6000 systems and the "PowerPC" instruction set is the
9968 architecture of the Motorola MPC5xx, MPC6xx, MPC8xx
9969 microprocessors, and the IBM 4xx microprocessors.
9971 Neither architecture is a subset of the other. However there is a
9972 large common subset of instructions supported by both. An MQ
9973 register is included in processors supporting the POWER
9976 You use these options to specify which instructions are available
9977 on the processor you are using. The default value of these
9978 options is determined when configuring GCC. Specifying the
9979 `-mcpu=CPU_TYPE' overrides the specification of these options. We
9980 recommend you use the `-mcpu=CPU_TYPE' option rather than the
9981 options listed above.
9983 The `-mpower' option allows GCC to generate instructions that are
9984 found only in the POWER architecture and to use the MQ register.
9985 Specifying `-mpower2' implies `-power' and also allows GCC to
9986 generate instructions that are present in the POWER2 architecture
9987 but not the original POWER architecture.
9989 The `-mpowerpc' option allows GCC to generate instructions that
9990 are found only in the 32-bit subset of the PowerPC architecture.
9991 Specifying `-mpowerpc-gpopt' implies `-mpowerpc' and also allows
9992 GCC to use the optional PowerPC architecture instructions in the
9993 General Purpose group, including floating-point square root.
9994 Specifying `-mpowerpc-gfxopt' implies `-mpowerpc' and also allows
9995 GCC to use the optional PowerPC architecture instructions in the
9996 Graphics group, including floating-point select.
9998 The `-mpowerpc64' option allows GCC to generate the additional
9999 64-bit instructions that are found in the full PowerPC64
10000 architecture and to treat GPRs as 64-bit, doubleword quantities.
10001 GCC defaults to `-mno-powerpc64'.
10003 If you specify both `-mno-power' and `-mno-powerpc', GCC will use
10004 only the instructions in the common subset of both architectures
10005 plus some special AIX common-mode calls, and will not use the MQ
10006 register. Specifying both `-mpower' and `-mpowerpc' permits GCC
10007 to use any instruction from either architecture and to allow use
10008 of the MQ register; specify this for the Motorola MPC601.
10012 Select which mnemonics to use in the generated assembler code.
10013 With `-mnew-mnemonics', GCC uses the assembler mnemonics defined
10014 for the PowerPC architecture. With `-mold-mnemonics' it uses the
10015 assembler mnemonics defined for the POWER architecture.
10016 Instructions defined in only one architecture have only one
10017 mnemonic; GCC uses that mnemonic irrespective of which of these
10018 options is specified.
10020 GCC defaults to the mnemonics appropriate for the architecture in
10021 use. Specifying `-mcpu=CPU_TYPE' sometimes overrides the value of
10022 these option. Unless you are building a cross-compiler, you
10023 should normally not specify either `-mnew-mnemonics' or
10024 `-mold-mnemonics', but should instead accept the default.
10027 Set architecture type, register usage, choice of mnemonics, and
10028 instruction scheduling parameters for machine type CPU_TYPE.
10029 Supported values for CPU_TYPE are `401', `403', `405', `405fp',
10030 `440', `440fp', `505', `601', `602', `603', `603e', `604', `604e',
10031 `620', `630', `740', `7400', `7450', `750', `801', `821', `823',
10032 `860', `970', `8540', `common', `ec603e', `G3', `G4', `G5',
10033 `power', `power2', `power3', `power4', `power5', `powerpc',
10034 `powerpc64', `rios', `rios1', `rios2', `rsc', and `rs64a'.
10036 `-mcpu=common' selects a completely generic processor. Code
10037 generated under this option will run on any POWER or PowerPC
10038 processor. GCC will use only the instructions in the common
10039 subset of both architectures, and will not use the MQ register.
10040 GCC assumes a generic processor model for scheduling purposes.
10042 `-mcpu=power', `-mcpu=power2', `-mcpu=powerpc', and
10043 `-mcpu=powerpc64' specify generic POWER, POWER2, pure 32-bit
10044 PowerPC (i.e., not MPC601), and 64-bit PowerPC architecture machine
10045 types, with an appropriate, generic processor model assumed for
10046 scheduling purposes.
10048 The other options specify a specific processor. Code generated
10049 under those options will run best on that processor, and may not
10050 run at all on others.
10052 The `-mcpu' options automatically enable or disable the following
10053 options: `-maltivec', `-mhard-float', `-mmfcrf', `-mmultiple',
10054 `-mnew-mnemonics', `-mpower', `-mpower2', `-mpowerpc64',
10055 `-mpowerpc-gpopt', `-mpowerpc-gfxopt', `-mstring'. The particular
10056 options set for any particular CPU will vary between compiler
10057 versions, depending on what setting seems to produce optimal code
10058 for that CPU; it doesn't necessarily reflect the actual hardware's
10059 capabilities. If you wish to set an individual option to a
10060 particular value, you may specify it after the `-mcpu' option,
10061 like `-mcpu=970 -mno-altivec'.
10063 On AIX, the `-maltivec' and `-mpowerpc64' options are not enabled
10064 or disabled by the `-mcpu' option at present, since AIX does not
10065 have full support for these options. You may still enable or
10066 disable them individually if you're sure it'll work in your
10070 Set the instruction scheduling parameters for machine type
10071 CPU_TYPE, but do not set the architecture type, register usage, or
10072 choice of mnemonics, as `-mcpu=CPU_TYPE' would. The same values
10073 for CPU_TYPE are used for `-mtune' as for `-mcpu'. If both are
10074 specified, the code generated will use the architecture,
10075 registers, and mnemonics set by `-mcpu', but the scheduling
10076 parameters set by `-mtune'.
10080 Generate code that uses (does not use) AltiVec instructions, and
10081 also enable the use of built-in functions that allow more direct
10082 access to the AltiVec instruction set. You may also need to set
10083 `-mabi=altivec' to adjust the current ABI with AltiVec ABI
10087 Extend the current ABI with SPE ABI extensions. This does not
10088 change the default ABI, instead it adds the SPE ABI extensions to
10092 Disable Booke SPE ABI extensions for the current ABI.
10096 This switch enables or disables the generation of ISEL
10101 This switch enables or disables the generation of SPE simd
10104 `-mfloat-gprs=YES/SINGLE/DOUBLE/NO'
10106 This switch enables or disables the generation of floating point
10107 operations on the general purpose registers for architectures that
10110 The argument YES or SINGLE enables the use of single-precision
10111 floating point operations.
10113 The argument DOUBLE enables the use of single and double-precision
10114 floating point operations.
10116 The argument NO disables floating point operations on the general
10119 This option is currently only available on the MPC854x.
10123 Generate code for 32-bit or 64-bit environments of Darwin and SVR4
10124 targets (including GNU/Linux). The 32-bit environment sets int,
10125 long and pointer to 32 bits and generates code that runs on any
10126 PowerPC variant. The 64-bit environment sets int to 32 bits and
10127 long and pointer to 64 bits, and generates code for PowerPC64, as
10134 Modify generation of the TOC (Table Of Contents), which is created
10135 for every executable file. The `-mfull-toc' option is selected by
10136 default. In that case, GCC will allocate at least one TOC entry
10137 for each unique non-automatic variable reference in your program.
10138 GCC will also place floating-point constants in the TOC. However,
10139 only 16,384 entries are available in the TOC.
10141 If you receive a linker error message that saying you have
10142 overflowed the available TOC space, you can reduce the amount of
10143 TOC space used with the `-mno-fp-in-toc' and `-mno-sum-in-toc'
10144 options. `-mno-fp-in-toc' prevents GCC from putting floating-point
10145 constants in the TOC and `-mno-sum-in-toc' forces GCC to generate
10146 code to calculate the sum of an address and a constant at run-time
10147 instead of putting that sum into the TOC. You may specify one or
10148 both of these options. Each causes GCC to produce very slightly
10149 slower and larger code at the expense of conserving TOC space.
10151 If you still run out of space in the TOC even when you specify
10152 both of these options, specify `-mminimal-toc' instead. This
10153 option causes GCC to make only one TOC entry for every file. When
10154 you specify this option, GCC will produce code that is slower and
10155 larger but which uses extremely little TOC space. You may wish to
10156 use this option only on files that contain less frequently
10161 Enable 64-bit AIX ABI and calling convention: 64-bit pointers,
10162 64-bit `long' type, and the infrastructure needed to support them.
10163 Specifying `-maix64' implies `-mpowerpc64' and `-mpowerpc', while
10164 `-maix32' disables the 64-bit ABI and implies `-mno-powerpc64'.
10165 GCC defaults to `-maix32'.
10169 Produce code that conforms more closely to IBM XLC semantics when
10170 using AIX-compatible ABI. Pass floating-point arguments to
10171 prototyped functions beyond the register save area (RSA) on the
10172 stack in addition to argument FPRs. Do not assume that most
10173 significant double in 128 bit long double value is properly
10174 rounded when comparing values.
10176 The AIX calling convention was extended but not initially
10177 documented to handle an obscure K&R C case of calling a function
10178 that takes the address of its arguments with fewer arguments than
10179 declared. AIX XL compilers access floating point arguments which
10180 do not fit in the RSA from the stack when a subroutine is compiled
10181 without optimization. Because always storing floating-point
10182 arguments on the stack is inefficient and rarely needed, this
10183 option is not enabled by default and only is necessary when
10184 calling subroutines compiled by AIX XL compilers without
10188 Support "IBM RS/6000 SP" "Parallel Environment" (PE). Link an
10189 application written to use message passing with special startup
10190 code to enable the application to run. The system must have PE
10191 installed in the standard location (`/usr/lpp/ppe.poe/'), or the
10192 `specs' file must be overridden with the `-specs=' option to
10193 specify the appropriate directory location. The Parallel
10194 Environment does not support threads, so the `-mpe' option and the
10195 `-pthread' option are incompatible.
10199 On AIX, 32-bit Darwin, and 64-bit PowerPC GNU/Linux, the option
10200 `-malign-natural' overrides the ABI-defined alignment of larger
10201 types, such as floating-point doubles, on their natural size-based
10202 boundary. The option `-malign-power' instructs GCC to follow the
10203 ABI-specified alignment rules. GCC defaults to the standard
10204 alignment defined in the ABI.
10206 On 64-bit Darwin, natural alignment is the default, and
10207 `-malign-power' is not supported.
10211 Generate code that does not use (uses) the floating-point register
10212 set. Software floating point emulation is provided if you use the
10213 `-msoft-float' option, and pass the option to GCC when linking.
10217 Generate code that uses (does not use) the load multiple word
10218 instructions and the store multiple word instructions. These
10219 instructions are generated by default on POWER systems, and not
10220 generated on PowerPC systems. Do not use `-mmultiple' on little
10221 endian PowerPC systems, since those instructions do not work when
10222 the processor is in little endian mode. The exceptions are PPC740
10223 and PPC750 which permit the instructions usage in little endian
10228 Generate code that uses (does not use) the load string instructions
10229 and the store string word instructions to save multiple registers
10230 and do small block moves. These instructions are generated by
10231 default on POWER systems, and not generated on PowerPC systems.
10232 Do not use `-mstring' on little endian PowerPC systems, since those
10233 instructions do not work when the processor is in little endian
10234 mode. The exceptions are PPC740 and PPC750 which permit the
10235 instructions usage in little endian mode.
10239 Generate code that uses (does not use) the load or store
10240 instructions that update the base register to the address of the
10241 calculated memory location. These instructions are generated by
10242 default. If you use `-mno-update', there is a small window
10243 between the time that the stack pointer is updated and the address
10244 of the previous frame is stored, which means code that walks the
10245 stack frame across interrupts or signals may get corrupted data.
10249 Generate code that uses (does not use) the floating point multiply
10250 and accumulate instructions. These instructions are generated by
10251 default if hardware floating is used.
10255 On System V.4 and embedded PowerPC systems do not (do) force
10256 structures and unions that contain bit-fields to be aligned to the
10257 base type of the bit-field.
10259 For example, by default a structure containing nothing but 8
10260 `unsigned' bit-fields of length 1 would be aligned to a 4 byte
10261 boundary and have a size of 4 bytes. By using `-mno-bit-align',
10262 the structure would be aligned to a 1 byte boundary and be one
10265 `-mno-strict-align'
10267 On System V.4 and embedded PowerPC systems do not (do) assume that
10268 unaligned memory references will be handled by the system.
10272 On embedded PowerPC systems generate code that allows (does not
10273 allow) the program to be relocated to a different address at
10274 runtime. If you use `-mrelocatable' on any module, all objects
10275 linked together must be compiled with `-mrelocatable' or
10276 `-mrelocatable-lib'.
10278 `-mrelocatable-lib'
10279 `-mno-relocatable-lib'
10280 On embedded PowerPC systems generate code that allows (does not
10281 allow) the program to be relocated to a different address at
10282 runtime. Modules compiled with `-mrelocatable-lib' can be linked
10283 with either modules compiled without `-mrelocatable' and
10284 `-mrelocatable-lib' or with modules compiled with the
10285 `-mrelocatable' options.
10289 On System V.4 and embedded PowerPC systems do not (do) assume that
10290 register 2 contains a pointer to a global area pointing to the
10291 addresses used in the program.
10295 On System V.4 and embedded PowerPC systems compile code for the
10296 processor in little endian mode. The `-mlittle-endian' option is
10297 the same as `-mlittle'.
10301 On System V.4 and embedded PowerPC systems compile code for the
10302 processor in big endian mode. The `-mbig-endian' option is the
10306 On Darwin and Mac OS X systems, compile code so that it is not
10307 relocatable, but that its external references are relocatable. The
10308 resulting code is suitable for applications, but not shared
10311 `-mprioritize-restricted-insns=PRIORITY'
10312 This option controls the priority that is assigned to
10313 dispatch-slot restricted instructions during the second scheduling
10314 pass. The argument PRIORITY takes the value 0/1/2 to assign
10315 NO/HIGHEST/SECOND-HIGHEST priority to dispatch slot restricted
10318 `-msched-costly-dep=DEPENDENCE_TYPE'
10319 This option controls which dependences are considered costly by
10320 the target during instruction scheduling. The argument
10321 DEPENDENCE_TYPE takes one of the following values: NO: no
10322 dependence is costly, ALL: all dependences are costly,
10323 TRUE_STORE_TO_LOAD: a true dependence from store to load is costly,
10324 STORE_TO_LOAD: any dependence from store to load is costly,
10325 NUMBER: any dependence which latency >= NUMBER is costly.
10327 `-minsert-sched-nops=SCHEME'
10328 This option controls which nop insertion scheme will be used during
10329 the second scheduling pass. The argument SCHEME takes one of the
10330 following values: NO: Don't insert nops. PAD: Pad with nops any
10331 dispatch group which has vacant issue slots, according to the
10332 scheduler's grouping. REGROUP_EXACT: Insert nops to force costly
10333 dependent insns into separate groups. Insert exactly as many nops
10334 as needed to force an insn to a new group, according to the
10335 estimated processor grouping. NUMBER: Insert nops to force costly
10336 dependent insns into separate groups. Insert NUMBER nops to force
10337 an insn to a new group.
10340 On System V.4 and embedded PowerPC systems compile code using
10341 calling conventions that adheres to the March 1995 draft of the
10342 System V Application Binary Interface, PowerPC processor
10343 supplement. This is the default unless you configured GCC using
10344 `powerpc-*-eabiaix'.
10347 Specify both `-mcall-sysv' and `-meabi' options.
10349 `-mcall-sysv-noeabi'
10350 Specify both `-mcall-sysv' and `-mno-eabi' options.
10353 On System V.4 and embedded PowerPC systems compile code for the
10354 Solaris operating system.
10357 On System V.4 and embedded PowerPC systems compile code for the
10358 Linux-based GNU system.
10361 On System V.4 and embedded PowerPC systems compile code for the
10362 Hurd-based GNU system.
10365 On System V.4 and embedded PowerPC systems compile code for the
10366 NetBSD operating system.
10368 `-maix-struct-return'
10369 Return all structures in memory (as specified by the AIX ABI).
10371 `-msvr4-struct-return'
10372 Return structures smaller than 8 bytes in registers (as specified
10376 Extend the current ABI with AltiVec ABI extensions. This does not
10377 change the default ABI, instead it adds the AltiVec ABI extensions
10378 to the current ABI.
10381 Disable AltiVec ABI extensions for the current ABI.
10385 On System V.4 and embedded PowerPC systems assume that all calls to
10386 variable argument functions are properly prototyped. Otherwise,
10387 the compiler must insert an instruction before every non
10388 prototyped call to set or clear bit 6 of the condition code
10389 register (CR) to indicate whether floating point values were
10390 passed in the floating point registers in case the function takes
10391 a variable arguments. With `-mprototype', only calls to
10392 prototyped variable argument functions will set or clear the bit.
10395 On embedded PowerPC systems, assume that the startup module is
10396 called `sim-crt0.o' and that the standard C libraries are
10397 `libsim.a' and `libc.a'. This is the default for
10398 `powerpc-*-eabisim'. configurations.
10401 On embedded PowerPC systems, assume that the startup module is
10402 called `crt0.o' and the standard C libraries are `libmvme.a' and
10406 On embedded PowerPC systems, assume that the startup module is
10407 called `crt0.o' and the standard C libraries are `libads.a' and
10411 On embedded PowerPC systems, assume that the startup module is
10412 called `crt0.o' and the standard C libraries are `libyk.a' and
10416 On System V.4 and embedded PowerPC systems, specify that you are
10417 compiling for a VxWorks system.
10420 Specify that you are compiling for the WindISS simulation
10424 On embedded PowerPC systems, set the PPC_EMB bit in the ELF flags
10425 header to indicate that `eabi' extended relocations are used.
10429 On System V.4 and embedded PowerPC systems do (do not) adhere to
10430 the Embedded Applications Binary Interface (eabi) which is a set of
10431 modifications to the System V.4 specifications. Selecting `-meabi'
10432 means that the stack is aligned to an 8 byte boundary, a function
10433 `__eabi' is called to from `main' to set up the eabi environment,
10434 and the `-msdata' option can use both `r2' and `r13' to point to
10435 two separate small data areas. Selecting `-mno-eabi' means that
10436 the stack is aligned to a 16 byte boundary, do not call an
10437 initialization function from `main', and the `-msdata' option will
10438 only use `r13' to point to a single small data area. The `-meabi'
10439 option is on by default if you configured GCC using one of the
10440 `powerpc*-*-eabi*' options.
10443 On System V.4 and embedded PowerPC systems, put small initialized
10444 `const' global and static data in the `.sdata2' section, which is
10445 pointed to by register `r2'. Put small initialized non-`const'
10446 global and static data in the `.sdata' section, which is pointed
10447 to by register `r13'. Put small uninitialized global and static
10448 data in the `.sbss' section, which is adjacent to the `.sdata'
10449 section. The `-msdata=eabi' option is incompatible with the
10450 `-mrelocatable' option. The `-msdata=eabi' option also sets the
10454 On System V.4 and embedded PowerPC systems, put small global and
10455 static data in the `.sdata' section, which is pointed to by
10456 register `r13'. Put small uninitialized global and static data in
10457 the `.sbss' section, which is adjacent to the `.sdata' section.
10458 The `-msdata=sysv' option is incompatible with the `-mrelocatable'
10463 On System V.4 and embedded PowerPC systems, if `-meabi' is used,
10464 compile code the same as `-msdata=eabi', otherwise compile code the
10465 same as `-msdata=sysv'.
10468 On System V.4 and embedded PowerPC systems, put small global and
10469 static data in the `.sdata' section. Put small uninitialized
10470 global and static data in the `.sbss' section. Do not use
10471 register `r13' to address small data however. This is the default
10472 behavior unless other `-msdata' options are used.
10476 On embedded PowerPC systems, put all initialized global and static
10477 data in the `.data' section, and all uninitialized data in the
10481 On embedded PowerPC systems, put global and static items less than
10482 or equal to NUM bytes into the small data or bss sections instead
10483 of the normal data or bss section. By default, NUM is 8. The `-G
10484 NUM' switch is also passed to the linker. All modules should be
10485 compiled with the same `-G NUM' value.
10489 On System V.4 and embedded PowerPC systems do (do not) emit
10490 register names in the assembly language output using symbolic
10495 Default to making all function calls indirectly, using a register,
10496 so that functions which reside further than 32 megabytes
10497 (33,554,432 bytes) from the current location can be called. This
10498 setting can be overridden by the `shortcall' function attribute,
10499 or by `#pragma longcall(0)'.
10501 Some linkers are capable of detecting out-of-range calls and
10502 generating glue code on the fly. On these systems, long calls are
10503 unnecessary and generate slower code. As of this writing, the AIX
10504 linker can do this, as can the GNU linker for PowerPC/64. It is
10505 planned to add this feature to the GNU linker for 32-bit PowerPC
10508 On Darwin/PPC systems, `#pragma longcall' will generate "jbsr
10509 callee, L42", plus a "branch island" (glue code). The two target
10510 addresses represent the callee and the "branch island". The
10511 Darwin/PPC linker will prefer the first address and generate a "bl
10512 callee" if the PPC "bl" instruction will reach the callee directly;
10513 otherwise, the linker will generate "bl L42" to call the "branch
10514 island". The "branch island" is appended to the body of the
10515 calling function; it computes the full 32-bit address of the callee
10518 On Mach-O (Darwin) systems, this option directs the compiler emit
10519 to the glue for every direct call, and the Darwin linker decides
10520 whether to use or discard it.
10522 In the future, we may cause GCC to ignore all longcall
10523 specifications when the linker is known to generate glue.
10526 Adds support for multithreading with the "pthreads" library. This
10527 option sets flags for both the preprocessor and linker.
10531 File: gcc.info, Node: S/390 and zSeries Options, Next: SH Options, Prev: RS/6000 and PowerPC Options, Up: Submodel Options
10533 3.17.25 S/390 and zSeries Options
10534 ---------------------------------
10536 These are the `-m' options defined for the S/390 and zSeries
10541 Use (do not use) the hardware floating-point instructions and
10542 registers for floating-point operations. When `-msoft-float' is
10543 specified, functions in `libgcc.a' will be used to perform
10544 floating-point operations. When `-mhard-float' is specified, the
10545 compiler generates IEEE floating-point instructions. This is the
10550 Store (do not store) the address of the caller's frame as
10551 backchain pointer into the callee's stack frame. A backchain may
10552 be needed to allow debugging using tools that do not understand
10553 DWARF-2 call frame information. When `-mno-packed-stack' is in
10554 effect, the backchain pointer is stored at the bottom of the stack
10555 frame; when `-mpacked-stack' is in effect, the backchain is placed
10556 into the topmost word of the 96/160 byte register save area.
10558 In general, code compiled with `-mbackchain' is call-compatible
10559 with code compiled with `-mmo-backchain'; however, use of the
10560 backchain for debugging purposes usually requires that the whole
10561 binary is built with `-mbackchain'. Note that the combination of
10562 `-mbackchain', `-mpacked-stack' and `-mhard-float' is not
10563 supported. In order to build a linux kernel use `-msoft-float'.
10565 The default is to not maintain the backchain.
10569 `-mno-packed-stack'
10570 Use (do not use) the packed stack layout. When
10571 `-mno-packed-stack' is specified, the compiler uses the all fields
10572 of the 96/160 byte register save area only for their default
10573 purpose; unused fields still take up stack space. When
10574 `-mpacked-stack' is specified, register save slots are densely
10575 packed at the top of the register save area; unused space is
10576 reused for other purposes, allowing for more efficient use of the
10577 available stack space. However, when `-mbackchain' is also in
10578 effect, the topmost word of the save area is always used to store
10579 the backchain, and the return address register is always saved two
10580 words below the backchain.
10582 As long as the stack frame backchain is not used, code generated
10583 with `-mpacked-stack' is call-compatible with code generated with
10584 `-mno-packed-stack'. Note that some non-FSF releases of GCC 2.95
10585 for S/390 or zSeries generated code that uses the stack frame
10586 backchain at run time, not just for debugging purposes. Such code
10587 is not call-compatible with code compiled with `-mpacked-stack'.
10588 Also, note that the combination of `-mbackchain', `-mpacked-stack'
10589 and `-mhard-float' is not supported. In order to build a linux
10590 kernel use `-msoft-float'.
10592 The default is to not use the packed stack layout.
10596 Generate (or do not generate) code using the `bras' instruction to
10597 do subroutine calls. This only works reliably if the total
10598 executable size does not exceed 64k. The default is to use the
10599 `basr' instruction instead, which does not have this limitation.
10603 When `-m31' is specified, generate code compliant to the GNU/Linux
10604 for S/390 ABI. When `-m64' is specified, generate code compliant
10605 to the GNU/Linux for zSeries ABI. This allows GCC in particular
10606 to generate 64-bit instructions. For the `s390' targets, the
10607 default is `-m31', while the `s390x' targets default to `-m64'.
10611 When `-mzarch' is specified, generate code using the instructions
10612 available on z/Architecture. When `-mesa' is specified, generate
10613 code using the instructions available on ESA/390. Note that
10614 `-mesa' is not possible with `-m64'. When generating code
10615 compliant to the GNU/Linux for S/390 ABI, the default is `-mesa'.
10616 When generating code compliant to the GNU/Linux for zSeries ABI,
10617 the default is `-mzarch'.
10621 Generate (or do not generate) code using the `mvcle' instruction
10622 to perform block moves. When `-mno-mvcle' is specified, use a
10623 `mvc' loop instead. This is the default.
10627 Print (or do not print) additional debug information when
10628 compiling. The default is to not print debug information.
10631 Generate code that will run on CPU-TYPE, which is the name of a
10632 system representing a certain processor type. Possible values for
10633 CPU-TYPE are `g5', `g6', `z900', and `z990'. When generating code
10634 using the instructions available on z/Architecture, the default is
10635 `-march=z900'. Otherwise, the default is `-march=g5'.
10638 Tune to CPU-TYPE everything applicable about the generated code,
10639 except for the ABI and the set of available instructions. The
10640 list of CPU-TYPE values is the same as for `-march'. The default
10641 is the value used for `-march'.
10645 Generate code that adds (does not add) in TPF OS specific branches
10646 to trace routines in the operating system. This option is off by
10647 default, even when compiling for the TPF OS.
10651 Generate code that uses (does not use) the floating point multiply
10652 and accumulate instructions. These instructions are generated by
10653 default if hardware floating point is used.
10655 `-mwarn-framesize=FRAMESIZE'
10656 Emit a warning if the current function exceeds the given frame
10657 size. Because this is a compile time check it doesn't need to be
10658 a real problem when the program runs. It is intended to identify
10659 functions which most probably cause a stack overflow. It is
10660 useful to be used in an environment with limited stack size e.g.
10663 `-mwarn-dynamicstack'
10664 Emit a warning if the function calls alloca or uses dynamically
10665 sized arrays. This is generally a bad idea with a limited stack
10668 `-mstack-guard=STACK-GUARD'
10670 `-mstack-size=STACK-SIZE'
10671 These arguments always have to be used in conjunction. If they
10672 are present the s390 back end emits additional instructions in the
10673 function prologue which trigger a trap if the stack size is
10674 STACK-GUARD bytes above the STACK-SIZE (remember that the stack on
10675 s390 grows downward). These options are intended to be used to
10676 help debugging stack overflow problems. The additionally emitted
10677 code cause only little overhead and hence can also be used in
10678 production like systems without greater performance degradation.
10679 The given values have to be exact powers of 2 and STACK-SIZE has
10680 to be greater than STACK-GUARD. In order to be efficient the
10681 extra code makes the assumption that the stack starts at an
10682 address aligned to the value given by STACK-SIZE.
10685 File: gcc.info, Node: SH Options, Next: SPARC Options, Prev: S/390 and zSeries Options, Up: Submodel Options
10690 These `-m' options are defined for the SH implementations:
10693 Generate code for the SH1.
10696 Generate code for the SH2.
10699 Generate code for the SH2e.
10702 Generate code for the SH3.
10705 Generate code for the SH3e.
10708 Generate code for the SH4 without a floating-point unit.
10711 Generate code for the SH4 with a floating-point unit that only
10712 supports single-precision arithmetic.
10715 Generate code for the SH4 assuming the floating-point unit is in
10716 single-precision mode by default.
10719 Generate code for the SH4.
10722 Generate code for the SH4al-dsp, or for a SH4a in such a way that
10723 the floating-point unit is not used.
10726 Generate code for the SH4a, in such a way that no double-precision
10727 floating point operations are used.
10730 Generate code for the SH4a assuming the floating-point unit is in
10731 single-precision mode by default.
10734 Generate code for the SH4a.
10737 Same as `-m4a-nofpu', except that it implicitly passes `-dsp' to
10738 the assembler. GCC doesn't generate any DSP instructions at the
10742 Compile code for the processor in big endian mode.
10745 Compile code for the processor in little endian mode.
10748 Align doubles at 64-bit boundaries. Note that this changes the
10749 calling conventions, and thus some functions from the standard C
10750 library will not work unless you recompile it first with
10754 Shorten some address references at link time, when possible; uses
10755 the linker option `-relax'.
10758 Use 32-bit offsets in `switch' tables. The default is to use
10762 Enable the use of the instruction `fmovd'.
10765 Comply with the calling conventions defined by Renesas.
10768 Comply with the calling conventions defined by Renesas.
10771 Comply with the calling conventions defined for GCC before the
10772 Renesas conventions were available. This option is the default
10773 for all targets of the SH toolchain except for `sh-symbianelf'.
10776 Mark the `MAC' register as call-clobbered, even if `-mhitachi' is
10780 Increase IEEE-compliance of floating-point code.
10783 Dump instruction size and location in the assembly code.
10786 This option is deprecated. It pads structures to multiple of 4
10787 bytes, which is incompatible with the SH ABI.
10790 Optimize for space instead of speed. Implied by `-Os'.
10793 When generating position-independent code, emit function calls
10794 using the Global Offset Table instead of the Procedure Linkage
10798 Generate a library function call to invalidate instruction cache
10799 entries, after fixing up a trampoline. This library function call
10800 doesn't assume it can write to the whole memory address space.
10801 This is the default when the target is `sh-*-linux*'.
10804 File: gcc.info, Node: SPARC Options, Next: System V Options, Prev: SH Options, Up: Submodel Options
10806 3.17.27 SPARC Options
10807 ---------------------
10809 These `-m' options are supported on the SPARC:
10813 Specify `-mapp-regs' to generate output using the global registers
10814 2 through 4, which the SPARC SVR4 ABI reserves for applications.
10815 This is the default.
10817 To be fully SVR4 ABI compliant at the cost of some performance
10818 loss, specify `-mno-app-regs'. You should compile libraries and
10819 system software with this option.
10823 Generate output containing floating point instructions. This is
10828 Generate output containing library calls for floating point.
10829 *Warning:* the requisite libraries are not available for all SPARC
10830 targets. Normally the facilities of the machine's usual C
10831 compiler are used, but this cannot be done directly in
10832 cross-compilation. You must make your own arrangements to provide
10833 suitable library functions for cross-compilation. The embedded
10834 targets `sparc-*-aout' and `sparclite-*-*' do provide software
10835 floating point support.
10837 `-msoft-float' changes the calling convention in the output file;
10838 therefore, it is only useful if you compile _all_ of a program with
10839 this option. In particular, you need to compile `libgcc.a', the
10840 library that comes with GCC, with `-msoft-float' in order for this
10843 `-mhard-quad-float'
10844 Generate output containing quad-word (long double) floating point
10847 `-msoft-quad-float'
10848 Generate output containing library calls for quad-word (long
10849 double) floating point instructions. The functions called are
10850 those specified in the SPARC ABI. This is the default.
10852 As of this writing, there are no SPARC implementations that have
10853 hardware support for the quad-word floating point instructions.
10854 They all invoke a trap handler for one of these instructions, and
10855 then the trap handler emulates the effect of the instruction.
10856 Because of the trap handler overhead, this is much slower than
10857 calling the ABI library routines. Thus the `-msoft-quad-float'
10858 option is the default.
10860 `-mno-unaligned-doubles'
10861 `-munaligned-doubles'
10862 Assume that doubles have 8 byte alignment. This is the default.
10864 With `-munaligned-doubles', GCC assumes that doubles have 8 byte
10865 alignment only if they are contained in another type, or if they
10866 have an absolute address. Otherwise, it assumes they have 4 byte
10867 alignment. Specifying this option avoids some rare compatibility
10868 problems with code generated by other compilers. It is not the
10869 default because it results in a performance loss, especially for
10870 floating point code.
10872 `-mno-faster-structs'
10874 With `-mfaster-structs', the compiler assumes that structures
10875 should have 8 byte alignment. This enables the use of pairs of
10876 `ldd' and `std' instructions for copies in structure assignment,
10877 in place of twice as many `ld' and `st' pairs. However, the use
10878 of this changed alignment directly violates the SPARC ABI. Thus,
10879 it's intended only for use on targets where the developer
10880 acknowledges that their resulting code will not be directly in
10881 line with the rules of the ABI.
10884 `-mimpure-text', used in addition to `-shared', tells the compiler
10885 to not pass `-z text' to the linker when linking a shared object.
10886 Using this option, you can link position-dependent code into a
10889 `-mimpure-text' suppresses the "relocations remain against
10890 allocatable but non-writable sections" linker error message.
10891 However, the necessary relocations will trigger copy-on-write, and
10892 the shared object is not actually shared across processes.
10893 Instead of using `-mimpure-text', you should compile all source
10894 code with `-fpic' or `-fPIC'.
10896 This option is only available on SunOS and Solaris.
10899 Set the instruction set, register set, and instruction scheduling
10900 parameters for machine type CPU_TYPE. Supported values for
10901 CPU_TYPE are `v7', `cypress', `v8', `supersparc', `sparclite',
10902 `f930', `f934', `hypersparc', `sparclite86x', `sparclet',
10903 `tsc701', `v9', `ultrasparc', and `ultrasparc3'.
10905 Default instruction scheduling parameters are used for values that
10906 select an architecture and not an implementation. These are `v7',
10907 `v8', `sparclite', `sparclet', `v9'.
10909 Here is a list of each supported architecture and their supported
10913 v8: supersparc, hypersparc
10914 sparclite: f930, f934, sparclite86x
10916 v9: ultrasparc, ultrasparc3
10918 By default (unless configured otherwise), GCC generates code for
10919 the V7 variant of the SPARC architecture. With `-mcpu=cypress',
10920 the compiler additionally optimizes it for the Cypress CY7C602
10921 chip, as used in the SPARCStation/SPARCServer 3xx series. This is
10922 also appropriate for the older SPARCStation 1, 2, IPX etc.
10924 With `-mcpu=v8', GCC generates code for the V8 variant of the SPARC
10925 architecture. The only difference from V7 code is that the
10926 compiler emits the integer multiply and integer divide
10927 instructions which exist in SPARC-V8 but not in SPARC-V7. With
10928 `-mcpu=supersparc', the compiler additionally optimizes it for the
10929 SuperSPARC chip, as used in the SPARCStation 10, 1000 and 2000
10932 With `-mcpu=sparclite', GCC generates code for the SPARClite
10933 variant of the SPARC architecture. This adds the integer
10934 multiply, integer divide step and scan (`ffs') instructions which
10935 exist in SPARClite but not in SPARC-V7. With `-mcpu=f930', the
10936 compiler additionally optimizes it for the Fujitsu MB86930 chip,
10937 which is the original SPARClite, with no FPU. With `-mcpu=f934',
10938 the compiler additionally optimizes it for the Fujitsu MB86934
10939 chip, which is the more recent SPARClite with FPU.
10941 With `-mcpu=sparclet', GCC generates code for the SPARClet variant
10942 of the SPARC architecture. This adds the integer multiply,
10943 multiply/accumulate, integer divide step and scan (`ffs')
10944 instructions which exist in SPARClet but not in SPARC-V7. With
10945 `-mcpu=tsc701', the compiler additionally optimizes it for the
10946 TEMIC SPARClet chip.
10948 With `-mcpu=v9', GCC generates code for the V9 variant of the SPARC
10949 architecture. This adds 64-bit integer and floating-point move
10950 instructions, 3 additional floating-point condition code registers
10951 and conditional move instructions. With `-mcpu=ultrasparc', the
10952 compiler additionally optimizes it for the Sun UltraSPARC I/II
10953 chips. With `-mcpu=ultrasparc3', the compiler additionally
10954 optimizes it for the Sun UltraSPARC III chip.
10957 Set the instruction scheduling parameters for machine type
10958 CPU_TYPE, but do not set the instruction set or register set that
10959 the option `-mcpu=CPU_TYPE' would.
10961 The same values for `-mcpu=CPU_TYPE' can be used for
10962 `-mtune=CPU_TYPE', but the only useful values are those that
10963 select a particular cpu implementation. Those are `cypress',
10964 `supersparc', `hypersparc', `f930', `f934', `sparclite86x',
10965 `tsc701', `ultrasparc', and `ultrasparc3'.
10969 With `-mv8plus', GCC generates code for the SPARC-V8+ ABI. The
10970 difference from the V8 ABI is that the global and out registers are
10971 considered 64-bit wide. This is enabled by default on Solaris in
10972 32-bit mode for all SPARC-V9 processors.
10976 With `-mvis', GCC generates code that takes advantage of the
10977 UltraSPARC Visual Instruction Set extensions. The default is
10980 These `-m' options are supported in addition to the above on SPARC-V9
10981 processors in 64-bit environments:
10984 Generate code for a processor running in little-endian mode. It
10985 is only available for a few configurations and most notably not on
10990 Generate code for a 32-bit or 64-bit environment. The 32-bit
10991 environment sets int, long and pointer to 32 bits. The 64-bit
10992 environment sets int to 32 bits and long and pointer to 64 bits.
10995 Generate code for the Medium/Low code model: 64-bit addresses,
10996 programs must be linked in the low 32 bits of memory. Programs
10997 can be statically or dynamically linked.
11000 Generate code for the Medium/Middle code model: 64-bit addresses,
11001 programs must be linked in the low 44 bits of memory, the text and
11002 data segments must be less than 2GB in size and the data segment
11003 must be located within 2GB of the text segment.
11006 Generate code for the Medium/Anywhere code model: 64-bit
11007 addresses, programs may be linked anywhere in memory, the text and
11008 data segments must be less than 2GB in size and the data segment
11009 must be located within 2GB of the text segment.
11011 `-mcmodel=embmedany'
11012 Generate code for the Medium/Anywhere code model for embedded
11013 systems: 64-bit addresses, the text and data segments must be less
11014 than 2GB in size, both starting anywhere in memory (determined at
11015 link time). The global register %g4 points to the base of the
11016 data segment. Programs are statically linked and PIC is not
11021 With `-mstack-bias', GCC assumes that the stack pointer, and frame
11022 pointer if present, are offset by -2047 which must be added back
11023 when making stack frame references. This is the default in 64-bit
11024 mode. Otherwise, assume no such offset is present.
11026 These switches are supported in addition to the above on Solaris:
11029 Add support for multithreading using the Solaris threads library.
11030 This option sets flags for both the preprocessor and linker. This
11031 option does not affect the thread safety of object code produced
11032 by the compiler or that of libraries supplied with it.
11035 Add support for multithreading using the POSIX threads library.
11036 This option sets flags for both the preprocessor and linker. This
11037 option does not affect the thread safety of object code produced
11038 by the compiler or that of libraries supplied with it.
11041 File: gcc.info, Node: System V Options, Next: TMS320C3x/C4x Options, Prev: SPARC Options, Up: Submodel Options
11043 3.17.28 Options for System V
11044 ----------------------------
11046 These additional options are available on System V Release 4 for
11047 compatibility with other compilers on those systems:
11050 Create a shared object. It is recommended that `-symbolic' or
11051 `-shared' be used instead.
11054 Identify the versions of each tool used by the compiler, in a
11055 `.ident' assembler directive in the output.
11058 Refrain from adding `.ident' directives to the output file (this is
11062 Search the directories DIRS, and no others, for libraries
11063 specified with `-l'.
11066 Look in the directory DIR to find the M4 preprocessor. The
11067 assembler uses this option.
11070 File: gcc.info, Node: TMS320C3x/C4x Options, Next: V850 Options, Prev: System V Options, Up: Submodel Options
11072 3.17.29 TMS320C3x/C4x Options
11073 -----------------------------
11075 These `-m' options are defined for TMS320C3x/C4x implementations:
11078 Set the instruction set, register set, and instruction scheduling
11079 parameters for machine type CPU_TYPE. Supported values for
11080 CPU_TYPE are `c30', `c31', `c32', `c40', and `c44'. The default
11081 is `c40' to generate code for the TMS320C40.
11087 Generates code for the big or small memory model. The small memory
11088 model assumed that all data fits into one 64K word page. At
11089 run-time the data page (DP) register must be set to point to the
11090 64K page containing the .bss and .data program sections. The big
11091 memory model is the default and requires reloading of the DP
11092 register for every direct memory access.
11096 Allow (disallow) allocation of general integer operands into the
11097 block count register BK.
11101 Enable (disable) generation of code using decrement and branch,
11102 DBcond(D), instructions. This is enabled by default for the C4x.
11103 To be on the safe side, this is disabled for the C3x, since the
11104 maximum iteration count on the C3x is 2^23 + 1 (but who iterates
11105 loops more than 2^23 times on the C3x?). Note that GCC will try
11106 to reverse a loop so that it can utilize the decrement and branch
11107 instruction, but will give up if there is more than one memory
11108 reference in the loop. Thus a loop where the loop counter is
11109 decremented can generate slightly more efficient code, in cases
11110 where the RPTB instruction cannot be utilized.
11114 Force the DP register to be saved on entry to an interrupt service
11115 routine (ISR), reloaded to point to the data section, and restored
11116 on exit from the ISR. This should not be required unless someone
11117 has violated the small memory model by modifying the DP register,
11118 say within an object library.
11122 For the C3x use the 24-bit MPYI instruction for integer multiplies
11123 instead of a library call to guarantee 32-bit results. Note that
11124 if one of the operands is a constant, then the multiplication will
11125 be performed using shifts and adds. If the `-mmpyi' option is not
11126 specified for the C3x, then squaring operations are performed
11127 inline instead of a library call.
11131 The C3x/C4x FIX instruction to convert a floating point value to an
11132 integer value chooses the nearest integer less than or equal to the
11133 floating point value rather than to the nearest integer. Thus if
11134 the floating point number is negative, the result will be
11135 incorrectly truncated an additional code is necessary to detect
11136 and correct this case. This option can be used to disable
11137 generation of the additional code required to correct the result.
11141 Enable (disable) generation of repeat block sequences using the
11142 RPTB instruction for zero overhead looping. The RPTB construct is
11143 only used for innermost loops that do not call functions or jump
11144 across the loop boundaries. There is no advantage having nested
11145 RPTB loops due to the overhead required to save and restore the
11146 RC, RS, and RE registers. This is enabled by default with `-O2'.
11150 Enable (disable) the use of the single instruction repeat
11151 instruction RPTS. If a repeat block contains a single
11152 instruction, and the loop count can be guaranteed to be less than
11153 the value COUNT, GCC will emit a RPTS instruction instead of a
11154 RPTB. If no value is specified, then a RPTS will be emitted even
11155 if the loop count cannot be determined at compile time. Note that
11156 the repeated instruction following RPTS does not have to be
11157 reloaded from memory each iteration, thus freeing up the CPU buses
11158 for operands. However, since interrupts are blocked by this
11159 instruction, it is disabled by default.
11162 `-mno-loop-unsigned'
11163 The maximum iteration count when using RPTS and RPTB (and DB on
11164 the C40) is 2^31 + 1 since these instructions test if the
11165 iteration count is negative to terminate the loop. If the
11166 iteration count is unsigned there is a possibility than the 2^31 +
11167 1 maximum iteration count may be exceeded. This switch allows an
11168 unsigned iteration count.
11171 Try to emit an assembler syntax that the TI assembler (asm30) is
11172 happy with. This also enforces compatibility with the API
11173 employed by the TI C3x C compiler. For example, long doubles are
11174 passed as structures rather than in floating point registers.
11178 Generate code that uses registers (stack) for passing arguments to
11179 functions. By default, arguments are passed in registers where
11180 possible rather than by pushing arguments on to the stack.
11183 `-mno-parallel-insns'
11184 Allow the generation of parallel instructions. This is enabled by
11185 default with `-O2'.
11188 `-mno-parallel-mpy'
11189 Allow the generation of MPY||ADD and MPY||SUB parallel
11190 instructions, provided `-mparallel-insns' is also specified.
11191 These instructions have tight register constraints which can
11192 pessimize the code generation of large functions.
11196 File: gcc.info, Node: V850 Options, Next: VAX Options, Prev: TMS320C3x/C4x Options, Up: Submodel Options
11198 3.17.30 V850 Options
11199 --------------------
11201 These `-m' options are defined for V850 implementations:
11205 Treat all calls as being far away (near). If calls are assumed to
11206 be far away, the compiler will always load the functions address
11207 up into a register, and call indirect through the pointer.
11211 Do not optimize (do optimize) basic blocks that use the same index
11212 pointer 4 or more times to copy pointer into the `ep' register, and
11213 use the shorter `sld' and `sst' instructions. The `-mep' option
11214 is on by default if you optimize.
11216 `-mno-prolog-function'
11217 `-mprolog-function'
11218 Do not use (do use) external functions to save and restore
11219 registers at the prologue and epilogue of a function. The
11220 external functions are slower, but use less code space if more
11221 than one function saves the same number of registers. The
11222 `-mprolog-function' option is on by default if you optimize.
11225 Try to make the code as small as possible. At present, this just
11226 turns on the `-mep' and `-mprolog-function' options.
11229 Put static or global variables whose size is N bytes or less into
11230 the tiny data area that register `ep' points to. The tiny data
11231 area can hold up to 256 bytes in total (128 bytes for byte
11235 Put static or global variables whose size is N bytes or less into
11236 the small data area that register `gp' points to. The small data
11237 area can hold up to 64 kilobytes.
11240 Put static or global variables whose size is N bytes or less into
11241 the first 32 kilobytes of memory.
11244 Specify that the target processor is the V850.
11247 Generate code suitable for big switch tables. Use this option
11248 only if the assembler/linker complain about out of range branches
11249 within a switch table.
11252 This option will cause r2 and r5 to be used in the code generated
11253 by the compiler. This setting is the default.
11256 This option will cause r2 and r5 to be treated as fixed registers.
11259 Specify that the target processor is the V850E1. The preprocessor
11260 constants `__v850e1__' and `__v850e__' will be defined if this
11264 Specify that the target processor is the V850E. The preprocessor
11265 constant `__v850e__' will be defined if this option is used.
11267 If neither `-mv850' nor `-mv850e' nor `-mv850e1' are defined then
11268 a default target processor will be chosen and the relevant
11269 `__v850*__' preprocessor constant will be defined.
11271 The preprocessor constants `__v850' and `__v851__' are always
11272 defined, regardless of which processor variant is the target.
11275 This option will suppress generation of the CALLT instruction for
11276 the v850e and v850e1 flavors of the v850 architecture. The
11277 default is `-mno-disable-callt' which allows the CALLT instruction
11282 File: gcc.info, Node: VAX Options, Next: x86-64 Options, Prev: V850 Options, Up: Submodel Options
11284 3.17.31 VAX Options
11285 -------------------
11287 These `-m' options are defined for the VAX:
11290 Do not output certain jump instructions (`aobleq' and so on) that
11291 the Unix assembler for the VAX cannot handle across long ranges.
11294 Do output those jump instructions, on the assumption that you will
11295 assemble with the GNU assembler.
11298 Output code for g-format floating point numbers instead of
11302 File: gcc.info, Node: x86-64 Options, Next: Xstormy16 Options, Prev: VAX Options, Up: Submodel Options
11304 3.17.32 x86-64 Options
11305 ----------------------
11307 These are listed under *Note i386 and x86-64 Options::.
11310 File: gcc.info, Node: Xstormy16 Options, Next: Xtensa Options, Prev: x86-64 Options, Up: Submodel Options
11312 3.17.33 Xstormy16 Options
11313 -------------------------
11315 These options are defined for Xstormy16:
11318 Choose startup files and linker script suitable for the simulator.
11321 File: gcc.info, Node: Xtensa Options, Next: zSeries Options, Prev: Xstormy16 Options, Up: Submodel Options
11323 3.17.34 Xtensa Options
11324 ----------------------
11326 These options are supported for Xtensa targets:
11330 Enable or disable use of `CONST16' instructions for loading
11331 constant values. The `CONST16' instruction is currently not a
11332 standard option from Tensilica. When enabled, `CONST16'
11333 instructions are always used in place of the standard `L32R'
11334 instructions. The use of `CONST16' is enabled by default only if
11335 the `L32R' instruction is not available.
11339 Enable or disable use of fused multiply/add and multiply/subtract
11340 instructions in the floating-point option. This has no effect if
11341 the floating-point option is not also enabled. Disabling fused
11342 multiply/add and multiply/subtract instructions forces the
11343 compiler to use separate instructions for the multiply and
11344 add/subtract operations. This may be desirable in some cases
11345 where strict IEEE 754-compliant results are required: the fused
11346 multiply add/subtract instructions do not round the intermediate
11347 result, thereby producing results with _more_ bits of precision
11348 than specified by the IEEE standard. Disabling fused multiply
11349 add/subtract instructions also ensures that the program output is
11350 not sensitive to the compiler's ability to combine multiply and
11351 add/subtract operations.
11353 `-mtext-section-literals'
11354 `-mno-text-section-literals'
11355 Control the treatment of literal pools. The default is
11356 `-mno-text-section-literals', which places literals in a separate
11357 section in the output file. This allows the literal pool to be
11358 placed in a data RAM/ROM, and it also allows the linker to combine
11359 literal pools from separate object files to remove redundant
11360 literals and improve code size. With `-mtext-section-literals',
11361 the literals are interspersed in the text section in order to keep
11362 them as close as possible to their references. This may be
11363 necessary for large assembly files.
11366 `-mno-target-align'
11367 When this option is enabled, GCC instructs the assembler to
11368 automatically align instructions to reduce branch penalties at the
11369 expense of some code density. The assembler attempts to widen
11370 density instructions to align branch targets and the instructions
11371 following call instructions. If there are not enough preceding
11372 safe density instructions to align a target, no widening will be
11373 performed. The default is `-mtarget-align'. These options do not
11374 affect the treatment of auto-aligned instructions like `LOOP',
11375 which the assembler will always align, either by widening density
11376 instructions or by inserting no-op instructions.
11380 When this option is enabled, GCC instructs the assembler to
11381 translate direct calls to indirect calls unless it can determine
11382 that the target of a direct call is in the range allowed by the
11383 call instruction. This translation typically occurs for calls to
11384 functions in other source files. Specifically, the assembler
11385 translates a direct `CALL' instruction into an `L32R' followed by
11386 a `CALLX' instruction. The default is `-mno-longcalls'. This
11387 option should be used in programs where the call target can
11388 potentially be out of range. This option is implemented in the
11389 assembler, not the compiler, so the assembly code generated by GCC
11390 will still show direct call instructions--look at the disassembled
11391 object code to see the actual instructions. Note that the
11392 assembler will use an indirect call for every cross-file call, not
11393 just those that really will be out of range.
11396 File: gcc.info, Node: zSeries Options, Prev: Xtensa Options, Up: Submodel Options
11398 3.17.35 zSeries Options
11399 -----------------------
11401 These are listed under *Note S/390 and zSeries Options::.
11404 File: gcc.info, Node: Code Gen Options, Next: Environment Variables, Prev: Submodel Options, Up: Invoking GCC
11406 3.18 Options for Code Generation Conventions
11407 ============================================
11409 These machine-independent options control the interface conventions
11410 used in code generation.
11412 Most of them have both positive and negative forms; the negative form
11413 of `-ffoo' would be `-fno-foo'. In the table below, only one of the
11414 forms is listed--the one which is not the default. You can figure out
11415 the other form by either removing `no-' or adding it.
11418 For front-ends that support it, generate additional code to check
11419 that indices used to access arrays are within the declared range.
11420 This is currently only supported by the Java and Fortran 77
11421 front-ends, where this option defaults to true and false
11425 This option generates traps for signed overflow on addition,
11426 subtraction, multiplication operations.
11429 This option instructs the compiler to assume that signed arithmetic
11430 overflow of addition, subtraction and multiplication wraps around
11431 using twos-complement representation. This flag enables some
11432 optimizations and disables other. This option is enabled by
11433 default for the Java front-end, as required by the Java language
11437 Enable exception handling. Generates extra code needed to
11438 propagate exceptions. For some targets, this implies GCC will
11439 generate frame unwind information for all functions, which can
11440 produce significant data size overhead, although it does not
11441 affect execution. If you do not specify this option, GCC will
11442 enable it by default for languages like C++ which normally require
11443 exception handling, and disable it for languages like C that do
11444 not normally require it. However, you may need to enable this
11445 option when compiling C code that needs to interoperate properly
11446 with exception handlers written in C++. You may also wish to
11447 disable this option if you are compiling older C++ programs that
11448 don't use exception handling.
11450 `-fnon-call-exceptions'
11451 Generate code that allows trapping instructions to throw
11452 exceptions. Note that this requires platform-specific runtime
11453 support that does not exist everywhere. Moreover, it only allows
11454 _trapping_ instructions to throw exceptions, i.e. memory
11455 references or floating point instructions. It does not allow
11456 exceptions to be thrown from arbitrary signal handlers such as
11460 Similar to `-fexceptions', except that it will just generate any
11461 needed static data, but will not affect the generated code in any
11462 other way. You will normally not enable this option; instead, a
11463 language processor that needs this handling would enable it on
11466 `-fasynchronous-unwind-tables'
11467 Generate unwind table in dwarf2 format, if supported by target
11468 machine. The table is exact at each instruction boundary, so it
11469 can be used for stack unwinding from asynchronous events (such as
11470 debugger or garbage collector).
11472 `-fpcc-struct-return'
11473 Return "short" `struct' and `union' values in memory like longer
11474 ones, rather than in registers. This convention is less
11475 efficient, but it has the advantage of allowing intercallability
11476 between GCC-compiled files and files compiled with other
11477 compilers, particularly the Portable C Compiler (pcc).
11479 The precise convention for returning structures in memory depends
11480 on the target configuration macros.
11482 Short structures and unions are those whose size and alignment
11483 match that of some integer type.
11485 *Warning:* code compiled with the `-fpcc-struct-return' switch is
11486 not binary compatible with code compiled with the
11487 `-freg-struct-return' switch. Use it to conform to a non-default
11488 application binary interface.
11490 `-freg-struct-return'
11491 Return `struct' and `union' values in registers when possible.
11492 This is more efficient for small structures than
11493 `-fpcc-struct-return'.
11495 If you specify neither `-fpcc-struct-return' nor
11496 `-freg-struct-return', GCC defaults to whichever convention is
11497 standard for the target. If there is no standard convention, GCC
11498 defaults to `-fpcc-struct-return', except on targets where GCC is
11499 the principal compiler. In those cases, we can choose the
11500 standard, and we chose the more efficient register return
11503 *Warning:* code compiled with the `-freg-struct-return' switch is
11504 not binary compatible with code compiled with the
11505 `-fpcc-struct-return' switch. Use it to conform to a non-default
11506 application binary interface.
11509 Allocate to an `enum' type only as many bytes as it needs for the
11510 declared range of possible values. Specifically, the `enum' type
11511 will be equivalent to the smallest integer type which has enough
11514 *Warning:* the `-fshort-enums' switch causes GCC to generate code
11515 that is not binary compatible with code generated without that
11516 switch. Use it to conform to a non-default application binary
11520 Use the same size for `double' as for `float'.
11522 *Warning:* the `-fshort-double' switch causes GCC to generate code
11523 that is not binary compatible with code generated without that
11524 switch. Use it to conform to a non-default application binary
11528 Override the underlying type for `wchar_t' to be `short unsigned
11529 int' instead of the default for the target. This option is useful
11530 for building programs to run under WINE.
11532 *Warning:* the `-fshort-wchar' switch causes GCC to generate code
11533 that is not binary compatible with code generated without that
11534 switch. Use it to conform to a non-default application binary
11538 Requests that the data and non-`const' variables of this
11539 compilation be shared data rather than private data. The
11540 distinction makes sense only on certain operating systems, where
11541 shared data is shared between processes running the same program,
11542 while private data exists in one copy per process.
11545 In C, allocate even uninitialized global variables in the data
11546 section of the object file, rather than generating them as common
11547 blocks. This has the effect that if the same variable is declared
11548 (without `extern') in two different compilations, you will get an
11549 error when you link them. The only reason this might be useful is
11550 if you wish to verify that the program will work on other systems
11551 which always work this way.
11554 Ignore the `#ident' directive.
11556 `-finhibit-size-directive'
11557 Don't output a `.size' assembler directive, or anything else that
11558 would cause trouble if the function is split in the middle, and the
11559 two halves are placed at locations far apart in memory. This
11560 option is used when compiling `crtstuff.c'; you should not need to
11561 use it for anything else.
11564 Put extra commentary information in the generated assembly code to
11565 make it more readable. This option is generally only of use to
11566 those who actually need to read the generated assembly code
11567 (perhaps while debugging the compiler itself).
11569 `-fno-verbose-asm', the default, causes the extra information to
11570 be omitted and is useful when comparing two assembler files.
11573 Generate position-independent code (PIC) suitable for use in a
11574 shared library, if supported for the target machine. Such code
11575 accesses all constant addresses through a global offset table
11576 (GOT). The dynamic loader resolves the GOT entries when the
11577 program starts (the dynamic loader is not part of GCC; it is part
11578 of the operating system). If the GOT size for the linked
11579 executable exceeds a machine-specific maximum size, you get an
11580 error message from the linker indicating that `-fpic' does not
11581 work; in that case, recompile with `-fPIC' instead. (These
11582 maximums are 8k on the SPARC and 32k on the m68k and RS/6000. The
11583 386 has no such limit.)
11585 Position-independent code requires special support, and therefore
11586 works only on certain machines. For the 386, GCC supports PIC for
11587 System V but not for the Sun 386i. Code generated for the IBM
11588 RS/6000 is always position-independent.
11591 If supported for the target machine, emit position-independent
11592 code, suitable for dynamic linking and avoiding any limit on the
11593 size of the global offset table. This option makes a difference
11594 on the m68k, PowerPC and SPARC.
11596 Position-independent code requires special support, and therefore
11597 works only on certain machines.
11601 These options are similar to `-fpic' and `-fPIC', but generated
11602 position independent code can be only linked into executables.
11603 Usually these options are used when `-pie' GCC option will be used
11607 Treat the register named REG as a fixed register; generated code
11608 should never refer to it (except perhaps as a stack pointer, frame
11609 pointer or in some other fixed role).
11611 REG must be the name of a register. The register names accepted
11612 are machine-specific and are defined in the `REGISTER_NAMES' macro
11613 in the machine description macro file.
11615 This flag does not have a negative form, because it specifies a
11619 Treat the register named REG as an allocable register that is
11620 clobbered by function calls. It may be allocated for temporaries
11621 or variables that do not live across a call. Functions compiled
11622 this way will not save and restore the register REG.
11624 It is an error to used this flag with the frame pointer or stack
11625 pointer. Use of this flag for other registers that have fixed
11626 pervasive roles in the machine's execution model will produce
11627 disastrous results.
11629 This flag does not have a negative form, because it specifies a
11633 Treat the register named REG as an allocable register saved by
11634 functions. It may be allocated even for temporaries or variables
11635 that live across a call. Functions compiled this way will save
11636 and restore the register REG if they use it.
11638 It is an error to used this flag with the frame pointer or stack
11639 pointer. Use of this flag for other registers that have fixed
11640 pervasive roles in the machine's execution model will produce
11641 disastrous results.
11643 A different sort of disaster will result from the use of this flag
11644 for a register in which function values may be returned.
11646 This flag does not have a negative form, because it specifies a
11649 `-fpack-struct[=N]'
11650 Without a value specified, pack all structure members together
11651 without holes. When a value is specified (which must be a small
11652 power of two), pack structure members according to this value,
11653 representing the maximum alignment (that is, objects with default
11654 alignment requirements larger than this will be output potentially
11655 unaligned at the next fitting location.
11657 *Warning:* the `-fpack-struct' switch causes GCC to generate code
11658 that is not binary compatible with code generated without that
11659 switch. Additionally, it makes the code suboptimal. Use it to
11660 conform to a non-default application binary interface.
11662 `-finstrument-functions'
11663 Generate instrumentation calls for entry and exit to functions.
11664 Just after function entry and just before function exit, the
11665 following profiling functions will be called with the address of
11666 the current function and its call site. (On some platforms,
11667 `__builtin_return_address' does not work beyond the current
11668 function, so the call site information may not be available to the
11669 profiling functions otherwise.)
11671 void __cyg_profile_func_enter (void *this_fn,
11673 void __cyg_profile_func_exit (void *this_fn,
11676 The first argument is the address of the start of the current
11677 function, which may be looked up exactly in the symbol table.
11679 This instrumentation is also done for functions expanded inline in
11680 other functions. The profiling calls will indicate where,
11681 conceptually, the inline function is entered and exited. This
11682 means that addressable versions of such functions must be
11683 available. If all your uses of a function are expanded inline,
11684 this may mean an additional expansion of code size. If you use
11685 `extern inline' in your C code, an addressable version of such
11686 functions must be provided. (This is normally the case anyways,
11687 but if you get lucky and the optimizer always expands the
11688 functions inline, you might have gotten away without providing
11691 A function may be given the attribute `no_instrument_function', in
11692 which case this instrumentation will not be done. This can be
11693 used, for example, for the profiling functions listed above,
11694 high-priority interrupt routines, and any functions from which the
11695 profiling functions cannot safely be called (perhaps signal
11696 handlers, if the profiling routines generate output or allocate
11700 Generate code to verify that you do not go beyond the boundary of
11701 the stack. You should specify this flag if you are running in an
11702 environment with multiple threads, but only rarely need to specify
11703 it in a single-threaded environment since stack overflow is
11704 automatically detected on nearly all systems if there is only one
11707 Note that this switch does not actually cause checking to be done;
11708 the operating system must do that. The switch causes generation
11709 of code to ensure that the operating system sees the stack being
11712 `-fstack-limit-register=REG'
11713 `-fstack-limit-symbol=SYM'
11715 Generate code to ensure that the stack does not grow beyond a
11716 certain value, either the value of a register or the address of a
11717 symbol. If the stack would grow beyond the value, a signal is
11718 raised. For most targets, the signal is raised before the stack
11719 overruns the boundary, so it is possible to catch the signal
11720 without taking special precautions.
11722 For instance, if the stack starts at absolute address `0x80000000'
11723 and grows downwards, you can use the flags
11724 `-fstack-limit-symbol=__stack_limit' and
11725 `-Wl,--defsym,__stack_limit=0x7ffe0000' to enforce a stack limit
11726 of 128KB. Note that this may only work with the GNU linker.
11729 `-fargument-noalias'
11730 `-fargument-noalias-global'
11731 Specify the possible relationships among parameters and between
11732 parameters and global data.
11734 `-fargument-alias' specifies that arguments (parameters) may alias
11735 each other and may alias global storage.
11736 `-fargument-noalias' specifies that arguments do not alias each
11737 other, but may alias global storage.
11738 `-fargument-noalias-global' specifies that arguments do not alias
11739 each other and do not alias global storage.
11741 Each language will automatically use whatever option is required by
11742 the language standard. You should not need to use these options
11745 `-fleading-underscore'
11746 This option and its counterpart, `-fno-leading-underscore',
11747 forcibly change the way C symbols are represented in the object
11748 file. One use is to help link with legacy assembly code.
11750 *Warning:* the `-fleading-underscore' switch causes GCC to
11751 generate code that is not binary compatible with code generated
11752 without that switch. Use it to conform to a non-default
11753 application binary interface. Not all targets provide complete
11754 support for this switch.
11756 `-ftls-model=MODEL'
11757 Alter the thread-local storage model to be used (*note
11758 Thread-Local::). The MODEL argument should be one of
11759 `global-dynamic', `local-dynamic', `initial-exec' or `local-exec'.
11761 The default without `-fpic' is `initial-exec'; with `-fpic' the
11762 default is `global-dynamic'.
11764 `-fvisibility=DEFAULT|INTERNAL|HIDDEN|PROTECTED'
11765 Set the default ELF image symbol visibility to the specified
11766 option--all symbols will be marked with this unless overridden
11767 within the code. Using this feature can very substantially
11768 improve linking and load times of shared object libraries, produce
11769 more optimized code, provide near-perfect API export and prevent
11770 symbol clashes. It is *strongly* recommended that you use this in
11771 any shared objects you distribute.
11773 Despite the nomenclature, `default' always means public ie;
11774 available to be linked against from outside the shared object.
11775 `protected' and `internal' are pretty useless in real-world usage
11776 so the only other commonly used option will be `hidden'. The
11777 default if `-fvisibility' isn't specified is `default', i.e., make
11778 every symbol public--this causes the same behavior as previous
11781 A good explanation of the benefits offered by ensuring ELF symbols
11782 have the correct visibility is given by "How To Write Shared
11783 Libraries" by Ulrich Drepper (which can be found at
11784 `http://people.redhat.com/~drepper/')--however a superior solution
11785 made possible by this option to marking things hidden when the
11786 default is public is to make the default hidden and mark things
11787 public. This is the norm with DLL's on Windows and with
11788 `-fvisibility=hidden' and `__attribute__
11789 ((visibility("default")))' instead of `__declspec(dllexport)' you
11790 get almost identical semantics with identical syntax. This is a
11791 great boon to those working with cross-platform projects.
11793 For those adding visibility support to existing code, you may find
11794 `#pragma GCC visibility' of use. This works by you enclosing the
11795 declarations you wish to set visibility for with (for example)
11796 `#pragma GCC visibility push(hidden)' and `#pragma GCC visibility
11797 pop'. These can be nested up to sixteen times. Bear in mind that
11798 symbol visibility should be viewed *as part of the API interface
11799 contract* and thus all new code should always specify visibility
11800 when it is not the default ie; declarations only for use within
11801 the local DSO should *always* be marked explicitly as hidden as so
11802 to avoid PLT indirection overheads--making this abundantly clear
11803 also aids readability and self-documentation of the code. Note
11804 that due to ISO C++ specification requirements, operator new and
11805 operator delete must always be of default visibility.
11807 An overview of these techniques, their benefits and how to use them
11808 is at `http://www.nedprod.com/programs/gccvisibility.html'.
11812 File: gcc.info, Node: Environment Variables, Next: Precompiled Headers, Prev: Code Gen Options, Up: Invoking GCC
11814 3.19 Environment Variables Affecting GCC
11815 ========================================
11817 This section describes several environment variables that affect how GCC
11818 operates. Some of them work by specifying directories or prefixes to
11819 use when searching for various kinds of files. Some are used to
11820 specify other aspects of the compilation environment.
11822 Note that you can also specify places to search using options such as
11823 `-B', `-I' and `-L' (*note Directory Options::). These take precedence
11824 over places specified using environment variables, which in turn take
11825 precedence over those specified by the configuration of GCC. *Note
11826 Controlling the Compilation Driver `gcc': (gccint)Driver.
11832 These environment variables control the way that GCC uses
11833 localization information that allow GCC to work with different
11834 national conventions. GCC inspects the locale categories
11835 `LC_CTYPE' and `LC_MESSAGES' if it has been configured to do so.
11836 These locale categories can be set to any value supported by your
11837 installation. A typical value is `en_GB.UTF-8' for English in the
11838 United Kingdom encoded in UTF-8.
11840 The `LC_CTYPE' environment variable specifies character
11841 classification. GCC uses it to determine the character boundaries
11842 in a string; this is needed for some multibyte encodings that
11843 contain quote and escape characters that would otherwise be
11844 interpreted as a string end or escape.
11846 The `LC_MESSAGES' environment variable specifies the language to
11847 use in diagnostic messages.
11849 If the `LC_ALL' environment variable is set, it overrides the value
11850 of `LC_CTYPE' and `LC_MESSAGES'; otherwise, `LC_CTYPE' and
11851 `LC_MESSAGES' default to the value of the `LANG' environment
11852 variable. If none of these variables are set, GCC defaults to
11853 traditional C English behavior.
11856 If `TMPDIR' is set, it specifies the directory to use for temporary
11857 files. GCC uses temporary files to hold the output of one stage of
11858 compilation which is to be used as input to the next stage: for
11859 example, the output of the preprocessor, which is the input to the
11863 If `GCC_EXEC_PREFIX' is set, it specifies a prefix to use in the
11864 names of the subprograms executed by the compiler. No slash is
11865 added when this prefix is combined with the name of a subprogram,
11866 but you can specify a prefix that ends with a slash if you wish.
11868 If `GCC_EXEC_PREFIX' is not set, GCC will attempt to figure out an
11869 appropriate prefix to use based on the pathname it was invoked
11872 If GCC cannot find the subprogram using the specified prefix, it
11873 tries looking in the usual places for the subprogram.
11875 The default value of `GCC_EXEC_PREFIX' is `PREFIX/lib/gcc/' where
11876 PREFIX is the value of `prefix' when you ran the `configure'
11879 Other prefixes specified with `-B' take precedence over this
11882 This prefix is also used for finding files such as `crt0.o' that
11883 are used for linking.
11885 In addition, the prefix is used in an unusual way in finding the
11886 directories to search for header files. For each of the standard
11887 directories whose name normally begins with `/usr/local/lib/gcc'
11888 (more precisely, with the value of `GCC_INCLUDE_DIR'), GCC tries
11889 replacing that beginning with the specified prefix to produce an
11890 alternate directory name. Thus, with `-Bfoo/', GCC will search
11891 `foo/bar' where it would normally search `/usr/local/lib/bar'.
11892 These alternate directories are searched first; the standard
11893 directories come next.
11896 The value of `COMPILER_PATH' is a colon-separated list of
11897 directories, much like `PATH'. GCC tries the directories thus
11898 specified when searching for subprograms, if it can't find the
11899 subprograms using `GCC_EXEC_PREFIX'.
11902 The value of `LIBRARY_PATH' is a colon-separated list of
11903 directories, much like `PATH'. When configured as a native
11904 compiler, GCC tries the directories thus specified when searching
11905 for special linker files, if it can't find them using
11906 `GCC_EXEC_PREFIX'. Linking using GCC also uses these directories
11907 when searching for ordinary libraries for the `-l' option (but
11908 directories specified with `-L' come first).
11911 This variable is used to pass locale information to the compiler.
11912 One way in which this information is used is to determine the
11913 character set to be used when character literals, string literals
11914 and comments are parsed in C and C++. When the compiler is
11915 configured to allow multibyte characters, the following values for
11916 `LANG' are recognized:
11919 Recognize JIS characters.
11922 Recognize SJIS characters.
11925 Recognize EUCJP characters.
11927 If `LANG' is not defined, or if it has some other value, then the
11928 compiler will use mblen and mbtowc as defined by the default
11929 locale to recognize and translate multibyte characters.
11931 Some additional environments variables affect the behavior of the
11936 `CPLUS_INCLUDE_PATH'
11937 `OBJC_INCLUDE_PATH'
11938 Each variable's value is a list of directories separated by a
11939 special character, much like `PATH', in which to look for header
11940 files. The special character, `PATH_SEPARATOR', is
11941 target-dependent and determined at GCC build time. For Microsoft
11942 Windows-based targets it is a semicolon, and for almost all other
11943 targets it is a colon.
11945 `CPATH' specifies a list of directories to be searched as if
11946 specified with `-I', but after any paths given with `-I' options
11947 on the command line. This environment variable is used regardless
11948 of which language is being preprocessed.
11950 The remaining environment variables apply only when preprocessing
11951 the particular language indicated. Each specifies a list of
11952 directories to be searched as if specified with `-isystem', but
11953 after any paths given with `-isystem' options on the command line.
11955 In all these variables, an empty element instructs the compiler to
11956 search its current working directory. Empty elements can appear
11957 at the beginning or end of a path. For instance, if the value of
11958 `CPATH' is `:/special/include', that has the same effect as
11959 `-I. -I/special/include'.
11961 `DEPENDENCIES_OUTPUT'
11962 If this variable is set, its value specifies how to output
11963 dependencies for Make based on the non-system header files
11964 processed by the compiler. System header files are ignored in the
11967 The value of `DEPENDENCIES_OUTPUT' can be just a file name, in
11968 which case the Make rules are written to that file, guessing the
11969 target name from the source file name. Or the value can have the
11970 form `FILE TARGET', in which case the rules are written to file
11971 FILE using TARGET as the target name.
11973 In other words, this environment variable is equivalent to
11974 combining the options `-MM' and `-MF' (*note Preprocessor
11975 Options::), with an optional `-MT' switch too.
11977 `SUNPRO_DEPENDENCIES'
11978 This variable is the same as `DEPENDENCIES_OUTPUT' (see above),
11979 except that system header files are not ignored, so it implies
11980 `-M' rather than `-MM'. However, the dependence on the main input
11981 file is omitted. *Note Preprocessor Options::.
11984 File: gcc.info, Node: Precompiled Headers, Next: Running Protoize, Prev: Environment Variables, Up: Invoking GCC
11986 3.20 Using Precompiled Headers
11987 ==============================
11989 Often large projects have many header files that are included in every
11990 source file. The time the compiler takes to process these header files
11991 over and over again can account for nearly all of the time required to
11992 build the project. To make builds faster, GCC allows users to
11993 `precompile' a header file; then, if builds can use the precompiled
11994 header file they will be much faster.
11996 *Caution:* There are a few known situations where GCC will crash when
11997 trying to use a precompiled header. If you have trouble with a
11998 precompiled header, you should remove the precompiled header and
11999 compile without it. In addition, please use GCC's on-line
12000 defect-tracking system to report any problems you encounter with
12001 precompiled headers. *Note Bugs::.
12003 To create a precompiled header file, simply compile it as you would any
12004 other file, if necessary using the `-x' option to make the driver treat
12005 it as a C or C++ header file. You will probably want to use a tool
12006 like `make' to keep the precompiled header up-to-date when the headers
12007 it contains change.
12009 A precompiled header file will be searched for when `#include' is seen
12010 in the compilation. As it searches for the included file (*note Search
12011 Path: (cpp)Search Path.) the compiler looks for a precompiled header in
12012 each directory just before it looks for the include file in that
12013 directory. The name searched for is the name specified in the
12014 `#include' with `.gch' appended. If the precompiled header file can't
12015 be used, it is ignored.
12017 For instance, if you have `#include "all.h"', and you have `all.h.gch'
12018 in the same directory as `all.h', then the precompiled header file will
12019 be used if possible, and the original header will be used otherwise.
12021 Alternatively, you might decide to put the precompiled header file in a
12022 directory and use `-I' to ensure that directory is searched before (or
12023 instead of) the directory containing the original header. Then, if you
12024 want to check that the precompiled header file is always used, you can
12025 put a file of the same name as the original header in this directory
12026 containing an `#error' command.
12028 This also works with `-include'. So yet another way to use
12029 precompiled headers, good for projects not designed with precompiled
12030 header files in mind, is to simply take most of the header files used by
12031 a project, include them from another header file, precompile that header
12032 file, and `-include' the precompiled header. If the header files have
12033 guards against multiple inclusion, they will be skipped because they've
12034 already been included (in the precompiled header).
12036 If you need to precompile the same header file for different
12037 languages, targets, or compiler options, you can instead make a
12038 _directory_ named like `all.h.gch', and put each precompiled header in
12039 the directory, perhaps using `-o'. It doesn't matter what you call the
12040 files in the directory, every precompiled header in the directory will
12041 be considered. The first precompiled header encountered in the
12042 directory that is valid for this compilation will be used; they're
12043 searched in no particular order.
12045 There are many other possibilities, limited only by your imagination,
12046 good sense, and the constraints of your build system.
12048 A precompiled header file can be used only when these conditions apply:
12050 * Only one precompiled header can be used in a particular
12053 * A precompiled header can't be used once the first C token is seen.
12054 You can have preprocessor directives before a precompiled header;
12055 you can even include a precompiled header from inside another
12056 header, so long as there are no C tokens before the `#include'.
12058 * The precompiled header file must be produced for the same language
12059 as the current compilation. You can't use a C precompiled header
12060 for a C++ compilation.
12062 * The precompiled header file must be produced by the same compiler
12063 version and configuration as the current compilation is using.
12064 The easiest way to guarantee this is to use the same compiler
12065 binary for creating and using precompiled headers.
12067 * Any macros defined before the precompiled header is included must
12068 either be defined in the same way as when the precompiled header
12069 was generated, or must not affect the precompiled header, which
12070 usually means that they don't appear in the precompiled header at
12073 The `-D' option is one way to define a macro before a precompiled
12074 header is included; using a `#define' can also do it. There are
12075 also some options that define macros implicitly, like `-O' and
12076 `-Wdeprecated'; the same rule applies to macros defined this way.
12078 * If debugging information is output when using the precompiled
12079 header, using `-g' or similar, the same kind of debugging
12080 information must have been output when building the precompiled
12081 header. However, a precompiled header built using `-g' can be
12082 used in a compilation when no debugging information is being
12085 * The same `-m' options must generally be used when building and
12086 using the precompiled header. *Note Submodel Options::, for any
12087 cases where this rule is relaxed.
12089 * Each of the following options must be the same when building and
12090 using the precompiled header:
12092 -fexceptions -funit-at-a-time
12094 * Some other command-line options starting with `-f', `-p', or `-O'
12095 must be defined in the same way as when the precompiled header was
12096 generated. At present, it's not clear which options are safe to
12097 change and which are not; the safest choice is to use exactly the
12098 same options when generating and using the precompiled header.
12099 The following are known to be safe:
12101 -fpreprocessed -pedantic-errors
12104 For all of these except the last, the compiler will automatically
12105 ignore the precompiled header if the conditions aren't met. If you
12106 find an option combination that doesn't work and doesn't cause the
12107 precompiled header to be ignored, please consider filing a bug report,
12110 If you do use differing options when generating and using the
12111 precompiled header, the actual behavior will be a mixture of the
12112 behavior for the options. For instance, if you use `-g' to generate
12113 the precompiled header but not when using it, you may or may not get
12114 debugging information for routines in the precompiled header.
12117 File: gcc.info, Node: Running Protoize, Prev: Precompiled Headers, Up: Invoking GCC
12119 3.21 Running Protoize
12120 =====================
12122 The program `protoize' is an optional part of GCC. You can use it to
12123 add prototypes to a program, thus converting the program to ISO C in
12124 one respect. The companion program `unprotoize' does the reverse: it
12125 removes argument types from any prototypes that are found.
12127 When you run these programs, you must specify a set of source files as
12128 command line arguments. The conversion programs start out by compiling
12129 these files to see what functions they define. The information gathered
12130 about a file FOO is saved in a file named `FOO.X'.
12132 After scanning comes actual conversion. The specified files are all
12133 eligible to be converted; any files they include (whether sources or
12134 just headers) are eligible as well.
12136 But not all the eligible files are converted. By default, `protoize'
12137 and `unprotoize' convert only source and header files in the current
12138 directory. You can specify additional directories whose files should
12139 be converted with the `-d DIRECTORY' option. You can also specify
12140 particular files to exclude with the `-x FILE' option. A file is
12141 converted if it is eligible, its directory name matches one of the
12142 specified directory names, and its name within the directory has not
12145 Basic conversion with `protoize' consists of rewriting most function
12146 definitions and function declarations to specify the types of the
12147 arguments. The only ones not rewritten are those for varargs functions.
12149 `protoize' optionally inserts prototype declarations at the beginning
12150 of the source file, to make them available for any calls that precede
12151 the function's definition. Or it can insert prototype declarations
12152 with block scope in the blocks where undeclared functions are called.
12154 Basic conversion with `unprotoize' consists of rewriting most function
12155 declarations to remove any argument types, and rewriting function
12156 definitions to the old-style pre-ISO form.
12158 Both conversion programs print a warning for any function declaration
12159 or definition that they can't convert. You can suppress these warnings
12162 The output from `protoize' or `unprotoize' replaces the original
12163 source file. The original file is renamed to a name ending with
12164 `.save' (for DOS, the saved filename ends in `.sav' without the
12165 original `.c' suffix). If the `.save' (`.sav' for DOS) file already
12166 exists, then the source file is simply discarded.
12168 `protoize' and `unprotoize' both depend on GCC itself to scan the
12169 program and collect information about the functions it uses. So
12170 neither of these programs will work until GCC is installed.
12172 Here is a table of the options you can use with `protoize' and
12173 `unprotoize'. Each option works with both programs unless otherwise
12177 Look for the file `SYSCALLS.c.X' in DIRECTORY, instead of the
12178 usual directory (normally `/usr/local/lib'). This file contains
12179 prototype information about standard system functions. This option
12180 applies only to `protoize'.
12182 `-c COMPILATION-OPTIONS'
12183 Use COMPILATION-OPTIONS as the options when running `gcc' to
12184 produce the `.X' files. The special option `-aux-info' is always
12185 passed in addition, to tell `gcc' to write a `.X' file.
12187 Note that the compilation options must be given as a single
12188 argument to `protoize' or `unprotoize'. If you want to specify
12189 several `gcc' options, you must quote the entire set of
12190 compilation options to make them a single word in the shell.
12192 There are certain `gcc' arguments that you cannot use, because they
12193 would produce the wrong kind of output. These include `-g', `-O',
12194 `-c', `-S', and `-o' If you include these in the
12195 COMPILATION-OPTIONS, they are ignored.
12198 Rename files to end in `.C' (`.cc' for DOS-based file systems)
12199 instead of `.c'. This is convenient if you are converting a C
12200 program to C++. This option applies only to `protoize'.
12203 Add explicit global declarations. This means inserting explicit
12204 declarations at the beginning of each source file for each function
12205 that is called in the file and was not declared. These
12206 declarations precede the first function definition that contains a
12207 call to an undeclared function. This option applies only to
12211 Indent old-style parameter declarations with the string STRING.
12212 This option applies only to `protoize'.
12214 `unprotoize' converts prototyped function definitions to old-style
12215 function definitions, where the arguments are declared between the
12216 argument list and the initial `{'. By default, `unprotoize' uses
12217 five spaces as the indentation. If you want to indent with just
12218 one space instead, use `-i " "'.
12221 Keep the `.X' files. Normally, they are deleted after conversion
12225 Add explicit local declarations. `protoize' with `-l' inserts a
12226 prototype declaration for each function in each block which calls
12227 the function without any declaration. This option applies only to
12231 Make no real changes. This mode just prints information about the
12232 conversions that would have been done without `-n'.
12235 Make no `.save' files. The original files are simply deleted.
12236 Use this option with caution.
12239 Use the program PROGRAM as the compiler. Normally, the name `gcc'
12243 Work quietly. Most warnings are suppressed.
12246 Print the version number, just like `-v' for `gcc'.
12248 If you need special compiler options to compile one of your program's
12249 source files, then you should generate that file's `.X' file specially,
12250 by running `gcc' on that source file with the appropriate options and
12251 the option `-aux-info'. Then run `protoize' on the entire set of
12252 files. `protoize' will use the existing `.X' file because it is newer
12253 than the source file. For example:
12255 gcc -Dfoo=bar file1.c -aux-info file1.X
12258 You need to include the special files along with the rest in the
12259 `protoize' command, even though their `.X' files already exist, because
12260 otherwise they won't get converted.
12262 *Note Protoize Caveats::, for more information on how to use
12263 `protoize' successfully.
12266 File: gcc.info, Node: C Implementation, Next: C Extensions, Prev: Invoking GCC, Up: Top
12268 4 C Implementation-defined behavior
12269 ***********************************
12271 A conforming implementation of ISO C is required to document its choice
12272 of behavior in each of the areas that are designated "implementation
12273 defined". The following lists all such areas, along with the section
12274 numbers from the ISO/IEC 9899:1990 and ISO/IEC 9899:1999 standards.
12275 Some areas are only implementation-defined in one version of the
12278 Some choices depend on the externally determined ABI for the platform
12279 (including standard character encodings) which GCC follows; these are
12280 listed as "determined by ABI" below. *Note Binary Compatibility:
12281 Compatibility, and `http://gcc.gnu.org/readings.html'. Some choices
12282 are documented in the preprocessor manual. *Note
12283 Implementation-defined behavior: (cpp)Implementation-defined behavior.
12284 Some choices are made by the library and operating system (or other
12285 environment when compiling for a freestanding environment); refer to
12286 their documentation for details.
12290 * Translation implementation::
12291 * Environment implementation::
12292 * Identifiers implementation::
12293 * Characters implementation::
12294 * Integers implementation::
12295 * Floating point implementation::
12296 * Arrays and pointers implementation::
12297 * Hints implementation::
12298 * Structures unions enumerations and bit-fields implementation::
12299 * Qualifiers implementation::
12300 * Declarators implementation::
12301 * Statements implementation::
12302 * Preprocessing directives implementation::
12303 * Library functions implementation::
12304 * Architecture implementation::
12305 * Locale-specific behavior implementation::
12308 File: gcc.info, Node: Translation implementation, Next: Environment implementation, Up: C Implementation
12313 * `How a diagnostic is identified (C90 3.7, C99 3.10, C90 and C99
12316 Diagnostics consist of all the output sent to stderr by GCC.
12318 * `Whether each nonempty sequence of white-space characters other
12319 than new-line is retained or replaced by one space character in
12320 translation phase 3 (C90 and C99 5.1.1.2).'
12322 *Note Implementation-defined behavior: (cpp)Implementation-defined
12327 File: gcc.info, Node: Environment implementation, Next: Identifiers implementation, Prev: Translation implementation, Up: C Implementation
12332 The behavior of most of these points are dependent on the implementation
12333 of the C library, and are not defined by GCC itself.
12335 * `The mapping between physical source file multibyte characters and
12336 the source character set in translation phase 1 (C90 and C99
12339 *Note Implementation-defined behavior: (cpp)Implementation-defined
12344 File: gcc.info, Node: Identifiers implementation, Next: Characters implementation, Prev: Environment implementation, Up: C Implementation
12349 * `Which additional multibyte characters may appear in identifiers
12350 and their correspondence to universal character names (C99 6.4.2).'
12352 *Note Implementation-defined behavior: (cpp)Implementation-defined
12355 * `The number of significant initial characters in an identifier
12356 (C90 6.1.2, C90 and C99 5.2.4.1, C99 6.4.2).'
12358 For internal names, all characters are significant. For external
12359 names, the number of significant characters are defined by the
12360 linker; for almost all targets, all characters are significant.
12362 * `Whether case distinctions are significant in an identifier with
12363 external linkage (C90 6.1.2).'
12365 This is a property of the linker. C99 requires that case
12366 distinctions are always significant in identifiers with external
12367 linkage and systems without this property are not supported by GCC.
12371 File: gcc.info, Node: Characters implementation, Next: Integers implementation, Prev: Identifiers implementation, Up: C Implementation
12376 * `The number of bits in a byte (C90 3.4, C99 3.6).'
12380 * `The values of the members of the execution character set (C90 and
12385 * `The unique value of the member of the execution character set
12386 produced for each of the standard alphabetic escape sequences (C90
12391 * `The value of a `char' object into which has been stored any
12392 character other than a member of the basic execution character set
12393 (C90 6.1.2.5, C99 6.2.5).'
12397 * `Which of `signed char' or `unsigned char' has the same range,
12398 representation, and behavior as "plain" `char' (C90 6.1.2.5, C90
12399 6.2.1.1, C99 6.2.5, C99 6.3.1.1).'
12401 Determined by ABI. The options `-funsigned-char' and
12402 `-fsigned-char' change the default. *Note Options Controlling C
12403 Dialect: C Dialect Options.
12405 * `The mapping of members of the source character set (in character
12406 constants and string literals) to members of the execution
12407 character set (C90 6.1.3.4, C99 6.4.4.4, C90 and C99 5.1.1.2).'
12411 * `The value of an integer character constant containing more than
12412 one character or containing a character or escape sequence that
12413 does not map to a single-byte execution character (C90 6.1.3.4,
12416 *Note Implementation-defined behavior: (cpp)Implementation-defined
12419 * `The value of a wide character constant containing more than one
12420 multibyte character, or containing a multibyte character or escape
12421 sequence not represented in the extended execution character set
12422 (C90 6.1.3.4, C99 6.4.4.4).'
12424 *Note Implementation-defined behavior: (cpp)Implementation-defined
12427 * `The current locale used to convert a wide character constant
12428 consisting of a single multibyte character that maps to a member
12429 of the extended execution character set into a corresponding wide
12430 character code (C90 6.1.3.4, C99 6.4.4.4).'
12432 *Note Implementation-defined behavior: (cpp)Implementation-defined
12435 * `The current locale used to convert a wide string literal into
12436 corresponding wide character codes (C90 6.1.4, C99 6.4.5).'
12438 *Note Implementation-defined behavior: (cpp)Implementation-defined
12441 * `The value of a string literal containing a multibyte character or
12442 escape sequence not represented in the execution character set
12443 (C90 6.1.4, C99 6.4.5).'
12445 *Note Implementation-defined behavior: (cpp)Implementation-defined
12449 File: gcc.info, Node: Integers implementation, Next: Floating point implementation, Prev: Characters implementation, Up: C Implementation
12454 * `Any extended integer types that exist in the implementation (C99
12457 GCC does not support any extended integer types.
12459 * `Whether signed integer types are represented using sign and
12460 magnitude, two's complement, or one's complement, and whether the
12461 extraordinary value is a trap representation or an ordinary value
12464 GCC supports only two's complement integer types, and all bit
12465 patterns are ordinary values.
12467 * `The rank of any extended integer type relative to another extended
12468 integer type with the same precision (C99 6.3.1.1).'
12470 GCC does not support any extended integer types.
12472 * `The result of, or the signal raised by, converting an integer to a
12473 signed integer type when the value cannot be represented in an
12474 object of that type (C90 6.2.1.2, C99 6.3.1.3).'
12476 For conversion to a type of width N, the value is reduced modulo
12477 2^N to be within range of the type; no signal is raised.
12479 * `The results of some bitwise operations on signed integers (C90
12482 Bitwise operators act on the representation of the value including
12483 both the sign and value bits, where the sign bit is considered
12484 immediately above the highest-value value bit. Signed `>>' acts
12485 on negative numbers by sign extension.
12487 GCC does not use the latitude given in C99 only to treat certain
12488 aspects of signed `<<' as undefined, but this is subject to change.
12490 * `The sign of the remainder on integer division (C90 6.3.5).'
12492 GCC always follows the C99 requirement that the result of division
12493 is truncated towards zero.
12497 File: gcc.info, Node: Floating point implementation, Next: Arrays and pointers implementation, Prev: Integers implementation, Up: C Implementation
12502 * `The accuracy of the floating-point operations and of the library
12503 functions in `<math.h>' and `<complex.h>' that return
12504 floating-point results (C90 and C99 5.2.4.2.2).'
12506 The accuracy is unknown.
12508 * `The rounding behaviors characterized by non-standard values of
12509 `FLT_ROUNDS' (C90 and C99 5.2.4.2.2).'
12511 GCC does not use such values.
12513 * `The evaluation methods characterized by non-standard negative
12514 values of `FLT_EVAL_METHOD' (C99 5.2.4.2.2).'
12516 GCC does not use such values.
12518 * `The direction of rounding when an integer is converted to a
12519 floating-point number that cannot exactly represent the original
12520 value (C90 6.2.1.3, C99 6.3.1.4).'
12522 C99 Annex F is followed.
12524 * `The direction of rounding when a floating-point number is
12525 converted to a narrower floating-point number (C90 6.2.1.4, C99
12528 C99 Annex F is followed.
12530 * `How the nearest representable value or the larger or smaller
12531 representable value immediately adjacent to the nearest
12532 representable value is chosen for certain floating constants (C90
12533 6.1.3.1, C99 6.4.4.2).'
12535 C99 Annex F is followed.
12537 * `Whether and how floating expressions are contracted when not
12538 disallowed by the `FP_CONTRACT' pragma (C99 6.5).'
12540 Expressions are currently only contracted if
12541 `-funsafe-math-optimizations' or `-ffast-math' are used. This is
12544 * `The default state for the `FENV_ACCESS' pragma (C99 7.6.1).'
12546 This pragma is not implemented, but the default is to "off" unless
12547 `-frounding-math' is used in which case it is "on".
12549 * `Additional floating-point exceptions, rounding modes,
12550 environments, and classifications, and their macro names (C99 7.6,
12553 This is dependent on the implementation of the C library, and is
12554 not defined by GCC itself.
12556 * `The default state for the `FP_CONTRACT' pragma (C99 7.12.2).'
12558 This pragma is not implemented. Expressions are currently only
12559 contracted if `-funsafe-math-optimizations' or `-ffast-math' are
12560 used. This is subject to change.
12562 * `Whether the "inexact" floating-point exception can be raised when
12563 the rounded result actually does equal the mathematical result in
12564 an IEC 60559 conformant implementation (C99 F.9).'
12566 This is dependent on the implementation of the C library, and is
12567 not defined by GCC itself.
12569 * `Whether the "underflow" (and "inexact") floating-point exception
12570 can be raised when a result is tiny but not inexact in an IEC
12571 60559 conformant implementation (C99 F.9).'
12573 This is dependent on the implementation of the C library, and is
12574 not defined by GCC itself.
12578 File: gcc.info, Node: Arrays and pointers implementation, Next: Hints implementation, Prev: Floating point implementation, Up: C Implementation
12580 4.7 Arrays and pointers
12581 =======================
12583 * `The result of converting a pointer to an integer or vice versa
12584 (C90 6.3.4, C99 6.3.2.3).'
12586 A cast from pointer to integer discards most-significant bits if
12587 the pointer representation is larger than the integer type,
12588 sign-extends(1) if the pointer representation is smaller than the
12589 integer type, otherwise the bits are unchanged.
12591 A cast from integer to pointer discards most-significant bits if
12592 the pointer representation is smaller than the integer type,
12593 extends according to the signedness of the integer type if the
12594 pointer representation is larger than the integer type, otherwise
12595 the bits are unchanged.
12597 When casting from pointer to integer and back again, the resulting
12598 pointer must reference the same object as the original pointer,
12599 otherwise the behavior is undefined. That is, one may not use
12600 integer arithmetic to avoid the undefined behavior of pointer
12601 arithmetic as proscribed in C99 6.5.6/8.
12603 * `The size of the result of subtracting two pointers to elements of
12604 the same array (C90 6.3.6, C99 6.5.6).'
12606 The value is as specified in the standard and the type is
12607 determined by the ABI.
12610 ---------- Footnotes ----------
12612 (1) Future versions of GCC may zero-extend, or use a target-defined
12613 `ptr_extend' pattern. Do not rely on sign extension.
12616 File: gcc.info, Node: Hints implementation, Next: Structures unions enumerations and bit-fields implementation, Prev: Arrays and pointers implementation, Up: C Implementation
12621 * `The extent to which suggestions made by using the `register'
12622 storage-class specifier are effective (C90 6.5.1, C99 6.7.1).'
12624 The `register' specifier affects code generation only in these
12627 * When used as part of the register variable extension, see
12628 *Note Explicit Reg Vars::.
12630 * When `-O0' is in use, the compiler allocates distinct stack
12631 memory for all variables that do not have the `register'
12632 storage-class specifier; if `register' is specified, the
12633 variable may have a shorter lifespan than the code would
12634 indicate and may never be placed in memory.
12636 * On some rare x86 targets, `setjmp' doesn't save the registers
12637 in all circumstances. In those cases, GCC doesn't allocate
12638 any variables in registers unless they are marked `register'.
12641 * `The extent to which suggestions made by using the inline function
12642 specifier are effective (C99 6.7.4).'
12644 GCC will not inline any functions if the `-fno-inline' option is
12645 used or if `-O0' is used. Otherwise, GCC may still be unable to
12646 inline a function for many reasons; the `-Winline' option may be
12647 used to determine if a function has not been inlined and why not.
12651 File: gcc.info, Node: Structures unions enumerations and bit-fields implementation, Next: Qualifiers implementation, Prev: Hints implementation, Up: C Implementation
12653 4.9 Structures, unions, enumerations, and bit-fields
12654 ====================================================
12656 * `A member of a union object is accessed using a member of a
12657 different type (C90 6.3.2.3).'
12659 The relevant bytes of the representation of the object are treated
12660 as an object of the type used for the access. This may be a trap
12663 * `Whether a "plain" `int' bit-field is treated as a `signed int'
12664 bit-field or as an `unsigned int' bit-field (C90 6.5.2, C90
12665 6.5.2.1, C99 6.7.2, C99 6.7.2.1).'
12667 By default it is treated as `signed int' but this may be changed
12668 by the `-funsigned-bitfields' option.
12670 * `Allowable bit-field types other than `_Bool', `signed int', and
12671 `unsigned int' (C99 6.7.2.1).'
12673 No other types are permitted in strictly conforming mode.
12675 * `Whether a bit-field can straddle a storage-unit boundary (C90
12676 6.5.2.1, C99 6.7.2.1).'
12680 * `The order of allocation of bit-fields within a unit (C90 6.5.2.1,
12685 * `The alignment of non-bit-field members of structures (C90
12686 6.5.2.1, C99 6.7.2.1).'
12690 * `The integer type compatible with each enumerated type (C90
12691 6.5.2.2, C99 6.7.2.2).'
12693 Normally, the type is `unsigned int' if there are no negative
12694 values in the enumeration, otherwise `int'. If `-fshort-enums' is
12695 specified, then if there are negative values it is the first of
12696 `signed char', `short' and `int' that can represent all the
12697 values, otherwise it is the first of `unsigned char', `unsigned
12698 short' and `unsigned int' that can represent all the values.
12700 On some targets, `-fshort-enums' is the default; this is
12701 determined by the ABI.
12705 File: gcc.info, Node: Qualifiers implementation, Next: Declarators implementation, Prev: Structures unions enumerations and bit-fields implementation, Up: C Implementation
12710 * `What constitutes an access to an object that has
12711 volatile-qualified type (C90 6.5.3, C99 6.7.3).'
12713 *Note When is a Volatile Object Accessed?: Volatiles.
12717 File: gcc.info, Node: Declarators implementation, Next: Statements implementation, Prev: Qualifiers implementation, Up: C Implementation
12722 * `The maximum number of declarators that may modify an arithmetic,
12723 structure or union type (C90 6.5.4).'
12725 GCC is only limited by available memory.
12729 File: gcc.info, Node: Statements implementation, Next: Preprocessing directives implementation, Prev: Declarators implementation, Up: C Implementation
12734 * `The maximum number of `case' values in a `switch' statement (C90
12737 GCC is only limited by available memory.
12741 File: gcc.info, Node: Preprocessing directives implementation, Next: Library functions implementation, Prev: Statements implementation, Up: C Implementation
12743 4.13 Preprocessing directives
12744 =============================
12746 *Note Implementation-defined behavior: (cpp)Implementation-defined
12747 behavior, for details of these aspects of implementation-defined
12750 * `How sequences in both forms of header names are mapped to headers
12751 or external source file names (C90 6.1.7, C99 6.4.7).'
12753 * `Whether the value of a character constant in a constant expression
12754 that controls conditional inclusion matches the value of the same
12755 character constant in the execution character set (C90 6.8.1, C99
12758 * `Whether the value of a single-character character constant in a
12759 constant expression that controls conditional inclusion may have a
12760 negative value (C90 6.8.1, C99 6.10.1).'
12762 * `The places that are searched for an included `<>' delimited
12763 header, and how the places are specified or the header is
12764 identified (C90 6.8.2, C99 6.10.2).'
12766 * `How the named source file is searched for in an included `""'
12767 delimited header (C90 6.8.2, C99 6.10.2).'
12769 * `The method by which preprocessing tokens (possibly resulting from
12770 macro expansion) in a `#include' directive are combined into a
12771 header name (C90 6.8.2, C99 6.10.2).'
12773 * `The nesting limit for `#include' processing (C90 6.8.2, C99
12776 * `Whether the `#' operator inserts a `\' character before the `\'
12777 character that begins a universal character name in a character
12778 constant or string literal (C99 6.10.3.2).'
12780 * `The behavior on each recognized non-`STDC #pragma' directive (C90
12781 6.8.6, C99 6.10.6).'
12783 *Note Pragmas: (cpp)Pragmas, for details of pragmas accepted by
12784 GCC on all targets. *Note Pragmas Accepted by GCC: Pragmas, for
12785 details of target-specific pragmas.
12787 * `The definitions for `__DATE__' and `__TIME__' when respectively,
12788 the date and time of translation are not available (C90 6.8.8, C99
12793 File: gcc.info, Node: Library functions implementation, Next: Architecture implementation, Prev: Preprocessing directives implementation, Up: C Implementation
12795 4.14 Library functions
12796 ======================
12798 The behavior of most of these points are dependent on the implementation
12799 of the C library, and are not defined by GCC itself.
12801 * `The null pointer constant to which the macro `NULL' expands (C90
12804 In `<stddef.h>', `NULL' expands to `((void *)0)'. GCC does not
12805 provide the other headers which define `NULL' and some library
12806 implementations may use other definitions in those headers.
12810 File: gcc.info, Node: Architecture implementation, Next: Locale-specific behavior implementation, Prev: Library functions implementation, Up: C Implementation
12815 * `The values or expressions assigned to the macros specified in the
12816 headers `<float.h>', `<limits.h>', and `<stdint.h>' (C90 and C99
12817 5.2.4.2, C99 7.18.2, C99 7.18.3).'
12821 * `The number, order, and encoding of bytes in any object (when not
12822 explicitly specified in this International Standard) (C99
12827 * `The value of the result of the `sizeof' operator (C90 6.3.3.4,
12834 File: gcc.info, Node: Locale-specific behavior implementation, Prev: Architecture implementation, Up: C Implementation
12836 4.16 Locale-specific behavior
12837 =============================
12839 The behavior of these points are dependent on the implementation of the
12840 C library, and are not defined by GCC itself.
12843 File: gcc.info, Node: C Extensions, Next: C++ Extensions, Prev: C Implementation, Up: Top
12845 5 Extensions to the C Language Family
12846 *************************************
12848 GNU C provides several language features not found in ISO standard C.
12849 (The `-pedantic' option directs GCC to print a warning message if any
12850 of these features is used.) To test for the availability of these
12851 features in conditional compilation, check for a predefined macro
12852 `__GNUC__', which is always defined under GCC.
12854 These extensions are available in C and Objective-C. Most of them are
12855 also available in C++. *Note Extensions to the C++ Language: C++
12856 Extensions, for extensions that apply _only_ to C++.
12858 Some features that are in ISO C99 but not C89 or C++ are also, as
12859 extensions, accepted by GCC in C89 mode and in C++.
12863 * Statement Exprs:: Putting statements and declarations inside expressions.
12864 * Local Labels:: Labels local to a block.
12865 * Labels as Values:: Getting pointers to labels, and computed gotos.
12866 * Nested Functions:: As in Algol and Pascal, lexical scoping of functions.
12867 * Constructing Calls:: Dispatching a call to another function.
12868 * Typeof:: `typeof': referring to the type of an expression.
12869 * Conditionals:: Omitting the middle operand of a `?:' expression.
12870 * Long Long:: Double-word integers---`long long int'.
12871 * Complex:: Data types for complex numbers.
12872 * Hex Floats:: Hexadecimal floating-point constants.
12873 * Zero Length:: Zero-length arrays.
12874 * Variable Length:: Arrays whose length is computed at run time.
12875 * Empty Structures:: Structures with no members.
12876 * Variadic Macros:: Macros with a variable number of arguments.
12877 * Escaped Newlines:: Slightly looser rules for escaped newlines.
12878 * Subscripting:: Any array can be subscripted, even if not an lvalue.
12879 * Pointer Arith:: Arithmetic on `void'-pointers and function pointers.
12880 * Initializers:: Non-constant initializers.
12881 * Compound Literals:: Compound literals give structures, unions
12882 or arrays as values.
12883 * Designated Inits:: Labeling elements of initializers.
12884 * Cast to Union:: Casting to union type from any member of the union.
12885 * Case Ranges:: `case 1 ... 9' and such.
12886 * Mixed Declarations:: Mixing declarations and code.
12887 * Function Attributes:: Declaring that functions have no side effects,
12888 or that they can never return.
12889 * Attribute Syntax:: Formal syntax for attributes.
12890 * Function Prototypes:: Prototype declarations and old-style definitions.
12891 * C++ Comments:: C++ comments are recognized.
12892 * Dollar Signs:: Dollar sign is allowed in identifiers.
12893 * Character Escapes:: `\e' stands for the character <ESC>.
12894 * Variable Attributes:: Specifying attributes of variables.
12895 * Type Attributes:: Specifying attributes of types.
12896 * Alignment:: Inquiring about the alignment of a type or variable.
12897 * Inline:: Defining inline functions (as fast as macros).
12898 * Extended Asm:: Assembler instructions with C expressions as operands.
12899 (With them you can define ``built-in'' functions.)
12900 * Constraints:: Constraints for asm operands
12901 * Asm Labels:: Specifying the assembler name to use for a C symbol.
12902 * Explicit Reg Vars:: Defining variables residing in specified registers.
12903 * Alternate Keywords:: `__const__', `__asm__', etc., for header files.
12904 * Incomplete Enums:: `enum foo;', with details to follow.
12905 * Function Names:: Printable strings which are the name of the current
12907 * Return Address:: Getting the return or frame address of a function.
12908 * Vector Extensions:: Using vector instructions through built-in functions.
12909 * Offsetof:: Special syntax for implementing `offsetof'.
12910 * Other Builtins:: Other built-in functions.
12911 * Target Builtins:: Built-in functions specific to particular targets.
12912 * Target Format Checks:: Format checks specific to particular targets.
12913 * Pragmas:: Pragmas accepted by GCC.
12914 * Unnamed Fields:: Unnamed struct/union fields within structs/unions.
12915 * Thread-Local:: Per-thread variables.
12918 File: gcc.info, Node: Statement Exprs, Next: Local Labels, Up: C Extensions
12920 5.1 Statements and Declarations in Expressions
12921 ==============================================
12923 A compound statement enclosed in parentheses may appear as an expression
12924 in GNU C. This allows you to use loops, switches, and local variables
12925 within an expression.
12927 Recall that a compound statement is a sequence of statements surrounded
12928 by braces; in this construct, parentheses go around the braces. For
12931 ({ int y = foo (); int z;
12936 is a valid (though slightly more complex than necessary) expression for
12937 the absolute value of `foo ()'.
12939 The last thing in the compound statement should be an expression
12940 followed by a semicolon; the value of this subexpression serves as the
12941 value of the entire construct. (If you use some other kind of statement
12942 last within the braces, the construct has type `void', and thus
12943 effectively no value.)
12945 This feature is especially useful in making macro definitions "safe"
12946 (so that they evaluate each operand exactly once). For example, the
12947 "maximum" function is commonly defined as a macro in standard C as
12950 #define max(a,b) ((a) > (b) ? (a) : (b))
12952 But this definition computes either A or B twice, with bad results if
12953 the operand has side effects. In GNU C, if you know the type of the
12954 operands (here taken as `int'), you can define the macro safely as
12957 #define maxint(a,b) \
12958 ({int _a = (a), _b = (b); _a > _b ? _a : _b; })
12960 Embedded statements are not allowed in constant expressions, such as
12961 the value of an enumeration constant, the width of a bit-field, or the
12962 initial value of a static variable.
12964 If you don't know the type of the operand, you can still do this, but
12965 you must use `typeof' (*note Typeof::).
12967 In G++, the result value of a statement expression undergoes array and
12968 function pointer decay, and is returned by value to the enclosing
12969 expression. For instance, if `A' is a class, then
12975 will construct a temporary `A' object to hold the result of the
12976 statement expression, and that will be used to invoke `Foo'. Therefore
12977 the `this' pointer observed by `Foo' will not be the address of `a'.
12979 Any temporaries created within a statement within a statement
12980 expression will be destroyed at the statement's end. This makes
12981 statement expressions inside macros slightly different from function
12982 calls. In the latter case temporaries introduced during argument
12983 evaluation will be destroyed at the end of the statement that includes
12984 the function call. In the statement expression case they will be
12985 destroyed during the statement expression. For instance,
12987 #define macro(a) ({__typeof__(a) b = (a); b + 3; })
12988 template<typename T> T function(T a) { T b = a; return b + 3; }
12996 will have different places where temporaries are destroyed. For the
12997 `macro' case, the temporary `X' will be destroyed just after the
12998 initialization of `b'. In the `function' case that temporary will be
12999 destroyed when the function returns.
13001 These considerations mean that it is probably a bad idea to use
13002 statement-expressions of this form in header files that are designed to
13003 work with C++. (Note that some versions of the GNU C Library contained
13004 header files using statement-expression that lead to precisely this
13007 Jumping into a statement expression with `goto' or using a `switch'
13008 statement outside the statement expression with a `case' or `default'
13009 label inside the statement expression is not permitted. Jumping into a
13010 statement expression with a computed `goto' (*note Labels as Values::)
13011 yields undefined behavior. Jumping out of a statement expression is
13012 permitted, but if the statement expression is part of a larger
13013 expression then it is unspecified which other subexpressions of that
13014 expression have been evaluated except where the language definition
13015 requires certain subexpressions to be evaluated before or after the
13016 statement expression. In any case, as with a function call the
13017 evaluation of a statement expression is not interleaved with the
13018 evaluation of other parts of the containing expression. For example,
13020 foo (), (({ bar1 (); goto a; 0; }) + bar2 ()), baz();
13022 will call `foo' and `bar1' and will not call `baz' but may or may not
13023 call `bar2'. If `bar2' is called, it will be called after `foo' and
13027 File: gcc.info, Node: Local Labels, Next: Labels as Values, Prev: Statement Exprs, Up: C Extensions
13029 5.2 Locally Declared Labels
13030 ===========================
13032 GCC allows you to declare "local labels" in any nested block scope. A
13033 local label is just like an ordinary label, but you can only reference
13034 it (with a `goto' statement, or by taking its address) within the block
13035 in which it was declared.
13037 A local label declaration looks like this:
13043 __label__ LABEL1, LABEL2, /* ... */;
13045 Local label declarations must come at the beginning of the block,
13046 before any ordinary declarations or statements.
13048 The label declaration defines the label _name_, but does not define
13049 the label itself. You must do this in the usual way, with `LABEL:',
13050 within the statements of the statement expression.
13052 The local label feature is useful for complex macros. If a macro
13053 contains nested loops, a `goto' can be useful for breaking out of them.
13054 However, an ordinary label whose scope is the whole function cannot be
13055 used: if the macro can be expanded several times in one function, the
13056 label will be multiply defined in that function. A local label avoids
13057 this problem. For example:
13059 #define SEARCH(value, array, target) \
13062 typeof (target) _SEARCH_target = (target); \
13063 typeof (*(array)) *_SEARCH_array = (array); \
13066 for (i = 0; i < max; i++) \
13067 for (j = 0; j < max; j++) \
13068 if (_SEARCH_array[i][j] == _SEARCH_target) \
13069 { (value) = i; goto found; } \
13074 This could also be written using a statement-expression:
13076 #define SEARCH(array, target) \
13079 typeof (target) _SEARCH_target = (target); \
13080 typeof (*(array)) *_SEARCH_array = (array); \
13083 for (i = 0; i < max; i++) \
13084 for (j = 0; j < max; j++) \
13085 if (_SEARCH_array[i][j] == _SEARCH_target) \
13086 { value = i; goto found; } \
13092 Local label declarations also make the labels they declare visible to
13093 nested functions, if there are any. *Note Nested Functions::, for
13097 File: gcc.info, Node: Labels as Values, Next: Nested Functions, Prev: Local Labels, Up: C Extensions
13099 5.3 Labels as Values
13100 ====================
13102 You can get the address of a label defined in the current function (or
13103 a containing function) with the unary operator `&&'. The value has
13104 type `void *'. This value is a constant and can be used wherever a
13105 constant of that type is valid. For example:
13111 To use these values, you need to be able to jump to one. This is done
13112 with the computed goto statement(1), `goto *EXP;'. For example,
13116 Any expression of type `void *' is allowed.
13118 One way of using these constants is in initializing a static array that
13119 will serve as a jump table:
13121 static void *array[] = { &&foo, &&bar, &&hack };
13123 Then you can select a label with indexing, like this:
13127 Note that this does not check whether the subscript is in bounds--array
13128 indexing in C never does that.
13130 Such an array of label values serves a purpose much like that of the
13131 `switch' statement. The `switch' statement is cleaner, so use that
13132 rather than an array unless the problem does not fit a `switch'
13133 statement very well.
13135 Another use of label values is in an interpreter for threaded code.
13136 The labels within the interpreter function can be stored in the
13137 threaded code for super-fast dispatching.
13139 You may not use this mechanism to jump to code in a different function.
13140 If you do that, totally unpredictable things will happen. The best way
13141 to avoid this is to store the label address only in automatic variables
13142 and never pass it as an argument.
13144 An alternate way to write the above example is
13146 static const int array[] = { &&foo - &&foo, &&bar - &&foo,
13148 goto *(&&foo + array[i]);
13150 This is more friendly to code living in shared libraries, as it reduces
13151 the number of dynamic relocations that are needed, and by consequence,
13152 allows the data to be read-only.
13154 ---------- Footnotes ----------
13156 (1) The analogous feature in Fortran is called an assigned goto, but
13157 that name seems inappropriate in C, where one can do more than simply
13158 store label addresses in label variables.
13161 File: gcc.info, Node: Nested Functions, Next: Constructing Calls, Prev: Labels as Values, Up: C Extensions
13163 5.4 Nested Functions
13164 ====================
13166 A "nested function" is a function defined inside another function.
13167 (Nested functions are not supported for GNU C++.) The nested function's
13168 name is local to the block where it is defined. For example, here we
13169 define a nested function named `square', and call it twice:
13171 foo (double a, double b)
13173 double square (double z) { return z * z; }
13175 return square (a) + square (b);
13178 The nested function can access all the variables of the containing
13179 function that are visible at the point of its definition. This is
13180 called "lexical scoping". For example, here we show a nested function
13181 which uses an inherited variable named `offset':
13183 bar (int *array, int offset, int size)
13185 int access (int *array, int index)
13186 { return array[index + offset]; }
13189 for (i = 0; i < size; i++)
13190 /* ... */ access (array, i) /* ... */
13193 Nested function definitions are permitted within functions in the
13194 places where variable definitions are allowed; that is, in any block,
13195 mixed with the other declarations and statements in the block.
13197 It is possible to call the nested function from outside the scope of
13198 its name by storing its address or passing the address to another
13201 hack (int *array, int size)
13203 void store (int index, int value)
13204 { array[index] = value; }
13206 intermediate (store, size);
13209 Here, the function `intermediate' receives the address of `store' as
13210 an argument. If `intermediate' calls `store', the arguments given to
13211 `store' are used to store into `array'. But this technique works only
13212 so long as the containing function (`hack', in this example) does not
13215 If you try to call the nested function through its address after the
13216 containing function has exited, all hell will break loose. If you try
13217 to call it after a containing scope level has exited, and if it refers
13218 to some of the variables that are no longer in scope, you may be lucky,
13219 but it's not wise to take the risk. If, however, the nested function
13220 does not refer to anything that has gone out of scope, you should be
13223 GCC implements taking the address of a nested function using a
13224 technique called "trampolines". A paper describing them is available as
13226 `http://people.debian.org/~aaronl/Usenix88-lexic.pdf'.
13228 A nested function can jump to a label inherited from a containing
13229 function, provided the label was explicitly declared in the containing
13230 function (*note Local Labels::). Such a jump returns instantly to the
13231 containing function, exiting the nested function which did the `goto'
13232 and any intermediate functions as well. Here is an example:
13234 bar (int *array, int offset, int size)
13237 int access (int *array, int index)
13241 return array[index + offset];
13245 for (i = 0; i < size; i++)
13246 /* ... */ access (array, i) /* ... */
13250 /* Control comes here from `access'
13251 if it detects an error. */
13256 A nested function always has no linkage. Declaring one with `extern'
13257 or `static' is erroneous. If you need to declare the nested function
13258 before its definition, use `auto' (which is otherwise meaningless for
13259 function declarations).
13261 bar (int *array, int offset, int size)
13264 auto int access (int *, int);
13266 int access (int *array, int index)
13270 return array[index + offset];
13276 File: gcc.info, Node: Constructing Calls, Next: Typeof, Prev: Nested Functions, Up: C Extensions
13278 5.5 Constructing Function Calls
13279 ===============================
13281 Using the built-in functions described below, you can record the
13282 arguments a function received, and call another function with the same
13283 arguments, without knowing the number or types of the arguments.
13285 You can also record the return value of that function call, and later
13286 return that value, without knowing what data type the function tried to
13287 return (as long as your caller expects that data type).
13289 However, these built-in functions may interact badly with some
13290 sophisticated features or other extensions of the language. It is,
13291 therefore, not recommended to use them outside very simple functions
13292 acting as mere forwarders for their arguments.
13294 -- Built-in Function: void * __builtin_apply_args ()
13295 This built-in function returns a pointer to data describing how to
13296 perform a call with the same arguments as were passed to the
13299 The function saves the arg pointer register, structure value
13300 address, and all registers that might be used to pass arguments to
13301 a function into a block of memory allocated on the stack. Then it
13302 returns the address of that block.
13304 -- Built-in Function: void * __builtin_apply (void (*FUNCTION)(), void
13305 *ARGUMENTS, size_t SIZE)
13306 This built-in function invokes FUNCTION with a copy of the
13307 parameters described by ARGUMENTS and SIZE.
13309 The value of ARGUMENTS should be the value returned by
13310 `__builtin_apply_args'. The argument SIZE specifies the size of
13311 the stack argument data, in bytes.
13313 This function returns a pointer to data describing how to return
13314 whatever value was returned by FUNCTION. The data is saved in a
13315 block of memory allocated on the stack.
13317 It is not always simple to compute the proper value for SIZE. The
13318 value is used by `__builtin_apply' to compute the amount of data
13319 that should be pushed on the stack and copied from the incoming
13322 -- Built-in Function: void __builtin_return (void *RESULT)
13323 This built-in function returns the value described by RESULT from
13324 the containing function. You should specify, for RESULT, a value
13325 returned by `__builtin_apply'.
13328 File: gcc.info, Node: Typeof, Next: Conditionals, Prev: Constructing Calls, Up: C Extensions
13330 5.6 Referring to a Type with `typeof'
13331 =====================================
13333 Another way to refer to the type of an expression is with `typeof'.
13334 The syntax of using of this keyword looks like `sizeof', but the
13335 construct acts semantically like a type name defined with `typedef'.
13337 There are two ways of writing the argument to `typeof': with an
13338 expression or with a type. Here is an example with an expression:
13342 This assumes that `x' is an array of pointers to functions; the type
13343 described is that of the values of the functions.
13345 Here is an example with a typename as the argument:
13349 Here the type described is that of pointers to `int'.
13351 If you are writing a header file that must work when included in ISO C
13352 programs, write `__typeof__' instead of `typeof'. *Note Alternate
13355 A `typeof'-construct can be used anywhere a typedef name could be
13356 used. For example, you can use it in a declaration, in a cast, or
13357 inside of `sizeof' or `typeof'.
13359 `typeof' is often useful in conjunction with the
13360 statements-within-expressions feature. Here is how the two together can
13361 be used to define a safe "maximum" macro that operates on any
13362 arithmetic type and evaluates each of its arguments exactly once:
13365 ({ typeof (a) _a = (a); \
13366 typeof (b) _b = (b); \
13367 _a > _b ? _a : _b; })
13369 The reason for using names that start with underscores for the local
13370 variables is to avoid conflicts with variable names that occur within
13371 the expressions that are substituted for `a' and `b'. Eventually we
13372 hope to design a new form of declaration syntax that allows you to
13373 declare variables whose scopes start only after their initializers;
13374 this will be a more reliable way to prevent such conflicts.
13376 Some more examples of the use of `typeof':
13378 * This declares `y' with the type of what `x' points to.
13382 * This declares `y' as an array of such values.
13386 * This declares `y' as an array of pointers to characters:
13388 typeof (typeof (char *)[4]) y;
13390 It is equivalent to the following traditional C declaration:
13394 To see the meaning of the declaration using `typeof', and why it
13395 might be a useful way to write, rewrite it with these macros:
13397 #define pointer(T) typeof(T *)
13398 #define array(T, N) typeof(T [N])
13400 Now the declaration can be rewritten this way:
13402 array (pointer (char), 4) y;
13404 Thus, `array (pointer (char), 4)' is the type of arrays of 4
13405 pointers to `char'.
13407 _Compatibility Note:_ In addition to `typeof', GCC 2 supported a more
13408 limited extension which permitted one to write
13412 with the effect of declaring T to have the type of the expression EXPR.
13413 This extension does not work with GCC 3 (versions between 3.0 and 3.2
13414 will crash; 3.2.1 and later give an error). Code which relies on it
13415 should be rewritten to use `typeof':
13417 typedef typeof(EXPR) T;
13419 This will work with all versions of GCC.
13422 File: gcc.info, Node: Conditionals, Next: Long Long, Prev: Typeof, Up: C Extensions
13424 5.7 Conditionals with Omitted Operands
13425 ======================================
13427 The middle operand in a conditional expression may be omitted. Then if
13428 the first operand is nonzero, its value is the value of the conditional
13431 Therefore, the expression
13435 has the value of `x' if that is nonzero; otherwise, the value of `y'.
13437 This example is perfectly equivalent to
13441 In this simple case, the ability to omit the middle operand is not
13442 especially useful. When it becomes useful is when the first operand
13443 does, or may (if it is a macro argument), contain a side effect. Then
13444 repeating the operand in the middle would perform the side effect
13445 twice. Omitting the middle operand uses the value already computed
13446 without the undesirable effects of recomputing it.
13449 File: gcc.info, Node: Long Long, Next: Complex, Prev: Conditionals, Up: C Extensions
13451 5.8 Double-Word Integers
13452 ========================
13454 ISO C99 supports data types for integers that are at least 64 bits wide,
13455 and as an extension GCC supports them in C89 mode and in C++. Simply
13456 write `long long int' for a signed integer, or `unsigned long long int'
13457 for an unsigned integer. To make an integer constant of type `long
13458 long int', add the suffix `LL' to the integer. To make an integer
13459 constant of type `unsigned long long int', add the suffix `ULL' to the
13462 You can use these types in arithmetic like any other integer types.
13463 Addition, subtraction, and bitwise boolean operations on these types
13464 are open-coded on all types of machines. Multiplication is open-coded
13465 if the machine supports fullword-to-doubleword a widening multiply
13466 instruction. Division and shifts are open-coded only on machines that
13467 provide special support. The operations that are not open-coded use
13468 special library routines that come with GCC.
13470 There may be pitfalls when you use `long long' types for function
13471 arguments, unless you declare function prototypes. If a function
13472 expects type `int' for its argument, and you pass a value of type `long
13473 long int', confusion will result because the caller and the subroutine
13474 will disagree about the number of bytes for the argument. Likewise, if
13475 the function expects `long long int' and you pass `int'. The best way
13476 to avoid such problems is to use prototypes.
13479 File: gcc.info, Node: Complex, Next: Hex Floats, Prev: Long Long, Up: C Extensions
13481 5.9 Complex Numbers
13482 ===================
13484 ISO C99 supports complex floating data types, and as an extension GCC
13485 supports them in C89 mode and in C++, and supports complex integer data
13486 types which are not part of ISO C99. You can declare complex types
13487 using the keyword `_Complex'. As an extension, the older GNU keyword
13488 `__complex__' is also supported.
13490 For example, `_Complex double x;' declares `x' as a variable whose
13491 real part and imaginary part are both of type `double'. `_Complex
13492 short int y;' declares `y' to have real and imaginary parts of type
13493 `short int'; this is not likely to be useful, but it shows that the set
13494 of complex types is complete.
13496 To write a constant with a complex data type, use the suffix `i' or
13497 `j' (either one; they are equivalent). For example, `2.5fi' has type
13498 `_Complex float' and `3i' has type `_Complex int'. Such a constant
13499 always has a pure imaginary value, but you can form any complex value
13500 you like by adding one to a real constant. This is a GNU extension; if
13501 you have an ISO C99 conforming C library (such as GNU libc), and want
13502 to construct complex constants of floating type, you should include
13503 `<complex.h>' and use the macros `I' or `_Complex_I' instead.
13505 To extract the real part of a complex-valued expression EXP, write
13506 `__real__ EXP'. Likewise, use `__imag__' to extract the imaginary
13507 part. This is a GNU extension; for values of floating type, you should
13508 use the ISO C99 functions `crealf', `creal', `creall', `cimagf',
13509 `cimag' and `cimagl', declared in `<complex.h>' and also provided as
13510 built-in functions by GCC.
13512 The operator `~' performs complex conjugation when used on a value
13513 with a complex type. This is a GNU extension; for values of floating
13514 type, you should use the ISO C99 functions `conjf', `conj' and `conjl',
13515 declared in `<complex.h>' and also provided as built-in functions by
13518 GCC can allocate complex automatic variables in a noncontiguous
13519 fashion; it's even possible for the real part to be in a register while
13520 the imaginary part is on the stack (or vice-versa). Only the DWARF2
13521 debug info format can represent this, so use of DWARF2 is recommended.
13522 If you are using the stabs debug info format, GCC describes a
13523 noncontiguous complex variable as if it were two separate variables of
13524 noncomplex type. If the variable's actual name is `foo', the two
13525 fictitious variables are named `foo$real' and `foo$imag'. You can
13526 examine and set these two fictitious variables with your debugger.
13529 File: gcc.info, Node: Hex Floats, Next: Zero Length, Prev: Complex, Up: C Extensions
13534 ISO C99 supports floating-point numbers written not only in the usual
13535 decimal notation, such as `1.55e1', but also numbers such as `0x1.fp3'
13536 written in hexadecimal format. As a GNU extension, GCC supports this
13537 in C89 mode (except in some cases when strictly conforming) and in C++.
13538 In that format the `0x' hex introducer and the `p' or `P' exponent
13539 field are mandatory. The exponent is a decimal number that indicates
13540 the power of 2 by which the significant part will be multiplied. Thus
13541 `0x1.f' is 1 15/16, `p3' multiplies it by 8, and the value of `0x1.fp3'
13542 is the same as `1.55e1'.
13544 Unlike for floating-point numbers in the decimal notation the exponent
13545 is always required in the hexadecimal notation. Otherwise the compiler
13546 would not be able to resolve the ambiguity of, e.g., `0x1.f'. This
13547 could mean `1.0f' or `1.9375' since `f' is also the extension for
13548 floating-point constants of type `float'.
13551 File: gcc.info, Node: Zero Length, Next: Variable Length, Prev: Hex Floats, Up: C Extensions
13553 5.11 Arrays of Length Zero
13554 ==========================
13556 Zero-length arrays are allowed in GNU C. They are very useful as the
13557 last element of a structure which is really a header for a
13558 variable-length object:
13565 struct line *thisline = (struct line *)
13566 malloc (sizeof (struct line) + this_length);
13567 thisline->length = this_length;
13569 In ISO C90, you would have to give `contents' a length of 1, which
13570 means either you waste space or complicate the argument to `malloc'.
13572 In ISO C99, you would use a "flexible array member", which is slightly
13573 different in syntax and semantics:
13575 * Flexible array members are written as `contents[]' without the `0'.
13577 * Flexible array members have incomplete type, and so the `sizeof'
13578 operator may not be applied. As a quirk of the original
13579 implementation of zero-length arrays, `sizeof' evaluates to zero.
13581 * Flexible array members may only appear as the last member of a
13582 `struct' that is otherwise non-empty.
13584 * A structure containing a flexible array member, or a union
13585 containing such a structure (possibly recursively), may not be a
13586 member of a structure or an element of an array. (However, these
13587 uses are permitted by GCC as extensions.)
13589 GCC versions before 3.0 allowed zero-length arrays to be statically
13590 initialized, as if they were flexible arrays. In addition to those
13591 cases that were useful, it also allowed initializations in situations
13592 that would corrupt later data. Non-empty initialization of zero-length
13593 arrays is now treated like any case where there are more initializer
13594 elements than the array holds, in that a suitable warning about "excess
13595 elements in array" is given, and the excess elements (all of them, in
13596 this case) are ignored.
13598 Instead GCC allows static initialization of flexible array members.
13599 This is equivalent to defining a new structure containing the original
13600 structure followed by an array of sufficient size to contain the data.
13601 I.e. in the following, `f1' is constructed as if it were declared like
13606 } f1 = { 1, { 2, 3, 4 } };
13609 struct f1 f1; int data[3];
13610 } f2 = { { 1 }, { 2, 3, 4 } };
13612 The convenience of this extension is that `f1' has the desired type,
13613 eliminating the need to consistently refer to `f2.f1'.
13615 This has symmetry with normal static arrays, in that an array of
13616 unknown size is also written with `[]'.
13618 Of course, this extension only makes sense if the extra data comes at
13619 the end of a top-level object, as otherwise we would be overwriting
13620 data at subsequent offsets. To avoid undue complication and confusion
13621 with initialization of deeply nested arrays, we simply disallow any
13622 non-empty initialization except when the structure is the top-level
13623 object. For example:
13625 struct foo { int x; int y[]; };
13626 struct bar { struct foo z; };
13628 struct foo a = { 1, { 2, 3, 4 } }; // Valid.
13629 struct bar b = { { 1, { 2, 3, 4 } } }; // Invalid.
13630 struct bar c = { { 1, { } } }; // Valid.
13631 struct foo d[1] = { { 1 { 2, 3, 4 } } }; // Invalid.
13634 File: gcc.info, Node: Empty Structures, Next: Variadic Macros, Prev: Variable Length, Up: C Extensions
13636 5.12 Structures With No Members
13637 ===============================
13639 GCC permits a C structure to have no members:
13644 The structure will have size zero. In C++, empty structures are part
13645 of the language. G++ treats empty structures as if they had a single
13646 member of type `char'.
13649 File: gcc.info, Node: Variable Length, Next: Empty Structures, Prev: Zero Length, Up: C Extensions
13651 5.13 Arrays of Variable Length
13652 ==============================
13654 Variable-length automatic arrays are allowed in ISO C99, and as an
13655 extension GCC accepts them in C89 mode and in C++. (However, GCC's
13656 implementation of variable-length arrays does not yet conform in detail
13657 to the ISO C99 standard.) These arrays are declared like any other
13658 automatic arrays, but with a length that is not a constant expression.
13659 The storage is allocated at the point of declaration and deallocated
13660 when the brace-level is exited. For example:
13663 concat_fopen (char *s1, char *s2, char *mode)
13665 char str[strlen (s1) + strlen (s2) + 1];
13668 return fopen (str, mode);
13671 Jumping or breaking out of the scope of the array name deallocates the
13672 storage. Jumping into the scope is not allowed; you get an error
13675 You can use the function `alloca' to get an effect much like
13676 variable-length arrays. The function `alloca' is available in many
13677 other C implementations (but not in all). On the other hand,
13678 variable-length arrays are more elegant.
13680 There are other differences between these two methods. Space allocated
13681 with `alloca' exists until the containing _function_ returns. The
13682 space for a variable-length array is deallocated as soon as the array
13683 name's scope ends. (If you use both variable-length arrays and
13684 `alloca' in the same function, deallocation of a variable-length array
13685 will also deallocate anything more recently allocated with `alloca'.)
13687 You can also use variable-length arrays as arguments to functions:
13690 tester (int len, char data[len][len])
13695 The length of an array is computed once when the storage is allocated
13696 and is remembered for the scope of the array in case you access it with
13699 If you want to pass the array first and the length afterward, you can
13700 use a forward declaration in the parameter list--another GNU extension.
13703 tester (int len; char data[len][len], int len)
13708 The `int len' before the semicolon is a "parameter forward
13709 declaration", and it serves the purpose of making the name `len' known
13710 when the declaration of `data' is parsed.
13712 You can write any number of such parameter forward declarations in the
13713 parameter list. They can be separated by commas or semicolons, but the
13714 last one must end with a semicolon, which is followed by the "real"
13715 parameter declarations. Each forward declaration must match a "real"
13716 declaration in parameter name and data type. ISO C99 does not support
13717 parameter forward declarations.
13720 File: gcc.info, Node: Variadic Macros, Next: Escaped Newlines, Prev: Empty Structures, Up: C Extensions
13722 5.14 Macros with a Variable Number of Arguments.
13723 ================================================
13725 In the ISO C standard of 1999, a macro can be declared to accept a
13726 variable number of arguments much as a function can. The syntax for
13727 defining the macro is similar to that of a function. Here is an
13730 #define debug(format, ...) fprintf (stderr, format, __VA_ARGS__)
13732 Here `...' is a "variable argument". In the invocation of such a
13733 macro, it represents the zero or more tokens until the closing
13734 parenthesis that ends the invocation, including any commas. This set of
13735 tokens replaces the identifier `__VA_ARGS__' in the macro body wherever
13736 it appears. See the CPP manual for more information.
13738 GCC has long supported variadic macros, and used a different syntax
13739 that allowed you to give a name to the variable arguments just like any
13740 other argument. Here is an example:
13742 #define debug(format, args...) fprintf (stderr, format, args)
13744 This is in all ways equivalent to the ISO C example above, but arguably
13745 more readable and descriptive.
13747 GNU CPP has two further variadic macro extensions, and permits them to
13748 be used with either of the above forms of macro definition.
13750 In standard C, you are not allowed to leave the variable argument out
13751 entirely; but you are allowed to pass an empty argument. For example,
13752 this invocation is invalid in ISO C, because there is no comma after
13755 debug ("A message")
13757 GNU CPP permits you to completely omit the variable arguments in this
13758 way. In the above examples, the compiler would complain, though since
13759 the expansion of the macro still has the extra comma after the format
13762 To help solve this problem, CPP behaves specially for variable
13763 arguments used with the token paste operator, `##'. If instead you
13766 #define debug(format, ...) fprintf (stderr, format, ## __VA_ARGS__)
13768 and if the variable arguments are omitted or empty, the `##' operator
13769 causes the preprocessor to remove the comma before it. If you do
13770 provide some variable arguments in your macro invocation, GNU CPP does
13771 not complain about the paste operation and instead places the variable
13772 arguments after the comma. Just like any other pasted macro argument,
13773 these arguments are not macro expanded.
13776 File: gcc.info, Node: Escaped Newlines, Next: Subscripting, Prev: Variadic Macros, Up: C Extensions
13778 5.15 Slightly Looser Rules for Escaped Newlines
13779 ===============================================
13781 Recently, the preprocessor has relaxed its treatment of escaped
13782 newlines. Previously, the newline had to immediately follow a
13783 backslash. The current implementation allows whitespace in the form of
13784 spaces, horizontal and vertical tabs, and form feeds between the
13785 backslash and the subsequent newline. The preprocessor issues a
13786 warning, but treats it as a valid escaped newline and combines the two
13787 lines to form a single logical line. This works within comments and
13788 tokens, as well as between tokens. Comments are _not_ treated as
13789 whitespace for the purposes of this relaxation, since they have not yet
13790 been replaced with spaces.
13793 File: gcc.info, Node: Subscripting, Next: Pointer Arith, Prev: Escaped Newlines, Up: C Extensions
13795 5.16 Non-Lvalue Arrays May Have Subscripts
13796 ==========================================
13798 In ISO C99, arrays that are not lvalues still decay to pointers, and
13799 may be subscripted, although they may not be modified or used after the
13800 next sequence point and the unary `&' operator may not be applied to
13801 them. As an extension, GCC allows such arrays to be subscripted in C89
13802 mode, though otherwise they do not decay to pointers outside C99 mode.
13803 For example, this is valid in GNU C though not valid in C89:
13805 struct foo {int a[4];};
13811 return f().a[index];
13815 File: gcc.info, Node: Pointer Arith, Next: Initializers, Prev: Subscripting, Up: C Extensions
13817 5.17 Arithmetic on `void'- and Function-Pointers
13818 ================================================
13820 In GNU C, addition and subtraction operations are supported on pointers
13821 to `void' and on pointers to functions. This is done by treating the
13822 size of a `void' or of a function as 1.
13824 A consequence of this is that `sizeof' is also allowed on `void' and
13825 on function types, and returns 1.
13827 The option `-Wpointer-arith' requests a warning if these extensions
13831 File: gcc.info, Node: Initializers, Next: Compound Literals, Prev: Pointer Arith, Up: C Extensions
13833 5.18 Non-Constant Initializers
13834 ==============================
13836 As in standard C++ and ISO C99, the elements of an aggregate
13837 initializer for an automatic variable are not required to be constant
13838 expressions in GNU C. Here is an example of an initializer with
13839 run-time varying elements:
13841 foo (float f, float g)
13843 float beat_freqs[2] = { f-g, f+g };
13848 File: gcc.info, Node: Compound Literals, Next: Designated Inits, Prev: Initializers, Up: C Extensions
13850 5.19 Compound Literals
13851 ======================
13853 ISO C99 supports compound literals. A compound literal looks like a
13854 cast containing an initializer. Its value is an object of the type
13855 specified in the cast, containing the elements specified in the
13856 initializer; it is an lvalue. As an extension, GCC supports compound
13857 literals in C89 mode and in C++.
13859 Usually, the specified type is a structure. Assume that `struct foo'
13860 and `structure' are declared as shown:
13862 struct foo {int a; char b[2];} structure;
13864 Here is an example of constructing a `struct foo' with a compound
13867 structure = ((struct foo) {x + y, 'a', 0});
13869 This is equivalent to writing the following:
13872 struct foo temp = {x + y, 'a', 0};
13876 You can also construct an array. If all the elements of the compound
13877 literal are (made up of) simple constant expressions, suitable for use
13878 in initializers of objects of static storage duration, then the compound
13879 literal can be coerced to a pointer to its first element and used in
13880 such an initializer, as shown here:
13882 char **foo = (char *[]) { "x", "y", "z" };
13884 Compound literals for scalar types and union types are is also
13885 allowed, but then the compound literal is equivalent to a cast.
13887 As a GNU extension, GCC allows initialization of objects with static
13888 storage duration by compound literals (which is not possible in ISO
13889 C99, because the initializer is not a constant). It is handled as if
13890 the object was initialized only with the bracket enclosed list if
13891 compound literal's and object types match. The initializer list of the
13892 compound literal must be constant. If the object being initialized has
13893 array type of unknown size, the size is determined by compound literal
13896 static struct foo x = (struct foo) {1, 'a', 'b'};
13897 static int y[] = (int []) {1, 2, 3};
13898 static int z[] = (int [3]) {1};
13900 The above lines are equivalent to the following:
13901 static struct foo x = {1, 'a', 'b'};
13902 static int y[] = {1, 2, 3};
13903 static int z[] = {1, 0, 0};
13906 File: gcc.info, Node: Designated Inits, Next: Cast to Union, Prev: Compound Literals, Up: C Extensions
13908 5.20 Designated Initializers
13909 ============================
13911 Standard C89 requires the elements of an initializer to appear in a
13912 fixed order, the same as the order of the elements in the array or
13913 structure being initialized.
13915 In ISO C99 you can give the elements in any order, specifying the array
13916 indices or structure field names they apply to, and GNU C allows this as
13917 an extension in C89 mode as well. This extension is not implemented in
13920 To specify an array index, write `[INDEX] =' before the element value.
13923 int a[6] = { [4] = 29, [2] = 15 };
13927 int a[6] = { 0, 0, 15, 0, 29, 0 };
13929 The index values must be constant expressions, even if the array being
13930 initialized is automatic.
13932 An alternative syntax for this which has been obsolete since GCC 2.5
13933 but GCC still accepts is to write `[INDEX]' before the element value,
13936 To initialize a range of elements to the same value, write `[FIRST ...
13937 LAST] = VALUE'. This is a GNU extension. For example,
13939 int widths[] = { [0 ... 9] = 1, [10 ... 99] = 2, [100] = 3 };
13941 If the value in it has side-effects, the side-effects will happen only
13942 once, not for each initialized field by the range initializer.
13944 Note that the length of the array is the highest value specified plus
13947 In a structure initializer, specify the name of a field to initialize
13948 with `.FIELDNAME =' before the element value. For example, given the
13949 following structure,
13951 struct point { int x, y; };
13953 the following initialization
13955 struct point p = { .y = yvalue, .x = xvalue };
13959 struct point p = { xvalue, yvalue };
13961 Another syntax which has the same meaning, obsolete since GCC 2.5, is
13962 `FIELDNAME:', as shown here:
13964 struct point p = { y: yvalue, x: xvalue };
13966 The `[INDEX]' or `.FIELDNAME' is known as a "designator". You can
13967 also use a designator (or the obsolete colon syntax) when initializing
13968 a union, to specify which element of the union should be used. For
13971 union foo { int i; double d; };
13973 union foo f = { .d = 4 };
13975 will convert 4 to a `double' to store it in the union using the second
13976 element. By contrast, casting 4 to type `union foo' would store it
13977 into the union as the integer `i', since it is an integer. (*Note Cast
13980 You can combine this technique of naming elements with ordinary C
13981 initialization of successive elements. Each initializer element that
13982 does not have a designator applies to the next consecutive element of
13983 the array or structure. For example,
13985 int a[6] = { [1] = v1, v2, [4] = v4 };
13989 int a[6] = { 0, v1, v2, 0, v4, 0 };
13991 Labeling the elements of an array initializer is especially useful
13992 when the indices are characters or belong to an `enum' type. For
13995 int whitespace[256]
13996 = { [' '] = 1, ['\t'] = 1, ['\h'] = 1,
13997 ['\f'] = 1, ['\n'] = 1, ['\r'] = 1 };
13999 You can also write a series of `.FIELDNAME' and `[INDEX]' designators
14000 before an `=' to specify a nested subobject to initialize; the list is
14001 taken relative to the subobject corresponding to the closest
14002 surrounding brace pair. For example, with the `struct point'
14005 struct point ptarray[10] = { [2].y = yv2, [2].x = xv2, [0].x = xv0 };
14007 If the same field is initialized multiple times, it will have value from
14008 the last initialization. If any such overridden initialization has
14009 side-effect, it is unspecified whether the side-effect happens or not.
14010 Currently, GCC will discard them and issue a warning.
14013 File: gcc.info, Node: Case Ranges, Next: Mixed Declarations, Prev: Cast to Union, Up: C Extensions
14018 You can specify a range of consecutive values in a single `case' label,
14023 This has the same effect as the proper number of individual `case'
14024 labels, one for each integer value from LOW to HIGH, inclusive.
14026 This feature is especially useful for ranges of ASCII character codes:
14030 *Be careful:* Write spaces around the `...', for otherwise it may be
14031 parsed wrong when you use it with integer values. For example, write
14041 File: gcc.info, Node: Cast to Union, Next: Case Ranges, Prev: Designated Inits, Up: C Extensions
14043 5.22 Cast to a Union Type
14044 =========================
14046 A cast to union type is similar to other casts, except that the type
14047 specified is a union type. You can specify the type either with `union
14048 TAG' or with a typedef name. A cast to union is actually a constructor
14049 though, not a cast, and hence does not yield an lvalue like normal
14050 casts. (*Note Compound Literals::.)
14052 The types that may be cast to the union type are those of the members
14053 of the union. Thus, given the following union and variables:
14055 union foo { int i; double d; };
14059 both `x' and `y' can be cast to type `union foo'.
14061 Using the cast as the right-hand side of an assignment to a variable of
14062 union type is equivalent to storing in a member of the union:
14066 u = (union foo) x == u.i = x
14067 u = (union foo) y == u.d = y
14069 You can also use the union cast as a function argument:
14071 void hack (union foo);
14073 hack ((union foo) x);
14076 File: gcc.info, Node: Mixed Declarations, Next: Function Attributes, Prev: Case Ranges, Up: C Extensions
14078 5.23 Mixed Declarations and Code
14079 ================================
14081 ISO C99 and ISO C++ allow declarations and code to be freely mixed
14082 within compound statements. As an extension, GCC also allows this in
14083 C89 mode. For example, you could do:
14090 Each identifier is visible from where it is declared until the end of
14091 the enclosing block.
14094 File: gcc.info, Node: Function Attributes, Next: Attribute Syntax, Prev: Mixed Declarations, Up: C Extensions
14096 5.24 Declaring Attributes of Functions
14097 ======================================
14099 In GNU C, you declare certain things about functions called in your
14100 program which help the compiler optimize function calls and check your
14101 code more carefully.
14103 The keyword `__attribute__' allows you to specify special attributes
14104 when making a declaration. This keyword is followed by an attribute
14105 specification inside double parentheses. The following attributes are
14106 currently defined for functions on all targets: `noreturn', `noinline',
14107 `always_inline', `pure', `const', `nothrow', `sentinel', `format',
14108 `format_arg', `no_instrument_function', `section', `constructor',
14109 `destructor', `used', `unused', `deprecated', `weak', `malloc',
14110 `alias', `warn_unused_result' and `nonnull'. Several other attributes
14111 are defined for functions on particular target systems. Other
14112 attributes, including `section' are supported for variables declarations
14113 (*note Variable Attributes::) and for types (*note Type Attributes::).
14115 You may also specify attributes with `__' preceding and following each
14116 keyword. This allows you to use them in header files without being
14117 concerned about a possible macro of the same name. For example, you
14118 may use `__noreturn__' instead of `noreturn'.
14120 *Note Attribute Syntax::, for details of the exact syntax for using
14124 The `alias' attribute causes the declaration to be emitted as an
14125 alias for another symbol, which must be specified. For instance,
14127 void __f () { /* Do something. */; }
14128 void f () __attribute__ ((weak, alias ("__f")));
14130 declares `f' to be a weak alias for `__f'. In C++, the mangled
14131 name for the target must be used. It is an error if `__f' is not
14132 defined in the same translation unit.
14134 Not all target machines support this attribute.
14137 Generally, functions are not inlined unless optimization is
14138 specified. For functions declared inline, this attribute inlines
14139 the function even if no optimization level was specified.
14142 On the Intel 386, the `cdecl' attribute causes the compiler to
14143 assume that the calling function will pop off the stack space used
14144 to pass arguments. This is useful to override the effects of the
14148 Many functions do not examine any values except their arguments,
14149 and have no effects except the return value. Basically this is
14150 just slightly more strict class than the `pure' attribute below,
14151 since function is not allowed to read global memory.
14153 Note that a function that has pointer arguments and examines the
14154 data pointed to must _not_ be declared `const'. Likewise, a
14155 function that calls a non-`const' function usually must not be
14156 `const'. It does not make sense for a `const' function to return
14159 The attribute `const' is not implemented in GCC versions earlier
14160 than 2.5. An alternative way to declare that a function has no
14161 side effects, which works in the current version and in some older
14162 versions, is as follows:
14164 typedef int intfn ();
14166 extern const intfn square;
14168 This approach does not work in GNU C++ from 2.6.0 on, since the
14169 language specifies that the `const' must be attached to the return
14174 The `constructor' attribute causes the function to be called
14175 automatically before execution enters `main ()'. Similarly, the
14176 `destructor' attribute causes the function to be called
14177 automatically after `main ()' has completed or `exit ()' has been
14178 called. Functions with these attributes are useful for
14179 initializing data that will be used implicitly during the
14180 execution of the program.
14182 These attributes are not currently implemented for Objective-C.
14185 The `deprecated' attribute results in a warning if the function is
14186 used anywhere in the source file. This is useful when identifying
14187 functions that are expected to be removed in a future version of a
14188 program. The warning also includes the location of the declaration
14189 of the deprecated function, to enable users to easily find further
14190 information about why the function is deprecated, or what they
14191 should do instead. Note that the warnings only occurs for uses:
14193 int old_fn () __attribute__ ((deprecated));
14195 int (*fn_ptr)() = old_fn;
14197 results in a warning on line 3 but not line 2.
14199 The `deprecated' attribute can also be used for variables and
14200 types (*note Variable Attributes::, *note Type Attributes::.)
14203 On Microsoft Windows targets and Symbian OS targets the
14204 `dllexport' attribute causes the compiler to provide a global
14205 pointer to a pointer in a DLL, so that it can be referenced with
14206 the `dllimport' attribute. On Microsoft Windows targets, the
14207 pointer name is formed by combining `_imp__' and the function or
14210 You can use `__declspec(dllexport)' as a synonym for
14211 `__attribute__ ((dllexport))' for compatibility with other
14214 On systems that support the `visibility' attribute, this attribute
14215 also implies "default" visibility, unless a `visibility' attribute
14216 is explicitly specified. You should avoid the use of `dllexport'
14217 with "hidden" or "internal" visibility; in the future GCC may
14218 issue an error for those cases.
14220 Currently, the `dllexport' attribute is ignored for inlined
14221 functions, unless the `-fkeep-inline-functions' flag has been
14222 used. The attribute is also ignored for undefined symbols.
14224 When applied to C++ classes, the attribute marks defined
14225 non-inlined member functions and static data members as exports.
14226 Static consts initialized in-class are not marked unless they are
14227 also defined out-of-class.
14229 For Microsoft Windows targets there are alternative methods for
14230 including the symbol in the DLL's export table such as using a
14231 `.def' file with an `EXPORTS' section or, with GNU ld, using the
14232 `--export-all' linker flag.
14235 On Microsoft Windows and Symbian OS targets, the `dllimport'
14236 attribute causes the compiler to reference a function or variable
14237 via a global pointer to a pointer that is set up by the DLL
14238 exporting the symbol. The attribute implies `extern' storage. On
14239 Microsoft Windows targets, the pointer name is formed by combining
14240 `_imp__' and the function or variable name.
14242 You can use `__declspec(dllimport)' as a synonym for
14243 `__attribute__ ((dllimport))' for compatibility with other
14246 Currently, the attribute is ignored for inlined functions. If the
14247 attribute is applied to a symbol _definition_, an error is
14248 reported. If a symbol previously declared `dllimport' is later
14249 defined, the attribute is ignored in subsequent references, and a
14250 warning is emitted. The attribute is also overridden by a
14251 subsequent declaration as `dllexport'.
14253 When applied to C++ classes, the attribute marks non-inlined
14254 member functions and static data members as imports. However, the
14255 attribute is ignored for virtual methods to allow creation of
14256 vtables using thunks.
14258 On the SH Symbian OS target the `dllimport' attribute also has
14259 another affect--it can cause the vtable and run-time type
14260 information for a class to be exported. This happens when the
14261 class has a dllimport'ed constructor or a non-inline, non-pure
14262 virtual function and, for either of those two conditions, the
14263 class also has a inline constructor or destructor and has a key
14264 function that is defined in the current translation unit.
14266 For Microsoft Windows based targets the use of the `dllimport'
14267 attribute on functions is not necessary, but provides a small
14268 performance benefit by eliminating a thunk in the DLL. The use of
14269 the `dllimport' attribute on imported variables was required on
14270 older versions of the GNU linker, but can now be avoided by
14271 passing the `--enable-auto-import' switch to the GNU linker. As
14272 with functions, using the attribute for a variable eliminates a
14275 One drawback to using this attribute is that a pointer to a
14276 function or variable marked as `dllimport' cannot be used as a
14277 constant address. On Microsoft Windows targets, the attribute can
14278 be disabled for functions by setting the `-mnop-fun-dllimport'
14282 Use this attribute on the H8/300, H8/300H, and H8S to indicate
14283 that the specified variable should be placed into the eight bit
14284 data section. The compiler will generate more efficient code for
14285 certain operations on data in the eight bit data area. Note the
14286 eight bit data area is limited to 256 bytes of data.
14288 You must use GAS and GLD from GNU binutils version 2.7 or later for
14289 this attribute to work correctly.
14291 `exception_handler'
14292 Use this attribute on the Blackfin to indicate that the specified
14293 function is an exception handler. The compiler will generate
14294 function entry and exit sequences suitable for use in an exception
14295 handler when this attribute is present.
14298 On 68HC11 and 68HC12 the `far' attribute causes the compiler to
14299 use a calling convention that takes care of switching memory banks
14300 when entering and leaving a function. This calling convention is
14301 also the default when using the `-mlong-calls' option.
14303 On 68HC12 the compiler will use the `call' and `rtc' instructions
14304 to call and return from a function.
14306 On 68HC11 the compiler will generate a sequence of instructions to
14307 invoke a board-specific routine to switch the memory bank and call
14308 the real function. The board-specific routine simulates a `call'.
14309 At the end of a function, it will jump to a board-specific routine
14310 instead of using `rts'. The board-specific return routine
14311 simulates the `rtc'.
14314 On the Intel 386, the `fastcall' attribute causes the compiler to
14315 pass the first two arguments in the registers ECX and EDX.
14316 Subsequent arguments are passed on the stack. The called function
14317 will pop the arguments off the stack. If the number of arguments
14318 is variable all arguments are pushed on the stack.
14320 `format (ARCHETYPE, STRING-INDEX, FIRST-TO-CHECK)'
14321 The `format' attribute specifies that a function takes `printf',
14322 `scanf', `strftime' or `strfmon' style arguments which should be
14323 type-checked against a format string. For example, the
14327 my_printf (void *my_object, const char *my_format, ...)
14328 __attribute__ ((format (printf, 2, 3)));
14330 causes the compiler to check the arguments in calls to `my_printf'
14331 for consistency with the `printf' style format string argument
14334 The parameter ARCHETYPE determines how the format string is
14335 interpreted, and should be `printf', `scanf', `strftime' or
14336 `strfmon'. (You can also use `__printf__', `__scanf__',
14337 `__strftime__' or `__strfmon__'.) The parameter STRING-INDEX
14338 specifies which argument is the format string argument (starting
14339 from 1), while FIRST-TO-CHECK is the number of the first argument
14340 to check against the format string. For functions where the
14341 arguments are not available to be checked (such as `vprintf'),
14342 specify the third parameter as zero. In this case the compiler
14343 only checks the format string for consistency. For `strftime'
14344 formats, the third parameter is required to be zero. Since
14345 non-static C++ methods have an implicit `this' argument, the
14346 arguments of such methods should be counted from two, not one, when
14347 giving values for STRING-INDEX and FIRST-TO-CHECK.
14349 In the example above, the format string (`my_format') is the second
14350 argument of the function `my_print', and the arguments to check
14351 start with the third argument, so the correct parameters for the
14352 format attribute are 2 and 3.
14354 The `format' attribute allows you to identify your own functions
14355 which take format strings as arguments, so that GCC can check the
14356 calls to these functions for errors. The compiler always (unless
14357 `-ffreestanding' or `-fno-builtin' is used) checks formats for the
14358 standard library functions `printf', `fprintf', `sprintf',
14359 `scanf', `fscanf', `sscanf', `strftime', `vprintf', `vfprintf' and
14360 `vsprintf' whenever such warnings are requested (using
14361 `-Wformat'), so there is no need to modify the header file
14362 `stdio.h'. In C99 mode, the functions `snprintf', `vsnprintf',
14363 `vscanf', `vfscanf' and `vsscanf' are also checked. Except in
14364 strictly conforming C standard modes, the X/Open function
14365 `strfmon' is also checked as are `printf_unlocked' and
14366 `fprintf_unlocked'. *Note Options Controlling C Dialect: C
14369 The target may provide additional types of format checks. *Note
14370 Format Checks Specific to Particular Target Machines: Target
14373 `format_arg (STRING-INDEX)'
14374 The `format_arg' attribute specifies that a function takes a format
14375 string for a `printf', `scanf', `strftime' or `strfmon' style
14376 function and modifies it (for example, to translate it into
14377 another language), so the result can be passed to a `printf',
14378 `scanf', `strftime' or `strfmon' style function (with the
14379 remaining arguments to the format function the same as they would
14380 have been for the unmodified string). For example, the
14384 my_dgettext (char *my_domain, const char *my_format)
14385 __attribute__ ((format_arg (2)));
14387 causes the compiler to check the arguments in calls to a `printf',
14388 `scanf', `strftime' or `strfmon' type function, whose format
14389 string argument is a call to the `my_dgettext' function, for
14390 consistency with the format string argument `my_format'. If the
14391 `format_arg' attribute had not been specified, all the compiler
14392 could tell in such calls to format functions would be that the
14393 format string argument is not constant; this would generate a
14394 warning when `-Wformat-nonliteral' is used, but the calls could
14395 not be checked without the attribute.
14397 The parameter STRING-INDEX specifies which argument is the format
14398 string argument (starting from one). Since non-static C++ methods
14399 have an implicit `this' argument, the arguments of such methods
14400 should be counted from two.
14402 The `format-arg' attribute allows you to identify your own
14403 functions which modify format strings, so that GCC can check the
14404 calls to `printf', `scanf', `strftime' or `strfmon' type function
14405 whose operands are a call to one of your own function. The
14406 compiler always treats `gettext', `dgettext', and `dcgettext' in
14407 this manner except when strict ISO C support is requested by
14408 `-ansi' or an appropriate `-std' option, or `-ffreestanding' or
14409 `-fno-builtin' is used. *Note Options Controlling C Dialect: C
14413 Use this attribute on the H8/300, H8/300H, and H8S to indicate
14414 that the specified function should be called through the function
14415 vector. Calling a function through the function vector will
14416 reduce code size, however; the function vector has a limited size
14417 (maximum 128 entries on the H8/300 and 64 entries on the H8/300H
14418 and H8S) and shares space with the interrupt vector.
14420 You must use GAS and GLD from GNU binutils version 2.7 or later for
14421 this attribute to work correctly.
14424 Use this attribute on the ARM, AVR, C4x, M32R/D and Xstormy16
14425 ports to indicate that the specified function is an interrupt
14426 handler. The compiler will generate function entry and exit
14427 sequences suitable for use in an interrupt handler when this
14428 attribute is present.
14430 Note, interrupt handlers for the Blackfin, m68k, H8/300, H8/300H,
14431 H8S, and SH processors can be specified via the
14432 `interrupt_handler' attribute.
14434 Note, on the AVR, interrupts will be enabled inside the function.
14436 Note, for the ARM, you can specify the kind of interrupt to be
14437 handled by adding an optional parameter to the interrupt attribute
14440 void f () __attribute__ ((interrupt ("IRQ")));
14442 Permissible values for this parameter are: IRQ, FIQ, SWI, ABORT
14445 `interrupt_handler'
14446 Use this attribute on the Blackfin, m68k, H8/300, H8/300H, H8S,
14447 and SH to indicate that the specified function is an interrupt
14448 handler. The compiler will generate function entry and exit
14449 sequences suitable for use in an interrupt handler when this
14450 attribute is present.
14453 When used together with `interrupt_handler', `exception_handler'
14454 or `nmi_handler', code will be generated to load the stack pointer
14455 from the USP register in the function prologue.
14457 `long_call/short_call'
14458 This attribute specifies how a particular function is called on
14459 ARM. Both attributes override the `-mlong-calls' (*note ARM
14460 Options::) command line switch and `#pragma long_calls' settings.
14461 The `long_call' attribute causes the compiler to always call the
14462 function by first loading its address into a register and then
14463 using the contents of that register. The `short_call' attribute
14464 always places the offset to the function from the call site into
14465 the `BL' instruction directly.
14467 `longcall/shortcall'
14468 On the RS/6000 and PowerPC, the `longcall' attribute causes the
14469 compiler to always call this function via a pointer, just as it
14470 would if the `-mlongcall' option had been specified. The
14471 `shortcall' attribute causes the compiler not to do this. These
14472 attributes override both the `-mlongcall' switch and the `#pragma
14475 *Note RS/6000 and PowerPC Options::, for more information on
14476 whether long calls are necessary.
14479 The `malloc' attribute is used to tell the compiler that a function
14480 may be treated as if any non-`NULL' pointer it returns cannot
14481 alias any other pointer valid when the function returns. This
14482 will often improve optimization. Standard functions with this
14483 property include `malloc' and `calloc'. `realloc'-like functions
14484 have this property as long as the old pointer is never referred to
14485 (including comparing it to the new pointer) after the function
14486 returns a non-`NULL' value.
14488 `model (MODEL-NAME)'
14489 On the M32R/D, use this attribute to set the addressability of an
14490 object, and of the code generated for a function. The identifier
14491 MODEL-NAME is one of `small', `medium', or `large', representing
14492 each of the code models.
14494 Small model objects live in the lower 16MB of memory (so that their
14495 addresses can be loaded with the `ld24' instruction), and are
14496 callable with the `bl' instruction.
14498 Medium model objects may live anywhere in the 32-bit address space
14499 (the compiler will generate `seth/add3' instructions to load their
14500 addresses), and are callable with the `bl' instruction.
14502 Large model objects may live anywhere in the 32-bit address space
14503 (the compiler will generate `seth/add3' instructions to load their
14504 addresses), and may not be reachable with the `bl' instruction
14505 (the compiler will generate the much slower `seth/add3/jl'
14506 instruction sequence).
14508 On IA-64, use this attribute to set the addressability of an
14509 object. At present, the only supported identifier for MODEL-NAME
14510 is `small', indicating addressability via "small" (22-bit)
14511 addresses (so that their addresses can be loaded with the `addl'
14512 instruction). Caveat: such addressing is by definition not
14513 position independent and hence this attribute must not be used for
14514 objects defined by shared libraries.
14517 Use this attribute on the ARM, AVR, C4x and IP2K ports to indicate
14518 that the specified function does not need prologue/epilogue
14519 sequences generated by the compiler. It is up to the programmer
14520 to provide these sequences.
14523 On 68HC11 and 68HC12 the `near' attribute causes the compiler to
14524 use the normal calling convention based on `jsr' and `rts'. This
14525 attribute can be used to cancel the effect of the `-mlong-calls'
14529 Use this attribute together with `interrupt_handler',
14530 `exception_handler' or `nmi_handler' to indicate that the function
14531 entry code should enable nested interrupts or exceptions.
14534 Use this attribute on the Blackfin to indicate that the specified
14535 function is an NMI handler. The compiler will generate function
14536 entry and exit sequences suitable for use in an NMI handler when
14537 this attribute is present.
14539 `no_instrument_function'
14540 If `-finstrument-functions' is given, profiling function calls will
14541 be generated at entry and exit of most user-compiled functions.
14542 Functions with this attribute will not be so instrumented.
14545 This function attribute prevents a function from being considered
14548 `nonnull (ARG-INDEX, ...)'
14549 The `nonnull' attribute specifies that some function parameters
14550 should be non-null pointers. For instance, the declaration:
14553 my_memcpy (void *dest, const void *src, size_t len)
14554 __attribute__((nonnull (1, 2)));
14556 causes the compiler to check that, in calls to `my_memcpy',
14557 arguments DEST and SRC are non-null. If the compiler determines
14558 that a null pointer is passed in an argument slot marked as
14559 non-null, and the `-Wnonnull' option is enabled, a warning is
14560 issued. The compiler may also choose to make optimizations based
14561 on the knowledge that certain function arguments will not be null.
14563 If no argument index list is given to the `nonnull' attribute, all
14564 pointer arguments are marked as non-null. To illustrate, the
14565 following declaration is equivalent to the previous example:
14568 my_memcpy (void *dest, const void *src, size_t len)
14569 __attribute__((nonnull));
14572 A few standard library functions, such as `abort' and `exit',
14573 cannot return. GCC knows this automatically. Some programs define
14574 their own functions that never return. You can declare them
14575 `noreturn' to tell the compiler this fact. For example,
14577 void fatal () __attribute__ ((noreturn));
14582 /* ... */ /* Print error message. */ /* ... */
14586 The `noreturn' keyword tells the compiler to assume that `fatal'
14587 cannot return. It can then optimize without regard to what would
14588 happen if `fatal' ever did return. This makes slightly better
14589 code. More importantly, it helps avoid spurious warnings of
14590 uninitialized variables.
14592 The `noreturn' keyword does not affect the exceptional path when
14593 that applies: a `noreturn'-marked function may still return to the
14594 caller by throwing an exception or calling `longjmp'.
14596 Do not assume that registers saved by the calling function are
14597 restored before calling the `noreturn' function.
14599 It does not make sense for a `noreturn' function to have a return
14600 type other than `void'.
14602 The attribute `noreturn' is not implemented in GCC versions
14603 earlier than 2.5. An alternative way to declare that a function
14604 does not return, which works in the current version and in some
14605 older versions, is as follows:
14607 typedef void voidfn ();
14609 volatile voidfn fatal;
14611 This approach does not work in GNU C++.
14614 The `nothrow' attribute is used to inform the compiler that a
14615 function cannot throw an exception. For example, most functions in
14616 the standard C library can be guaranteed not to throw an exception
14617 with the notable exceptions of `qsort' and `bsearch' that take
14618 function pointer arguments. The `nothrow' attribute is not
14619 implemented in GCC versions earlier than 3.3.
14622 Many functions have no effects except the return value and their
14623 return value depends only on the parameters and/or global
14624 variables. Such a function can be subject to common subexpression
14625 elimination and loop optimization just as an arithmetic operator
14626 would be. These functions should be declared with the attribute
14627 `pure'. For example,
14629 int square (int) __attribute__ ((pure));
14631 says that the hypothetical function `square' is safe to call fewer
14632 times than the program says.
14634 Some of common examples of pure functions are `strlen' or `memcmp'.
14635 Interesting non-pure functions are functions with infinite loops
14636 or those depending on volatile memory or other system resource,
14637 that may change between two consecutive calls (such as `feof' in a
14638 multithreading environment).
14640 The attribute `pure' is not implemented in GCC versions earlier
14644 On the Intel 386, the `regparm' attribute causes the compiler to
14645 pass up to NUMBER integer arguments in registers EAX, EDX, and ECX
14646 instead of on the stack. Functions that take a variable number of
14647 arguments will continue to be passed all of their arguments on the
14650 Beware that on some ELF systems this attribute is unsuitable for
14651 global functions in shared libraries with lazy binding (which is
14652 the default). Lazy binding will send the first call via resolving
14653 code in the loader, which might assume EAX, EDX and ECX can be
14654 clobbered, as per the standard calling conventions. Solaris 8 is
14655 affected by this. GNU systems with GLIBC 2.1 or higher, and
14656 FreeBSD, are believed to be safe since the loaders there save all
14657 registers. (Lazy binding can be disabled with the linker or the
14658 loader if desired, to avoid the problem.)
14661 Use this attribute on the Blackfin, H8/300, H8/300H, and H8S to
14662 indicate that all registers except the stack pointer should be
14663 saved in the prologue regardless of whether they are used or not.
14665 `section ("SECTION-NAME")'
14666 Normally, the compiler places the code it generates in the `text'
14667 section. Sometimes, however, you need additional sections, or you
14668 need certain particular functions to appear in special sections.
14669 The `section' attribute specifies that a function lives in a
14670 particular section. For example, the declaration:
14672 extern void foobar (void) __attribute__ ((section ("bar")));
14674 puts the function `foobar' in the `bar' section.
14676 Some file formats do not support arbitrary sections so the
14677 `section' attribute is not available on all platforms. If you
14678 need to map the entire contents of a module to a particular
14679 section, consider using the facilities of the linker instead.
14682 This function attribute ensures that a parameter in a function
14683 call is an explicit `NULL'. The attribute is only valid on
14684 variadic functions. By default, the sentinel is located at
14685 position zero, the last parameter of the function call. If an
14686 optional integer position argument P is supplied to the attribute,
14687 the sentinel must be located at position P counting backwards from
14688 the end of the argument list.
14690 __attribute__ ((sentinel))
14692 __attribute__ ((sentinel(0)))
14694 The attribute is automatically set with a position of 0 for the
14695 built-in functions `execl' and `execlp'. The built-in function
14696 `execle' has the attribute set with a position of 1.
14698 A valid `NULL' in this context is defined as zero with any pointer
14699 type. If your system defines the `NULL' macro with an integer type
14700 then you need to add an explicit cast. GCC replaces `stddef.h'
14701 with a copy that redefines NULL appropriately.
14703 The warnings for missing or incorrect sentinels are enabled with
14707 See long_call/short_call.
14710 See longcall/shortcall.
14713 Use this attribute on the AVR to indicate that the specified
14714 function is a signal handler. The compiler will generate function
14715 entry and exit sequences suitable for use in a signal handler when
14716 this attribute is present. Interrupts will be disabled inside the
14720 Use this attribute on the SH to indicate an `interrupt_handler'
14721 function should switch to an alternate stack. It expects a string
14722 argument that names a global variable holding the address of the
14726 void f () __attribute__ ((interrupt_handler,
14727 sp_switch ("alt_stack")));
14730 On the Intel 386, the `stdcall' attribute causes the compiler to
14731 assume that the called function will pop off the stack space used
14732 to pass arguments, unless it takes a variable number of arguments.
14735 Use this attribute on the H8/300H and H8S to indicate that the
14736 specified variable should be placed into the tiny data section.
14737 The compiler will generate more efficient code for loads and stores
14738 on data in the tiny data section. Note the tiny data area is
14739 limited to slightly under 32kbytes of data.
14742 Use this attribute on the SH for an `interrupt_handler' to return
14743 using `trapa' instead of `rte'. This attribute expects an integer
14744 argument specifying the trap number to be used.
14747 This attribute, attached to a function, means that the function is
14748 meant to be possibly unused. GCC will not produce a warning for
14752 This attribute, attached to a function, means that code must be
14753 emitted for the function even if it appears that the function is
14754 not referenced. This is useful, for example, when the function is
14755 referenced only in inline assembly.
14757 `visibility ("VISIBILITY_TYPE")'
14758 The `visibility' attribute on ELF targets causes the declaration
14759 to be emitted with default, hidden, protected or internal
14762 void __attribute__ ((visibility ("protected")))
14763 f () { /* Do something. */; }
14764 int i __attribute__ ((visibility ("hidden")));
14766 See the ELF gABI for complete details, but the short story is:
14769 Default visibility is the normal case for ELF. This value is
14770 available for the visibility attribute to override other
14771 options that may change the assumed visibility of symbols.
14774 Hidden visibility indicates that the symbol will not be
14775 placed into the dynamic symbol table, so no other "module"
14776 (executable or shared library) can reference it directly.
14779 Internal visibility is like hidden visibility, but with
14780 additional processor specific semantics. Unless otherwise
14781 specified by the psABI, GCC defines internal visibility to
14782 mean that the function is _never_ called from another module.
14783 Note that hidden symbols, while they cannot be referenced
14784 directly by other modules, can be referenced indirectly via
14785 function pointers. By indicating that a symbol cannot be
14786 called from outside the module, GCC may for instance omit the
14787 load of a PIC register since it is known that the calling
14788 function loaded the correct value.
14791 Protected visibility indicates that the symbol will be placed
14792 in the dynamic symbol table, but that references within the
14793 defining module will bind to the local symbol. That is, the
14794 symbol cannot be overridden by another module.
14797 Not all ELF targets support this attribute.
14799 `warn_unused_result'
14800 The `warn_unused_result' attribute causes a warning to be emitted
14801 if a caller of the function with this attribute does not use its
14802 return value. This is useful for functions where not checking the
14803 result is either a security problem or always a bug, such as
14806 int fn () __attribute__ ((warn_unused_result));
14809 if (fn () < 0) return -1;
14814 results in warning on line 5.
14817 The `weak' attribute causes the declaration to be emitted as a weak
14818 symbol rather than a global. This is primarily useful in defining
14819 library functions which can be overridden in user code, though it
14820 can also be used with non-function declarations. Weak symbols are
14821 supported for ELF targets, and also for a.out targets when using
14822 the GNU assembler and linker.
14825 You can specify multiple attributes in a declaration by separating them
14826 by commas within the double parentheses or by immediately following an
14827 attribute declaration with another attribute declaration.
14829 Some people object to the `__attribute__' feature, suggesting that ISO
14830 C's `#pragma' should be used instead. At the time `__attribute__' was
14831 designed, there were two reasons for not doing this.
14833 1. It is impossible to generate `#pragma' commands from a macro.
14835 2. There is no telling what the same `#pragma' might mean in another
14838 These two reasons applied to almost any application that might have
14839 been proposed for `#pragma'. It was basically a mistake to use
14840 `#pragma' for _anything_.
14842 The ISO C99 standard includes `_Pragma', which now allows pragmas to
14843 be generated from macros. In addition, a `#pragma GCC' namespace is
14844 now in use for GCC-specific pragmas. However, it has been found
14845 convenient to use `__attribute__' to achieve a natural attachment of
14846 attributes to their corresponding declarations, whereas `#pragma GCC'
14847 is of use for constructs that do not naturally form part of the
14848 grammar. *Note Miscellaneous Preprocessing Directives: (cpp)Other
14852 File: gcc.info, Node: Attribute Syntax, Next: Function Prototypes, Prev: Function Attributes, Up: C Extensions
14854 5.25 Attribute Syntax
14855 =====================
14857 This section describes the syntax with which `__attribute__' may be
14858 used, and the constructs to which attribute specifiers bind, for the C
14859 language. Some details may vary for C++ and Objective-C. Because of
14860 infelicities in the grammar for attributes, some forms described here
14861 may not be successfully parsed in all cases.
14863 There are some problems with the semantics of attributes in C++. For
14864 example, there are no manglings for attributes, although they may affect
14865 code generation, so problems may arise when attributed types are used in
14866 conjunction with templates or overloading. Similarly, `typeid' does
14867 not distinguish between types with different attributes. Support for
14868 attributes in C++ may be restricted in future to attributes on
14869 declarations only, but not on nested declarators.
14871 *Note Function Attributes::, for details of the semantics of attributes
14872 applying to functions. *Note Variable Attributes::, for details of the
14873 semantics of attributes applying to variables. *Note Type Attributes::,
14874 for details of the semantics of attributes applying to structure, union
14875 and enumerated types.
14877 An "attribute specifier" is of the form `__attribute__
14878 ((ATTRIBUTE-LIST))'. An "attribute list" is a possibly empty
14879 comma-separated sequence of "attributes", where each attribute is one
14882 * Empty. Empty attributes are ignored.
14884 * A word (which may be an identifier such as `unused', or a reserved
14885 word such as `const').
14887 * A word, followed by, in parentheses, parameters for the attribute.
14888 These parameters take one of the following forms:
14890 * An identifier. For example, `mode' attributes use this form.
14892 * An identifier followed by a comma and a non-empty
14893 comma-separated list of expressions. For example, `format'
14894 attributes use this form.
14896 * A possibly empty comma-separated list of expressions. For
14897 example, `format_arg' attributes use this form with the list
14898 being a single integer constant expression, and `alias'
14899 attributes use this form with the list being a single string
14902 An "attribute specifier list" is a sequence of one or more attribute
14903 specifiers, not separated by any other tokens.
14905 In GNU C, an attribute specifier list may appear after the colon
14906 following a label, other than a `case' or `default' label. The only
14907 attribute it makes sense to use after a label is `unused'. This
14908 feature is intended for code generated by programs which contains labels
14909 that may be unused but which is compiled with `-Wall'. It would not
14910 normally be appropriate to use in it human-written code, though it
14911 could be useful in cases where the code that jumps to the label is
14912 contained within an `#ifdef' conditional. GNU C++ does not permit such
14913 placement of attribute lists, as it is permissible for a declaration,
14914 which could begin with an attribute list, to be labelled in C++.
14915 Declarations cannot be labelled in C90 or C99, so the ambiguity does
14918 An attribute specifier list may appear as part of a `struct', `union'
14919 or `enum' specifier. It may go either immediately after the `struct',
14920 `union' or `enum' keyword, or after the closing brace. It is ignored
14921 if the content of the structure, union or enumerated type is not
14922 defined in the specifier in which the attribute specifier list is
14923 used--that is, in usages such as `struct __attribute__((foo)) bar' with
14924 no following opening brace. Where attribute specifiers follow the
14925 closing brace, they are considered to relate to the structure, union or
14926 enumerated type defined, not to any enclosing declaration the type
14927 specifier appears in, and the type defined is not complete until after
14928 the attribute specifiers.
14930 Otherwise, an attribute specifier appears as part of a declaration,
14931 counting declarations of unnamed parameters and type names, and relates
14932 to that declaration (which may be nested in another declaration, for
14933 example in the case of a parameter declaration), or to a particular
14934 declarator within a declaration. Where an attribute specifier is
14935 applied to a parameter declared as a function or an array, it should
14936 apply to the function or array rather than the pointer to which the
14937 parameter is implicitly converted, but this is not yet correctly
14940 Any list of specifiers and qualifiers at the start of a declaration may
14941 contain attribute specifiers, whether or not such a list may in that
14942 context contain storage class specifiers. (Some attributes, however,
14943 are essentially in the nature of storage class specifiers, and only make
14944 sense where storage class specifiers may be used; for example,
14945 `section'.) There is one necessary limitation to this syntax: the
14946 first old-style parameter declaration in a function definition cannot
14947 begin with an attribute specifier, because such an attribute applies to
14948 the function instead by syntax described below (which, however, is not
14949 yet implemented in this case). In some other cases, attribute
14950 specifiers are permitted by this grammar but not yet supported by the
14951 compiler. All attribute specifiers in this place relate to the
14952 declaration as a whole. In the obsolescent usage where a type of `int'
14953 is implied by the absence of type specifiers, such a list of specifiers
14954 and qualifiers may be an attribute specifier list with no other
14955 specifiers or qualifiers.
14957 At present, the first parameter in a function prototype must have some
14958 type specifier which is not an attribute specifier; this resolves an
14959 ambiguity in the interpretation of `void f(int (__attribute__((foo))
14960 x))', but is subject to change. At present, if the parentheses of a
14961 function declarator contain only attributes then those attributes are
14962 ignored, rather than yielding an error or warning or implying a single
14963 parameter of type int, but this is subject to change.
14965 An attribute specifier list may appear immediately before a declarator
14966 (other than the first) in a comma-separated list of declarators in a
14967 declaration of more than one identifier using a single list of
14968 specifiers and qualifiers. Such attribute specifiers apply only to the
14969 identifier before whose declarator they appear. For example, in
14971 __attribute__((noreturn)) void d0 (void),
14972 __attribute__((format(printf, 1, 2))) d1 (const char *, ...),
14975 the `noreturn' attribute applies to all the functions declared; the
14976 `format' attribute only applies to `d1'.
14978 An attribute specifier list may appear immediately before the comma,
14979 `=' or semicolon terminating the declaration of an identifier other
14980 than a function definition. At present, such attribute specifiers apply
14981 to the declared object or function, but in future they may attach to the
14982 outermost adjacent declarator. In simple cases there is no difference,
14983 but, for example, in
14985 void (****f)(void) __attribute__((noreturn));
14987 at present the `noreturn' attribute applies to `f', which causes a
14988 warning since `f' is not a function, but in future it may apply to the
14989 function `****f'. The precise semantics of what attributes in such
14990 cases will apply to are not yet specified. Where an assembler name for
14991 an object or function is specified (*note Asm Labels::), at present the
14992 attribute must follow the `asm' specification; in future, attributes
14993 before the `asm' specification may apply to the adjacent declarator,
14994 and those after it to the declared object or function.
14996 An attribute specifier list may, in future, be permitted to appear
14997 after the declarator in a function definition (before any old-style
14998 parameter declarations or the function body).
15000 Attribute specifiers may be mixed with type qualifiers appearing inside
15001 the `[]' of a parameter array declarator, in the C99 construct by which
15002 such qualifiers are applied to the pointer to which the array is
15003 implicitly converted. Such attribute specifiers apply to the pointer,
15004 not to the array, but at present this is not implemented and they are
15007 An attribute specifier list may appear at the start of a nested
15008 declarator. At present, there are some limitations in this usage: the
15009 attributes correctly apply to the declarator, but for most individual
15010 attributes the semantics this implies are not implemented. When
15011 attribute specifiers follow the `*' of a pointer declarator, they may
15012 be mixed with any type qualifiers present. The following describes the
15013 formal semantics of this syntax. It will make the most sense if you
15014 are familiar with the formal specification of declarators in the ISO C
15017 Consider (as in C99 subclause 6.7.5 paragraph 4) a declaration `T D1',
15018 where `T' contains declaration specifiers that specify a type TYPE
15019 (such as `int') and `D1' is a declarator that contains an identifier
15020 IDENT. The type specified for IDENT for derived declarators whose type
15021 does not include an attribute specifier is as in the ISO C standard.
15023 If `D1' has the form `( ATTRIBUTE-SPECIFIER-LIST D )', and the
15024 declaration `T D' specifies the type "DERIVED-DECLARATOR-TYPE-LIST
15025 TYPE" for IDENT, then `T D1' specifies the type
15026 "DERIVED-DECLARATOR-TYPE-LIST ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
15028 If `D1' has the form `* TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST
15029 D', and the declaration `T D' specifies the type
15030 "DERIVED-DECLARATOR-TYPE-LIST TYPE" for IDENT, then `T D1' specifies
15031 the type "DERIVED-DECLARATOR-TYPE-LIST
15032 TYPE-QUALIFIER-AND-ATTRIBUTE-SPECIFIER-LIST TYPE" for IDENT.
15036 void (__attribute__((noreturn)) ****f) (void);
15038 specifies the type "pointer to pointer to pointer to pointer to
15039 non-returning function returning `void'". As another example,
15041 char *__attribute__((aligned(8))) *f;
15043 specifies the type "pointer to 8-byte-aligned pointer to `char'". Note
15044 again that this does not work with most attributes; for example, the
15045 usage of `aligned' and `noreturn' attributes given above is not yet
15048 For compatibility with existing code written for compiler versions that
15049 did not implement attributes on nested declarators, some laxity is
15050 allowed in the placing of attributes. If an attribute that only applies
15051 to types is applied to a declaration, it will be treated as applying to
15052 the type of that declaration. If an attribute that only applies to
15053 declarations is applied to the type of a declaration, it will be treated
15054 as applying to that declaration; and, for compatibility with code
15055 placing the attributes immediately before the identifier declared, such
15056 an attribute applied to a function return type will be treated as
15057 applying to the function type, and such an attribute applied to an array
15058 element type will be treated as applying to the array type. If an
15059 attribute that only applies to function types is applied to a
15060 pointer-to-function type, it will be treated as applying to the pointer
15061 target type; if such an attribute is applied to a function return type
15062 that is not a pointer-to-function type, it will be treated as applying
15063 to the function type.
15066 File: gcc.info, Node: Function Prototypes, Next: C++ Comments, Prev: Attribute Syntax, Up: C Extensions
15068 5.26 Prototypes and Old-Style Function Definitions
15069 ==================================================
15071 GNU C extends ISO C to allow a function prototype to override a later
15072 old-style non-prototype definition. Consider the following example:
15074 /* Use prototypes unless the compiler is old-fashioned. */
15081 /* Prototype function declaration. */
15082 int isroot P((uid_t));
15084 /* Old-style function definition. */
15086 isroot (x) /* ??? lossage here ??? */
15092 Suppose the type `uid_t' happens to be `short'. ISO C does not allow
15093 this example, because subword arguments in old-style non-prototype
15094 definitions are promoted. Therefore in this example the function
15095 definition's argument is really an `int', which does not match the
15096 prototype argument type of `short'.
15098 This restriction of ISO C makes it hard to write code that is portable
15099 to traditional C compilers, because the programmer does not know
15100 whether the `uid_t' type is `short', `int', or `long'. Therefore, in
15101 cases like these GNU C allows a prototype to override a later old-style
15102 definition. More precisely, in GNU C, a function prototype argument
15103 type overrides the argument type specified by a later old-style
15104 definition if the former type is the same as the latter type before
15105 promotion. Thus in GNU C the above example is equivalent to the
15108 int isroot (uid_t);
15116 GNU C++ does not support old-style function definitions, so this
15117 extension is irrelevant.
15120 File: gcc.info, Node: C++ Comments, Next: Dollar Signs, Prev: Function Prototypes, Up: C Extensions
15122 5.27 C++ Style Comments
15123 =======================
15125 In GNU C, you may use C++ style comments, which start with `//' and
15126 continue until the end of the line. Many other C implementations allow
15127 such comments, and they are included in the 1999 C standard. However,
15128 C++ style comments are not recognized if you specify an `-std' option
15129 specifying a version of ISO C before C99, or `-ansi' (equivalent to
15133 File: gcc.info, Node: Dollar Signs, Next: Character Escapes, Prev: C++ Comments, Up: C Extensions
15135 5.28 Dollar Signs in Identifier Names
15136 =====================================
15138 In GNU C, you may normally use dollar signs in identifier names. This
15139 is because many traditional C implementations allow such identifiers.
15140 However, dollar signs in identifiers are not supported on a few target
15141 machines, typically because the target assembler does not allow them.
15144 File: gcc.info, Node: Character Escapes, Next: Variable Attributes, Prev: Dollar Signs, Up: C Extensions
15146 5.29 The Character <ESC> in Constants
15147 =====================================
15149 You can use the sequence `\e' in a string or character constant to
15150 stand for the ASCII character <ESC>.
15153 File: gcc.info, Node: Alignment, Next: Inline, Prev: Type Attributes, Up: C Extensions
15155 5.30 Inquiring on Alignment of Types or Variables
15156 =================================================
15158 The keyword `__alignof__' allows you to inquire about how an object is
15159 aligned, or the minimum alignment usually required by a type. Its
15160 syntax is just like `sizeof'.
15162 For example, if the target machine requires a `double' value to be
15163 aligned on an 8-byte boundary, then `__alignof__ (double)' is 8. This
15164 is true on many RISC machines. On more traditional machine designs,
15165 `__alignof__ (double)' is 4 or even 2.
15167 Some machines never actually require alignment; they allow reference
15168 to any data type even at an odd address. For these machines,
15169 `__alignof__' reports the _recommended_ alignment of a type.
15171 If the operand of `__alignof__' is an lvalue rather than a type, its
15172 value is the required alignment for its type, taking into account any
15173 minimum alignment specified with GCC's `__attribute__' extension (*note
15174 Variable Attributes::). For example, after this declaration:
15176 struct foo { int x; char y; } foo1;
15178 the value of `__alignof__ (foo1.y)' is 1, even though its actual
15179 alignment is probably 2 or 4, the same as `__alignof__ (int)'.
15181 It is an error to ask for the alignment of an incomplete type.
15184 File: gcc.info, Node: Variable Attributes, Next: Type Attributes, Prev: Character Escapes, Up: C Extensions
15186 5.31 Specifying Attributes of Variables
15187 =======================================
15189 The keyword `__attribute__' allows you to specify special attributes of
15190 variables or structure fields. This keyword is followed by an
15191 attribute specification inside double parentheses. Some attributes are
15192 currently defined generically for variables. Other attributes are
15193 defined for variables on particular target systems. Other attributes
15194 are available for functions (*note Function Attributes::) and for types
15195 (*note Type Attributes::). Other front ends might define more
15196 attributes (*note Extensions to the C++ Language: C++ Extensions.).
15198 You may also specify attributes with `__' preceding and following each
15199 keyword. This allows you to use them in header files without being
15200 concerned about a possible macro of the same name. For example, you
15201 may use `__aligned__' instead of `aligned'.
15203 *Note Attribute Syntax::, for details of the exact syntax for using
15206 `aligned (ALIGNMENT)'
15207 This attribute specifies a minimum alignment for the variable or
15208 structure field, measured in bytes. For example, the declaration:
15210 int x __attribute__ ((aligned (16))) = 0;
15212 causes the compiler to allocate the global variable `x' on a
15213 16-byte boundary. On a 68040, this could be used in conjunction
15214 with an `asm' expression to access the `move16' instruction which
15215 requires 16-byte aligned operands.
15217 You can also specify the alignment of structure fields. For
15218 example, to create a double-word aligned `int' pair, you could
15221 struct foo { int x[2] __attribute__ ((aligned (8))); };
15223 This is an alternative to creating a union with a `double' member
15224 that forces the union to be double-word aligned.
15226 As in the preceding examples, you can explicitly specify the
15227 alignment (in bytes) that you wish the compiler to use for a given
15228 variable or structure field. Alternatively, you can leave out the
15229 alignment factor and just ask the compiler to align a variable or
15230 field to the maximum useful alignment for the target machine you
15231 are compiling for. For example, you could write:
15233 short array[3] __attribute__ ((aligned));
15235 Whenever you leave out the alignment factor in an `aligned'
15236 attribute specification, the compiler automatically sets the
15237 alignment for the declared variable or field to the largest
15238 alignment which is ever used for any data type on the target
15239 machine you are compiling for. Doing this can often make copy
15240 operations more efficient, because the compiler can use whatever
15241 instructions copy the biggest chunks of memory when performing
15242 copies to or from the variables or fields that you have aligned
15245 The `aligned' attribute can only increase the alignment; but you
15246 can decrease it by specifying `packed' as well. See below.
15248 Note that the effectiveness of `aligned' attributes may be limited
15249 by inherent limitations in your linker. On many systems, the
15250 linker is only able to arrange for variables to be aligned up to a
15251 certain maximum alignment. (For some linkers, the maximum
15252 supported alignment may be very very small.) If your linker is
15253 only able to align variables up to a maximum of 8 byte alignment,
15254 then specifying `aligned(16)' in an `__attribute__' will still
15255 only provide you with 8 byte alignment. See your linker
15256 documentation for further information.
15258 `cleanup (CLEANUP_FUNCTION)'
15259 The `cleanup' attribute runs a function when the variable goes out
15260 of scope. This attribute can only be applied to auto function
15261 scope variables; it may not be applied to parameters or variables
15262 with static storage duration. The function must take one
15263 parameter, a pointer to a type compatible with the variable. The
15264 return value of the function (if any) is ignored.
15266 If `-fexceptions' is enabled, then CLEANUP_FUNCTION will be run
15267 during the stack unwinding that happens during the processing of
15268 the exception. Note that the `cleanup' attribute does not allow
15269 the exception to be caught, only to perform an action. It is
15270 undefined what happens if CLEANUP_FUNCTION does not return
15275 The `common' attribute requests GCC to place a variable in
15276 "common" storage. The `nocommon' attribute requests the
15277 opposite--to allocate space for it directly.
15279 These attributes override the default chosen by the `-fno-common'
15280 and `-fcommon' flags respectively.
15283 The `deprecated' attribute results in a warning if the variable is
15284 used anywhere in the source file. This is useful when identifying
15285 variables that are expected to be removed in a future version of a
15286 program. The warning also includes the location of the declaration
15287 of the deprecated variable, to enable users to easily find further
15288 information about why the variable is deprecated, or what they
15289 should do instead. Note that the warning only occurs for uses:
15291 extern int old_var __attribute__ ((deprecated));
15292 extern int old_var;
15293 int new_fn () { return old_var; }
15295 results in a warning on line 3 but not line 2.
15297 The `deprecated' attribute can also be used for functions and
15298 types (*note Function Attributes::, *note Type Attributes::.)
15301 This attribute specifies the data type for the
15302 declaration--whichever type corresponds to the mode MODE. This in
15303 effect lets you request an integer or floating point type
15304 according to its width.
15306 You may also specify a mode of `byte' or `__byte__' to indicate
15307 the mode corresponding to a one-byte integer, `word' or `__word__'
15308 for the mode of a one-word integer, and `pointer' or `__pointer__'
15309 for the mode used to represent pointers.
15312 The `packed' attribute specifies that a variable or structure field
15313 should have the smallest possible alignment--one byte for a
15314 variable, and one bit for a field, unless you specify a larger
15315 value with the `aligned' attribute.
15317 Here is a structure in which the field `x' is packed, so that it
15318 immediately follows `a':
15323 int x[2] __attribute__ ((packed));
15326 `section ("SECTION-NAME")'
15327 Normally, the compiler places the objects it generates in sections
15328 like `data' and `bss'. Sometimes, however, you need additional
15329 sections, or you need certain particular variables to appear in
15330 special sections, for example to map to special hardware. The
15331 `section' attribute specifies that a variable (or function) lives
15332 in a particular section. For example, this small program uses
15333 several specific section names:
15335 struct duart a __attribute__ ((section ("DUART_A"))) = { 0 };
15336 struct duart b __attribute__ ((section ("DUART_B"))) = { 0 };
15337 char stack[10000] __attribute__ ((section ("STACK"))) = { 0 };
15338 int init_data __attribute__ ((section ("INITDATA"))) = 0;
15342 /* Initialize stack pointer */
15343 init_sp (stack + sizeof (stack));
15345 /* Initialize initialized data */
15346 memcpy (&init_data, &data, &edata - &data);
15348 /* Turn on the serial ports */
15353 Use the `section' attribute with an _initialized_ definition of a
15354 _global_ variable, as shown in the example. GCC issues a warning
15355 and otherwise ignores the `section' attribute in uninitialized
15356 variable declarations.
15358 You may only use the `section' attribute with a fully initialized
15359 global definition because of the way linkers work. The linker
15360 requires each object be defined once, with the exception that
15361 uninitialized variables tentatively go in the `common' (or `bss')
15362 section and can be multiply "defined". You can force a variable
15363 to be initialized with the `-fno-common' flag or the `nocommon'
15366 Some file formats do not support arbitrary sections so the
15367 `section' attribute is not available on all platforms. If you
15368 need to map the entire contents of a module to a particular
15369 section, consider using the facilities of the linker instead.
15372 On Microsoft Windows, in addition to putting variable definitions
15373 in a named section, the section can also be shared among all
15374 running copies of an executable or DLL. For example, this small
15375 program defines shared data by putting it in a named section
15376 `shared' and marking the section shareable:
15378 int foo __attribute__((section ("shared"), shared)) = 0;
15383 /* Read and write foo. All running
15384 copies see the same value. */
15388 You may only use the `shared' attribute along with `section'
15389 attribute with a fully initialized global definition because of
15390 the way linkers work. See `section' attribute for more
15393 The `shared' attribute is only available on Microsoft Windows.
15395 `tls_model ("TLS_MODEL")'
15396 The `tls_model' attribute sets thread-local storage model (*note
15397 Thread-Local::) of a particular `__thread' variable, overriding
15398 `-ftls-model=' command line switch on a per-variable basis. The
15399 TLS_MODEL argument should be one of `global-dynamic',
15400 `local-dynamic', `initial-exec' or `local-exec'.
15402 Not all targets support this attribute.
15404 `transparent_union'
15405 This attribute, attached to a function parameter which is a union,
15406 means that the corresponding argument may have the type of any
15407 union member, but the argument is passed as if its type were that
15408 of the first union member. For more details see *Note Type
15409 Attributes::. You can also use this attribute on a `typedef' for
15410 a union data type; then it applies to all function parameters with
15414 This attribute, attached to a variable, means that the variable is
15415 meant to be possibly unused. GCC will not produce a warning for
15418 `vector_size (BYTES)'
15419 This attribute specifies the vector size for the variable,
15420 measured in bytes. For example, the declaration:
15422 int foo __attribute__ ((vector_size (16)));
15424 causes the compiler to set the mode for `foo', to be 16 bytes,
15425 divided into `int' sized units. Assuming a 32-bit int (a vector of
15426 4 units of 4 bytes), the corresponding mode of `foo' will be V4SI.
15428 This attribute is only applicable to integral and float scalars,
15429 although arrays, pointers, and function return values are allowed
15430 in conjunction with this construct.
15432 Aggregates with this attribute are invalid, even if they are of
15433 the same size as a corresponding scalar. For example, the
15436 struct S { int a; };
15437 struct S __attribute__ ((vector_size (16))) foo;
15439 is invalid even if the size of the structure is the same as the
15443 The `weak' attribute is described in *Note Function Attributes::.
15446 The `dllimport' attribute is described in *Note Function
15450 The `dllexport' attribute is described in *Note Function
15454 5.31.1 M32R/D Variable Attributes
15455 ---------------------------------
15457 One attribute is currently defined for the M32R/D.
15459 `model (MODEL-NAME)'
15460 Use this attribute on the M32R/D to set the addressability of an
15461 object. The identifier MODEL-NAME is one of `small', `medium', or
15462 `large', representing each of the code models.
15464 Small model objects live in the lower 16MB of memory (so that their
15465 addresses can be loaded with the `ld24' instruction).
15467 Medium and large model objects may live anywhere in the 32-bit
15468 address space (the compiler will generate `seth/add3' instructions
15469 to load their addresses).
15471 5.31.2 i386 Variable Attributes
15472 -------------------------------
15474 Two attributes are currently defined for i386 configurations:
15475 `ms_struct' and `gcc_struct'
15479 If `packed' is used on a structure, or if bit-fields are used it
15480 may be that the Microsoft ABI packs them differently than GCC
15481 would normally pack them. Particularly when moving packed data
15482 between functions compiled with GCC and the native Microsoft
15483 compiler (either via function call or as data in a file), it may
15484 be necessary to access either format.
15486 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
15487 Windows X86 compilers to match the native Microsoft compiler.
15489 5.31.3 Xstormy16 Variable Attributes
15490 ------------------------------------
15492 One attribute is currently defined for xstormy16 configurations:
15496 If a variable has the `below100' attribute (`BELOW100' is allowed
15497 also), GCC will place the variable in the first 0x100 bytes of
15498 memory and use special opcodes to access it. Such variables will
15499 be placed in either the `.bss_below100' section or the
15500 `.data_below100' section.
15504 File: gcc.info, Node: Type Attributes, Next: Alignment, Prev: Variable Attributes, Up: C Extensions
15506 5.32 Specifying Attributes of Types
15507 ===================================
15509 The keyword `__attribute__' allows you to specify special attributes of
15510 `struct' and `union' types when you define such types. This keyword is
15511 followed by an attribute specification inside double parentheses. Six
15512 attributes are currently defined for types: `aligned', `packed',
15513 `transparent_union', `unused', `deprecated' and `may_alias'. Other
15514 attributes are defined for functions (*note Function Attributes::) and
15515 for variables (*note Variable Attributes::).
15517 You may also specify any one of these attributes with `__' preceding
15518 and following its keyword. This allows you to use these attributes in
15519 header files without being concerned about a possible macro of the same
15520 name. For example, you may use `__aligned__' instead of `aligned'.
15522 You may specify the `aligned' and `transparent_union' attributes
15523 either in a `typedef' declaration or just past the closing curly brace
15524 of a complete enum, struct or union type _definition_ and the `packed'
15525 attribute only past the closing brace of a definition.
15527 You may also specify attributes between the enum, struct or union tag
15528 and the name of the type rather than after the closing brace.
15530 *Note Attribute Syntax::, for details of the exact syntax for using
15533 `aligned (ALIGNMENT)'
15534 This attribute specifies a minimum alignment (in bytes) for
15535 variables of the specified type. For example, the declarations:
15537 struct S { short f[3]; } __attribute__ ((aligned (8)));
15538 typedef int more_aligned_int __attribute__ ((aligned (8)));
15540 force the compiler to insure (as far as it can) that each variable
15541 whose type is `struct S' or `more_aligned_int' will be allocated
15542 and aligned _at least_ on a 8-byte boundary. On a SPARC, having
15543 all variables of type `struct S' aligned to 8-byte boundaries
15544 allows the compiler to use the `ldd' and `std' (doubleword load and
15545 store) instructions when copying one variable of type `struct S' to
15546 another, thus improving run-time efficiency.
15548 Note that the alignment of any given `struct' or `union' type is
15549 required by the ISO C standard to be at least a perfect multiple of
15550 the lowest common multiple of the alignments of all of the members
15551 of the `struct' or `union' in question. This means that you _can_
15552 effectively adjust the alignment of a `struct' or `union' type by
15553 attaching an `aligned' attribute to any one of the members of such
15554 a type, but the notation illustrated in the example above is a
15555 more obvious, intuitive, and readable way to request the compiler
15556 to adjust the alignment of an entire `struct' or `union' type.
15558 As in the preceding example, you can explicitly specify the
15559 alignment (in bytes) that you wish the compiler to use for a given
15560 `struct' or `union' type. Alternatively, you can leave out the
15561 alignment factor and just ask the compiler to align a type to the
15562 maximum useful alignment for the target machine you are compiling
15563 for. For example, you could write:
15565 struct S { short f[3]; } __attribute__ ((aligned));
15567 Whenever you leave out the alignment factor in an `aligned'
15568 attribute specification, the compiler automatically sets the
15569 alignment for the type to the largest alignment which is ever used
15570 for any data type on the target machine you are compiling for.
15571 Doing this can often make copy operations more efficient, because
15572 the compiler can use whatever instructions copy the biggest chunks
15573 of memory when performing copies to or from the variables which
15574 have types that you have aligned this way.
15576 In the example above, if the size of each `short' is 2 bytes, then
15577 the size of the entire `struct S' type is 6 bytes. The smallest
15578 power of two which is greater than or equal to that is 8, so the
15579 compiler sets the alignment for the entire `struct S' type to 8
15582 Note that although you can ask the compiler to select a
15583 time-efficient alignment for a given type and then declare only
15584 individual stand-alone objects of that type, the compiler's
15585 ability to select a time-efficient alignment is primarily useful
15586 only when you plan to create arrays of variables having the
15587 relevant (efficiently aligned) type. If you declare or use arrays
15588 of variables of an efficiently-aligned type, then it is likely
15589 that your program will also be doing pointer arithmetic (or
15590 subscripting, which amounts to the same thing) on pointers to the
15591 relevant type, and the code that the compiler generates for these
15592 pointer arithmetic operations will often be more efficient for
15593 efficiently-aligned types than for other types.
15595 The `aligned' attribute can only increase the alignment; but you
15596 can decrease it by specifying `packed' as well. See below.
15598 Note that the effectiveness of `aligned' attributes may be limited
15599 by inherent limitations in your linker. On many systems, the
15600 linker is only able to arrange for variables to be aligned up to a
15601 certain maximum alignment. (For some linkers, the maximum
15602 supported alignment may be very very small.) If your linker is
15603 only able to align variables up to a maximum of 8 byte alignment,
15604 then specifying `aligned(16)' in an `__attribute__' will still
15605 only provide you with 8 byte alignment. See your linker
15606 documentation for further information.
15609 This attribute, attached to `struct' or `union' type definition,
15610 specifies that each member of the structure or union is placed to
15611 minimize the memory required. When attached to an `enum'
15612 definition, it indicates that the smallest integral type should be
15615 Specifying this attribute for `struct' and `union' types is
15616 equivalent to specifying the `packed' attribute on each of the
15617 structure or union members. Specifying the `-fshort-enums' flag
15618 on the line is equivalent to specifying the `packed' attribute on
15619 all `enum' definitions.
15621 In the following example `struct my_packed_struct''s members are
15622 packed closely together, but the internal layout of its `s' member
15623 is not packed--to do that, `struct my_unpacked_struct' would need
15626 struct my_unpacked_struct
15632 struct __attribute__ ((__packed__)) my_packed_struct
15636 struct my_unpacked_struct s;
15639 You may only specify this attribute on the definition of a `enum',
15640 `struct' or `union', not on a `typedef' which does not also define
15641 the enumerated type, structure or union.
15643 `transparent_union'
15644 This attribute, attached to a `union' type definition, indicates
15645 that any function parameter having that union type causes calls to
15646 that function to be treated in a special way.
15648 First, the argument corresponding to a transparent union type can
15649 be of any type in the union; no cast is required. Also, if the
15650 union contains a pointer type, the corresponding argument can be a
15651 null pointer constant or a void pointer expression; and if the
15652 union contains a void pointer type, the corresponding argument can
15653 be any pointer expression. If the union member type is a pointer,
15654 qualifiers like `const' on the referenced type must be respected,
15655 just as with normal pointer conversions.
15657 Second, the argument is passed to the function using the calling
15658 conventions of the first member of the transparent union, not the
15659 calling conventions of the union itself. All members of the union
15660 must have the same machine representation; this is necessary for
15661 this argument passing to work properly.
15663 Transparent unions are designed for library functions that have
15664 multiple interfaces for compatibility reasons. For example,
15665 suppose the `wait' function must accept either a value of type
15666 `int *' to comply with Posix, or a value of type `union wait *' to
15667 comply with the 4.1BSD interface. If `wait''s parameter were
15668 `void *', `wait' would accept both kinds of arguments, but it
15669 would also accept any other pointer type and this would make
15670 argument type checking less useful. Instead, `<sys/wait.h>' might
15671 define the interface as follows:
15677 } wait_status_ptr_t __attribute__ ((__transparent_union__));
15679 pid_t wait (wait_status_ptr_t);
15681 This interface allows either `int *' or `union wait *' arguments
15682 to be passed, using the `int *' calling convention. The program
15683 can call `wait' with arguments of either type:
15685 int w1 () { int w; return wait (&w); }
15686 int w2 () { union wait w; return wait (&w); }
15688 With this interface, `wait''s implementation might look like this:
15690 pid_t wait (wait_status_ptr_t p)
15692 return waitpid (-1, p.__ip, 0);
15696 When attached to a type (including a `union' or a `struct'), this
15697 attribute means that variables of that type are meant to appear
15698 possibly unused. GCC will not produce a warning for any variables
15699 of that type, even if the variable appears to do nothing. This is
15700 often the case with lock or thread classes, which are usually
15701 defined and then not referenced, but contain constructors and
15702 destructors that have nontrivial bookkeeping functions.
15705 The `deprecated' attribute results in a warning if the type is
15706 used anywhere in the source file. This is useful when identifying
15707 types that are expected to be removed in a future version of a
15708 program. If possible, the warning also includes the location of
15709 the declaration of the deprecated type, to enable users to easily
15710 find further information about why the type is deprecated, or what
15711 they should do instead. Note that the warnings only occur for
15712 uses and then only if the type is being applied to an identifier
15713 that itself is not being declared as deprecated.
15715 typedef int T1 __attribute__ ((deprecated));
15719 typedef T1 T3 __attribute__ ((deprecated));
15720 T3 z __attribute__ ((deprecated));
15722 results in a warning on line 2 and 3 but not lines 4, 5, or 6. No
15723 warning is issued for line 4 because T2 is not explicitly
15724 deprecated. Line 5 has no warning because T3 is explicitly
15725 deprecated. Similarly for line 6.
15727 The `deprecated' attribute can also be used for functions and
15728 variables (*note Function Attributes::, *note Variable
15732 Accesses to objects with types with this attribute are not
15733 subjected to type-based alias analysis, but are instead assumed to
15734 be able to alias any other type of objects, just like the `char'
15735 type. See `-fstrict-aliasing' for more information on aliasing
15740 typedef short __attribute__((__may_alias__)) short_a;
15745 int a = 0x12345678;
15746 short_a *b = (short_a *) &a;
15750 if (a == 0x12345678)
15756 If you replaced `short_a' with `short' in the variable
15757 declaration, the above program would abort when compiled with
15758 `-fstrict-aliasing', which is on by default at `-O2' or above in
15759 recent GCC versions.
15761 5.32.1 ARM Type Attributes
15762 --------------------------
15764 On those ARM targets that support `dllimport' (such as Symbian
15765 OS), you can use the `notshared' attribute to indicate that the virtual
15766 table and other similar data for a class should not be exported from a
15769 class __declspec(notshared) C {
15771 __declspec(dllimport) C();
15775 __declspec(dllexport)
15778 In this code, `C::C' is exported from the current DLL, but the
15779 virtual table for `C' is not exported. (You can use `__attribute__'
15780 instead of `__declspec' if you prefer, but most Symbian OS code uses
15783 5.32.2 i386 Type Attributes
15784 ---------------------------
15786 Two attributes are currently defined for i386 configurations:
15787 `ms_struct' and `gcc_struct'
15791 If `packed' is used on a structure, or if bit-fields are used it
15792 may be that the Microsoft ABI packs them differently than GCC
15793 would normally pack them. Particularly when moving packed data
15794 between functions compiled with GCC and the native Microsoft
15795 compiler (either via function call or as data in a file), it may
15796 be necessary to access either format.
15798 Currently `-m[no-]ms-bitfields' is provided for the Microsoft
15799 Windows X86 compilers to match the native Microsoft compiler.
15801 To specify multiple attributes, separate them by commas within the
15802 double parentheses: for example, `__attribute__ ((aligned (16),
15806 File: gcc.info, Node: Inline, Next: Extended Asm, Prev: Alignment, Up: C Extensions
15808 5.33 An Inline Function is As Fast As a Macro
15809 =============================================
15811 By declaring a function `inline', you can direct GCC to integrate that
15812 function's code into the code for its callers. This makes execution
15813 faster by eliminating the function-call overhead; in addition, if any
15814 of the actual argument values are constant, their known values may
15815 permit simplifications at compile time so that not all of the inline
15816 function's code needs to be included. The effect on code size is less
15817 predictable; object code may be larger or smaller with function
15818 inlining, depending on the particular case. Inlining of functions is an
15819 optimization and it really "works" only in optimizing compilation. If
15820 you don't use `-O', no function is really inline.
15822 Inline functions are included in the ISO C99 standard, but there are
15823 currently substantial differences between what GCC implements and what
15824 the ISO C99 standard requires.
15826 To declare a function inline, use the `inline' keyword in its
15827 declaration, like this:
15835 (If you are writing a header file to be included in ISO C programs,
15836 write `__inline__' instead of `inline'. *Note Alternate Keywords::.)
15837 You can also make all "simple enough" functions inline with the option
15838 `-finline-functions'.
15840 Note that certain usages in a function definition can make it
15841 unsuitable for inline substitution. Among these usages are: use of
15842 varargs, use of alloca, use of variable sized data types (*note
15843 Variable Length::), use of computed goto (*note Labels as Values::),
15844 use of nonlocal goto, and nested functions (*note Nested Functions::).
15845 Using `-Winline' will warn when a function marked `inline' could not be
15846 substituted, and will give the reason for the failure.
15848 Note that in C and Objective-C, unlike C++, the `inline' keyword does
15849 not affect the linkage of the function.
15851 GCC automatically inlines member functions defined within the class
15852 body of C++ programs even if they are not explicitly declared `inline'.
15853 (You can override this with `-fno-default-inline'; *note Options
15854 Controlling C++ Dialect: C++ Dialect Options.)
15856 When a function is both inline and `static', if all calls to the
15857 function are integrated into the caller, and the function's address is
15858 never used, then the function's own assembler code is never referenced.
15859 In this case, GCC does not actually output assembler code for the
15860 function, unless you specify the option `-fkeep-inline-functions'.
15861 Some calls cannot be integrated for various reasons (in particular,
15862 calls that precede the function's definition cannot be integrated, and
15863 neither can recursive calls within the definition). If there is a
15864 nonintegrated call, then the function is compiled to assembler code as
15865 usual. The function must also be compiled as usual if the program
15866 refers to its address, because that can't be inlined.
15868 When an inline function is not `static', then the compiler must assume
15869 that there may be calls from other source files; since a global symbol
15870 can be defined only once in any program, the function must not be
15871 defined in the other source files, so the calls therein cannot be
15872 integrated. Therefore, a non-`static' inline function is always
15873 compiled on its own in the usual fashion.
15875 If you specify both `inline' and `extern' in the function definition,
15876 then the definition is used only for inlining. In no case is the
15877 function compiled on its own, not even if you refer to its address
15878 explicitly. Such an address becomes an external reference, as if you
15879 had only declared the function, and had not defined it.
15881 This combination of `inline' and `extern' has almost the effect of a
15882 macro. The way to use it is to put a function definition in a header
15883 file with these keywords, and put another copy of the definition
15884 (lacking `inline' and `extern') in a library file. The definition in
15885 the header file will cause most calls to the function to be inlined.
15886 If any uses of the function remain, they will refer to the single copy
15889 Since GCC eventually will implement ISO C99 semantics for inline
15890 functions, it is best to use `static inline' only to guarantee
15891 compatibility. (The existing semantics will remain available when
15892 `-std=gnu89' is specified, but eventually the default will be
15893 `-std=gnu99' and that will implement the C99 semantics, though it does
15896 GCC does not inline any functions when not optimizing unless you
15897 specify the `always_inline' attribute for the function, like this:
15900 inline void foo (const char) __attribute__((always_inline));
15903 File: gcc.info, Node: Extended Asm, Next: Constraints, Prev: Inline, Up: C Extensions
15905 5.34 Assembler Instructions with C Expression Operands
15906 ======================================================
15908 In an assembler instruction using `asm', you can specify the operands
15909 of the instruction using C expressions. This means you need not guess
15910 which registers or memory locations will contain the data you want to
15913 You must specify an assembler instruction template much like what
15914 appears in a machine description, plus an operand constraint string for
15917 For example, here is how to use the 68881's `fsinx' instruction:
15919 asm ("fsinx %1,%0" : "=f" (result) : "f" (angle));
15921 Here `angle' is the C expression for the input operand while `result'
15922 is that of the output operand. Each has `"f"' as its operand
15923 constraint, saying that a floating point register is required. The `='
15924 in `=f' indicates that the operand is an output; all output operands'
15925 constraints must use `='. The constraints use the same language used
15926 in the machine description (*note Constraints::).
15928 Each operand is described by an operand-constraint string followed by
15929 the C expression in parentheses. A colon separates the assembler
15930 template from the first output operand and another separates the last
15931 output operand from the first input, if any. Commas separate the
15932 operands within each group. The total number of operands is currently
15933 limited to 30; this limitation may be lifted in some future version of
15936 If there are no output operands but there are input operands, you must
15937 place two consecutive colons surrounding the place where the output
15940 As of GCC version 3.1, it is also possible to specify input and output
15941 operands using symbolic names which can be referenced within the
15942 assembler code. These names are specified inside square brackets
15943 preceding the constraint string, and can be referenced inside the
15944 assembler code using `%[NAME]' instead of a percentage sign followed by
15945 the operand number. Using named operands the above example could look
15948 asm ("fsinx %[angle],%[output]"
15949 : [output] "=f" (result)
15950 : [angle] "f" (angle));
15952 Note that the symbolic operand names have no relation whatsoever to
15953 other C identifiers. You may use any name you like, even those of
15954 existing C symbols, but you must ensure that no two operands within the
15955 same assembler construct use the same symbolic name.
15957 Output operand expressions must be lvalues; the compiler can check
15958 this. The input operands need not be lvalues. The compiler cannot
15959 check whether the operands have data types that are reasonable for the
15960 instruction being executed. It does not parse the assembler instruction
15961 template and does not know what it means or even whether it is valid
15962 assembler input. The extended `asm' feature is most often used for
15963 machine instructions the compiler itself does not know exist. If the
15964 output expression cannot be directly addressed (for example, it is a
15965 bit-field), your constraint must allow a register. In that case, GCC
15966 will use the register as the output of the `asm', and then store that
15967 register into the output.
15969 The ordinary output operands must be write-only; GCC will assume that
15970 the values in these operands before the instruction are dead and need
15971 not be generated. Extended asm supports input-output or read-write
15972 operands. Use the constraint character `+' to indicate such an operand
15973 and list it with the output operands. You should only use read-write
15974 operands when the constraints for the operand (or the operand in which
15975 only some of the bits are to be changed) allow a register.
15977 You may, as an alternative, logically split its function into two
15978 separate operands, one input operand and one write-only output operand.
15979 The connection between them is expressed by constraints which say they
15980 need to be in the same location when the instruction executes. You can
15981 use the same C expression for both operands, or different expressions.
15982 For example, here we write the (fictitious) `combine' instruction with
15983 `bar' as its read-only source operand and `foo' as its read-write
15986 asm ("combine %2,%0" : "=r" (foo) : "0" (foo), "g" (bar));
15988 The constraint `"0"' for operand 1 says that it must occupy the same
15989 location as operand 0. A number in constraint is allowed only in an
15990 input operand and it must refer to an output operand.
15992 Only a number in the constraint can guarantee that one operand will be
15993 in the same place as another. The mere fact that `foo' is the value of
15994 both operands is not enough to guarantee that they will be in the same
15995 place in the generated assembler code. The following would not work
15998 asm ("combine %2,%0" : "=r" (foo) : "r" (foo), "g" (bar));
16000 Various optimizations or reloading could cause operands 0 and 1 to be
16001 in different registers; GCC knows no reason not to do so. For example,
16002 the compiler might find a copy of the value of `foo' in one register and
16003 use it for operand 1, but generate the output operand 0 in a different
16004 register (copying it afterward to `foo''s own address). Of course,
16005 since the register for operand 1 is not even mentioned in the assembler
16006 code, the result will not work, but GCC can't tell that.
16008 As of GCC version 3.1, one may write `[NAME]' instead of the operand
16009 number for a matching constraint. For example:
16011 asm ("cmoveq %1,%2,%[result]"
16012 : [result] "=r"(result)
16013 : "r" (test), "r"(new), "[result]"(old));
16015 Sometimes you need to make an `asm' operand be a specific register,
16016 but there's no matching constraint letter for that register _by
16017 itself_. To force the operand into that register, use a local variable
16018 for the operand and specify the register in the variable declaration.
16019 *Note Explicit Reg Vars::. Then for the `asm' operand, use any
16020 register constraint letter that matches the register:
16022 register int *p1 asm ("r0") = ...;
16023 register int *p2 asm ("r1") = ...;
16024 register int *result asm ("r0");
16025 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
16027 In the above example, beware that a register that is call-clobbered by
16028 the target ABI will be overwritten by any function call in the
16029 assignment, including library calls for arithmetic operators. Assuming
16030 it is a call-clobbered register, this may happen to `r0' above by the
16031 assignment to `p2'. If you have to use such a register, use temporary
16032 variables for expressions between the register assignment and use:
16035 register int *p1 asm ("r0") = ...;
16036 register int *p2 asm ("r1") = t1;
16037 register int *result asm ("r0");
16038 asm ("sysint" : "=r" (result) : "0" (p1), "r" (p2));
16040 Some instructions clobber specific hard registers. To describe this,
16041 write a third colon after the input operands, followed by the names of
16042 the clobbered hard registers (given as strings). Here is a realistic
16043 example for the VAX:
16045 asm volatile ("movc3 %0,%1,%2"
16047 : "g" (from), "g" (to), "g" (count)
16048 : "r0", "r1", "r2", "r3", "r4", "r5");
16050 You may not write a clobber description in a way that overlaps with an
16051 input or output operand. For example, you may not have an operand
16052 describing a register class with one member if you mention that register
16053 in the clobber list. Variables declared to live in specific registers
16054 (*note Explicit Reg Vars::), and used as asm input or output operands
16055 must have no part mentioned in the clobber description. There is no
16056 way for you to specify that an input operand is modified without also
16057 specifying it as an output operand. Note that if all the output
16058 operands you specify are for this purpose (and hence unused), you will
16059 then also need to specify `volatile' for the `asm' construct, as
16060 described below, to prevent GCC from deleting the `asm' statement as
16063 If you refer to a particular hardware register from the assembler code,
16064 you will probably have to list the register after the third colon to
16065 tell the compiler the register's value is modified. In some assemblers,
16066 the register names begin with `%'; to produce one `%' in the assembler
16067 code, you must write `%%' in the input.
16069 If your assembler instruction can alter the condition code register,
16070 add `cc' to the list of clobbered registers. GCC on some machines
16071 represents the condition codes as a specific hardware register; `cc'
16072 serves to name this register. On other machines, the condition code is
16073 handled differently, and specifying `cc' has no effect. But it is
16074 valid no matter what the machine.
16076 If your assembler instructions access memory in an unpredictable
16077 fashion, add `memory' to the list of clobbered registers. This will
16078 cause GCC to not keep memory values cached in registers across the
16079 assembler instruction and not optimize stores or loads to that memory.
16080 You will also want to add the `volatile' keyword if the memory affected
16081 is not listed in the inputs or outputs of the `asm', as the `memory'
16082 clobber does not count as a side-effect of the `asm'. If you know how
16083 large the accessed memory is, you can add it as input or output but if
16084 this is not known, you should add `memory'. As an example, if you
16085 access ten bytes of a string, you can use a memory input like:
16087 {"m"( ({ struct { char x[10]; } *p = (void *)ptr ; *p; }) )}.
16089 Note that in the following example the memory input is necessary,
16090 otherwise GCC might optimize the store to `x' away:
16096 asm ("magic stuff accessing an 'int' pointed to by '%1'"
16097 "=&d" (r) : "a" (y), "m" (*y));
16101 You can put multiple assembler instructions together in a single `asm'
16102 template, separated by the characters normally used in assembly code
16103 for the system. A combination that works in most places is a newline
16104 to break the line, plus a tab character to move to the instruction field
16105 (written as `\n\t'). Sometimes semicolons can be used, if the
16106 assembler allows semicolons as a line-breaking character. Note that
16107 some assembler dialects use semicolons to start a comment. The input
16108 operands are guaranteed not to use any of the clobbered registers, and
16109 neither will the output operands' addresses, so you can read and write
16110 the clobbered registers as many times as you like. Here is an example
16111 of multiple instructions in a template; it assumes the subroutine
16112 `_foo' accepts arguments in registers 9 and 10:
16114 asm ("movl %0,r9\n\tmovl %1,r10\n\tcall _foo"
16116 : "g" (from), "g" (to)
16119 Unless an output operand has the `&' constraint modifier, GCC may
16120 allocate it in the same register as an unrelated input operand, on the
16121 assumption the inputs are consumed before the outputs are produced.
16122 This assumption may be false if the assembler code actually consists of
16123 more than one instruction. In such a case, use `&' for each output
16124 operand that may not overlap an input. *Note Modifiers::.
16126 If you want to test the condition code produced by an assembler
16127 instruction, you must include a branch and a label in the `asm'
16128 construct, as follows:
16130 asm ("clr %0\n\tfrob %1\n\tbeq 0f\n\tmov #1,%0\n0:"
16134 This assumes your assembler supports local labels, as the GNU assembler
16135 and most Unix assemblers do.
16137 Speaking of labels, jumps from one `asm' to another are not supported.
16138 The compiler's optimizers do not know about these jumps, and therefore
16139 they cannot take account of them when deciding how to optimize.
16141 Usually the most convenient way to use these `asm' instructions is to
16142 encapsulate them in macros that look like functions. For example,
16145 ({ double __value, __arg = (x); \
16146 asm ("fsinx %1,%0": "=f" (__value): "f" (__arg)); \
16149 Here the variable `__arg' is used to make sure that the instruction
16150 operates on a proper `double' value, and to accept only those arguments
16151 `x' which can convert automatically to a `double'.
16153 Another way to make sure the instruction operates on the correct data
16154 type is to use a cast in the `asm'. This is different from using a
16155 variable `__arg' in that it converts more different types. For
16156 example, if the desired type were `int', casting the argument to `int'
16157 would accept a pointer with no complaint, while assigning the argument
16158 to an `int' variable named `__arg' would warn about using a pointer
16159 unless the caller explicitly casts it.
16161 If an `asm' has output operands, GCC assumes for optimization purposes
16162 the instruction has no side effects except to change the output
16163 operands. This does not mean instructions with a side effect cannot be
16164 used, but you must be careful, because the compiler may eliminate them
16165 if the output operands aren't used, or move them out of loops, or
16166 replace two with one if they constitute a common subexpression. Also,
16167 if your instruction does have a side effect on a variable that otherwise
16168 appears not to change, the old value of the variable may be reused later
16169 if it happens to be found in a register.
16171 You can prevent an `asm' instruction from being deleted by writing the
16172 keyword `volatile' after the `asm'. For example:
16174 #define get_and_set_priority(new) \
16176 asm volatile ("get_and_set_priority %0, %1" \
16177 : "=g" (__old) : "g" (new)); \
16180 The `volatile' keyword indicates that the instruction has important
16181 side-effects. GCC will not delete a volatile `asm' if it is reachable.
16182 (The instruction can still be deleted if GCC can prove that
16183 control-flow will never reach the location of the instruction.) Note
16184 that even a volatile `asm' instruction can be moved relative to other
16185 code, including across jump instructions. For example, on many targets
16186 there is a system register which can be set to control the rounding
16187 mode of floating point operations. You might try setting it with a
16188 volatile `asm', like this PowerPC example:
16190 asm volatile("mtfsf 255,%0" : : "f" (fpenv));
16193 This will not work reliably, as the compiler may move the addition back
16194 before the volatile `asm'. To make it work you need to add an
16195 artificial dependency to the `asm' referencing a variable in the code
16196 you don't want moved, for example:
16198 asm volatile ("mtfsf 255,%1" : "=X"(sum): "f"(fpenv));
16201 Similarly, you can't expect a sequence of volatile `asm' instructions
16202 to remain perfectly consecutive. If you want consecutive output, use a
16203 single `asm'. Also, GCC will perform some optimizations across a
16204 volatile `asm' instruction; GCC does not "forget everything" when it
16205 encounters a volatile `asm' instruction the way some other compilers do.
16207 An `asm' instruction without any output operands will be treated
16208 identically to a volatile `asm' instruction.
16210 It is a natural idea to look for a way to give access to the condition
16211 code left by the assembler instruction. However, when we attempted to
16212 implement this, we found no way to make it work reliably. The problem
16213 is that output operands might need reloading, which would result in
16214 additional following "store" instructions. On most machines, these
16215 instructions would alter the condition code before there was time to
16216 test it. This problem doesn't arise for ordinary "test" and "compare"
16217 instructions because they don't have any output operands.
16219 For reasons similar to those described above, it is not possible to
16220 give an assembler instruction access to the condition code left by
16221 previous instructions.
16223 If you are writing a header file that should be includable in ISO C
16224 programs, write `__asm__' instead of `asm'. *Note Alternate Keywords::.
16226 5.34.1 Size of an `asm'
16227 -----------------------
16229 Some targets require that GCC track the size of each instruction used in
16230 order to generate correct code. Because the final length of an `asm'
16231 is only known by the assembler, GCC must make an estimate as to how big
16232 it will be. The estimate is formed by counting the number of
16233 statements in the pattern of the `asm' and multiplying that by the
16234 length of the longest instruction on that processor. Statements in the
16235 `asm' are identified by newline characters and whatever statement
16236 separator characters are supported by the assembler; on most processors
16237 this is the ``;'' character.
16239 Normally, GCC's estimate is perfectly adequate to ensure that correct
16240 code is generated, but it is possible to confuse the compiler if you use
16241 pseudo instructions or assembler macros that expand into multiple real
16242 instructions or if you use assembler directives that expand to more
16243 space in the object file than would be needed for a single instruction.
16244 If this happens then the assembler will produce a diagnostic saying that
16245 a label is unreachable.
16247 5.34.2 i386 floating point asm operands
16248 ---------------------------------------
16250 There are several rules on the usage of stack-like regs in asm_operands
16251 insns. These rules apply only to the operands that are stack-like regs:
16253 1. Given a set of input regs that die in an asm_operands, it is
16254 necessary to know which are implicitly popped by the asm, and
16255 which must be explicitly popped by gcc.
16257 An input reg that is implicitly popped by the asm must be
16258 explicitly clobbered, unless it is constrained to match an output
16261 2. For any input reg that is implicitly popped by an asm, it is
16262 necessary to know how to adjust the stack to compensate for the
16263 pop. If any non-popped input is closer to the top of the
16264 reg-stack than the implicitly popped reg, it would not be possible
16265 to know what the stack looked like--it's not clear how the rest of
16266 the stack "slides up".
16268 All implicitly popped input regs must be closer to the top of the
16269 reg-stack than any input that is not implicitly popped.
16271 It is possible that if an input dies in an insn, reload might use
16272 the input reg for an output reload. Consider this example:
16274 asm ("foo" : "=t" (a) : "f" (b));
16276 This asm says that input B is not popped by the asm, and that the
16277 asm pushes a result onto the reg-stack, i.e., the stack is one
16278 deeper after the asm than it was before. But, it is possible that
16279 reload will think that it can use the same reg for both the input
16280 and the output, if input B dies in this insn.
16282 If any input operand uses the `f' constraint, all output reg
16283 constraints must use the `&' earlyclobber.
16285 The asm above would be written as
16287 asm ("foo" : "=&t" (a) : "f" (b));
16289 3. Some operands need to be in particular places on the stack. All
16290 output operands fall in this category--there is no other way to
16291 know which regs the outputs appear in unless the user indicates
16292 this in the constraints.
16294 Output operands must specifically indicate which reg an output
16295 appears in after an asm. `=f' is not allowed: the operand
16296 constraints must select a class with a single reg.
16298 4. Output operands may not be "inserted" between existing stack regs.
16299 Since no 387 opcode uses a read/write operand, all output operands
16300 are dead before the asm_operands, and are pushed by the
16301 asm_operands. It makes no sense to push anywhere but the top of
16304 Output operands must start at the top of the reg-stack: output
16305 operands may not "skip" a reg.
16307 5. Some asm statements may need extra stack space for internal
16308 calculations. This can be guaranteed by clobbering stack registers
16309 unrelated to the inputs and outputs.
16312 Here are a couple of reasonable asms to want to write. This asm takes
16313 one input, which is internally popped, and produces two outputs.
16315 asm ("fsincos" : "=t" (cos), "=u" (sin) : "0" (inp));
16317 This asm takes two inputs, which are popped by the `fyl2xp1' opcode,
16318 and replaces them with one output. The user must code the `st(1)'
16319 clobber for reg-stack.c to know that `fyl2xp1' pops both inputs.
16321 asm ("fyl2xp1" : "=t" (result) : "0" (x), "u" (y) : "st(1)");
16324 File: gcc.info, Node: Constraints, Next: Asm Labels, Prev: Extended Asm, Up: C Extensions
16326 5.35 Constraints for `asm' Operands
16327 ===================================
16329 Here are specific details on what constraint letters you can use with
16330 `asm' operands. Constraints can say whether an operand may be in a
16331 register, and which kinds of register; whether the operand can be a
16332 memory reference, and which kinds of address; whether the operand may
16333 be an immediate constant, and which possible values it may have.
16334 Constraints can also require two operands to match.
16338 * Simple Constraints:: Basic use of constraints.
16339 * Multi-Alternative:: When an insn has two alternative constraint-patterns.
16340 * Modifiers:: More precise control over effects of constraints.
16341 * Machine Constraints:: Special constraints for some particular machines.
16344 File: gcc.info, Node: Simple Constraints, Next: Multi-Alternative, Up: Constraints
16346 5.35.1 Simple Constraints
16347 -------------------------
16349 The simplest kind of constraint is a string full of letters, each of
16350 which describes one kind of operand that is permitted. Here are the
16351 letters that are allowed:
16354 Whitespace characters are ignored and can be inserted at any
16355 position except the first. This enables each alternative for
16356 different operands to be visually aligned in the machine
16357 description even if they have different number of constraints and
16361 A memory operand is allowed, with any kind of address that the
16362 machine supports in general.
16365 A memory operand is allowed, but only if the address is
16366 "offsettable". This means that adding a small integer (actually,
16367 the width in bytes of the operand, as determined by its machine
16368 mode) may be added to the address and the result is also a valid
16371 For example, an address which is constant is offsettable; so is an
16372 address that is the sum of a register and a constant (as long as a
16373 slightly larger constant is also within the range of
16374 address-offsets supported by the machine); but an autoincrement or
16375 autodecrement address is not offsettable. More complicated
16376 indirect/indexed addresses may or may not be offsettable depending
16377 on the other addressing modes that the machine supports.
16379 Note that in an output operand which can be matched by another
16380 operand, the constraint letter `o' is valid only when accompanied
16381 by both `<' (if the target machine has predecrement addressing)
16382 and `>' (if the target machine has preincrement addressing).
16385 A memory operand that is not offsettable. In other words,
16386 anything that would fit the `m' constraint but not the `o'
16390 A memory operand with autodecrement addressing (either
16391 predecrement or postdecrement) is allowed.
16394 A memory operand with autoincrement addressing (either
16395 preincrement or postincrement) is allowed.
16398 A register operand is allowed provided that it is in a general
16402 An immediate integer operand (one with constant value) is allowed.
16403 This includes symbolic constants whose values will be known only at
16404 assembly time or later.
16407 An immediate integer operand with a known numeric value is allowed.
16408 Many systems cannot support assembly-time constants for operands
16409 less than a word wide. Constraints for these operands should use
16410 `n' rather than `i'.
16412 `I', `J', `K', ... `P'
16413 Other letters in the range `I' through `P' may be defined in a
16414 machine-dependent fashion to permit immediate integer operands with
16415 explicit integer values in specified ranges. For example, on the
16416 68000, `I' is defined to stand for the range of values 1 to 8.
16417 This is the range permitted as a shift count in the shift
16421 An immediate floating operand (expression code `const_double') is
16422 allowed, but only if the target floating point format is the same
16423 as that of the host machine (on which the compiler is running).
16426 An immediate floating operand (expression code `const_double' or
16427 `const_vector') is allowed.
16430 `G' and `H' may be defined in a machine-dependent fashion to
16431 permit immediate floating operands in particular ranges of values.
16434 An immediate integer operand whose value is not an explicit
16435 integer is allowed.
16437 This might appear strange; if an insn allows a constant operand
16438 with a value not known at compile time, it certainly must allow
16439 any known value. So why use `s' instead of `i'? Sometimes it
16440 allows better code to be generated.
16442 For example, on the 68000 in a fullword instruction it is possible
16443 to use an immediate operand; but if the immediate value is between
16444 -128 and 127, better code results from loading the value into a
16445 register and using the register. This is because the load into
16446 the register can be done with a `moveq' instruction. We arrange
16447 for this to happen by defining the letter `K' to mean "any integer
16448 outside the range -128 to 127", and then specifying `Ks' in the
16449 operand constraints.
16452 Any register, memory or immediate integer operand is allowed,
16453 except for registers that are not general registers.
16456 Any operand whatsoever is allowed.
16458 `0', `1', `2', ... `9'
16459 An operand that matches the specified operand number is allowed.
16460 If a digit is used together with letters within the same
16461 alternative, the digit should come last.
16463 This number is allowed to be more than a single digit. If multiple
16464 digits are encountered consecutively, they are interpreted as a
16465 single decimal integer. There is scant chance for ambiguity,
16466 since to-date it has never been desirable that `10' be interpreted
16467 as matching either operand 1 _or_ operand 0. Should this be
16468 desired, one can use multiple alternatives instead.
16470 This is called a "matching constraint" and what it really means is
16471 that the assembler has only a single operand that fills two roles
16472 which `asm' distinguishes. For example, an add instruction uses
16473 two input operands and an output operand, but on most CISC
16474 machines an add instruction really has only two operands, one of
16475 them an input-output operand:
16479 Matching constraints are used in these circumstances. More
16480 precisely, the two operands that match must include one input-only
16481 operand and one output-only operand. Moreover, the digit must be a
16482 smaller number than the number of the operand that uses it in the
16486 An operand that is a valid memory address is allowed. This is for
16487 "load address" and "push address" instructions.
16489 `p' in the constraint must be accompanied by `address_operand' as
16490 the predicate in the `match_operand'. This predicate interprets
16491 the mode specified in the `match_operand' as the mode of the memory
16492 reference for which the address would be valid.
16495 Other letters can be defined in machine-dependent fashion to stand
16496 for particular classes of registers or other arbitrary operand
16497 types. `d', `a' and `f' are defined on the 68000/68020 to stand
16498 for data, address and floating point registers.
16502 File: gcc.info, Node: Multi-Alternative, Next: Modifiers, Prev: Simple Constraints, Up: Constraints
16504 5.35.2 Multiple Alternative Constraints
16505 ---------------------------------------
16507 Sometimes a single instruction has multiple alternative sets of possible
16508 operands. For example, on the 68000, a logical-or instruction can
16509 combine register or an immediate value into memory, or it can combine
16510 any kind of operand into a register; but it cannot combine one memory
16511 location into another.
16513 These constraints are represented as multiple alternatives. An
16514 alternative can be described by a series of letters for each operand.
16515 The overall constraint for an operand is made from the letters for this
16516 operand from the first alternative, a comma, the letters for this
16517 operand from the second alternative, a comma, and so on until the last
16520 If all the operands fit any one alternative, the instruction is valid.
16521 Otherwise, for each alternative, the compiler counts how many
16522 instructions must be added to copy the operands so that that
16523 alternative applies. The alternative requiring the least copying is
16524 chosen. If two alternatives need the same amount of copying, the one
16525 that comes first is chosen. These choices can be altered with the `?'
16526 and `!' characters:
16529 Disparage slightly the alternative that the `?' appears in, as a
16530 choice when no alternative applies exactly. The compiler regards
16531 this alternative as one unit more costly for each `?' that appears
16535 Disparage severely the alternative that the `!' appears in. This
16536 alternative can still be used if it fits without reloading, but if
16537 reloading is needed, some other alternative will be used.
16540 File: gcc.info, Node: Modifiers, Next: Machine Constraints, Prev: Multi-Alternative, Up: Constraints
16542 5.35.3 Constraint Modifier Characters
16543 -------------------------------------
16545 Here are constraint modifier characters.
16548 Means that this operand is write-only for this instruction: the
16549 previous value is discarded and replaced by output data.
16552 Means that this operand is both read and written by the
16555 When the compiler fixes up the operands to satisfy the constraints,
16556 it needs to know which operands are inputs to the instruction and
16557 which are outputs from it. `=' identifies an output; `+'
16558 identifies an operand that is both input and output; all other
16559 operands are assumed to be input only.
16561 If you specify `=' or `+' in a constraint, you put it in the first
16562 character of the constraint string.
16565 Means (in a particular alternative) that this operand is an
16566 "earlyclobber" operand, which is modified before the instruction is
16567 finished using the input operands. Therefore, this operand may
16568 not lie in a register that is used as an input operand or as part
16569 of any memory address.
16571 `&' applies only to the alternative in which it is written. In
16572 constraints with multiple alternatives, sometimes one alternative
16573 requires `&' while others do not. See, for example, the `movdf'
16576 An input operand can be tied to an earlyclobber operand if its only
16577 use as an input occurs before the early result is written. Adding
16578 alternatives of this form often allows GCC to produce better code
16579 when only some of the inputs can be affected by the earlyclobber.
16580 See, for example, the `mulsi3' insn of the ARM.
16582 `&' does not obviate the need to write `='.
16585 Declares the instruction to be commutative for this operand and the
16586 following operand. This means that the compiler may interchange
16587 the two operands if that is the cheapest way to make all operands
16588 fit the constraints. GCC can only handle one commutative pair in
16589 an asm; if you use more, the compiler may fail. Note that you
16590 need not use the modifier if the two alternatives are strictly
16591 identical; this would only waste time in the reload pass.
16594 Says that all following characters, up to the next comma, are to be
16595 ignored as a constraint. They are significant only for choosing
16596 register preferences.
16599 Says that the following character should be ignored when choosing
16600 register preferences. `*' has no effect on the meaning of the
16601 constraint as a constraint, and no effect on reloading.
16605 File: gcc.info, Node: Machine Constraints, Prev: Modifiers, Up: Constraints
16607 5.35.4 Constraints for Particular Machines
16608 ------------------------------------------
16610 Whenever possible, you should use the general-purpose constraint letters
16611 in `asm' arguments, since they will convey meaning more readily to
16612 people reading your code. Failing that, use the constraint letters
16613 that usually have very similar meanings across architectures. The most
16614 commonly used constraints are `m' and `r' (for memory and
16615 general-purpose registers respectively; *note Simple Constraints::), and
16616 `I', usually the letter indicating the most common immediate-constant
16619 For each machine architecture, the `config/MACHINE/MACHINE.h' file
16620 defines additional constraints. These constraints are used by the
16621 compiler itself for instruction generation, as well as for `asm'
16622 statements; therefore, some of the constraints are not particularly
16623 interesting for `asm'. The constraints are defined through these
16626 `REG_CLASS_FROM_LETTER'
16627 Register class constraints (usually lowercase).
16629 `CONST_OK_FOR_LETTER_P'
16630 Immediate constant constraints, for non-floating point constants of
16631 word size or smaller precision (usually uppercase).
16633 `CONST_DOUBLE_OK_FOR_LETTER_P'
16634 Immediate constant constraints, for all floating point constants
16635 and for constants of greater than word size precision (usually
16639 Special cases of registers or memory. This macro is not required,
16640 and is only defined for some machines.
16642 Inspecting these macro definitions in the compiler source for your
16643 machine is the best way to be certain you have the right constraints.
16644 However, here is a summary of the machine-dependent constraints
16645 available on some particular machines.
16647 _ARM family--`arm.h'_
16650 Floating-point register
16653 VFP floating-point register
16656 One of the floating-point constants 0.0, 0.5, 1.0, 2.0, 3.0,
16660 Floating-point constant that would satisfy the constraint `F'
16664 Integer that is valid as an immediate operand in a data
16665 processing instruction. That is, an integer in the range 0
16666 to 255 rotated by a multiple of 2
16669 Integer in the range -4095 to 4095
16672 Integer that satisfies constraint `I' when inverted (ones
16676 Integer that satisfies constraint `I' when negated (twos
16680 Integer in the range 0 to 32
16683 A memory reference where the exact address is in a single
16684 register (``m'' is preferable for `asm' statements)
16687 An item in the constant pool
16690 A symbol in the text segment of the current file
16693 A memory reference suitable for VFP load/store insns (reg+constant
16697 A memory reference suitable for iWMMXt load/store instructions.
16700 A memory reference suitable for the ARMv4 ldrsb instruction.
16702 _AVR family--`avr.h'_
16705 Registers from r0 to r15
16708 Registers from r16 to r23
16711 Registers from r16 to r31
16714 Registers from r24 to r31. These registers can be used in
16718 Pointer register (r26-r31)
16721 Base pointer register (r28-r31)
16724 Stack pointer register (SPH:SPL)
16727 Temporary register r0
16730 Register pair X (r27:r26)
16733 Register pair Y (r29:r28)
16736 Register pair Z (r31:r30)
16739 Constant greater than -1, less than 64
16742 Constant greater than -64, less than 1
16751 Constant that fits in 8 bits
16754 Constant integer -1
16757 Constant integer 8, 16, or 24
16763 A floating point constant 0.0
16765 _PowerPC and IBM RS6000--`rs6000.h'_
16768 Address base register
16771 Floating point register
16777 `MQ', `CTR', or `LINK' register
16789 `CR' register (condition register) number 0
16792 `CR' register (condition register)
16795 `FPMEM' stack memory for FPR-GPR transfers
16798 Signed 16-bit constant
16801 Unsigned 16-bit constant shifted left 16 bits (use `L'
16802 instead for `SImode' constants)
16805 Unsigned 16-bit constant
16808 Signed 16-bit constant shifted left 16 bits
16811 Constant larger than 31
16820 Constant whose negation is a signed 16-bit constant
16823 Floating point constant that can be loaded into a register
16824 with one instruction per word
16827 Memory operand that is an offset from a register (`m' is
16828 preferable for `asm' statements)
16834 Constant suitable as a 64-bit mask operand
16837 Constant suitable as a 32-bit mask operand
16840 System V Release 4 small data area reference
16842 _Intel 386--`i386.h'_
16845 `a', `b', `c', or `d' register for the i386. For x86-64 it
16846 is equivalent to `r' class (for 8-bit instructions that do
16847 not use upper halves).
16850 `a', `b', `c', or `d' register (for 8-bit instructions, that
16851 do use upper halves).
16854 Legacy register--equivalent to `r' class in i386 mode. (for
16855 non-8-bit registers used together with 8-bit upper halves in
16856 a single instruction)
16859 Specifies the `a' or `d' registers. This is primarily useful
16860 for 64-bit integer values (when in 32-bit mode) intended to
16861 be returned with the `d' register holding the most
16862 significant bits and the `a' register holding the least
16866 Floating point register
16869 First (top of stack) floating point register
16872 Second floating point register
16884 Specifies constant that can be easily constructed in SSE
16885 register without loading it from memory.
16903 Constant in range 0 to 31 (for 32-bit shifts)
16906 Constant in range 0 to 63 (for 64-bit shifts)
16915 0, 1, 2, or 3 (shifts for `lea' instruction)
16918 Constant in range 0 to 255 (for `out' instruction)
16921 Constant in range 0 to `0xffffffff' or symbolic reference
16922 known to fit specified range. (for using immediates in zero
16923 extending 32-bit to 64-bit x86-64 instructions)
16926 Constant in range -2147483648 to 2147483647 or symbolic
16927 reference known to fit specified range. (for using
16928 immediates in 64-bit x86-64 instructions)
16931 Standard 80387 floating point constant
16933 _Intel IA-64--`ia64.h'_
16936 General register `r0' to `r3' for `addl' instruction
16942 Predicate register (`c' as in "conditional")
16945 Application register residing in M-unit
16948 Application register residing in I-unit
16951 Floating-point register
16954 Memory operand. Remember that `m' allows postincrement and
16955 postdecrement which require printing with `%Pn' on IA-64.
16956 Use `S' to disallow postincrement and postdecrement.
16959 Floating-point constant 0.0 or 1.0
16962 14-bit signed integer constant
16965 22-bit signed integer constant
16968 8-bit signed integer constant for logical instructions
16971 8-bit adjusted signed integer constant for compare pseudo-ops
16974 6-bit unsigned integer constant for shift counts
16977 9-bit signed integer constant for load and store
16984 0 or -1 for `dep' instruction
16987 Non-volatile memory for floating-point loads and stores
16990 Integer constant in the range 1 to 4 for `shladd' instruction
16993 Memory operand except postincrement and postdecrement
16998 Register in the class `ACC_REGS' (`acc0' to `acc7').
17001 Register in the class `EVEN_ACC_REGS' (`acc0' to `acc7').
17004 Register in the class `CC_REGS' (`fcc0' to `fcc3' and `icc0'
17008 Register in the class `GPR_REGS' (`gr0' to `gr63').
17011 Register in the class `EVEN_REGS' (`gr0' to `gr63'). Odd
17012 registers are excluded not in the class but through the use
17013 of a machine mode larger than 4 bytes.
17016 Register in the class `FPR_REGS' (`fr0' to `fr63').
17019 Register in the class `FEVEN_REGS' (`fr0' to `fr63'). Odd
17020 registers are excluded not in the class but through the use
17021 of a machine mode larger than 4 bytes.
17024 Register in the class `LR_REG' (the `lr' register).
17027 Register in the class `QUAD_REGS' (`gr2' to `gr63').
17028 Register numbers not divisible by 4 are excluded not in the
17029 class but through the use of a machine mode larger than 8
17033 Register in the class `ICC_REGS' (`icc0' to `icc3').
17036 Register in the class `FCC_REGS' (`fcc0' to `fcc3').
17039 Register in the class `ICR_REGS' (`cc4' to `cc7').
17042 Register in the class `FCR_REGS' (`cc0' to `cc3').
17045 Register in the class `QUAD_FPR_REGS' (`fr0' to `fr63').
17046 Register numbers not divisible by 4 are excluded not in the
17047 class but through the use of a machine mode larger than 8
17051 Register in the class `SPR_REGS' (`lcr' and `lr').
17054 Register in the class `QUAD_ACC_REGS' (`acc0' to `acc7').
17057 Register in the class `ACCG_REGS' (`accg0' to `accg7').
17060 Register in the class `CR_REGS' (`cc0' to `cc7').
17063 Floating point constant zero
17066 6-bit signed integer constant
17069 10-bit signed integer constant
17072 16-bit signed integer constant
17075 16-bit unsigned integer constant
17078 12-bit signed integer constant that is negative--i.e. in the
17079 range of -2048 to -1
17085 12-bit signed integer constant that is greater than
17086 zero--i.e. in the range of 1 to 2047.
17089 _Blackfin family--`bfin.h'_
17098 A call clobbered P register.
17101 Even-numbered D register
17104 Odd-numbered D register
17107 Accumulator register.
17110 Even-numbered accumulator register.
17113 Odd-numbered accumulator register.
17125 Registers used for circular buffering, i.e. I, B, or L
17132 Any D, P, B, M, I or L register.
17135 Additional registers typically used only in prologues and
17136 epilogues: RETS, RETN, RETI, RETX, RETE, ASTAT, SEQSTAT and
17140 Any register except accumulators or CC.
17143 Signed 16 bit integer (in the range -32768 to 32767)
17146 Unsigned 16 bit integer (in the range 0 to 65535)
17149 Signed 7 bit integer (in the range -64 to 63)
17152 Unsigned 7 bit integer (in the range 0 to 127)
17155 Unsigned 5 bit integer (in the range 0 to 31)
17158 Signed 4 bit integer (in the range -8 to 7)
17161 Signed 3 bit integer (in the range -3 to 4)
17164 Unsigned 3 bit integer (in the range 0 to 7)
17167 Constant N, where N is a single-digit constant in the range 0
17177 An integer constant with exactly a single bit set.
17180 An integer constant with all bits set except exactly one.
17190 `DP' or `IP' registers (general address)
17214 `DP' or `SP' registers (offsettable address)
17217 Non-pointer registers (not `SP', `DP', `IP')
17220 Non-SP registers (everything except `SP')
17223 Indirect through `IP'--Avoid this except for `QImode', since
17224 we can't access extra bytes
17227 Indirect through `SP' or `DP' with short displacement (0..127)
17230 Data-section immediate value
17233 Integers from -255 to -1
17236 Integers from 0 to 7--valid bit number in a register
17239 Integers from 0 to 127--valid displacement for addressing mode
17242 Integers from 1 to 127
17254 Integers from 0 to 255
17259 General-purpose integer register
17262 Floating-point register (if available)
17271 `Hi' or `Lo' register
17274 General-purpose integer register
17277 Floating-point status register
17280 Signed 16-bit constant (for arithmetic instructions)
17286 Zero-extended 16-bit constant (for logic instructions)
17289 Constant with low 16 bits zero (can be loaded with `lui')
17292 32-bit constant which requires two instructions to load (a
17293 constant which is not `I', `K', or `L')
17296 Negative 16-bit constant
17302 Positive 16-bit constant
17305 Floating point zero
17308 Memory reference that can be loaded with more than one
17309 instruction (`m' is preferable for `asm' statements)
17312 Memory reference that can be loaded with one instruction (`m'
17313 is preferable for `asm' statements)
17316 Memory reference in external OSF/rose PIC format (`m' is
17317 preferable for `asm' statements)
17319 _Motorola 680x0--`m68k.h'_
17328 68881 floating-point register, if available
17331 Integer in the range 1 to 8
17334 16-bit signed number
17337 Signed number whose magnitude is greater than 0x80
17340 Integer in the range -8 to -1
17343 Signed number whose magnitude is greater than 0x100
17346 Floating point constant that is not a 68881 constant
17348 _Motorola 68HC11 & 68HC12 families--`m68hc11.h'_
17363 Temporary soft register _.tmp
17366 A soft register _.d1 to _.d31
17369 Stack pointer register
17378 Pseudo register `z' (replaced by `x' or `y' at the end)
17381 An address register: x, y or z
17384 An address register: x or y
17387 Register pair (x:d) to form a 32-bit value
17390 Constants in the range -65536 to 65535
17393 Constants whose 16-bit low part is zero
17396 Constant integer 1 or -1
17399 Constant integer 16
17402 Constants in the range -8 to 2
17408 Floating-point register on the SPARC-V8 architecture and
17409 lower floating-point register on the SPARC-V9 architecture.
17412 Floating-point register. It is equivalent to `f' on the
17413 SPARC-V8 architecture and contains both lower and upper
17414 floating-point registers on the SPARC-V9 architecture.
17417 Floating-point condition code register.
17420 Lower floating-point register. It is only valid on the
17421 SPARC-V9 architecture when the Visual Instruction Set is
17425 Floating-point register. It is only valid on the SPARC-V9
17426 architecture when the Visual Instruction Set is available.
17429 64-bit global or out register for the SPARC-V8+ architecture.
17432 Signed 13-bit constant
17438 32-bit constant with the low 12 bits clear (a constant that
17439 can be loaded with the `sethi' instruction)
17442 A constant in the range supported by `movcc' instructions
17445 A constant in the range supported by `movrcc' instructions
17448 Same as `K', except that it verifies that bits that are not
17449 in the lower 32-bit range are all zero. Must be used instead
17450 of `K' for modes wider than `SImode'
17456 Floating-point zero
17459 Signed 13-bit constant, sign-extended to 32 or 64 bits
17462 Floating-point constant whose integral representation can be
17463 moved into an integer register using a single sethi
17467 Floating-point constant whose integral representation can be
17468 moved into an integer register using a single mov instruction
17471 Floating-point constant whose integral representation can be
17472 moved into an integer register using a high/lo_sum
17473 instruction sequence
17476 Memory address aligned to an 8-byte boundary
17482 Memory address for `e' constraint registers
17488 _TMS320C3x/C4x--`c4x.h'_
17491 Auxiliary (address) register (ar0-ar7)
17494 Stack pointer register (sp)
17497 Standard (32-bit) precision integer register
17500 Extended (40-bit) precision register (r0-r11)
17503 Block count register (bk)
17506 Extended (40-bit) precision low register (r0-r7)
17509 Extended (40-bit) precision register (r0-r1)
17512 Extended (40-bit) precision register (r2-r3)
17515 Repeat count register (rc)
17518 Index register (ir0-ir1)
17521 Status (condition code) register (st)
17524 Data page register (dp)
17527 Floating-point zero
17530 Immediate 16-bit floating-point constant
17533 Signed 16-bit constant
17536 Signed 8-bit constant
17539 Signed 5-bit constant
17542 Unsigned 16-bit constant
17545 Unsigned 8-bit constant
17548 Ones complement of unsigned 16-bit constant
17551 High 16-bit constant (32-bit constant with 16 LSBs zero)
17554 Indirect memory reference with signed 8-bit or index register
17558 Indirect memory reference with unsigned 5-bit displacement
17561 Indirect memory reference with 1 bit or index register
17565 Direct memory reference
17571 _S/390 and zSeries--`s390.h'_
17574 Address register (general purpose register except r0)
17577 Condition code register
17580 Data register (arbitrary general purpose register)
17583 Floating-point register
17586 Unsigned 8-bit constant (0-255)
17589 Unsigned 12-bit constant (0-4095)
17592 Signed 16-bit constant (-32768-32767)
17595 Value appropriate as displacement.
17597 for short displacement
17599 `(-524288..524287)'
17600 for long displacement
17603 Constant integer with a value of 0x7fffffff.
17606 Multiple letter constraint followed by 4 parameter letters.
17608 number of the part counting from most to least
17615 mode of the containing operand
17618 value of the other parts (F--all bits set)
17619 The constraint matches if the specified part of a constant
17620 has a value different from it's other parts.
17623 Memory reference without index register and with short
17627 Memory reference with index register and short displacement.
17630 Memory reference without index register but with long
17634 Memory reference with index register and long displacement.
17637 Pointer with short displacement.
17640 Pointer with long displacement.
17643 Shift count operand.
17646 _Xstormy16--`stormy16.h'_
17661 Registers r0 through r7.
17664 Registers r0 and r1.
17667 The carry register.
17670 Registers r8 and r9.
17673 A constant between 0 and 3 inclusive.
17676 A constant that has exactly one bit set.
17679 A constant that has exactly one bit clear.
17682 A constant between 0 and 255 inclusive.
17685 A constant between -255 and 0 inclusive.
17688 A constant between -3 and 0 inclusive.
17691 A constant between 1 and 4 inclusive.
17694 A constant between -4 and -1 inclusive.
17697 A memory reference that is a stack push.
17700 A memory reference that is a stack pop.
17703 A memory reference that refers to a constant address of known
17707 The register indicated by Rx (not implemented yet).
17710 A constant that is not between 2 and 15 inclusive.
17716 _Xtensa--`xtensa.h'_
17719 General-purpose 32-bit register
17722 One-bit boolean register
17725 MAC16 40-bit accumulator register
17728 Signed 12-bit integer constant, for use in MOVI instructions
17731 Signed 8-bit integer constant, for use in ADDI instructions
17734 Integer constant valid for BccI instructions
17737 Unsigned constant valid for BccUI instructions
17742 File: gcc.info, Node: Asm Labels, Next: Explicit Reg Vars, Prev: Constraints, Up: C Extensions
17744 5.36 Controlling Names Used in Assembler Code
17745 =============================================
17747 You can specify the name to be used in the assembler code for a C
17748 function or variable by writing the `asm' (or `__asm__') keyword after
17749 the declarator as follows:
17751 int foo asm ("myfoo") = 2;
17753 This specifies that the name to be used for the variable `foo' in the
17754 assembler code should be `myfoo' rather than the usual `_foo'.
17756 On systems where an underscore is normally prepended to the name of a C
17757 function or variable, this feature allows you to define names for the
17758 linker that do not start with an underscore.
17760 It does not make sense to use this feature with a non-static local
17761 variable since such variables do not have assembler names. If you are
17762 trying to put the variable in a particular register, see *Note Explicit
17763 Reg Vars::. GCC presently accepts such code with a warning, but will
17764 probably be changed to issue an error, rather than a warning, in the
17767 You cannot use `asm' in this way in a function _definition_; but you
17768 can get the same effect by writing a declaration for the function
17769 before its definition and putting `asm' there, like this:
17771 extern func () asm ("FUNC");
17777 It is up to you to make sure that the assembler names you choose do not
17778 conflict with any other assembler symbols. Also, you must not use a
17779 register name; that would produce completely invalid assembler code.
17780 GCC does not as yet have the ability to store static variables in
17781 registers. Perhaps that will be added.
17784 File: gcc.info, Node: Explicit Reg Vars, Next: Alternate Keywords, Prev: Asm Labels, Up: C Extensions
17786 5.37 Variables in Specified Registers
17787 =====================================
17789 GNU C allows you to put a few global variables into specified hardware
17790 registers. You can also specify the register in which an ordinary
17791 register variable should be allocated.
17793 * Global register variables reserve registers throughout the program.
17794 This may be useful in programs such as programming language
17795 interpreters which have a couple of global variables that are
17796 accessed very often.
17798 * Local register variables in specific registers do not reserve the
17799 registers, except at the point where they are used as input or
17800 output operands in an `asm' statement and the `asm' statement
17801 itself is not deleted. The compiler's data flow analysis is
17802 capable of determining where the specified registers contain live
17803 values, and where they are available for other uses. Stores into
17804 local register variables may be deleted when they appear to be
17805 dead according to dataflow analysis. References to local register
17806 variables may be deleted or moved or simplified.
17808 These local variables are sometimes convenient for use with the
17809 extended `asm' feature (*note Extended Asm::), if you want to
17810 write one output of the assembler instruction directly into a
17811 particular register. (This will work provided the register you
17812 specify fits the constraints specified for that operand in the
17817 * Global Reg Vars::
17821 File: gcc.info, Node: Global Reg Vars, Next: Local Reg Vars, Up: Explicit Reg Vars
17823 5.37.1 Defining Global Register Variables
17824 -----------------------------------------
17826 You can define a global register variable in GNU C like this:
17828 register int *foo asm ("a5");
17830 Here `a5' is the name of the register which should be used. Choose a
17831 register which is normally saved and restored by function calls on your
17832 machine, so that library routines will not clobber it.
17834 Naturally the register name is cpu-dependent, so you would need to
17835 conditionalize your program according to cpu type. The register `a5'
17836 would be a good choice on a 68000 for a variable of pointer type. On
17837 machines with register windows, be sure to choose a "global" register
17838 that is not affected magically by the function call mechanism.
17840 In addition, operating systems on one type of cpu may differ in how
17841 they name the registers; then you would need additional conditionals.
17842 For example, some 68000 operating systems call this register `%a5'.
17844 Eventually there may be a way of asking the compiler to choose a
17845 register automatically, but first we need to figure out how it should
17846 choose and how to enable you to guide the choice. No solution is
17849 Defining a global register variable in a certain register reserves that
17850 register entirely for this use, at least within the current compilation.
17851 The register will not be allocated for any other purpose in the
17852 functions in the current compilation. The register will not be saved
17853 and restored by these functions. Stores into this register are never
17854 deleted even if they would appear to be dead, but references may be
17855 deleted or moved or simplified.
17857 It is not safe to access the global register variables from signal
17858 handlers, or from more than one thread of control, because the system
17859 library routines may temporarily use the register for other things
17860 (unless you recompile them specially for the task at hand).
17862 It is not safe for one function that uses a global register variable to
17863 call another such function `foo' by way of a third function `lose' that
17864 was compiled without knowledge of this variable (i.e. in a different
17865 source file in which the variable wasn't declared). This is because
17866 `lose' might save the register and put some other value there. For
17867 example, you can't expect a global register variable to be available in
17868 the comparison-function that you pass to `qsort', since `qsort' might
17869 have put something else in that register. (If you are prepared to
17870 recompile `qsort' with the same global register variable, you can solve
17873 If you want to recompile `qsort' or other source files which do not
17874 actually use your global register variable, so that they will not use
17875 that register for any other purpose, then it suffices to specify the
17876 compiler option `-ffixed-REG'. You need not actually add a global
17877 register declaration to their source code.
17879 A function which can alter the value of a global register variable
17880 cannot safely be called from a function compiled without this variable,
17881 because it could clobber the value the caller expects to find there on
17882 return. Therefore, the function which is the entry point into the part
17883 of the program that uses the global register variable must explicitly
17884 save and restore the value which belongs to its caller.
17886 On most machines, `longjmp' will restore to each global register
17887 variable the value it had at the time of the `setjmp'. On some
17888 machines, however, `longjmp' will not change the value of global
17889 register variables. To be portable, the function that called `setjmp'
17890 should make other arrangements to save the values of the global register
17891 variables, and to restore them in a `longjmp'. This way, the same
17892 thing will happen regardless of what `longjmp' does.
17894 All global register variable declarations must precede all function
17895 definitions. If such a declaration could appear after function
17896 definitions, the declaration would be too late to prevent the register
17897 from being used for other purposes in the preceding functions.
17899 Global register variables may not have initial values, because an
17900 executable file has no means to supply initial contents for a register.
17902 On the SPARC, there are reports that g3 ... g7 are suitable registers,
17903 but certain library functions, such as `getwd', as well as the
17904 subroutines for division and remainder, modify g3 and g4. g1 and g2
17905 are local temporaries.
17907 On the 68000, a2 ... a5 should be suitable, as should d2 ... d7. Of
17908 course, it will not do to use more than a few of those.
17911 File: gcc.info, Node: Local Reg Vars, Prev: Global Reg Vars, Up: Explicit Reg Vars
17913 5.37.2 Specifying Registers for Local Variables
17914 -----------------------------------------------
17916 You can define a local register variable with a specified register like
17919 register int *foo asm ("a5");
17921 Here `a5' is the name of the register which should be used. Note that
17922 this is the same syntax used for defining global register variables,
17923 but for a local variable it would appear within a function.
17925 Naturally the register name is cpu-dependent, but this is not a
17926 problem, since specific registers are most often useful with explicit
17927 assembler instructions (*note Extended Asm::). Both of these things
17928 generally require that you conditionalize your program according to cpu
17931 In addition, operating systems on one type of cpu may differ in how
17932 they name the registers; then you would need additional conditionals.
17933 For example, some 68000 operating systems call this register `%a5'.
17935 Defining such a register variable does not reserve the register; it
17936 remains available for other uses in places where flow control determines
17937 the variable's value is not live.
17939 This option does not guarantee that GCC will generate code that has
17940 this variable in the register you specify at all times. You may not
17941 code an explicit reference to this register in the _assembler
17942 instruction template_ part of an `asm' statement and assume it will
17943 always refer to this variable. However, using the variable as an `asm'
17944 _operand_ guarantees that the specified register is used for the
17947 Stores into local register variables may be deleted when they appear
17948 to be dead according to dataflow analysis. References to local
17949 register variables may be deleted or moved or simplified.
17951 As for global register variables, it's recommended that you choose a
17952 register which is normally saved and restored by function calls on your
17953 machine, so that library routines will not clobber it. A common
17954 pitfall is to initialize multiple call-clobbered registers with
17955 arbitrary expressions, where a function call or library call for an
17956 arithmetic operator will overwrite a register value from a previous
17957 assignment, for example `r0' below:
17958 register int *p1 asm ("r0") = ...;
17959 register int *p2 asm ("r1") = ...;
17960 In those cases, a solution is to use a temporary variable for each
17961 arbitrary expression. *Note Example of asm with clobbered asm reg::.
17964 File: gcc.info, Node: Alternate Keywords, Next: Incomplete Enums, Prev: Explicit Reg Vars, Up: C Extensions
17966 5.38 Alternate Keywords
17967 =======================
17969 `-ansi' and the various `-std' options disable certain keywords. This
17970 causes trouble when you want to use GNU C extensions, or a
17971 general-purpose header file that should be usable by all programs,
17972 including ISO C programs. The keywords `asm', `typeof' and `inline'
17973 are not available in programs compiled with `-ansi' or `-std' (although
17974 `inline' can be used in a program compiled with `-std=c99'). The ISO
17975 C99 keyword `restrict' is only available when `-std=gnu99' (which will
17976 eventually be the default) or `-std=c99' (or the equivalent
17977 `-std=iso9899:1999') is used.
17979 The way to solve these problems is to put `__' at the beginning and
17980 end of each problematical keyword. For example, use `__asm__' instead
17981 of `asm', and `__inline__' instead of `inline'.
17983 Other C compilers won't accept these alternative keywords; if you want
17984 to compile with another compiler, you can define the alternate keywords
17985 as macros to replace them with the customary keywords. It looks like
17989 #define __asm__ asm
17992 `-pedantic' and other options cause warnings for many GNU C extensions.
17993 You can prevent such warnings within one expression by writing
17994 `__extension__' before the expression. `__extension__' has no effect
17998 File: gcc.info, Node: Incomplete Enums, Next: Function Names, Prev: Alternate Keywords, Up: C Extensions
18000 5.39 Incomplete `enum' Types
18001 ============================
18003 You can define an `enum' tag without specifying its possible values.
18004 This results in an incomplete type, much like what you get if you write
18005 `struct foo' without describing the elements. A later declaration
18006 which does specify the possible values completes the type.
18008 You can't allocate variables or storage using the type while it is
18009 incomplete. However, you can work with pointers to that type.
18011 This extension may not be very useful, but it makes the handling of
18012 `enum' more consistent with the way `struct' and `union' are handled.
18014 This extension is not supported by GNU C++.
18017 File: gcc.info, Node: Function Names, Next: Return Address, Prev: Incomplete Enums, Up: C Extensions
18019 5.40 Function Names as Strings
18020 ==============================
18022 GCC provides three magic variables which hold the name of the current
18023 function, as a string. The first of these is `__func__', which is part
18024 of the C99 standard:
18026 The identifier `__func__' is implicitly declared by the translator
18027 as if, immediately following the opening brace of each function
18028 definition, the declaration
18029 static const char __func__[] = "function-name";
18031 appeared, where function-name is the name of the lexically-enclosing
18032 function. This name is the unadorned name of the function.
18034 `__FUNCTION__' is another name for `__func__'. Older versions of GCC
18035 recognize only this name. However, it is not standardized. For
18036 maximum portability, we recommend you use `__func__', but provide a
18037 fallback definition with the preprocessor:
18039 #if __STDC_VERSION__ < 199901L
18041 # define __func__ __FUNCTION__
18043 # define __func__ "<unknown>"
18047 In C, `__PRETTY_FUNCTION__' is yet another name for `__func__'.
18048 However, in C++, `__PRETTY_FUNCTION__' contains the type signature of
18049 the function as well as its bare name. For example, this program:
18052 extern int printf (char *, ...);
18059 printf ("__FUNCTION__ = %s\n", __FUNCTION__);
18060 printf ("__PRETTY_FUNCTION__ = %s\n", __PRETTY_FUNCTION__);
18075 __PRETTY_FUNCTION__ = void a::sub(int)
18077 These identifiers are not preprocessor macros. In GCC 3.3 and
18078 earlier, in C only, `__FUNCTION__' and `__PRETTY_FUNCTION__' were
18079 treated as string literals; they could be used to initialize `char'
18080 arrays, and they could be concatenated with other string literals. GCC
18081 3.4 and later treat them as variables, like `__func__'. In C++,
18082 `__FUNCTION__' and `__PRETTY_FUNCTION__' have always been variables.
18085 File: gcc.info, Node: Return Address, Next: Vector Extensions, Prev: Function Names, Up: C Extensions
18087 5.41 Getting the Return or Frame Address of a Function
18088 ======================================================
18090 These functions may be used to get information about the callers of a
18093 -- Built-in Function: void * __builtin_return_address (unsigned int
18095 This function returns the return address of the current function,
18096 or of one of its callers. The LEVEL argument is number of frames
18097 to scan up the call stack. A value of `0' yields the return
18098 address of the current function, a value of `1' yields the return
18099 address of the caller of the current function, and so forth. When
18100 inlining the expected behavior is that the function will return
18101 the address of the function that will be returned to. To work
18102 around this behavior use the `noinline' function attribute.
18104 The LEVEL argument must be a constant integer.
18106 On some machines it may be impossible to determine the return
18107 address of any function other than the current one; in such cases,
18108 or when the top of the stack has been reached, this function will
18109 return `0' or a random value. In addition,
18110 `__builtin_frame_address' may be used to determine if the top of
18111 the stack has been reached.
18113 This function should only be used with a nonzero argument for
18114 debugging purposes.
18116 -- Built-in Function: void * __builtin_frame_address (unsigned int
18118 This function is similar to `__builtin_return_address', but it
18119 returns the address of the function frame rather than the return
18120 address of the function. Calling `__builtin_frame_address' with a
18121 value of `0' yields the frame address of the current function, a
18122 value of `1' yields the frame address of the caller of the current
18123 function, and so forth.
18125 The frame is the area on the stack which holds local variables and
18126 saved registers. The frame address is normally the address of the
18127 first word pushed on to the stack by the function. However, the
18128 exact definition depends upon the processor and the calling
18129 convention. If the processor has a dedicated frame pointer
18130 register, and the function has a frame, then
18131 `__builtin_frame_address' will return the value of the frame
18134 On some machines it may be impossible to determine the frame
18135 address of any function other than the current one; in such cases,
18136 or when the top of the stack has been reached, this function will
18137 return `0' if the first frame pointer is properly initialized by
18140 This function should only be used with a nonzero argument for
18141 debugging purposes.
18144 File: gcc.info, Node: Vector Extensions, Next: Offsetof, Prev: Return Address, Up: C Extensions
18146 5.42 Using vector instructions through built-in functions
18147 =========================================================
18149 On some targets, the instruction set contains SIMD vector instructions
18150 that operate on multiple values contained in one large register at the
18151 same time. For example, on the i386 the MMX, 3Dnow! and SSE extensions
18152 can be used this way.
18154 The first step in using these extensions is to provide the necessary
18155 data types. This should be done using an appropriate `typedef':
18157 typedef int v4si __attribute__ ((vector_size (16)));
18159 The `int' type specifies the base type, while the attribute specifies
18160 the vector size for the variable, measured in bytes. For example, the
18161 declaration above causes the compiler to set the mode for the `v4si'
18162 type to be 16 bytes wide and divided into `int' sized units. For a
18163 32-bit `int' this means a vector of 4 units of 4 bytes, and the
18164 corresponding mode of `foo' will be V4SI.
18166 The `vector_size' attribute is only applicable to integral and float
18167 scalars, although arrays, pointers, and function return values are
18168 allowed in conjunction with this construct.
18170 All the basic integer types can be used as base types, both as signed
18171 and as unsigned: `char', `short', `int', `long', `long long'. In
18172 addition, `float' and `double' can be used to build floating-point
18175 Specifying a combination that is not valid for the current architecture
18176 will cause GCC to synthesize the instructions using a narrower mode.
18177 For example, if you specify a variable of type `V4SI' and your
18178 architecture does not allow for this specific SIMD type, GCC will
18179 produce code that uses 4 `SIs'.
18181 The types defined in this manner can be used with a subset of normal C
18182 operations. Currently, GCC will allow using the following operators on
18183 these types: `+, -, *, /, unary minus, ^, |, &, ~'.
18185 The operations behave like C++ `valarrays'. Addition is defined as
18186 the addition of the corresponding elements of the operands. For
18187 example, in the code below, each of the 4 elements in A will be added
18188 to the corresponding 4 elements in B and the resulting vector will be
18191 typedef int v4si __attribute__ ((vector_size (16)));
18197 Subtraction, multiplication, division, and the logical operations
18198 operate in a similar manner. Likewise, the result of using the unary
18199 minus or complement operators on a vector type is a vector whose
18200 elements are the negative or complemented values of the corresponding
18201 elements in the operand.
18203 You can declare variables and use them in function calls and returns,
18204 as well as in assignments and some casts. You can specify a vector
18205 type as a return type for a function. Vector types can also be used as
18206 function arguments. It is possible to cast from one vector type to
18207 another, provided they are of the same size (in fact, you can also cast
18208 vectors to and from other datatypes of the same size).
18210 You cannot operate between vectors of different lengths or different
18211 signedness without a cast.
18213 A port that supports hardware vector operations, usually provides a set
18214 of built-in functions that can be used to operate on vectors. For
18215 example, a function to add two vectors and multiply the result by a
18216 third could look like this:
18218 v4si f (v4si a, v4si b, v4si c)
18220 v4si tmp = __builtin_addv4si (a, b);
18221 return __builtin_mulv4si (tmp, c);
18225 File: gcc.info, Node: Offsetof, Next: Other Builtins, Prev: Vector Extensions, Up: C Extensions
18230 GCC implements for both C and C++ a syntactic extension to implement
18231 the `offsetof' macro.
18234 "__builtin_offsetof" "(" `typename' "," offsetof_member_designator ")"
18236 offsetof_member_designator:
18238 | offsetof_member_designator "." `identifier'
18239 | offsetof_member_designator "[" `expr' "]"
18241 This extension is sufficient such that
18243 #define offsetof(TYPE, MEMBER) __builtin_offsetof (TYPE, MEMBER)
18245 is a suitable definition of the `offsetof' macro. In C++, TYPE may be
18246 dependent. In either case, MEMBER may consist of a single identifier,
18247 or a sequence of member accesses and array references.
18250 File: gcc.info, Node: Other Builtins, Next: Target Builtins, Prev: Offsetof, Up: C Extensions
18252 5.44 Other built-in functions provided by GCC
18253 =============================================
18255 GCC provides a large number of built-in functions other than the ones
18256 mentioned above. Some of these are for internal use in the processing
18257 of exceptions or variable-length argument lists and will not be
18258 documented here because they may change from time to time; we do not
18259 recommend general use of these functions.
18261 The remaining functions are provided for optimization purposes.
18263 GCC includes built-in versions of many of the functions in the standard
18264 C library. The versions prefixed with `__builtin_' will always be
18265 treated as having the same meaning as the C library function even if you
18266 specify the `-fno-builtin' option. (*note C Dialect Options::) Many of
18267 these functions are only optimized in certain cases; if they are not
18268 optimized in a particular case, a call to the library function will be
18271 Outside strict ISO C mode (`-ansi', `-std=c89' or `-std=c99'), the
18272 functions `_exit', `alloca', `bcmp', `bzero', `dcgettext', `dgettext',
18273 `dremf', `dreml', `drem', `exp10f', `exp10l', `exp10', `ffsll', `ffsl',
18274 `ffs', `fprintf_unlocked', `fputs_unlocked', `gammaf', `gammal',
18275 `gamma', `gettext', `index', `isascii', `j0f', `j0l', `j0', `j1f',
18276 `j1l', `j1', `jnf', `jnl', `jn', `mempcpy', `pow10f', `pow10l', `pow10',
18277 `printf_unlocked', `rindex', `scalbf', `scalbl', `scalb', `signbit',
18278 `signbitf', `signbitl', `significandf', `significandl', `significand',
18279 `sincosf', `sincosl', `sincos', `stpcpy', `strdup', `strfmon',
18280 `toascii', `y0f', `y0l', `y0', `y1f', `y1l', `y1', `ynf', `ynl' and `yn'
18281 may be handled as built-in functions. All these functions have
18282 corresponding versions prefixed with `__builtin_', which may be used
18283 even in strict C89 mode.
18285 The ISO C99 functions `_Exit', `acoshf', `acoshl', `acosh', `asinhf',
18286 `asinhl', `asinh', `atanhf', `atanhl', `atanh', `cabsf', `cabsl',
18287 `cabs', `cacosf', `cacoshf', `cacoshl', `cacosh', `cacosl', `cacos',
18288 `cargf', `cargl', `carg', `casinf', `casinhf', `casinhl', `casinh',
18289 `casinl', `casin', `catanf', `catanhf', `catanhl', `catanh', `catanl',
18290 `catan', `cbrtf', `cbrtl', `cbrt', `ccosf', `ccoshf', `ccoshl',
18291 `ccosh', `ccosl', `ccos', `cexpf', `cexpl', `cexp', `cimagf', `cimagl',
18292 `cimag', `conjf', `conjl', `conj', `copysignf', `copysignl',
18293 `copysign', `cpowf', `cpowl', `cpow', `cprojf', `cprojl', `cproj',
18294 `crealf', `creall', `creal', `csinf', `csinhf', `csinhl', `csinh',
18295 `csinl', `csin', `csqrtf', `csqrtl', `csqrt', `ctanf', `ctanhf',
18296 `ctanhl', `ctanh', `ctanl', `ctan', `erfcf', `erfcl', `erfc', `erff',
18297 `erfl', `erf', `exp2f', `exp2l', `exp2', `expm1f', `expm1l', `expm1',
18298 `fdimf', `fdiml', `fdim', `fmaf', `fmal', `fmaxf', `fmaxl', `fmax',
18299 `fma', `fminf', `fminl', `fmin', `hypotf', `hypotl', `hypot', `ilogbf',
18300 `ilogbl', `ilogb', `imaxabs', `isblank', `iswblank', `lgammaf',
18301 `lgammal', `lgamma', `llabs', `llrintf', `llrintl', `llrint',
18302 `llroundf', `llroundl', `llround', `log1pf', `log1pl', `log1p',
18303 `log2f', `log2l', `log2', `logbf', `logbl', `logb', `lrintf', `lrintl',
18304 `lrint', `lroundf', `lroundl', `lround', `nearbyintf', `nearbyintl',
18305 `nearbyint', `nextafterf', `nextafterl', `nextafter', `nexttowardf',
18306 `nexttowardl', `nexttoward', `remainderf', `remainderl', `remainder',
18307 `remquof', `remquol', `remquo', `rintf', `rintl', `rint', `roundf',
18308 `roundl', `round', `scalblnf', `scalblnl', `scalbln', `scalbnf',
18309 `scalbnl', `scalbn', `snprintf', `tgammaf', `tgammal', `tgamma',
18310 `truncf', `truncl', `trunc', `vfscanf', `vscanf', `vsnprintf' and
18311 `vsscanf' are handled as built-in functions except in strict ISO C90
18312 mode (`-ansi' or `-std=c89').
18314 There are also built-in versions of the ISO C99 functions `acosf',
18315 `acosl', `asinf', `asinl', `atan2f', `atan2l', `atanf', `atanl',
18316 `ceilf', `ceill', `cosf', `coshf', `coshl', `cosl', `expf', `expl',
18317 `fabsf', `fabsl', `floorf', `floorl', `fmodf', `fmodl', `frexpf',
18318 `frexpl', `ldexpf', `ldexpl', `log10f', `log10l', `logf', `logl',
18319 `modfl', `modf', `powf', `powl', `sinf', `sinhf', `sinhl', `sinl',
18320 `sqrtf', `sqrtl', `tanf', `tanhf', `tanhl' and `tanl' that are
18321 recognized in any mode since ISO C90 reserves these names for the
18322 purpose to which ISO C99 puts them. All these functions have
18323 corresponding versions prefixed with `__builtin_'.
18325 The ISO C94 functions `iswalnum', `iswalpha', `iswcntrl', `iswdigit',
18326 `iswgraph', `iswlower', `iswprint', `iswpunct', `iswspace', `iswupper',
18327 `iswxdigit', `towlower' and `towupper' are handled as built-in functions
18328 except in strict ISO C90 mode (`-ansi' or `-std=c89').
18330 The ISO C90 functions `abort', `abs', `acos', `asin', `atan2', `atan',
18331 `calloc', `ceil', `cosh', `cos', `exit', `exp', `fabs', `floor', `fmod',
18332 `fprintf', `fputs', `frexp', `fscanf', `isalnum', `isalpha', `iscntrl',
18333 `isdigit', `isgraph', `islower', `isprint', `ispunct', `isspace',
18334 `isupper', `isxdigit', `tolower', `toupper', `labs', `ldexp', `log10',
18335 `log', `malloc', `memcmp', `memcpy', `memset', `modf', `pow', `printf',
18336 `putchar', `puts', `scanf', `sinh', `sin', `snprintf', `sprintf',
18337 `sqrt', `sscanf', `strcat', `strchr', `strcmp', `strcpy', `strcspn',
18338 `strlen', `strncat', `strncmp', `strncpy', `strpbrk', `strrchr',
18339 `strspn', `strstr', `tanh', `tan', `vfprintf', `vprintf' and `vsprintf'
18340 are all recognized as built-in functions unless `-fno-builtin' is
18341 specified (or `-fno-builtin-FUNCTION' is specified for an individual
18342 function). All of these functions have corresponding versions prefixed
18345 GCC provides built-in versions of the ISO C99 floating point comparison
18346 macros that avoid raising exceptions for unordered operands. They have
18347 the same names as the standard macros ( `isgreater', `isgreaterequal',
18348 `isless', `islessequal', `islessgreater', and `isunordered') , with
18349 `__builtin_' prefixed. We intend for a library implementor to be able
18350 to simply `#define' each standard macro to its built-in equivalent.
18352 -- Built-in Function: int __builtin_types_compatible_p (TYPE1, TYPE2)
18353 You can use the built-in function `__builtin_types_compatible_p' to
18354 determine whether two types are the same.
18356 This built-in function returns 1 if the unqualified versions of the
18357 types TYPE1 and TYPE2 (which are types, not expressions) are
18358 compatible, 0 otherwise. The result of this built-in function can
18359 be used in integer constant expressions.
18361 This built-in function ignores top level qualifiers (e.g., `const',
18362 `volatile'). For example, `int' is equivalent to `const int'.
18364 The type `int[]' and `int[5]' are compatible. On the other hand,
18365 `int' and `char *' are not compatible, even if the size of their
18366 types, on the particular architecture are the same. Also, the
18367 amount of pointer indirection is taken into account when
18368 determining similarity. Consequently, `short *' is not similar to
18369 `short **'. Furthermore, two types that are typedefed are
18370 considered compatible if their underlying types are compatible.
18372 An `enum' type is not considered to be compatible with another
18373 `enum' type even if both are compatible with the same integer
18374 type; this is what the C standard specifies. For example, `enum
18375 {foo, bar}' is not similar to `enum {hot, dog}'.
18377 You would typically use this function in code whose execution
18378 varies depending on the arguments' types. For example:
18383 if (__builtin_types_compatible_p (typeof (x), long double)) \
18384 tmp = foo_long_double (tmp); \
18385 else if (__builtin_types_compatible_p (typeof (x), double)) \
18386 tmp = foo_double (tmp); \
18387 else if (__builtin_types_compatible_p (typeof (x), float)) \
18388 tmp = foo_float (tmp); \
18394 _Note:_ This construct is only available for C.
18397 -- Built-in Function: TYPE __builtin_choose_expr (CONST_EXP, EXP1,
18399 You can use the built-in function `__builtin_choose_expr' to
18400 evaluate code depending on the value of a constant expression.
18401 This built-in function returns EXP1 if CONST_EXP, which is a
18402 constant expression that must be able to be determined at compile
18403 time, is nonzero. Otherwise it returns 0.
18405 This built-in function is analogous to the `? :' operator in C,
18406 except that the expression returned has its type unaltered by
18407 promotion rules. Also, the built-in function does not evaluate
18408 the expression that was not chosen. For example, if CONST_EXP
18409 evaluates to true, EXP2 is not evaluated even if it has
18412 This built-in function can return an lvalue if the chosen argument
18415 If EXP1 is returned, the return type is the same as EXP1's type.
18416 Similarly, if EXP2 is returned, its return type is the same as
18422 __builtin_choose_expr ( \
18423 __builtin_types_compatible_p (typeof (x), double), \
18425 __builtin_choose_expr ( \
18426 __builtin_types_compatible_p (typeof (x), float), \
18428 /* The void expression results in a compile-time error \
18429 when assigning the result to something. */ \
18432 _Note:_ This construct is only available for C. Furthermore, the
18433 unused expression (EXP1 or EXP2 depending on the value of
18434 CONST_EXP) may still generate syntax errors. This may change in
18438 -- Built-in Function: int __builtin_constant_p (EXP)
18439 You can use the built-in function `__builtin_constant_p' to
18440 determine if a value is known to be constant at compile-time and
18441 hence that GCC can perform constant-folding on expressions
18442 involving that value. The argument of the function is the value
18443 to test. The function returns the integer 1 if the argument is
18444 known to be a compile-time constant and 0 if it is not known to be
18445 a compile-time constant. A return of 0 does not indicate that the
18446 value is _not_ a constant, but merely that GCC cannot prove it is
18447 a constant with the specified value of the `-O' option.
18449 You would typically use this function in an embedded application
18450 where memory was a critical resource. If you have some complex
18451 calculation, you may want it to be folded if it involves
18452 constants, but need to call a function if it does not. For
18455 #define Scale_Value(X) \
18456 (__builtin_constant_p (X) \
18457 ? ((X) * SCALE + OFFSET) : Scale (X))
18459 You may use this built-in function in either a macro or an inline
18460 function. However, if you use it in an inlined function and pass
18461 an argument of the function as the argument to the built-in, GCC
18462 will never return 1 when you call the inline function with a
18463 string constant or compound literal (*note Compound Literals::)
18464 and will not return 1 when you pass a constant numeric value to
18465 the inline function unless you specify the `-O' option.
18467 You may also use `__builtin_constant_p' in initializers for static
18468 data. For instance, you can write
18470 static const int table[] = {
18471 __builtin_constant_p (EXPRESSION) ? (EXPRESSION) : -1,
18475 This is an acceptable initializer even if EXPRESSION is not a
18476 constant expression. GCC must be more conservative about
18477 evaluating the built-in in this case, because it has no
18478 opportunity to perform optimization.
18480 Previous versions of GCC did not accept this built-in in data
18481 initializers. The earliest version where it is completely safe is
18484 -- Built-in Function: long __builtin_expect (long EXP, long C)
18485 You may use `__builtin_expect' to provide the compiler with branch
18486 prediction information. In general, you should prefer to use
18487 actual profile feedback for this (`-fprofile-arcs'), as
18488 programmers are notoriously bad at predicting how their programs
18489 actually perform. However, there are applications in which this
18490 data is hard to collect.
18492 The return value is the value of EXP, which should be an integral
18493 expression. The value of C must be a compile-time constant. The
18494 semantics of the built-in are that it is expected that EXP == C.
18497 if (__builtin_expect (x, 0))
18500 would indicate that we do not expect to call `foo', since we
18501 expect `x' to be zero. Since you are limited to integral
18502 expressions for EXP, you should use constructions such as
18504 if (__builtin_expect (ptr != NULL, 1))
18507 when testing pointer or floating-point values.
18509 -- Built-in Function: void __builtin_prefetch (const void *ADDR, ...)
18510 This function is used to minimize cache-miss latency by moving
18511 data into a cache before it is accessed. You can insert calls to
18512 `__builtin_prefetch' into code for which you know addresses of
18513 data in memory that is likely to be accessed soon. If the target
18514 supports them, data prefetch instructions will be generated. If
18515 the prefetch is done early enough before the access then the data
18516 will be in the cache by the time it is accessed.
18518 The value of ADDR is the address of the memory to prefetch. There
18519 are two optional arguments, RW and LOCALITY. The value of RW is a
18520 compile-time constant one or zero; one means that the prefetch is
18521 preparing for a write to the memory address and zero, the default,
18522 means that the prefetch is preparing for a read. The value
18523 LOCALITY must be a compile-time constant integer between zero and
18524 three. A value of zero means that the data has no temporal
18525 locality, so it need not be left in the cache after the access. A
18526 value of three means that the data has a high degree of temporal
18527 locality and should be left in all levels of cache possible.
18528 Values of one and two mean, respectively, a low or moderate degree
18529 of temporal locality. The default is three.
18531 for (i = 0; i < n; i++)
18533 a[i] = a[i] + b[i];
18534 __builtin_prefetch (&a[i+j], 1, 1);
18535 __builtin_prefetch (&b[i+j], 0, 1);
18539 Data prefetch does not generate faults if ADDR is invalid, but the
18540 address expression itself must be valid. For example, a prefetch
18541 of `p->next' will not fault if `p->next' is not a valid address,
18542 but evaluation will fault if `p' is not a valid address.
18544 If the target does not support data prefetch, the address
18545 expression is evaluated if it includes side effects but no other
18546 code is generated and GCC does not issue a warning.
18548 -- Built-in Function: double __builtin_huge_val (void)
18549 Returns a positive infinity, if supported by the floating-point
18550 format, else `DBL_MAX'. This function is suitable for
18551 implementing the ISO C macro `HUGE_VAL'.
18553 -- Built-in Function: float __builtin_huge_valf (void)
18554 Similar to `__builtin_huge_val', except the return type is `float'.
18556 -- Built-in Function: long double __builtin_huge_vall (void)
18557 Similar to `__builtin_huge_val', except the return type is `long
18560 -- Built-in Function: double __builtin_inf (void)
18561 Similar to `__builtin_huge_val', except a warning is generated if
18562 the target floating-point format does not support infinities.
18564 -- Built-in Function: float __builtin_inff (void)
18565 Similar to `__builtin_inf', except the return type is `float'.
18566 This function is suitable for implementing the ISO C99 macro
18569 -- Built-in Function: long double __builtin_infl (void)
18570 Similar to `__builtin_inf', except the return type is `long
18573 -- Built-in Function: double __builtin_nan (const char *str)
18574 This is an implementation of the ISO C99 function `nan'.
18576 Since ISO C99 defines this function in terms of `strtod', which we
18577 do not implement, a description of the parsing is in order. The
18578 string is parsed as by `strtol'; that is, the base is recognized by
18579 leading `0' or `0x' prefixes. The number parsed is placed in the
18580 significand such that the least significant bit of the number is
18581 at the least significant bit of the significand. The number is
18582 truncated to fit the significand field provided. The significand
18583 is forced to be a quiet NaN.
18585 This function, if given a string literal, is evaluated early enough
18586 that it is considered a compile-time constant.
18588 -- Built-in Function: float __builtin_nanf (const char *str)
18589 Similar to `__builtin_nan', except the return type is `float'.
18591 -- Built-in Function: long double __builtin_nanl (const char *str)
18592 Similar to `__builtin_nan', except the return type is `long
18595 -- Built-in Function: double __builtin_nans (const char *str)
18596 Similar to `__builtin_nan', except the significand is forced to be
18597 a signaling NaN. The `nans' function is proposed by WG14 N965.
18599 -- Built-in Function: float __builtin_nansf (const char *str)
18600 Similar to `__builtin_nans', except the return type is `float'.
18602 -- Built-in Function: long double __builtin_nansl (const char *str)
18603 Similar to `__builtin_nans', except the return type is `long
18606 -- Built-in Function: int __builtin_ffs (unsigned int x)
18607 Returns one plus the index of the least significant 1-bit of X, or
18608 if X is zero, returns zero.
18610 -- Built-in Function: int __builtin_clz (unsigned int x)
18611 Returns the number of leading 0-bits in X, starting at the most
18612 significant bit position. If X is 0, the result is undefined.
18614 -- Built-in Function: int __builtin_ctz (unsigned int x)
18615 Returns the number of trailing 0-bits in X, starting at the least
18616 significant bit position. If X is 0, the result is undefined.
18618 -- Built-in Function: int __builtin_popcount (unsigned int x)
18619 Returns the number of 1-bits in X.
18621 -- Built-in Function: int __builtin_parity (unsigned int x)
18622 Returns the parity of X, i.e. the number of 1-bits in X modulo 2.
18624 -- Built-in Function: int __builtin_ffsl (unsigned long)
18625 Similar to `__builtin_ffs', except the argument type is `unsigned
18628 -- Built-in Function: int __builtin_clzl (unsigned long)
18629 Similar to `__builtin_clz', except the argument type is `unsigned
18632 -- Built-in Function: int __builtin_ctzl (unsigned long)
18633 Similar to `__builtin_ctz', except the argument type is `unsigned
18636 -- Built-in Function: int __builtin_popcountl (unsigned long)
18637 Similar to `__builtin_popcount', except the argument type is
18640 -- Built-in Function: int __builtin_parityl (unsigned long)
18641 Similar to `__builtin_parity', except the argument type is
18644 -- Built-in Function: int __builtin_ffsll (unsigned long long)
18645 Similar to `__builtin_ffs', except the argument type is `unsigned
18648 -- Built-in Function: int __builtin_clzll (unsigned long long)
18649 Similar to `__builtin_clz', except the argument type is `unsigned
18652 -- Built-in Function: int __builtin_ctzll (unsigned long long)
18653 Similar to `__builtin_ctz', except the argument type is `unsigned
18656 -- Built-in Function: int __builtin_popcountll (unsigned long long)
18657 Similar to `__builtin_popcount', except the argument type is
18658 `unsigned long long'.
18660 -- Built-in Function: int __builtin_parityll (unsigned long long)
18661 Similar to `__builtin_parity', except the argument type is
18662 `unsigned long long'.
18664 -- Built-in Function: double __builtin_powi (double, int)
18665 Returns the first argument raised to the power of the second.
18666 Unlike the `pow' function no guarantees about precision and
18669 -- Built-in Function: float __builtin_powif (float, int)
18670 Similar to `__builtin_powi', except the argument and return types
18673 -- Built-in Function: long double __builtin_powil (long double, int)
18674 Similar to `__builtin_powi', except the argument and return types
18678 File: gcc.info, Node: Target Builtins, Next: Target Format Checks, Prev: Other Builtins, Up: C Extensions
18680 5.45 Built-in Functions Specific to Particular Target Machines
18681 ==============================================================
18683 On some target machines, GCC supports many built-in functions specific
18684 to those machines. Generally these generate calls to specific machine
18685 instructions, but allow the compiler to schedule those calls.
18689 * Alpha Built-in Functions::
18690 * ARM Built-in Functions::
18691 * FR-V Built-in Functions::
18692 * X86 Built-in Functions::
18693 * MIPS Paired-Single Support::
18694 * PowerPC AltiVec Built-in Functions::
18695 * SPARC VIS Built-in Functions::
18698 File: gcc.info, Node: Alpha Built-in Functions, Next: ARM Built-in Functions, Up: Target Builtins
18700 5.45.1 Alpha Built-in Functions
18701 -------------------------------
18703 These built-in functions are available for the Alpha family of
18704 processors, depending on the command-line switches used.
18706 The following built-in functions are always available. They all
18707 generate the machine instruction that is part of the name.
18709 long __builtin_alpha_implver (void)
18710 long __builtin_alpha_rpcc (void)
18711 long __builtin_alpha_amask (long)
18712 long __builtin_alpha_cmpbge (long, long)
18713 long __builtin_alpha_extbl (long, long)
18714 long __builtin_alpha_extwl (long, long)
18715 long __builtin_alpha_extll (long, long)
18716 long __builtin_alpha_extql (long, long)
18717 long __builtin_alpha_extwh (long, long)
18718 long __builtin_alpha_extlh (long, long)
18719 long __builtin_alpha_extqh (long, long)
18720 long __builtin_alpha_insbl (long, long)
18721 long __builtin_alpha_inswl (long, long)
18722 long __builtin_alpha_insll (long, long)
18723 long __builtin_alpha_insql (long, long)
18724 long __builtin_alpha_inswh (long, long)
18725 long __builtin_alpha_inslh (long, long)
18726 long __builtin_alpha_insqh (long, long)
18727 long __builtin_alpha_mskbl (long, long)
18728 long __builtin_alpha_mskwl (long, long)
18729 long __builtin_alpha_mskll (long, long)
18730 long __builtin_alpha_mskql (long, long)
18731 long __builtin_alpha_mskwh (long, long)
18732 long __builtin_alpha_msklh (long, long)
18733 long __builtin_alpha_mskqh (long, long)
18734 long __builtin_alpha_umulh (long, long)
18735 long __builtin_alpha_zap (long, long)
18736 long __builtin_alpha_zapnot (long, long)
18738 The following built-in functions are always with `-mmax' or
18739 `-mcpu=CPU' where CPU is `pca56' or later. They all generate the
18740 machine instruction that is part of the name.
18742 long __builtin_alpha_pklb (long)
18743 long __builtin_alpha_pkwb (long)
18744 long __builtin_alpha_unpkbl (long)
18745 long __builtin_alpha_unpkbw (long)
18746 long __builtin_alpha_minub8 (long, long)
18747 long __builtin_alpha_minsb8 (long, long)
18748 long __builtin_alpha_minuw4 (long, long)
18749 long __builtin_alpha_minsw4 (long, long)
18750 long __builtin_alpha_maxub8 (long, long)
18751 long __builtin_alpha_maxsb8 (long, long)
18752 long __builtin_alpha_maxuw4 (long, long)
18753 long __builtin_alpha_maxsw4 (long, long)
18754 long __builtin_alpha_perr (long, long)
18756 The following built-in functions are always with `-mcix' or
18757 `-mcpu=CPU' where CPU is `ev67' or later. They all generate the
18758 machine instruction that is part of the name.
18760 long __builtin_alpha_cttz (long)
18761 long __builtin_alpha_ctlz (long)
18762 long __builtin_alpha_ctpop (long)
18764 The following builtins are available on systems that use the OSF/1
18765 PALcode. Normally they invoke the `rduniq' and `wruniq' PAL calls, but
18766 when invoked with `-mtls-kernel', they invoke `rdval' and `wrval'.
18768 void *__builtin_thread_pointer (void)
18769 void __builtin_set_thread_pointer (void *)
18772 File: gcc.info, Node: ARM Built-in Functions, Next: FR-V Built-in Functions, Prev: Alpha Built-in Functions, Up: Target Builtins
18774 5.45.2 ARM Built-in Functions
18775 -----------------------------
18777 These built-in functions are available for the ARM family of
18778 processors, when the `-mcpu=iwmmxt' switch is used:
18780 typedef int v2si __attribute__ ((vector_size (8)));
18781 typedef short v4hi __attribute__ ((vector_size (8)));
18782 typedef char v8qi __attribute__ ((vector_size (8)));
18784 int __builtin_arm_getwcx (int)
18785 void __builtin_arm_setwcx (int, int)
18786 int __builtin_arm_textrmsb (v8qi, int)
18787 int __builtin_arm_textrmsh (v4hi, int)
18788 int __builtin_arm_textrmsw (v2si, int)
18789 int __builtin_arm_textrmub (v8qi, int)
18790 int __builtin_arm_textrmuh (v4hi, int)
18791 int __builtin_arm_textrmuw (v2si, int)
18792 v8qi __builtin_arm_tinsrb (v8qi, int)
18793 v4hi __builtin_arm_tinsrh (v4hi, int)
18794 v2si __builtin_arm_tinsrw (v2si, int)
18795 long long __builtin_arm_tmia (long long, int, int)
18796 long long __builtin_arm_tmiabb (long long, int, int)
18797 long long __builtin_arm_tmiabt (long long, int, int)
18798 long long __builtin_arm_tmiaph (long long, int, int)
18799 long long __builtin_arm_tmiatb (long long, int, int)
18800 long long __builtin_arm_tmiatt (long long, int, int)
18801 int __builtin_arm_tmovmskb (v8qi)
18802 int __builtin_arm_tmovmskh (v4hi)
18803 int __builtin_arm_tmovmskw (v2si)
18804 long long __builtin_arm_waccb (v8qi)
18805 long long __builtin_arm_wacch (v4hi)
18806 long long __builtin_arm_waccw (v2si)
18807 v8qi __builtin_arm_waddb (v8qi, v8qi)
18808 v8qi __builtin_arm_waddbss (v8qi, v8qi)
18809 v8qi __builtin_arm_waddbus (v8qi, v8qi)
18810 v4hi __builtin_arm_waddh (v4hi, v4hi)
18811 v4hi __builtin_arm_waddhss (v4hi, v4hi)
18812 v4hi __builtin_arm_waddhus (v4hi, v4hi)
18813 v2si __builtin_arm_waddw (v2si, v2si)
18814 v2si __builtin_arm_waddwss (v2si, v2si)
18815 v2si __builtin_arm_waddwus (v2si, v2si)
18816 v8qi __builtin_arm_walign (v8qi, v8qi, int)
18817 long long __builtin_arm_wand(long long, long long)
18818 long long __builtin_arm_wandn (long long, long long)
18819 v8qi __builtin_arm_wavg2b (v8qi, v8qi)
18820 v8qi __builtin_arm_wavg2br (v8qi, v8qi)
18821 v4hi __builtin_arm_wavg2h (v4hi, v4hi)
18822 v4hi __builtin_arm_wavg2hr (v4hi, v4hi)
18823 v8qi __builtin_arm_wcmpeqb (v8qi, v8qi)
18824 v4hi __builtin_arm_wcmpeqh (v4hi, v4hi)
18825 v2si __builtin_arm_wcmpeqw (v2si, v2si)
18826 v8qi __builtin_arm_wcmpgtsb (v8qi, v8qi)
18827 v4hi __builtin_arm_wcmpgtsh (v4hi, v4hi)
18828 v2si __builtin_arm_wcmpgtsw (v2si, v2si)
18829 v8qi __builtin_arm_wcmpgtub (v8qi, v8qi)
18830 v4hi __builtin_arm_wcmpgtuh (v4hi, v4hi)
18831 v2si __builtin_arm_wcmpgtuw (v2si, v2si)
18832 long long __builtin_arm_wmacs (long long, v4hi, v4hi)
18833 long long __builtin_arm_wmacsz (v4hi, v4hi)
18834 long long __builtin_arm_wmacu (long long, v4hi, v4hi)
18835 long long __builtin_arm_wmacuz (v4hi, v4hi)
18836 v4hi __builtin_arm_wmadds (v4hi, v4hi)
18837 v4hi __builtin_arm_wmaddu (v4hi, v4hi)
18838 v8qi __builtin_arm_wmaxsb (v8qi, v8qi)
18839 v4hi __builtin_arm_wmaxsh (v4hi, v4hi)
18840 v2si __builtin_arm_wmaxsw (v2si, v2si)
18841 v8qi __builtin_arm_wmaxub (v8qi, v8qi)
18842 v4hi __builtin_arm_wmaxuh (v4hi, v4hi)
18843 v2si __builtin_arm_wmaxuw (v2si, v2si)
18844 v8qi __builtin_arm_wminsb (v8qi, v8qi)
18845 v4hi __builtin_arm_wminsh (v4hi, v4hi)
18846 v2si __builtin_arm_wminsw (v2si, v2si)
18847 v8qi __builtin_arm_wminub (v8qi, v8qi)
18848 v4hi __builtin_arm_wminuh (v4hi, v4hi)
18849 v2si __builtin_arm_wminuw (v2si, v2si)
18850 v4hi __builtin_arm_wmulsm (v4hi, v4hi)
18851 v4hi __builtin_arm_wmulul (v4hi, v4hi)
18852 v4hi __builtin_arm_wmulum (v4hi, v4hi)
18853 long long __builtin_arm_wor (long long, long long)
18854 v2si __builtin_arm_wpackdss (long long, long long)
18855 v2si __builtin_arm_wpackdus (long long, long long)
18856 v8qi __builtin_arm_wpackhss (v4hi, v4hi)
18857 v8qi __builtin_arm_wpackhus (v4hi, v4hi)
18858 v4hi __builtin_arm_wpackwss (v2si, v2si)
18859 v4hi __builtin_arm_wpackwus (v2si, v2si)
18860 long long __builtin_arm_wrord (long long, long long)
18861 long long __builtin_arm_wrordi (long long, int)
18862 v4hi __builtin_arm_wrorh (v4hi, long long)
18863 v4hi __builtin_arm_wrorhi (v4hi, int)
18864 v2si __builtin_arm_wrorw (v2si, long long)
18865 v2si __builtin_arm_wrorwi (v2si, int)
18866 v2si __builtin_arm_wsadb (v8qi, v8qi)
18867 v2si __builtin_arm_wsadbz (v8qi, v8qi)
18868 v2si __builtin_arm_wsadh (v4hi, v4hi)
18869 v2si __builtin_arm_wsadhz (v4hi, v4hi)
18870 v4hi __builtin_arm_wshufh (v4hi, int)
18871 long long __builtin_arm_wslld (long long, long long)
18872 long long __builtin_arm_wslldi (long long, int)
18873 v4hi __builtin_arm_wsllh (v4hi, long long)
18874 v4hi __builtin_arm_wsllhi (v4hi, int)
18875 v2si __builtin_arm_wsllw (v2si, long long)
18876 v2si __builtin_arm_wsllwi (v2si, int)
18877 long long __builtin_arm_wsrad (long long, long long)
18878 long long __builtin_arm_wsradi (long long, int)
18879 v4hi __builtin_arm_wsrah (v4hi, long long)
18880 v4hi __builtin_arm_wsrahi (v4hi, int)
18881 v2si __builtin_arm_wsraw (v2si, long long)
18882 v2si __builtin_arm_wsrawi (v2si, int)
18883 long long __builtin_arm_wsrld (long long, long long)
18884 long long __builtin_arm_wsrldi (long long, int)
18885 v4hi __builtin_arm_wsrlh (v4hi, long long)
18886 v4hi __builtin_arm_wsrlhi (v4hi, int)
18887 v2si __builtin_arm_wsrlw (v2si, long long)
18888 v2si __builtin_arm_wsrlwi (v2si, int)
18889 v8qi __builtin_arm_wsubb (v8qi, v8qi)
18890 v8qi __builtin_arm_wsubbss (v8qi, v8qi)
18891 v8qi __builtin_arm_wsubbus (v8qi, v8qi)
18892 v4hi __builtin_arm_wsubh (v4hi, v4hi)
18893 v4hi __builtin_arm_wsubhss (v4hi, v4hi)
18894 v4hi __builtin_arm_wsubhus (v4hi, v4hi)
18895 v2si __builtin_arm_wsubw (v2si, v2si)
18896 v2si __builtin_arm_wsubwss (v2si, v2si)
18897 v2si __builtin_arm_wsubwus (v2si, v2si)
18898 v4hi __builtin_arm_wunpckehsb (v8qi)
18899 v2si __builtin_arm_wunpckehsh (v4hi)
18900 long long __builtin_arm_wunpckehsw (v2si)
18901 v4hi __builtin_arm_wunpckehub (v8qi)
18902 v2si __builtin_arm_wunpckehuh (v4hi)
18903 long long __builtin_arm_wunpckehuw (v2si)
18904 v4hi __builtin_arm_wunpckelsb (v8qi)
18905 v2si __builtin_arm_wunpckelsh (v4hi)
18906 long long __builtin_arm_wunpckelsw (v2si)
18907 v4hi __builtin_arm_wunpckelub (v8qi)
18908 v2si __builtin_arm_wunpckeluh (v4hi)
18909 long long __builtin_arm_wunpckeluw (v2si)
18910 v8qi __builtin_arm_wunpckihb (v8qi, v8qi)
18911 v4hi __builtin_arm_wunpckihh (v4hi, v4hi)
18912 v2si __builtin_arm_wunpckihw (v2si, v2si)
18913 v8qi __builtin_arm_wunpckilb (v8qi, v8qi)
18914 v4hi __builtin_arm_wunpckilh (v4hi, v4hi)
18915 v2si __builtin_arm_wunpckilw (v2si, v2si)
18916 long long __builtin_arm_wxor (long long, long long)
18917 long long __builtin_arm_wzero ()
18920 File: gcc.info, Node: FR-V Built-in Functions, Next: X86 Built-in Functions, Prev: ARM Built-in Functions, Up: Target Builtins
18922 5.45.3 FR-V Built-in Functions
18923 ------------------------------
18925 GCC provides many FR-V-specific built-in functions. In general, these
18926 functions are intended to be compatible with those described by `FR-V
18927 Family, Softune C/C++ Compiler Manual (V6), Fujitsu Semiconductor'.
18928 The two exceptions are `__MDUNPACKH' and `__MBTOHE', the gcc forms of
18929 which pass 128-bit values by pointer rather than by value.
18931 Most of the functions are named after specific FR-V instructions.
18932 Such functions are said to be "directly mapped" and are summarized here
18938 * Directly-mapped Integer Functions::
18939 * Directly-mapped Media Functions::
18940 * Other Built-in Functions::
18943 File: gcc.info, Node: Argument Types, Next: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
18945 5.45.3.1 Argument Types
18946 .......................
18948 The arguments to the built-in functions can be divided into three
18949 groups: register numbers, compile-time constants and run-time values.
18950 In order to make this classification clear at a glance, the arguments
18951 and return values are given the following pseudo types:
18953 Pseudo type Real C type Constant? Description
18954 `uh' `unsigned short' No an unsigned halfword
18955 `uw1' `unsigned int' No an unsigned word
18956 `sw1' `int' No a signed word
18957 `uw2' `unsigned long long' No an unsigned doubleword
18958 `sw2' `long long' No a signed doubleword
18959 `const' `int' Yes an integer constant
18960 `acc' `int' Yes an ACC register number
18961 `iacc' `int' Yes an IACC register number
18963 These pseudo types are not defined by GCC, they are simply a notational
18964 convenience used in this manual.
18966 Arguments of type `uh', `uw1', `sw1', `uw2' and `sw2' are evaluated at
18967 run time. They correspond to register operands in the underlying FR-V
18970 `const' arguments represent immediate operands in the underlying FR-V
18971 instructions. They must be compile-time constants.
18973 `acc' arguments are evaluated at compile time and specify the number
18974 of an accumulator register. For example, an `acc' argument of 2 will
18975 select the ACC2 register.
18977 `iacc' arguments are similar to `acc' arguments but specify the number
18978 of an IACC register. See *note Other Built-in Functions:: for more
18982 File: gcc.info, Node: Directly-mapped Integer Functions, Next: Directly-mapped Media Functions, Prev: Argument Types, Up: FR-V Built-in Functions
18984 5.45.3.2 Directly-mapped Integer Functions
18985 ..........................................
18987 The functions listed below map directly to FR-V I-type instructions.
18989 Function prototype Example usage Assembly output
18990 `sw1 __ADDSS (sw1, sw1)' `C = __ADDSS (A, B)' `ADDSS A,B,C'
18991 `sw1 __SCAN (sw1, sw1)' `C = __SCAN (A, B)' `SCAN A,B,C'
18992 `sw1 __SCUTSS (sw1)' `B = __SCUTSS (A)' `SCUTSS A,B'
18993 `sw1 __SLASS (sw1, sw1)' `C = __SLASS (A, B)' `SLASS A,B,C'
18994 `void __SMASS (sw1, sw1)' `__SMASS (A, B)' `SMASS A,B'
18995 `void __SMSSS (sw1, sw1)' `__SMSSS (A, B)' `SMSSS A,B'
18996 `void __SMU (sw1, sw1)' `__SMU (A, B)' `SMU A,B'
18997 `sw2 __SMUL (sw1, sw1)' `C = __SMUL (A, B)' `SMUL A,B,C'
18998 `sw1 __SUBSS (sw1, sw1)' `C = __SUBSS (A, B)' `SUBSS A,B,C'
18999 `uw2 __UMUL (uw1, uw1)' `C = __UMUL (A, B)' `UMUL A,B,C'
19002 File: gcc.info, Node: Directly-mapped Media Functions, Next: Other Built-in Functions, Prev: Directly-mapped Integer Functions, Up: FR-V Built-in Functions
19004 5.45.3.3 Directly-mapped Media Functions
19005 ........................................
19007 The functions listed below map directly to FR-V M-type instructions.
19009 Function prototype Example usage Assembly output
19010 `uw1 __MABSHS (sw1)' `B = __MABSHS (A)' `MABSHS A,B'
19011 `void __MADDACCS (acc, acc)' `__MADDACCS (B, A)' `MADDACCS A,B'
19012 `sw1 __MADDHSS (sw1, sw1)' `C = __MADDHSS (A, B)' `MADDHSS A,B,C'
19013 `uw1 __MADDHUS (uw1, uw1)' `C = __MADDHUS (A, B)' `MADDHUS A,B,C'
19014 `uw1 __MAND (uw1, uw1)' `C = __MAND (A, B)' `MAND A,B,C'
19015 `void __MASACCS (acc, acc)' `__MASACCS (B, A)' `MASACCS A,B'
19016 `uw1 __MAVEH (uw1, uw1)' `C = __MAVEH (A, B)' `MAVEH A,B,C'
19017 `uw2 __MBTOH (uw1)' `B = __MBTOH (A)' `MBTOH A,B'
19018 `void __MBTOHE (uw1 *, uw1)' `__MBTOHE (&B, A)' `MBTOHE A,B'
19019 `void __MCLRACC (acc)' `__MCLRACC (A)' `MCLRACC A'
19020 `void __MCLRACCA (void)' `__MCLRACCA ()' `MCLRACCA'
19021 `uw1 __Mcop1 (uw1, uw1)' `C = __Mcop1 (A, B)' `Mcop1 A,B,C'
19022 `uw1 __Mcop2 (uw1, uw1)' `C = __Mcop2 (A, B)' `Mcop2 A,B,C'
19023 `uw1 __MCPLHI (uw2, const)' `C = __MCPLHI (A, B)' `MCPLHI A,#B,C'
19024 `uw1 __MCPLI (uw2, const)' `C = __MCPLI (A, B)' `MCPLI A,#B,C'
19025 `void __MCPXIS (acc, sw1, sw1)' `__MCPXIS (C, A, B)' `MCPXIS A,B,C'
19026 `void __MCPXIU (acc, uw1, uw1)' `__MCPXIU (C, A, B)' `MCPXIU A,B,C'
19027 `void __MCPXRS (acc, sw1, sw1)' `__MCPXRS (C, A, B)' `MCPXRS A,B,C'
19028 `void __MCPXRU (acc, uw1, uw1)' `__MCPXRU (C, A, B)' `MCPXRU A,B,C'
19029 `uw1 __MCUT (acc, uw1)' `C = __MCUT (A, B)' `MCUT A,B,C'
19030 `uw1 __MCUTSS (acc, sw1)' `C = __MCUTSS (A, B)' `MCUTSS A,B,C'
19031 `void __MDADDACCS (acc, acc)' `__MDADDACCS (B, A)' `MDADDACCS A,B'
19032 `void __MDASACCS (acc, acc)' `__MDASACCS (B, A)' `MDASACCS A,B'
19033 `uw2 __MDCUTSSI (acc, const)' `C = __MDCUTSSI (A, B)' `MDCUTSSI A,#B,C'
19034 `uw2 __MDPACKH (uw2, uw2)' `C = __MDPACKH (A, B)' `MDPACKH A,B,C'
19035 `uw2 __MDROTLI (uw2, const)' `C = __MDROTLI (A, B)' `MDROTLI A,#B,C'
19036 `void __MDSUBACCS (acc, acc)' `__MDSUBACCS (B, A)' `MDSUBACCS A,B'
19037 `void __MDUNPACKH (uw1 *, uw2)' `__MDUNPACKH (&B, A)' `MDUNPACKH A,B'
19038 `uw2 __MEXPDHD (uw1, const)' `C = __MEXPDHD (A, B)' `MEXPDHD A,#B,C'
19039 `uw1 __MEXPDHW (uw1, const)' `C = __MEXPDHW (A, B)' `MEXPDHW A,#B,C'
19040 `uw1 __MHDSETH (uw1, const)' `C = __MHDSETH (A, B)' `MHDSETH A,#B,C'
19041 `sw1 __MHDSETS (const)' `B = __MHDSETS (A)' `MHDSETS #A,B'
19042 `uw1 __MHSETHIH (uw1, const)' `B = __MHSETHIH (B, A)' `MHSETHIH #A,B'
19043 `sw1 __MHSETHIS (sw1, const)' `B = __MHSETHIS (B, A)' `MHSETHIS #A,B'
19044 `uw1 __MHSETLOH (uw1, const)' `B = __MHSETLOH (B, A)' `MHSETLOH #A,B'
19045 `sw1 __MHSETLOS (sw1, const)' `B = __MHSETLOS (B, A)' `MHSETLOS #A,B'
19046 `uw1 __MHTOB (uw2)' `B = __MHTOB (A)' `MHTOB A,B'
19047 `void __MMACHS (acc, sw1, sw1)' `__MMACHS (C, A, B)' `MMACHS A,B,C'
19048 `void __MMACHU (acc, uw1, uw1)' `__MMACHU (C, A, B)' `MMACHU A,B,C'
19049 `void __MMRDHS (acc, sw1, sw1)' `__MMRDHS (C, A, B)' `MMRDHS A,B,C'
19050 `void __MMRDHU (acc, uw1, uw1)' `__MMRDHU (C, A, B)' `MMRDHU A,B,C'
19051 `void __MMULHS (acc, sw1, sw1)' `__MMULHS (C, A, B)' `MMULHS A,B,C'
19052 `void __MMULHU (acc, uw1, uw1)' `__MMULHU (C, A, B)' `MMULHU A,B,C'
19053 `void __MMULXHS (acc, sw1, sw1)' `__MMULXHS (C, A, B)' `MMULXHS A,B,C'
19054 `void __MMULXHU (acc, uw1, uw1)' `__MMULXHU (C, A, B)' `MMULXHU A,B,C'
19055 `uw1 __MNOT (uw1)' `B = __MNOT (A)' `MNOT A,B'
19056 `uw1 __MOR (uw1, uw1)' `C = __MOR (A, B)' `MOR A,B,C'
19057 `uw1 __MPACKH (uh, uh)' `C = __MPACKH (A, B)' `MPACKH A,B,C'
19058 `sw2 __MQADDHSS (sw2, sw2)' `C = __MQADDHSS (A, B)' `MQADDHSS A,B,C'
19059 `uw2 __MQADDHUS (uw2, uw2)' `C = __MQADDHUS (A, B)' `MQADDHUS A,B,C'
19060 `void __MQCPXIS (acc, sw2, sw2)' `__MQCPXIS (C, A, B)' `MQCPXIS A,B,C'
19061 `void __MQCPXIU (acc, uw2, uw2)' `__MQCPXIU (C, A, B)' `MQCPXIU A,B,C'
19062 `void __MQCPXRS (acc, sw2, sw2)' `__MQCPXRS (C, A, B)' `MQCPXRS A,B,C'
19063 `void __MQCPXRU (acc, uw2, uw2)' `__MQCPXRU (C, A, B)' `MQCPXRU A,B,C'
19064 `sw2 __MQLCLRHS (sw2, sw2)' `C = __MQLCLRHS (A, B)' `MQLCLRHS A,B,C'
19065 `sw2 __MQLMTHS (sw2, sw2)' `C = __MQLMTHS (A, B)' `MQLMTHS A,B,C'
19066 `void __MQMACHS (acc, sw2, sw2)' `__MQMACHS (C, A, B)' `MQMACHS A,B,C'
19067 `void __MQMACHU (acc, uw2, uw2)' `__MQMACHU (C, A, B)' `MQMACHU A,B,C'
19068 `void __MQMACXHS (acc, sw2, `__MQMACXHS (C, A, B)' `MQMACXHS A,B,C'
19070 `void __MQMULHS (acc, sw2, sw2)' `__MQMULHS (C, A, B)' `MQMULHS A,B,C'
19071 `void __MQMULHU (acc, uw2, uw2)' `__MQMULHU (C, A, B)' `MQMULHU A,B,C'
19072 `void __MQMULXHS (acc, sw2, `__MQMULXHS (C, A, B)' `MQMULXHS A,B,C'
19074 `void __MQMULXHU (acc, uw2, `__MQMULXHU (C, A, B)' `MQMULXHU A,B,C'
19076 `sw2 __MQSATHS (sw2, sw2)' `C = __MQSATHS (A, B)' `MQSATHS A,B,C'
19077 `uw2 __MQSLLHI (uw2, int)' `C = __MQSLLHI (A, B)' `MQSLLHI A,B,C'
19078 `sw2 __MQSRAHI (sw2, int)' `C = __MQSRAHI (A, B)' `MQSRAHI A,B,C'
19079 `sw2 __MQSUBHSS (sw2, sw2)' `C = __MQSUBHSS (A, B)' `MQSUBHSS A,B,C'
19080 `uw2 __MQSUBHUS (uw2, uw2)' `C = __MQSUBHUS (A, B)' `MQSUBHUS A,B,C'
19081 `void __MQXMACHS (acc, sw2, `__MQXMACHS (C, A, B)' `MQXMACHS A,B,C'
19083 `void __MQXMACXHS (acc, sw2, `__MQXMACXHS (C, A, B)' `MQXMACXHS A,B,C'
19085 `uw1 __MRDACC (acc)' `B = __MRDACC (A)' `MRDACC A,B'
19086 `uw1 __MRDACCG (acc)' `B = __MRDACCG (A)' `MRDACCG A,B'
19087 `uw1 __MROTLI (uw1, const)' `C = __MROTLI (A, B)' `MROTLI A,#B,C'
19088 `uw1 __MROTRI (uw1, const)' `C = __MROTRI (A, B)' `MROTRI A,#B,C'
19089 `sw1 __MSATHS (sw1, sw1)' `C = __MSATHS (A, B)' `MSATHS A,B,C'
19090 `uw1 __MSATHU (uw1, uw1)' `C = __MSATHU (A, B)' `MSATHU A,B,C'
19091 `uw1 __MSLLHI (uw1, const)' `C = __MSLLHI (A, B)' `MSLLHI A,#B,C'
19092 `sw1 __MSRAHI (sw1, const)' `C = __MSRAHI (A, B)' `MSRAHI A,#B,C'
19093 `uw1 __MSRLHI (uw1, const)' `C = __MSRLHI (A, B)' `MSRLHI A,#B,C'
19094 `void __MSUBACCS (acc, acc)' `__MSUBACCS (B, A)' `MSUBACCS A,B'
19095 `sw1 __MSUBHSS (sw1, sw1)' `C = __MSUBHSS (A, B)' `MSUBHSS A,B,C'
19096 `uw1 __MSUBHUS (uw1, uw1)' `C = __MSUBHUS (A, B)' `MSUBHUS A,B,C'
19097 `void __MTRAP (void)' `__MTRAP ()' `MTRAP'
19098 `uw2 __MUNPACKH (uw1)' `B = __MUNPACKH (A)' `MUNPACKH A,B'
19099 `uw1 __MWCUT (uw2, uw1)' `C = __MWCUT (A, B)' `MWCUT A,B,C'
19100 `void __MWTACC (acc, uw1)' `__MWTACC (B, A)' `MWTACC A,B'
19101 `void __MWTACCG (acc, uw1)' `__MWTACCG (B, A)' `MWTACCG A,B'
19102 `uw1 __MXOR (uw1, uw1)' `C = __MXOR (A, B)' `MXOR A,B,C'
19105 File: gcc.info, Node: Other Built-in Functions, Prev: Directly-mapped Media Functions, Up: FR-V Built-in Functions
19107 5.45.3.4 Other Built-in Functions
19108 .................................
19110 This section describes built-in functions that are not named after a
19111 specific FR-V instruction.
19113 `sw2 __IACCreadll (iacc REG)'
19114 Return the full 64-bit value of IACC0. The REG argument is
19115 reserved for future expansion and must be 0.
19117 `sw1 __IACCreadl (iacc REG)'
19118 Return the value of IACC0H if REG is 0 and IACC0L if REG is 1.
19119 Other values of REG are rejected as invalid.
19121 `void __IACCsetll (iacc REG, sw2 X)'
19122 Set the full 64-bit value of IACC0 to X. The REG argument is
19123 reserved for future expansion and must be 0.
19125 `void __IACCsetl (iacc REG, sw1 X)'
19126 Set IACC0H to X if REG is 0 and IACC0L to X if REG is 1. Other
19127 values of REG are rejected as invalid.
19129 `void __data_prefetch0 (const void *X)'
19130 Use the `dcpl' instruction to load the contents of address X into
19133 `void __data_prefetch (const void *X)'
19134 Use the `nldub' instruction to load the contents of address X into
19135 the data cache. The instruction will be issued in slot I1.
19138 File: gcc.info, Node: X86 Built-in Functions, Next: MIPS Paired-Single Support, Prev: FR-V Built-in Functions, Up: Target Builtins
19140 5.45.4 X86 Built-in Functions
19141 -----------------------------
19143 These built-in functions are available for the i386 and x86-64 family
19144 of computers, depending on the command-line switches used.
19146 The following machine modes are available for use with MMX built-in
19147 functions (*note Vector Extensions::): `V2SI' for a vector of two
19148 32-bit integers, `V4HI' for a vector of four 16-bit integers, and
19149 `V8QI' for a vector of eight 8-bit integers. Some of the built-in
19150 functions operate on MMX registers as a whole 64-bit entity, these use
19151 `DI' as their mode.
19153 If 3Dnow extensions are enabled, `V2SF' is used as a mode for a vector
19154 of two 32-bit floating point values.
19156 If SSE extensions are enabled, `V4SF' is used for a vector of four
19157 32-bit floating point values. Some instructions use a vector of four
19158 32-bit integers, these use `V4SI'. Finally, some instructions operate
19159 on an entire vector register, interpreting it as a 128-bit integer,
19160 these use mode `TI'.
19162 The following built-in functions are made available by `-mmmx'. All
19163 of them generate the machine instruction that is part of the name.
19165 v8qi __builtin_ia32_paddb (v8qi, v8qi)
19166 v4hi __builtin_ia32_paddw (v4hi, v4hi)
19167 v2si __builtin_ia32_paddd (v2si, v2si)
19168 v8qi __builtin_ia32_psubb (v8qi, v8qi)
19169 v4hi __builtin_ia32_psubw (v4hi, v4hi)
19170 v2si __builtin_ia32_psubd (v2si, v2si)
19171 v8qi __builtin_ia32_paddsb (v8qi, v8qi)
19172 v4hi __builtin_ia32_paddsw (v4hi, v4hi)
19173 v8qi __builtin_ia32_psubsb (v8qi, v8qi)
19174 v4hi __builtin_ia32_psubsw (v4hi, v4hi)
19175 v8qi __builtin_ia32_paddusb (v8qi, v8qi)
19176 v4hi __builtin_ia32_paddusw (v4hi, v4hi)
19177 v8qi __builtin_ia32_psubusb (v8qi, v8qi)
19178 v4hi __builtin_ia32_psubusw (v4hi, v4hi)
19179 v4hi __builtin_ia32_pmullw (v4hi, v4hi)
19180 v4hi __builtin_ia32_pmulhw (v4hi, v4hi)
19181 di __builtin_ia32_pand (di, di)
19182 di __builtin_ia32_pandn (di,di)
19183 di __builtin_ia32_por (di, di)
19184 di __builtin_ia32_pxor (di, di)
19185 v8qi __builtin_ia32_pcmpeqb (v8qi, v8qi)
19186 v4hi __builtin_ia32_pcmpeqw (v4hi, v4hi)
19187 v2si __builtin_ia32_pcmpeqd (v2si, v2si)
19188 v8qi __builtin_ia32_pcmpgtb (v8qi, v8qi)
19189 v4hi __builtin_ia32_pcmpgtw (v4hi, v4hi)
19190 v2si __builtin_ia32_pcmpgtd (v2si, v2si)
19191 v8qi __builtin_ia32_punpckhbw (v8qi, v8qi)
19192 v4hi __builtin_ia32_punpckhwd (v4hi, v4hi)
19193 v2si __builtin_ia32_punpckhdq (v2si, v2si)
19194 v8qi __builtin_ia32_punpcklbw (v8qi, v8qi)
19195 v4hi __builtin_ia32_punpcklwd (v4hi, v4hi)
19196 v2si __builtin_ia32_punpckldq (v2si, v2si)
19197 v8qi __builtin_ia32_packsswb (v4hi, v4hi)
19198 v4hi __builtin_ia32_packssdw (v2si, v2si)
19199 v8qi __builtin_ia32_packuswb (v4hi, v4hi)
19201 The following built-in functions are made available either with
19202 `-msse', or with a combination of `-m3dnow' and `-march=athlon'. All
19203 of them generate the machine instruction that is part of the name.
19205 v4hi __builtin_ia32_pmulhuw (v4hi, v4hi)
19206 v8qi __builtin_ia32_pavgb (v8qi, v8qi)
19207 v4hi __builtin_ia32_pavgw (v4hi, v4hi)
19208 v4hi __builtin_ia32_psadbw (v8qi, v8qi)
19209 v8qi __builtin_ia32_pmaxub (v8qi, v8qi)
19210 v4hi __builtin_ia32_pmaxsw (v4hi, v4hi)
19211 v8qi __builtin_ia32_pminub (v8qi, v8qi)
19212 v4hi __builtin_ia32_pminsw (v4hi, v4hi)
19213 int __builtin_ia32_pextrw (v4hi, int)
19214 v4hi __builtin_ia32_pinsrw (v4hi, int, int)
19215 int __builtin_ia32_pmovmskb (v8qi)
19216 void __builtin_ia32_maskmovq (v8qi, v8qi, char *)
19217 void __builtin_ia32_movntq (di *, di)
19218 void __builtin_ia32_sfence (void)
19220 The following built-in functions are available when `-msse' is used.
19221 All of them generate the machine instruction that is part of the name.
19223 int __builtin_ia32_comieq (v4sf, v4sf)
19224 int __builtin_ia32_comineq (v4sf, v4sf)
19225 int __builtin_ia32_comilt (v4sf, v4sf)
19226 int __builtin_ia32_comile (v4sf, v4sf)
19227 int __builtin_ia32_comigt (v4sf, v4sf)
19228 int __builtin_ia32_comige (v4sf, v4sf)
19229 int __builtin_ia32_ucomieq (v4sf, v4sf)
19230 int __builtin_ia32_ucomineq (v4sf, v4sf)
19231 int __builtin_ia32_ucomilt (v4sf, v4sf)
19232 int __builtin_ia32_ucomile (v4sf, v4sf)
19233 int __builtin_ia32_ucomigt (v4sf, v4sf)
19234 int __builtin_ia32_ucomige (v4sf, v4sf)
19235 v4sf __builtin_ia32_addps (v4sf, v4sf)
19236 v4sf __builtin_ia32_subps (v4sf, v4sf)
19237 v4sf __builtin_ia32_mulps (v4sf, v4sf)
19238 v4sf __builtin_ia32_divps (v4sf, v4sf)
19239 v4sf __builtin_ia32_addss (v4sf, v4sf)
19240 v4sf __builtin_ia32_subss (v4sf, v4sf)
19241 v4sf __builtin_ia32_mulss (v4sf, v4sf)
19242 v4sf __builtin_ia32_divss (v4sf, v4sf)
19243 v4si __builtin_ia32_cmpeqps (v4sf, v4sf)
19244 v4si __builtin_ia32_cmpltps (v4sf, v4sf)
19245 v4si __builtin_ia32_cmpleps (v4sf, v4sf)
19246 v4si __builtin_ia32_cmpgtps (v4sf, v4sf)
19247 v4si __builtin_ia32_cmpgeps (v4sf, v4sf)
19248 v4si __builtin_ia32_cmpunordps (v4sf, v4sf)
19249 v4si __builtin_ia32_cmpneqps (v4sf, v4sf)
19250 v4si __builtin_ia32_cmpnltps (v4sf, v4sf)
19251 v4si __builtin_ia32_cmpnleps (v4sf, v4sf)
19252 v4si __builtin_ia32_cmpngtps (v4sf, v4sf)
19253 v4si __builtin_ia32_cmpngeps (v4sf, v4sf)
19254 v4si __builtin_ia32_cmpordps (v4sf, v4sf)
19255 v4si __builtin_ia32_cmpeqss (v4sf, v4sf)
19256 v4si __builtin_ia32_cmpltss (v4sf, v4sf)
19257 v4si __builtin_ia32_cmpless (v4sf, v4sf)
19258 v4si __builtin_ia32_cmpunordss (v4sf, v4sf)
19259 v4si __builtin_ia32_cmpneqss (v4sf, v4sf)
19260 v4si __builtin_ia32_cmpnlts (v4sf, v4sf)
19261 v4si __builtin_ia32_cmpnless (v4sf, v4sf)
19262 v4si __builtin_ia32_cmpordss (v4sf, v4sf)
19263 v4sf __builtin_ia32_maxps (v4sf, v4sf)
19264 v4sf __builtin_ia32_maxss (v4sf, v4sf)
19265 v4sf __builtin_ia32_minps (v4sf, v4sf)
19266 v4sf __builtin_ia32_minss (v4sf, v4sf)
19267 v4sf __builtin_ia32_andps (v4sf, v4sf)
19268 v4sf __builtin_ia32_andnps (v4sf, v4sf)
19269 v4sf __builtin_ia32_orps (v4sf, v4sf)
19270 v4sf __builtin_ia32_xorps (v4sf, v4sf)
19271 v4sf __builtin_ia32_movss (v4sf, v4sf)
19272 v4sf __builtin_ia32_movhlps (v4sf, v4sf)
19273 v4sf __builtin_ia32_movlhps (v4sf, v4sf)
19274 v4sf __builtin_ia32_unpckhps (v4sf, v4sf)
19275 v4sf __builtin_ia32_unpcklps (v4sf, v4sf)
19276 v4sf __builtin_ia32_cvtpi2ps (v4sf, v2si)
19277 v4sf __builtin_ia32_cvtsi2ss (v4sf, int)
19278 v2si __builtin_ia32_cvtps2pi (v4sf)
19279 int __builtin_ia32_cvtss2si (v4sf)
19280 v2si __builtin_ia32_cvttps2pi (v4sf)
19281 int __builtin_ia32_cvttss2si (v4sf)
19282 v4sf __builtin_ia32_rcpps (v4sf)
19283 v4sf __builtin_ia32_rsqrtps (v4sf)
19284 v4sf __builtin_ia32_sqrtps (v4sf)
19285 v4sf __builtin_ia32_rcpss (v4sf)
19286 v4sf __builtin_ia32_rsqrtss (v4sf)
19287 v4sf __builtin_ia32_sqrtss (v4sf)
19288 v4sf __builtin_ia32_shufps (v4sf, v4sf, int)
19289 void __builtin_ia32_movntps (float *, v4sf)
19290 int __builtin_ia32_movmskps (v4sf)
19292 The following built-in functions are available when `-msse' is used.
19294 `v4sf __builtin_ia32_loadaps (float *)'
19295 Generates the `movaps' machine instruction as a load from memory.
19297 `void __builtin_ia32_storeaps (float *, v4sf)'
19298 Generates the `movaps' machine instruction as a store to memory.
19300 `v4sf __builtin_ia32_loadups (float *)'
19301 Generates the `movups' machine instruction as a load from memory.
19303 `void __builtin_ia32_storeups (float *, v4sf)'
19304 Generates the `movups' machine instruction as a store to memory.
19306 `v4sf __builtin_ia32_loadsss (float *)'
19307 Generates the `movss' machine instruction as a load from memory.
19309 `void __builtin_ia32_storess (float *, v4sf)'
19310 Generates the `movss' machine instruction as a store to memory.
19312 `v4sf __builtin_ia32_loadhps (v4sf, v2si *)'
19313 Generates the `movhps' machine instruction as a load from memory.
19315 `v4sf __builtin_ia32_loadlps (v4sf, v2si *)'
19316 Generates the `movlps' machine instruction as a load from memory
19318 `void __builtin_ia32_storehps (v4sf, v2si *)'
19319 Generates the `movhps' machine instruction as a store to memory.
19321 `void __builtin_ia32_storelps (v4sf, v2si *)'
19322 Generates the `movlps' machine instruction as a store to memory.
19324 The following built-in functions are available when `-msse3' is used.
19325 All of them generate the machine instruction that is part of the name.
19327 v2df __builtin_ia32_addsubpd (v2df, v2df)
19328 v2df __builtin_ia32_addsubps (v2df, v2df)
19329 v2df __builtin_ia32_haddpd (v2df, v2df)
19330 v2df __builtin_ia32_haddps (v2df, v2df)
19331 v2df __builtin_ia32_hsubpd (v2df, v2df)
19332 v2df __builtin_ia32_hsubps (v2df, v2df)
19333 v16qi __builtin_ia32_lddqu (char const *)
19334 void __builtin_ia32_monitor (void *, unsigned int, unsigned int)
19335 v2df __builtin_ia32_movddup (v2df)
19336 v4sf __builtin_ia32_movshdup (v4sf)
19337 v4sf __builtin_ia32_movsldup (v4sf)
19338 void __builtin_ia32_mwait (unsigned int, unsigned int)
19340 The following built-in functions are available when `-msse3' is used.
19342 `v2df __builtin_ia32_loadddup (double const *)'
19343 Generates the `movddup' machine instruction as a load from memory.
19345 The following built-in functions are available when `-m3dnow' is used.
19346 All of them generate the machine instruction that is part of the name.
19348 void __builtin_ia32_femms (void)
19349 v8qi __builtin_ia32_pavgusb (v8qi, v8qi)
19350 v2si __builtin_ia32_pf2id (v2sf)
19351 v2sf __builtin_ia32_pfacc (v2sf, v2sf)
19352 v2sf __builtin_ia32_pfadd (v2sf, v2sf)
19353 v2si __builtin_ia32_pfcmpeq (v2sf, v2sf)
19354 v2si __builtin_ia32_pfcmpge (v2sf, v2sf)
19355 v2si __builtin_ia32_pfcmpgt (v2sf, v2sf)
19356 v2sf __builtin_ia32_pfmax (v2sf, v2sf)
19357 v2sf __builtin_ia32_pfmin (v2sf, v2sf)
19358 v2sf __builtin_ia32_pfmul (v2sf, v2sf)
19359 v2sf __builtin_ia32_pfrcp (v2sf)
19360 v2sf __builtin_ia32_pfrcpit1 (v2sf, v2sf)
19361 v2sf __builtin_ia32_pfrcpit2 (v2sf, v2sf)
19362 v2sf __builtin_ia32_pfrsqrt (v2sf)
19363 v2sf __builtin_ia32_pfrsqrtit1 (v2sf, v2sf)
19364 v2sf __builtin_ia32_pfsub (v2sf, v2sf)
19365 v2sf __builtin_ia32_pfsubr (v2sf, v2sf)
19366 v2sf __builtin_ia32_pi2fd (v2si)
19367 v4hi __builtin_ia32_pmulhrw (v4hi, v4hi)
19369 The following built-in functions are available when both `-m3dnow' and
19370 `-march=athlon' are used. All of them generate the machine instruction
19371 that is part of the name.
19373 v2si __builtin_ia32_pf2iw (v2sf)
19374 v2sf __builtin_ia32_pfnacc (v2sf, v2sf)
19375 v2sf __builtin_ia32_pfpnacc (v2sf, v2sf)
19376 v2sf __builtin_ia32_pi2fw (v2si)
19377 v2sf __builtin_ia32_pswapdsf (v2sf)
19378 v2si __builtin_ia32_pswapdsi (v2si)
19381 File: gcc.info, Node: MIPS Paired-Single Support, Next: PowerPC AltiVec Built-in Functions, Prev: X86 Built-in Functions, Up: Target Builtins
19383 5.45.5 MIPS Paired-Single Support
19384 ---------------------------------
19386 The MIPS64 architecture includes a number of instructions that operate
19387 on pairs of single-precision floating-point values. Each pair is
19388 packed into a 64-bit floating-point register, with one element being
19389 designated the "upper half" and the other being designated the "lower
19392 GCC supports paired-single operations using both the generic vector
19393 extensions (*note Vector Extensions::) and a collection of
19394 MIPS-specific built-in functions. Both kinds of support are enabled by
19395 the `-mpaired-single' command-line option.
19397 The vector type associated with paired-single values is usually called
19398 `v2sf'. It can be defined in C as follows:
19400 typedef float v2sf __attribute__ ((vector_size (8)));
19402 `v2sf' values are initialized in the same way as aggregates. For
19405 v2sf a = {1.5, 9.1};
19410 _Note:_ The CPU's endianness determines which value is stored in the
19411 upper half of a register and which value is stored in the lower half.
19412 On little-endian targets, the first value is the lower one and the
19413 second value is the upper one. The opposite order applies to
19414 big-endian targets. For example, the code above will set the lower
19415 half of `a' to `1.5' on little-endian targets and `9.1' on big-endian
19420 * Paired-Single Arithmetic::
19421 * Paired-Single Built-in Functions::
19422 * MIPS-3D Built-in Functions::
19425 File: gcc.info, Node: Paired-Single Arithmetic, Next: Paired-Single Built-in Functions, Up: MIPS Paired-Single Support
19427 5.45.5.1 Paired-Single Arithmetic
19428 .................................
19430 The table below lists the `v2sf' operations for which hardware support
19431 exists. `a', `b' and `c' are `v2sf' values and `x' is an integral
19434 C code MIPS instruction
19439 `a * b + c' `madd.ps'
19440 `a * b - c' `msub.ps'
19441 `-(a * b + c)' `nmadd.ps'
19442 `-(a * b - c)' `nmsub.ps'
19443 `x ? a : b' `movn.ps'/`movz.ps'
19445 Note that the multiply-accumulate instructions can be disabled using
19446 the command-line option `-mno-fused-madd'.
19449 File: gcc.info, Node: Paired-Single Built-in Functions, Next: MIPS-3D Built-in Functions, Prev: Paired-Single Arithmetic, Up: MIPS Paired-Single Support
19451 5.45.5.2 Paired-Single Built-in Functions
19452 .........................................
19454 The following paired-single functions map directly to a particular MIPS
19455 instruction. Please refer to the architecture specification for
19456 details on what each instruction does.
19458 `v2sf __builtin_mips_pll_ps (v2sf, v2sf)'
19459 Pair lower lower (`pll.ps').
19461 `v2sf __builtin_mips_pul_ps (v2sf, v2sf)'
19462 Pair upper lower (`pul.ps').
19464 `v2sf __builtin_mips_plu_ps (v2sf, v2sf)'
19465 Pair lower upper (`plu.ps').
19467 `v2sf __builtin_mips_puu_ps (v2sf, v2sf)'
19468 Pair upper upper (`puu.ps').
19470 `v2sf __builtin_mips_cvt_ps_s (float, float)'
19471 Convert pair to paired single (`cvt.ps.s').
19473 `float __builtin_mips_cvt_s_pl (v2sf)'
19474 Convert pair lower to single (`cvt.s.pl').
19476 `float __builtin_mips_cvt_s_pu (v2sf)'
19477 Convert pair upper to single (`cvt.s.pu').
19479 `v2sf __builtin_mips_abs_ps (v2sf)'
19480 Absolute value (`abs.ps').
19482 `v2sf __builtin_mips_alnv_ps (v2sf, v2sf, int)'
19483 Align variable (`alnv.ps').
19485 _Note:_ The value of the third parameter must be 0 or 4 modulo 8,
19486 otherwise the result will be unpredictable. Please read the
19487 instruction description for details.
19489 The following multi-instruction functions are also available. In each
19490 case, COND can be any of the 16 floating-point conditions: `f', `un',
19491 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
19492 `lt', `nge', `le' or `ngt'.
19494 `v2sf __builtin_mips_movt_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
19495 `v2sf __builtin_mips_movf_c_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
19496 Conditional move based on floating point comparison (`c.COND.ps',
19497 `movt.ps'/`movf.ps').
19499 The `movt' functions return the value X computed by:
19505 The `movf' functions are similar but use `movf.ps' instead of
19508 `int __builtin_mips_upper_c_COND_ps (v2sf A, v2sf B)'
19509 `int __builtin_mips_lower_c_COND_ps (v2sf A, v2sf B)'
19510 Comparison of two paired-single values (`c.COND.ps',
19513 These functions compare A and B using `c.COND.ps' and return
19514 either the upper or lower half of the result. For example:
19517 if (__builtin_mips_upper_c_eq_ps (a, b))
19518 upper_halves_are_equal ();
19520 upper_halves_are_unequal ();
19522 if (__builtin_mips_lower_c_eq_ps (a, b))
19523 lower_halves_are_equal ();
19525 lower_halves_are_unequal ();
19528 File: gcc.info, Node: MIPS-3D Built-in Functions, Prev: Paired-Single Built-in Functions, Up: MIPS Paired-Single Support
19530 5.45.5.3 MIPS-3D Built-in Functions
19531 ...................................
19533 The MIPS-3D Application-Specific Extension (ASE) includes additional
19534 paired-single instructions that are designed to improve the performance
19535 of 3D graphics operations. Support for these instructions is controlled
19536 by the `-mips3d' command-line option.
19538 The functions listed below map directly to a particular MIPS-3D
19539 instruction. Please refer to the architecture specification for more
19540 details on what each instruction does.
19542 `v2sf __builtin_mips_addr_ps (v2sf, v2sf)'
19543 Reduction add (`addr.ps').
19545 `v2sf __builtin_mips_mulr_ps (v2sf, v2sf)'
19546 Reduction multiply (`mulr.ps').
19548 `v2sf __builtin_mips_cvt_pw_ps (v2sf)'
19549 Convert paired single to paired word (`cvt.pw.ps').
19551 `v2sf __builtin_mips_cvt_ps_pw (v2sf)'
19552 Convert paired word to paired single (`cvt.ps.pw').
19554 `float __builtin_mips_recip1_s (float)'
19555 `double __builtin_mips_recip1_d (double)'
19556 `v2sf __builtin_mips_recip1_ps (v2sf)'
19557 Reduced precision reciprocal (sequence step 1) (`recip1.FMT').
19559 `float __builtin_mips_recip2_s (float, float)'
19560 `double __builtin_mips_recip2_d (double, double)'
19561 `v2sf __builtin_mips_recip2_ps (v2sf, v2sf)'
19562 Reduced precision reciprocal (sequence step 2) (`recip2.FMT').
19564 `float __builtin_mips_rsqrt1_s (float)'
19565 `double __builtin_mips_rsqrt1_d (double)'
19566 `v2sf __builtin_mips_rsqrt1_ps (v2sf)'
19567 Reduced precision reciprocal square root (sequence step 1)
19570 `float __builtin_mips_rsqrt2_s (float, float)'
19571 `double __builtin_mips_rsqrt2_d (double, double)'
19572 `v2sf __builtin_mips_rsqrt2_ps (v2sf, v2sf)'
19573 Reduced precision reciprocal square root (sequence step 2)
19576 The following multi-instruction functions are also available. In each
19577 case, COND can be any of the 16 floating-point conditions: `f', `un',
19578 `eq', `ueq', `olt', `ult', `ole', `ule', `sf', `ngle', `seq', `ngl',
19579 `lt', `nge', `le' or `ngt'.
19581 `int __builtin_mips_cabs_COND_s (float A, float B)'
19582 `int __builtin_mips_cabs_COND_d (double A, double B)'
19583 Absolute comparison of two scalar values (`cabs.COND.FMT',
19586 These functions compare A and B using `cabs.COND.s' or
19587 `cabs.COND.d' and return the result as a boolean value. For
19591 if (__builtin_mips_cabs_eq_s (a, b))
19596 `int __builtin_mips_upper_cabs_COND_ps (v2sf A, v2sf B)'
19597 `int __builtin_mips_lower_cabs_COND_ps (v2sf A, v2sf B)'
19598 Absolute comparison of two paired-single values (`cabs.COND.ps',
19601 These functions compare A and B using `cabs.COND.ps' and return
19602 either the upper or lower half of the result. For example:
19605 if (__builtin_mips_upper_cabs_eq_ps (a, b))
19606 upper_halves_are_equal ();
19608 upper_halves_are_unequal ();
19610 if (__builtin_mips_lower_cabs_eq_ps (a, b))
19611 lower_halves_are_equal ();
19613 lower_halves_are_unequal ();
19615 `v2sf __builtin_mips_movt_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
19616 `v2sf __builtin_mips_movf_cabs_COND_ps (v2sf A, v2sf B, v2sf C, v2sf D)'
19617 Conditional move based on absolute comparison (`cabs.COND.ps',
19618 `movt.ps'/`movf.ps').
19620 The `movt' functions return the value X computed by:
19622 cabs.COND.ps CC,A,B
19626 The `movf' functions are similar but use `movf.ps' instead of
19629 `int __builtin_mips_any_c_COND_ps (v2sf A, v2sf B)'
19630 `int __builtin_mips_all_c_COND_ps (v2sf A, v2sf B)'
19631 `int __builtin_mips_any_cabs_COND_ps (v2sf A, v2sf B)'
19632 `int __builtin_mips_all_cabs_COND_ps (v2sf A, v2sf B)'
19633 Comparison of two paired-single values (`c.COND.ps'/`cabs.COND.ps',
19634 `bc1any2t'/`bc1any2f').
19636 These functions compare A and B using `c.COND.ps' or
19637 `cabs.COND.ps'. The `any' forms return true if either result is
19638 true and the `all' forms return true if both results are true.
19642 if (__builtin_mips_any_c_eq_ps (a, b))
19647 if (__builtin_mips_all_c_eq_ps (a, b))
19652 `int __builtin_mips_any_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
19653 `int __builtin_mips_all_c_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
19654 `int __builtin_mips_any_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
19655 `int __builtin_mips_all_cabs_COND_4s (v2sf A, v2sf B, v2sf C, v2sf D)'
19656 Comparison of four paired-single values
19657 (`c.COND.ps'/`cabs.COND.ps', `bc1any4t'/`bc1any4f').
19659 These functions use `c.COND.ps' or `cabs.COND.ps' to compare A
19660 with B and to compare C with D. The `any' forms return true if
19661 any of the four results are true and the `all' forms return true
19662 if all four results are true. For example:
19665 if (__builtin_mips_any_c_eq_4s (a, b, c, d))
19670 if (__builtin_mips_all_c_eq_4s (a, b, c, d))
19676 File: gcc.info, Node: PowerPC AltiVec Built-in Functions, Next: SPARC VIS Built-in Functions, Prev: MIPS Paired-Single Support, Up: Target Builtins
19678 5.45.6 PowerPC AltiVec Built-in Functions
19679 -----------------------------------------
19681 GCC provides an interface for the PowerPC family of processors to access
19682 the AltiVec operations described in Motorola's AltiVec Programming
19683 Interface Manual. The interface is made available by including
19684 `<altivec.h>' and using `-maltivec' and `-mabi=altivec'. The interface
19685 supports the following vector types.
19687 vector unsigned char
19691 vector unsigned short
19692 vector signed short
19696 vector unsigned int
19701 GCC's implementation of the high-level language interface available
19702 from C and C++ code differs from Motorola's documentation in several
19705 * A vector constant is a list of constant expressions within curly
19708 * A vector initializer requires no cast if the vector constant is of
19709 the same type as the variable it is initializing.
19711 * If `signed' or `unsigned' is omitted, the signedness of the vector
19712 type is the default signedness of the base type. The default
19713 varies depending on the operating system, so a portable program
19714 should always specify the signedness.
19716 * Compiling with `-maltivec' adds keywords `__vector', `__pixel',
19717 and `__bool'. Macros `vector', `pixel', and `bool' are defined in
19718 `<altivec.h>' and can be undefined.
19720 * GCC allows using a `typedef' name as the type specifier for a
19723 * For C, overloaded functions are implemented with macros so the
19724 following does not work:
19726 vec_add ((vector signed int){1, 2, 3, 4}, foo);
19728 Since `vec_add' is a macro, the vector constant in the example is
19729 treated as four separate arguments. Wrap the entire argument in
19730 parentheses for this to work.
19732 _Note:_ Only the `<altivec.h>' interface is supported. Internally,
19733 GCC uses built-in functions to achieve the functionality in the
19734 aforementioned header file, but they are not supported and are subject
19735 to change without notice.
19737 The following interfaces are supported for the generic and specific
19738 AltiVec operations and the AltiVec predicates. In cases where there is
19739 a direct mapping between generic and specific operations, only the
19740 generic names are shown here, although the specific operations can also
19743 Arguments that are documented as `const int' require literal integral
19744 values within the range required for that operation.
19746 vector signed char vec_abs (vector signed char);
19747 vector signed short vec_abs (vector signed short);
19748 vector signed int vec_abs (vector signed int);
19749 vector float vec_abs (vector float);
19751 vector signed char vec_abss (vector signed char);
19752 vector signed short vec_abss (vector signed short);
19753 vector signed int vec_abss (vector signed int);
19755 vector signed char vec_add (vector bool char, vector signed char);
19756 vector signed char vec_add (vector signed char, vector bool char);
19757 vector signed char vec_add (vector signed char, vector signed char);
19758 vector unsigned char vec_add (vector bool char, vector unsigned char);
19759 vector unsigned char vec_add (vector unsigned char, vector bool char);
19760 vector unsigned char vec_add (vector unsigned char,
19761 vector unsigned char);
19762 vector signed short vec_add (vector bool short, vector signed short);
19763 vector signed short vec_add (vector signed short, vector bool short);
19764 vector signed short vec_add (vector signed short, vector signed short);
19765 vector unsigned short vec_add (vector bool short,
19766 vector unsigned short);
19767 vector unsigned short vec_add (vector unsigned short,
19768 vector bool short);
19769 vector unsigned short vec_add (vector unsigned short,
19770 vector unsigned short);
19771 vector signed int vec_add (vector bool int, vector signed int);
19772 vector signed int vec_add (vector signed int, vector bool int);
19773 vector signed int vec_add (vector signed int, vector signed int);
19774 vector unsigned int vec_add (vector bool int, vector unsigned int);
19775 vector unsigned int vec_add (vector unsigned int, vector bool int);
19776 vector unsigned int vec_add (vector unsigned int, vector unsigned int);
19777 vector float vec_add (vector float, vector float);
19779 vector float vec_vaddfp (vector float, vector float);
19781 vector signed int vec_vadduwm (vector bool int, vector signed int);
19782 vector signed int vec_vadduwm (vector signed int, vector bool int);
19783 vector signed int vec_vadduwm (vector signed int, vector signed int);
19784 vector unsigned int vec_vadduwm (vector bool int, vector unsigned int);
19785 vector unsigned int vec_vadduwm (vector unsigned int, vector bool int);
19786 vector unsigned int vec_vadduwm (vector unsigned int,
19787 vector unsigned int);
19789 vector signed short vec_vadduhm (vector bool short,
19790 vector signed short);
19791 vector signed short vec_vadduhm (vector signed short,
19792 vector bool short);
19793 vector signed short vec_vadduhm (vector signed short,
19794 vector signed short);
19795 vector unsigned short vec_vadduhm (vector bool short,
19796 vector unsigned short);
19797 vector unsigned short vec_vadduhm (vector unsigned short,
19798 vector bool short);
19799 vector unsigned short vec_vadduhm (vector unsigned short,
19800 vector unsigned short);
19802 vector signed char vec_vaddubm (vector bool char, vector signed char);
19803 vector signed char vec_vaddubm (vector signed char, vector bool char);
19804 vector signed char vec_vaddubm (vector signed char, vector signed char);
19805 vector unsigned char vec_vaddubm (vector bool char,
19806 vector unsigned char);
19807 vector unsigned char vec_vaddubm (vector unsigned char,
19809 vector unsigned char vec_vaddubm (vector unsigned char,
19810 vector unsigned char);
19812 vector unsigned int vec_addc (vector unsigned int, vector unsigned int);
19814 vector unsigned char vec_adds (vector bool char, vector unsigned char);
19815 vector unsigned char vec_adds (vector unsigned char, vector bool char);
19816 vector unsigned char vec_adds (vector unsigned char,
19817 vector unsigned char);
19818 vector signed char vec_adds (vector bool char, vector signed char);
19819 vector signed char vec_adds (vector signed char, vector bool char);
19820 vector signed char vec_adds (vector signed char, vector signed char);
19821 vector unsigned short vec_adds (vector bool short,
19822 vector unsigned short);
19823 vector unsigned short vec_adds (vector unsigned short,
19824 vector bool short);
19825 vector unsigned short vec_adds (vector unsigned short,
19826 vector unsigned short);
19827 vector signed short vec_adds (vector bool short, vector signed short);
19828 vector signed short vec_adds (vector signed short, vector bool short);
19829 vector signed short vec_adds (vector signed short, vector signed short);
19830 vector unsigned int vec_adds (vector bool int, vector unsigned int);
19831 vector unsigned int vec_adds (vector unsigned int, vector bool int);
19832 vector unsigned int vec_adds (vector unsigned int, vector unsigned int);
19833 vector signed int vec_adds (vector bool int, vector signed int);
19834 vector signed int vec_adds (vector signed int, vector bool int);
19835 vector signed int vec_adds (vector signed int, vector signed int);
19837 vector signed int vec_vaddsws (vector bool int, vector signed int);
19838 vector signed int vec_vaddsws (vector signed int, vector bool int);
19839 vector signed int vec_vaddsws (vector signed int, vector signed int);
19841 vector unsigned int vec_vadduws (vector bool int, vector unsigned int);
19842 vector unsigned int vec_vadduws (vector unsigned int, vector bool int);
19843 vector unsigned int vec_vadduws (vector unsigned int,
19844 vector unsigned int);
19846 vector signed short vec_vaddshs (vector bool short,
19847 vector signed short);
19848 vector signed short vec_vaddshs (vector signed short,
19849 vector bool short);
19850 vector signed short vec_vaddshs (vector signed short,
19851 vector signed short);
19853 vector unsigned short vec_vadduhs (vector bool short,
19854 vector unsigned short);
19855 vector unsigned short vec_vadduhs (vector unsigned short,
19856 vector bool short);
19857 vector unsigned short vec_vadduhs (vector unsigned short,
19858 vector unsigned short);
19860 vector signed char vec_vaddsbs (vector bool char, vector signed char);
19861 vector signed char vec_vaddsbs (vector signed char, vector bool char);
19862 vector signed char vec_vaddsbs (vector signed char, vector signed char);
19864 vector unsigned char vec_vaddubs (vector bool char,
19865 vector unsigned char);
19866 vector unsigned char vec_vaddubs (vector unsigned char,
19868 vector unsigned char vec_vaddubs (vector unsigned char,
19869 vector unsigned char);
19871 vector float vec_and (vector float, vector float);
19872 vector float vec_and (vector float, vector bool int);
19873 vector float vec_and (vector bool int, vector float);
19874 vector bool int vec_and (vector bool int, vector bool int);
19875 vector signed int vec_and (vector bool int, vector signed int);
19876 vector signed int vec_and (vector signed int, vector bool int);
19877 vector signed int vec_and (vector signed int, vector signed int);
19878 vector unsigned int vec_and (vector bool int, vector unsigned int);
19879 vector unsigned int vec_and (vector unsigned int, vector bool int);
19880 vector unsigned int vec_and (vector unsigned int, vector unsigned int);
19881 vector bool short vec_and (vector bool short, vector bool short);
19882 vector signed short vec_and (vector bool short, vector signed short);
19883 vector signed short vec_and (vector signed short, vector bool short);
19884 vector signed short vec_and (vector signed short, vector signed short);
19885 vector unsigned short vec_and (vector bool short,
19886 vector unsigned short);
19887 vector unsigned short vec_and (vector unsigned short,
19888 vector bool short);
19889 vector unsigned short vec_and (vector unsigned short,
19890 vector unsigned short);
19891 vector signed char vec_and (vector bool char, vector signed char);
19892 vector bool char vec_and (vector bool char, vector bool char);
19893 vector signed char vec_and (vector signed char, vector bool char);
19894 vector signed char vec_and (vector signed char, vector signed char);
19895 vector unsigned char vec_and (vector bool char, vector unsigned char);
19896 vector unsigned char vec_and (vector unsigned char, vector bool char);
19897 vector unsigned char vec_and (vector unsigned char,
19898 vector unsigned char);
19900 vector float vec_andc (vector float, vector float);
19901 vector float vec_andc (vector float, vector bool int);
19902 vector float vec_andc (vector bool int, vector float);
19903 vector bool int vec_andc (vector bool int, vector bool int);
19904 vector signed int vec_andc (vector bool int, vector signed int);
19905 vector signed int vec_andc (vector signed int, vector bool int);
19906 vector signed int vec_andc (vector signed int, vector signed int);
19907 vector unsigned int vec_andc (vector bool int, vector unsigned int);
19908 vector unsigned int vec_andc (vector unsigned int, vector bool int);
19909 vector unsigned int vec_andc (vector unsigned int, vector unsigned int);
19910 vector bool short vec_andc (vector bool short, vector bool short);
19911 vector signed short vec_andc (vector bool short, vector signed short);
19912 vector signed short vec_andc (vector signed short, vector bool short);
19913 vector signed short vec_andc (vector signed short, vector signed short);
19914 vector unsigned short vec_andc (vector bool short,
19915 vector unsigned short);
19916 vector unsigned short vec_andc (vector unsigned short,
19917 vector bool short);
19918 vector unsigned short vec_andc (vector unsigned short,
19919 vector unsigned short);
19920 vector signed char vec_andc (vector bool char, vector signed char);
19921 vector bool char vec_andc (vector bool char, vector bool char);
19922 vector signed char vec_andc (vector signed char, vector bool char);
19923 vector signed char vec_andc (vector signed char, vector signed char);
19924 vector unsigned char vec_andc (vector bool char, vector unsigned char);
19925 vector unsigned char vec_andc (vector unsigned char, vector bool char);
19926 vector unsigned char vec_andc (vector unsigned char,
19927 vector unsigned char);
19929 vector unsigned char vec_avg (vector unsigned char,
19930 vector unsigned char);
19931 vector signed char vec_avg (vector signed char, vector signed char);
19932 vector unsigned short vec_avg (vector unsigned short,
19933 vector unsigned short);
19934 vector signed short vec_avg (vector signed short, vector signed short);
19935 vector unsigned int vec_avg (vector unsigned int, vector unsigned int);
19936 vector signed int vec_avg (vector signed int, vector signed int);
19938 vector signed int vec_vavgsw (vector signed int, vector signed int);
19940 vector unsigned int vec_vavguw (vector unsigned int,
19941 vector unsigned int);
19943 vector signed short vec_vavgsh (vector signed short,
19944 vector signed short);
19946 vector unsigned short vec_vavguh (vector unsigned short,
19947 vector unsigned short);
19949 vector signed char vec_vavgsb (vector signed char, vector signed char);
19951 vector unsigned char vec_vavgub (vector unsigned char,
19952 vector unsigned char);
19954 vector float vec_ceil (vector float);
19956 vector signed int vec_cmpb (vector float, vector float);
19958 vector bool char vec_cmpeq (vector signed char, vector signed char);
19959 vector bool char vec_cmpeq (vector unsigned char, vector unsigned char);
19960 vector bool short vec_cmpeq (vector signed short, vector signed short);
19961 vector bool short vec_cmpeq (vector unsigned short,
19962 vector unsigned short);
19963 vector bool int vec_cmpeq (vector signed int, vector signed int);
19964 vector bool int vec_cmpeq (vector unsigned int, vector unsigned int);
19965 vector bool int vec_cmpeq (vector float, vector float);
19967 vector bool int vec_vcmpeqfp (vector float, vector float);
19969 vector bool int vec_vcmpequw (vector signed int, vector signed int);
19970 vector bool int vec_vcmpequw (vector unsigned int, vector unsigned int);
19972 vector bool short vec_vcmpequh (vector signed short,
19973 vector signed short);
19974 vector bool short vec_vcmpequh (vector unsigned short,
19975 vector unsigned short);
19977 vector bool char vec_vcmpequb (vector signed char, vector signed char);
19978 vector bool char vec_vcmpequb (vector unsigned char,
19979 vector unsigned char);
19981 vector bool int vec_cmpge (vector float, vector float);
19983 vector bool char vec_cmpgt (vector unsigned char, vector unsigned char);
19984 vector bool char vec_cmpgt (vector signed char, vector signed char);
19985 vector bool short vec_cmpgt (vector unsigned short,
19986 vector unsigned short);
19987 vector bool short vec_cmpgt (vector signed short, vector signed short);
19988 vector bool int vec_cmpgt (vector unsigned int, vector unsigned int);
19989 vector bool int vec_cmpgt (vector signed int, vector signed int);
19990 vector bool int vec_cmpgt (vector float, vector float);
19992 vector bool int vec_vcmpgtfp (vector float, vector float);
19994 vector bool int vec_vcmpgtsw (vector signed int, vector signed int);
19996 vector bool int vec_vcmpgtuw (vector unsigned int, vector unsigned int);
19998 vector bool short vec_vcmpgtsh (vector signed short,
19999 vector signed short);
20001 vector bool short vec_vcmpgtuh (vector unsigned short,
20002 vector unsigned short);
20004 vector bool char vec_vcmpgtsb (vector signed char, vector signed char);
20006 vector bool char vec_vcmpgtub (vector unsigned char,
20007 vector unsigned char);
20009 vector bool int vec_cmple (vector float, vector float);
20011 vector bool char vec_cmplt (vector unsigned char, vector unsigned char);
20012 vector bool char vec_cmplt (vector signed char, vector signed char);
20013 vector bool short vec_cmplt (vector unsigned short,
20014 vector unsigned short);
20015 vector bool short vec_cmplt (vector signed short, vector signed short);
20016 vector bool int vec_cmplt (vector unsigned int, vector unsigned int);
20017 vector bool int vec_cmplt (vector signed int, vector signed int);
20018 vector bool int vec_cmplt (vector float, vector float);
20020 vector float vec_ctf (vector unsigned int, const int);
20021 vector float vec_ctf (vector signed int, const int);
20023 vector float vec_vcfsx (vector signed int, const int);
20025 vector float vec_vcfux (vector unsigned int, const int);
20027 vector signed int vec_cts (vector float, const int);
20029 vector unsigned int vec_ctu (vector float, const int);
20031 void vec_dss (const int);
20033 void vec_dssall (void);
20035 void vec_dst (const vector unsigned char *, int, const int);
20036 void vec_dst (const vector signed char *, int, const int);
20037 void vec_dst (const vector bool char *, int, const int);
20038 void vec_dst (const vector unsigned short *, int, const int);
20039 void vec_dst (const vector signed short *, int, const int);
20040 void vec_dst (const vector bool short *, int, const int);
20041 void vec_dst (const vector pixel *, int, const int);
20042 void vec_dst (const vector unsigned int *, int, const int);
20043 void vec_dst (const vector signed int *, int, const int);
20044 void vec_dst (const vector bool int *, int, const int);
20045 void vec_dst (const vector float *, int, const int);
20046 void vec_dst (const unsigned char *, int, const int);
20047 void vec_dst (const signed char *, int, const int);
20048 void vec_dst (const unsigned short *, int, const int);
20049 void vec_dst (const short *, int, const int);
20050 void vec_dst (const unsigned int *, int, const int);
20051 void vec_dst (const int *, int, const int);
20052 void vec_dst (const unsigned long *, int, const int);
20053 void vec_dst (const long *, int, const int);
20054 void vec_dst (const float *, int, const int);
20056 void vec_dstst (const vector unsigned char *, int, const int);
20057 void vec_dstst (const vector signed char *, int, const int);
20058 void vec_dstst (const vector bool char *, int, const int);
20059 void vec_dstst (const vector unsigned short *, int, const int);
20060 void vec_dstst (const vector signed short *, int, const int);
20061 void vec_dstst (const vector bool short *, int, const int);
20062 void vec_dstst (const vector pixel *, int, const int);
20063 void vec_dstst (const vector unsigned int *, int, const int);
20064 void vec_dstst (const vector signed int *, int, const int);
20065 void vec_dstst (const vector bool int *, int, const int);
20066 void vec_dstst (const vector float *, int, const int);
20067 void vec_dstst (const unsigned char *, int, const int);
20068 void vec_dstst (const signed char *, int, const int);
20069 void vec_dstst (const unsigned short *, int, const int);
20070 void vec_dstst (const short *, int, const int);
20071 void vec_dstst (const unsigned int *, int, const int);
20072 void vec_dstst (const int *, int, const int);
20073 void vec_dstst (const unsigned long *, int, const int);
20074 void vec_dstst (const long *, int, const int);
20075 void vec_dstst (const float *, int, const int);
20077 void vec_dststt (const vector unsigned char *, int, const int);
20078 void vec_dststt (const vector signed char *, int, const int);
20079 void vec_dststt (const vector bool char *, int, const int);
20080 void vec_dststt (const vector unsigned short *, int, const int);
20081 void vec_dststt (const vector signed short *, int, const int);
20082 void vec_dststt (const vector bool short *, int, const int);
20083 void vec_dststt (const vector pixel *, int, const int);
20084 void vec_dststt (const vector unsigned int *, int, const int);
20085 void vec_dststt (const vector signed int *, int, const int);
20086 void vec_dststt (const vector bool int *, int, const int);
20087 void vec_dststt (const vector float *, int, const int);
20088 void vec_dststt (const unsigned char *, int, const int);
20089 void vec_dststt (const signed char *, int, const int);
20090 void vec_dststt (const unsigned short *, int, const int);
20091 void vec_dststt (const short *, int, const int);
20092 void vec_dststt (const unsigned int *, int, const int);
20093 void vec_dststt (const int *, int, const int);
20094 void vec_dststt (const unsigned long *, int, const int);
20095 void vec_dststt (const long *, int, const int);
20096 void vec_dststt (const float *, int, const int);
20098 void vec_dstt (const vector unsigned char *, int, const int);
20099 void vec_dstt (const vector signed char *, int, const int);
20100 void vec_dstt (const vector bool char *, int, const int);
20101 void vec_dstt (const vector unsigned short *, int, const int);
20102 void vec_dstt (const vector signed short *, int, const int);
20103 void vec_dstt (const vector bool short *, int, const int);
20104 void vec_dstt (const vector pixel *, int, const int);
20105 void vec_dstt (const vector unsigned int *, int, const int);
20106 void vec_dstt (const vector signed int *, int, const int);
20107 void vec_dstt (const vector bool int *, int, const int);
20108 void vec_dstt (const vector float *, int, const int);
20109 void vec_dstt (const unsigned char *, int, const int);
20110 void vec_dstt (const signed char *, int, const int);
20111 void vec_dstt (const unsigned short *, int, const int);
20112 void vec_dstt (const short *, int, const int);
20113 void vec_dstt (const unsigned int *, int, const int);
20114 void vec_dstt (const int *, int, const int);
20115 void vec_dstt (const unsigned long *, int, const int);
20116 void vec_dstt (const long *, int, const int);
20117 void vec_dstt (const float *, int, const int);
20119 vector float vec_expte (vector float);
20121 vector float vec_floor (vector float);
20123 vector float vec_ld (int, const vector float *);
20124 vector float vec_ld (int, const float *);
20125 vector bool int vec_ld (int, const vector bool int *);
20126 vector signed int vec_ld (int, const vector signed int *);
20127 vector signed int vec_ld (int, const int *);
20128 vector signed int vec_ld (int, const long *);
20129 vector unsigned int vec_ld (int, const vector unsigned int *);
20130 vector unsigned int vec_ld (int, const unsigned int *);
20131 vector unsigned int vec_ld (int, const unsigned long *);
20132 vector bool short vec_ld (int, const vector bool short *);
20133 vector pixel vec_ld (int, const vector pixel *);
20134 vector signed short vec_ld (int, const vector signed short *);
20135 vector signed short vec_ld (int, const short *);
20136 vector unsigned short vec_ld (int, const vector unsigned short *);
20137 vector unsigned short vec_ld (int, const unsigned short *);
20138 vector bool char vec_ld (int, const vector bool char *);
20139 vector signed char vec_ld (int, const vector signed char *);
20140 vector signed char vec_ld (int, const signed char *);
20141 vector unsigned char vec_ld (int, const vector unsigned char *);
20142 vector unsigned char vec_ld (int, const unsigned char *);
20144 vector signed char vec_lde (int, const signed char *);
20145 vector unsigned char vec_lde (int, const unsigned char *);
20146 vector signed short vec_lde (int, const short *);
20147 vector unsigned short vec_lde (int, const unsigned short *);
20148 vector float vec_lde (int, const float *);
20149 vector signed int vec_lde (int, const int *);
20150 vector unsigned int vec_lde (int, const unsigned int *);
20151 vector signed int vec_lde (int, const long *);
20152 vector unsigned int vec_lde (int, const unsigned long *);
20154 vector float vec_lvewx (int, float *);
20155 vector signed int vec_lvewx (int, int *);
20156 vector unsigned int vec_lvewx (int, unsigned int *);
20157 vector signed int vec_lvewx (int, long *);
20158 vector unsigned int vec_lvewx (int, unsigned long *);
20160 vector signed short vec_lvehx (int, short *);
20161 vector unsigned short vec_lvehx (int, unsigned short *);
20163 vector signed char vec_lvebx (int, char *);
20164 vector unsigned char vec_lvebx (int, unsigned char *);
20166 vector float vec_ldl (int, const vector float *);
20167 vector float vec_ldl (int, const float *);
20168 vector bool int vec_ldl (int, const vector bool int *);
20169 vector signed int vec_ldl (int, const vector signed int *);
20170 vector signed int vec_ldl (int, const int *);
20171 vector signed int vec_ldl (int, const long *);
20172 vector unsigned int vec_ldl (int, const vector unsigned int *);
20173 vector unsigned int vec_ldl (int, const unsigned int *);
20174 vector unsigned int vec_ldl (int, const unsigned long *);
20175 vector bool short vec_ldl (int, const vector bool short *);
20176 vector pixel vec_ldl (int, const vector pixel *);
20177 vector signed short vec_ldl (int, const vector signed short *);
20178 vector signed short vec_ldl (int, const short *);
20179 vector unsigned short vec_ldl (int, const vector unsigned short *);
20180 vector unsigned short vec_ldl (int, const unsigned short *);
20181 vector bool char vec_ldl (int, const vector bool char *);
20182 vector signed char vec_ldl (int, const vector signed char *);
20183 vector signed char vec_ldl (int, const signed char *);
20184 vector unsigned char vec_ldl (int, const vector unsigned char *);
20185 vector unsigned char vec_ldl (int, const unsigned char *);
20187 vector float vec_loge (vector float);
20189 vector unsigned char vec_lvsl (int, const volatile unsigned char *);
20190 vector unsigned char vec_lvsl (int, const volatile signed char *);
20191 vector unsigned char vec_lvsl (int, const volatile unsigned short *);
20192 vector unsigned char vec_lvsl (int, const volatile short *);
20193 vector unsigned char vec_lvsl (int, const volatile unsigned int *);
20194 vector unsigned char vec_lvsl (int, const volatile int *);
20195 vector unsigned char vec_lvsl (int, const volatile unsigned long *);
20196 vector unsigned char vec_lvsl (int, const volatile long *);
20197 vector unsigned char vec_lvsl (int, const volatile float *);
20199 vector unsigned char vec_lvsr (int, const volatile unsigned char *);
20200 vector unsigned char vec_lvsr (int, const volatile signed char *);
20201 vector unsigned char vec_lvsr (int, const volatile unsigned short *);
20202 vector unsigned char vec_lvsr (int, const volatile short *);
20203 vector unsigned char vec_lvsr (int, const volatile unsigned int *);
20204 vector unsigned char vec_lvsr (int, const volatile int *);
20205 vector unsigned char vec_lvsr (int, const volatile unsigned long *);
20206 vector unsigned char vec_lvsr (int, const volatile long *);
20207 vector unsigned char vec_lvsr (int, const volatile float *);
20209 vector float vec_madd (vector float, vector float, vector float);
20211 vector signed short vec_madds (vector signed short,
20212 vector signed short,
20213 vector signed short);
20215 vector unsigned char vec_max (vector bool char, vector unsigned char);
20216 vector unsigned char vec_max (vector unsigned char, vector bool char);
20217 vector unsigned char vec_max (vector unsigned char,
20218 vector unsigned char);
20219 vector signed char vec_max (vector bool char, vector signed char);
20220 vector signed char vec_max (vector signed char, vector bool char);
20221 vector signed char vec_max (vector signed char, vector signed char);
20222 vector unsigned short vec_max (vector bool short,
20223 vector unsigned short);
20224 vector unsigned short vec_max (vector unsigned short,
20225 vector bool short);
20226 vector unsigned short vec_max (vector unsigned short,
20227 vector unsigned short);
20228 vector signed short vec_max (vector bool short, vector signed short);
20229 vector signed short vec_max (vector signed short, vector bool short);
20230 vector signed short vec_max (vector signed short, vector signed short);
20231 vector unsigned int vec_max (vector bool int, vector unsigned int);
20232 vector unsigned int vec_max (vector unsigned int, vector bool int);
20233 vector unsigned int vec_max (vector unsigned int, vector unsigned int);
20234 vector signed int vec_max (vector bool int, vector signed int);
20235 vector signed int vec_max (vector signed int, vector bool int);
20236 vector signed int vec_max (vector signed int, vector signed int);
20237 vector float vec_max (vector float, vector float);
20239 vector float vec_vmaxfp (vector float, vector float);
20241 vector signed int vec_vmaxsw (vector bool int, vector signed int);
20242 vector signed int vec_vmaxsw (vector signed int, vector bool int);
20243 vector signed int vec_vmaxsw (vector signed int, vector signed int);
20245 vector unsigned int vec_vmaxuw (vector bool int, vector unsigned int);
20246 vector unsigned int vec_vmaxuw (vector unsigned int, vector bool int);
20247 vector unsigned int vec_vmaxuw (vector unsigned int,
20248 vector unsigned int);
20250 vector signed short vec_vmaxsh (vector bool short, vector signed short);
20251 vector signed short vec_vmaxsh (vector signed short, vector bool short);
20252 vector signed short vec_vmaxsh (vector signed short,
20253 vector signed short);
20255 vector unsigned short vec_vmaxuh (vector bool short,
20256 vector unsigned short);
20257 vector unsigned short vec_vmaxuh (vector unsigned short,
20258 vector bool short);
20259 vector unsigned short vec_vmaxuh (vector unsigned short,
20260 vector unsigned short);
20262 vector signed char vec_vmaxsb (vector bool char, vector signed char);
20263 vector signed char vec_vmaxsb (vector signed char, vector bool char);
20264 vector signed char vec_vmaxsb (vector signed char, vector signed char);
20266 vector unsigned char vec_vmaxub (vector bool char,
20267 vector unsigned char);
20268 vector unsigned char vec_vmaxub (vector unsigned char,
20270 vector unsigned char vec_vmaxub (vector unsigned char,
20271 vector unsigned char);
20273 vector bool char vec_mergeh (vector bool char, vector bool char);
20274 vector signed char vec_mergeh (vector signed char, vector signed char);
20275 vector unsigned char vec_mergeh (vector unsigned char,
20276 vector unsigned char);
20277 vector bool short vec_mergeh (vector bool short, vector bool short);
20278 vector pixel vec_mergeh (vector pixel, vector pixel);
20279 vector signed short vec_mergeh (vector signed short,
20280 vector signed short);
20281 vector unsigned short vec_mergeh (vector unsigned short,
20282 vector unsigned short);
20283 vector float vec_mergeh (vector float, vector float);
20284 vector bool int vec_mergeh (vector bool int, vector bool int);
20285 vector signed int vec_mergeh (vector signed int, vector signed int);
20286 vector unsigned int vec_mergeh (vector unsigned int,
20287 vector unsigned int);
20289 vector float vec_vmrghw (vector float, vector float);
20290 vector bool int vec_vmrghw (vector bool int, vector bool int);
20291 vector signed int vec_vmrghw (vector signed int, vector signed int);
20292 vector unsigned int vec_vmrghw (vector unsigned int,
20293 vector unsigned int);
20295 vector bool short vec_vmrghh (vector bool short, vector bool short);
20296 vector signed short vec_vmrghh (vector signed short,
20297 vector signed short);
20298 vector unsigned short vec_vmrghh (vector unsigned short,
20299 vector unsigned short);
20300 vector pixel vec_vmrghh (vector pixel, vector pixel);
20302 vector bool char vec_vmrghb (vector bool char, vector bool char);
20303 vector signed char vec_vmrghb (vector signed char, vector signed char);
20304 vector unsigned char vec_vmrghb (vector unsigned char,
20305 vector unsigned char);
20307 vector bool char vec_mergel (vector bool char, vector bool char);
20308 vector signed char vec_mergel (vector signed char, vector signed char);
20309 vector unsigned char vec_mergel (vector unsigned char,
20310 vector unsigned char);
20311 vector bool short vec_mergel (vector bool short, vector bool short);
20312 vector pixel vec_mergel (vector pixel, vector pixel);
20313 vector signed short vec_mergel (vector signed short,
20314 vector signed short);
20315 vector unsigned short vec_mergel (vector unsigned short,
20316 vector unsigned short);
20317 vector float vec_mergel (vector float, vector float);
20318 vector bool int vec_mergel (vector bool int, vector bool int);
20319 vector signed int vec_mergel (vector signed int, vector signed int);
20320 vector unsigned int vec_mergel (vector unsigned int,
20321 vector unsigned int);
20323 vector float vec_vmrglw (vector float, vector float);
20324 vector signed int vec_vmrglw (vector signed int, vector signed int);
20325 vector unsigned int vec_vmrglw (vector unsigned int,
20326 vector unsigned int);
20327 vector bool int vec_vmrglw (vector bool int, vector bool int);
20329 vector bool short vec_vmrglh (vector bool short, vector bool short);
20330 vector signed short vec_vmrglh (vector signed short,
20331 vector signed short);
20332 vector unsigned short vec_vmrglh (vector unsigned short,
20333 vector unsigned short);
20334 vector pixel vec_vmrglh (vector pixel, vector pixel);
20336 vector bool char vec_vmrglb (vector bool char, vector bool char);
20337 vector signed char vec_vmrglb (vector signed char, vector signed char);
20338 vector unsigned char vec_vmrglb (vector unsigned char,
20339 vector unsigned char);
20341 vector unsigned short vec_mfvscr (void);
20343 vector unsigned char vec_min (vector bool char, vector unsigned char);
20344 vector unsigned char vec_min (vector unsigned char, vector bool char);
20345 vector unsigned char vec_min (vector unsigned char,
20346 vector unsigned char);
20347 vector signed char vec_min (vector bool char, vector signed char);
20348 vector signed char vec_min (vector signed char, vector bool char);
20349 vector signed char vec_min (vector signed char, vector signed char);
20350 vector unsigned short vec_min (vector bool short,
20351 vector unsigned short);
20352 vector unsigned short vec_min (vector unsigned short,
20353 vector bool short);
20354 vector unsigned short vec_min (vector unsigned short,
20355 vector unsigned short);
20356 vector signed short vec_min (vector bool short, vector signed short);
20357 vector signed short vec_min (vector signed short, vector bool short);
20358 vector signed short vec_min (vector signed short, vector signed short);
20359 vector unsigned int vec_min (vector bool int, vector unsigned int);
20360 vector unsigned int vec_min (vector unsigned int, vector bool int);
20361 vector unsigned int vec_min (vector unsigned int, vector unsigned int);
20362 vector signed int vec_min (vector bool int, vector signed int);
20363 vector signed int vec_min (vector signed int, vector bool int);
20364 vector signed int vec_min (vector signed int, vector signed int);
20365 vector float vec_min (vector float, vector float);
20367 vector float vec_vminfp (vector float, vector float);
20369 vector signed int vec_vminsw (vector bool int, vector signed int);
20370 vector signed int vec_vminsw (vector signed int, vector bool int);
20371 vector signed int vec_vminsw (vector signed int, vector signed int);
20373 vector unsigned int vec_vminuw (vector bool int, vector unsigned int);
20374 vector unsigned int vec_vminuw (vector unsigned int, vector bool int);
20375 vector unsigned int vec_vminuw (vector unsigned int,
20376 vector unsigned int);
20378 vector signed short vec_vminsh (vector bool short, vector signed short);
20379 vector signed short vec_vminsh (vector signed short, vector bool short);
20380 vector signed short vec_vminsh (vector signed short,
20381 vector signed short);
20383 vector unsigned short vec_vminuh (vector bool short,
20384 vector unsigned short);
20385 vector unsigned short vec_vminuh (vector unsigned short,
20386 vector bool short);
20387 vector unsigned short vec_vminuh (vector unsigned short,
20388 vector unsigned short);
20390 vector signed char vec_vminsb (vector bool char, vector signed char);
20391 vector signed char vec_vminsb (vector signed char, vector bool char);
20392 vector signed char vec_vminsb (vector signed char, vector signed char);
20394 vector unsigned char vec_vminub (vector bool char,
20395 vector unsigned char);
20396 vector unsigned char vec_vminub (vector unsigned char,
20398 vector unsigned char vec_vminub (vector unsigned char,
20399 vector unsigned char);
20401 vector signed short vec_mladd (vector signed short,
20402 vector signed short,
20403 vector signed short);
20404 vector signed short vec_mladd (vector signed short,
20405 vector unsigned short,
20406 vector unsigned short);
20407 vector signed short vec_mladd (vector unsigned short,
20408 vector signed short,
20409 vector signed short);
20410 vector unsigned short vec_mladd (vector unsigned short,
20411 vector unsigned short,
20412 vector unsigned short);
20414 vector signed short vec_mradds (vector signed short,
20415 vector signed short,
20416 vector signed short);
20418 vector unsigned int vec_msum (vector unsigned char,
20419 vector unsigned char,
20420 vector unsigned int);
20421 vector signed int vec_msum (vector signed char,
20422 vector unsigned char,
20423 vector signed int);
20424 vector unsigned int vec_msum (vector unsigned short,
20425 vector unsigned short,
20426 vector unsigned int);
20427 vector signed int vec_msum (vector signed short,
20428 vector signed short,
20429 vector signed int);
20431 vector signed int vec_vmsumshm (vector signed short,
20432 vector signed short,
20433 vector signed int);
20435 vector unsigned int vec_vmsumuhm (vector unsigned short,
20436 vector unsigned short,
20437 vector unsigned int);
20439 vector signed int vec_vmsummbm (vector signed char,
20440 vector unsigned char,
20441 vector signed int);
20443 vector unsigned int vec_vmsumubm (vector unsigned char,
20444 vector unsigned char,
20445 vector unsigned int);
20447 vector unsigned int vec_msums (vector unsigned short,
20448 vector unsigned short,
20449 vector unsigned int);
20450 vector signed int vec_msums (vector signed short,
20451 vector signed short,
20452 vector signed int);
20454 vector signed int vec_vmsumshs (vector signed short,
20455 vector signed short,
20456 vector signed int);
20458 vector unsigned int vec_vmsumuhs (vector unsigned short,
20459 vector unsigned short,
20460 vector unsigned int);
20462 void vec_mtvscr (vector signed int);
20463 void vec_mtvscr (vector unsigned int);
20464 void vec_mtvscr (vector bool int);
20465 void vec_mtvscr (vector signed short);
20466 void vec_mtvscr (vector unsigned short);
20467 void vec_mtvscr (vector bool short);
20468 void vec_mtvscr (vector pixel);
20469 void vec_mtvscr (vector signed char);
20470 void vec_mtvscr (vector unsigned char);
20471 void vec_mtvscr (vector bool char);
20473 vector unsigned short vec_mule (vector unsigned char,
20474 vector unsigned char);
20475 vector signed short vec_mule (vector signed char,
20476 vector signed char);
20477 vector unsigned int vec_mule (vector unsigned short,
20478 vector unsigned short);
20479 vector signed int vec_mule (vector signed short, vector signed short);
20481 vector signed int vec_vmulesh (vector signed short,
20482 vector signed short);
20484 vector unsigned int vec_vmuleuh (vector unsigned short,
20485 vector unsigned short);
20487 vector signed short vec_vmulesb (vector signed char,
20488 vector signed char);
20490 vector unsigned short vec_vmuleub (vector unsigned char,
20491 vector unsigned char);
20493 vector unsigned short vec_mulo (vector unsigned char,
20494 vector unsigned char);
20495 vector signed short vec_mulo (vector signed char, vector signed char);
20496 vector unsigned int vec_mulo (vector unsigned short,
20497 vector unsigned short);
20498 vector signed int vec_mulo (vector signed short, vector signed short);
20500 vector signed int vec_vmulosh (vector signed short,
20501 vector signed short);
20503 vector unsigned int vec_vmulouh (vector unsigned short,
20504 vector unsigned short);
20506 vector signed short vec_vmulosb (vector signed char,
20507 vector signed char);
20509 vector unsigned short vec_vmuloub (vector unsigned char,
20510 vector unsigned char);
20512 vector float vec_nmsub (vector float, vector float, vector float);
20514 vector float vec_nor (vector float, vector float);
20515 vector signed int vec_nor (vector signed int, vector signed int);
20516 vector unsigned int vec_nor (vector unsigned int, vector unsigned int);
20517 vector bool int vec_nor (vector bool int, vector bool int);
20518 vector signed short vec_nor (vector signed short, vector signed short);
20519 vector unsigned short vec_nor (vector unsigned short,
20520 vector unsigned short);
20521 vector bool short vec_nor (vector bool short, vector bool short);
20522 vector signed char vec_nor (vector signed char, vector signed char);
20523 vector unsigned char vec_nor (vector unsigned char,
20524 vector unsigned char);
20525 vector bool char vec_nor (vector bool char, vector bool char);
20527 vector float vec_or (vector float, vector float);
20528 vector float vec_or (vector float, vector bool int);
20529 vector float vec_or (vector bool int, vector float);
20530 vector bool int vec_or (vector bool int, vector bool int);
20531 vector signed int vec_or (vector bool int, vector signed int);
20532 vector signed int vec_or (vector signed int, vector bool int);
20533 vector signed int vec_or (vector signed int, vector signed int);
20534 vector unsigned int vec_or (vector bool int, vector unsigned int);
20535 vector unsigned int vec_or (vector unsigned int, vector bool int);
20536 vector unsigned int vec_or (vector unsigned int, vector unsigned int);
20537 vector bool short vec_or (vector bool short, vector bool short);
20538 vector signed short vec_or (vector bool short, vector signed short);
20539 vector signed short vec_or (vector signed short, vector bool short);
20540 vector signed short vec_or (vector signed short, vector signed short);
20541 vector unsigned short vec_or (vector bool short, vector unsigned short);
20542 vector unsigned short vec_or (vector unsigned short, vector bool short);
20543 vector unsigned short vec_or (vector unsigned short,
20544 vector unsigned short);
20545 vector signed char vec_or (vector bool char, vector signed char);
20546 vector bool char vec_or (vector bool char, vector bool char);
20547 vector signed char vec_or (vector signed char, vector bool char);
20548 vector signed char vec_or (vector signed char, vector signed char);
20549 vector unsigned char vec_or (vector bool char, vector unsigned char);
20550 vector unsigned char vec_or (vector unsigned char, vector bool char);
20551 vector unsigned char vec_or (vector unsigned char,
20552 vector unsigned char);
20554 vector signed char vec_pack (vector signed short, vector signed short);
20555 vector unsigned char vec_pack (vector unsigned short,
20556 vector unsigned short);
20557 vector bool char vec_pack (vector bool short, vector bool short);
20558 vector signed short vec_pack (vector signed int, vector signed int);
20559 vector unsigned short vec_pack (vector unsigned int,
20560 vector unsigned int);
20561 vector bool short vec_pack (vector bool int, vector bool int);
20563 vector bool short vec_vpkuwum (vector bool int, vector bool int);
20564 vector signed short vec_vpkuwum (vector signed int, vector signed int);
20565 vector unsigned short vec_vpkuwum (vector unsigned int,
20566 vector unsigned int);
20568 vector bool char vec_vpkuhum (vector bool short, vector bool short);
20569 vector signed char vec_vpkuhum (vector signed short,
20570 vector signed short);
20571 vector unsigned char vec_vpkuhum (vector unsigned short,
20572 vector unsigned short);
20574 vector pixel vec_packpx (vector unsigned int, vector unsigned int);
20576 vector unsigned char vec_packs (vector unsigned short,
20577 vector unsigned short);
20578 vector signed char vec_packs (vector signed short, vector signed short);
20579 vector unsigned short vec_packs (vector unsigned int,
20580 vector unsigned int);
20581 vector signed short vec_packs (vector signed int, vector signed int);
20583 vector signed short vec_vpkswss (vector signed int, vector signed int);
20585 vector unsigned short vec_vpkuwus (vector unsigned int,
20586 vector unsigned int);
20588 vector signed char vec_vpkshss (vector signed short,
20589 vector signed short);
20591 vector unsigned char vec_vpkuhus (vector unsigned short,
20592 vector unsigned short);
20594 vector unsigned char vec_packsu (vector unsigned short,
20595 vector unsigned short);
20596 vector unsigned char vec_packsu (vector signed short,
20597 vector signed short);
20598 vector unsigned short vec_packsu (vector unsigned int,
20599 vector unsigned int);
20600 vector unsigned short vec_packsu (vector signed int, vector signed int);
20602 vector unsigned short vec_vpkswus (vector signed int,
20603 vector signed int);
20605 vector unsigned char vec_vpkshus (vector signed short,
20606 vector signed short);
20608 vector float vec_perm (vector float,
20610 vector unsigned char);
20611 vector signed int vec_perm (vector signed int,
20613 vector unsigned char);
20614 vector unsigned int vec_perm (vector unsigned int,
20615 vector unsigned int,
20616 vector unsigned char);
20617 vector bool int vec_perm (vector bool int,
20619 vector unsigned char);
20620 vector signed short vec_perm (vector signed short,
20621 vector signed short,
20622 vector unsigned char);
20623 vector unsigned short vec_perm (vector unsigned short,
20624 vector unsigned short,
20625 vector unsigned char);
20626 vector bool short vec_perm (vector bool short,
20628 vector unsigned char);
20629 vector pixel vec_perm (vector pixel,
20631 vector unsigned char);
20632 vector signed char vec_perm (vector signed char,
20633 vector signed char,
20634 vector unsigned char);
20635 vector unsigned char vec_perm (vector unsigned char,
20636 vector unsigned char,
20637 vector unsigned char);
20638 vector bool char vec_perm (vector bool char,
20640 vector unsigned char);
20642 vector float vec_re (vector float);
20644 vector signed char vec_rl (vector signed char,
20645 vector unsigned char);
20646 vector unsigned char vec_rl (vector unsigned char,
20647 vector unsigned char);
20648 vector signed short vec_rl (vector signed short, vector unsigned short);
20649 vector unsigned short vec_rl (vector unsigned short,
20650 vector unsigned short);
20651 vector signed int vec_rl (vector signed int, vector unsigned int);
20652 vector unsigned int vec_rl (vector unsigned int, vector unsigned int);
20654 vector signed int vec_vrlw (vector signed int, vector unsigned int);
20655 vector unsigned int vec_vrlw (vector unsigned int, vector unsigned int);
20657 vector signed short vec_vrlh (vector signed short,
20658 vector unsigned short);
20659 vector unsigned short vec_vrlh (vector unsigned short,
20660 vector unsigned short);
20662 vector signed char vec_vrlb (vector signed char, vector unsigned char);
20663 vector unsigned char vec_vrlb (vector unsigned char,
20664 vector unsigned char);
20666 vector float vec_round (vector float);
20668 vector float vec_rsqrte (vector float);
20670 vector float vec_sel (vector float, vector float, vector bool int);
20671 vector float vec_sel (vector float, vector float, vector unsigned int);
20672 vector signed int vec_sel (vector signed int,
20675 vector signed int vec_sel (vector signed int,
20677 vector unsigned int);
20678 vector unsigned int vec_sel (vector unsigned int,
20679 vector unsigned int,
20681 vector unsigned int vec_sel (vector unsigned int,
20682 vector unsigned int,
20683 vector unsigned int);
20684 vector bool int vec_sel (vector bool int,
20687 vector bool int vec_sel (vector bool int,
20689 vector unsigned int);
20690 vector signed short vec_sel (vector signed short,
20691 vector signed short,
20692 vector bool short);
20693 vector signed short vec_sel (vector signed short,
20694 vector signed short,
20695 vector unsigned short);
20696 vector unsigned short vec_sel (vector unsigned short,
20697 vector unsigned short,
20698 vector bool short);
20699 vector unsigned short vec_sel (vector unsigned short,
20700 vector unsigned short,
20701 vector unsigned short);
20702 vector bool short vec_sel (vector bool short,
20704 vector bool short);
20705 vector bool short vec_sel (vector bool short,
20707 vector unsigned short);
20708 vector signed char vec_sel (vector signed char,
20709 vector signed char,
20711 vector signed char vec_sel (vector signed char,
20712 vector signed char,
20713 vector unsigned char);
20714 vector unsigned char vec_sel (vector unsigned char,
20715 vector unsigned char,
20717 vector unsigned char vec_sel (vector unsigned char,
20718 vector unsigned char,
20719 vector unsigned char);
20720 vector bool char vec_sel (vector bool char,
20723 vector bool char vec_sel (vector bool char,
20725 vector unsigned char);
20727 vector signed char vec_sl (vector signed char,
20728 vector unsigned char);
20729 vector unsigned char vec_sl (vector unsigned char,
20730 vector unsigned char);
20731 vector signed short vec_sl (vector signed short, vector unsigned short);
20732 vector unsigned short vec_sl (vector unsigned short,
20733 vector unsigned short);
20734 vector signed int vec_sl (vector signed int, vector unsigned int);
20735 vector unsigned int vec_sl (vector unsigned int, vector unsigned int);
20737 vector signed int vec_vslw (vector signed int, vector unsigned int);
20738 vector unsigned int vec_vslw (vector unsigned int, vector unsigned int);
20740 vector signed short vec_vslh (vector signed short,
20741 vector unsigned short);
20742 vector unsigned short vec_vslh (vector unsigned short,
20743 vector unsigned short);
20745 vector signed char vec_vslb (vector signed char, vector unsigned char);
20746 vector unsigned char vec_vslb (vector unsigned char,
20747 vector unsigned char);
20749 vector float vec_sld (vector float, vector float, const int);
20750 vector signed int vec_sld (vector signed int,
20753 vector unsigned int vec_sld (vector unsigned int,
20754 vector unsigned int,
20756 vector bool int vec_sld (vector bool int,
20759 vector signed short vec_sld (vector signed short,
20760 vector signed short,
20762 vector unsigned short vec_sld (vector unsigned short,
20763 vector unsigned short,
20765 vector bool short vec_sld (vector bool short,
20768 vector pixel vec_sld (vector pixel,
20771 vector signed char vec_sld (vector signed char,
20772 vector signed char,
20774 vector unsigned char vec_sld (vector unsigned char,
20775 vector unsigned char,
20777 vector bool char vec_sld (vector bool char,
20781 vector signed int vec_sll (vector signed int,
20782 vector unsigned int);
20783 vector signed int vec_sll (vector signed int,
20784 vector unsigned short);
20785 vector signed int vec_sll (vector signed int,
20786 vector unsigned char);
20787 vector unsigned int vec_sll (vector unsigned int,
20788 vector unsigned int);
20789 vector unsigned int vec_sll (vector unsigned int,
20790 vector unsigned short);
20791 vector unsigned int vec_sll (vector unsigned int,
20792 vector unsigned char);
20793 vector bool int vec_sll (vector bool int,
20794 vector unsigned int);
20795 vector bool int vec_sll (vector bool int,
20796 vector unsigned short);
20797 vector bool int vec_sll (vector bool int,
20798 vector unsigned char);
20799 vector signed short vec_sll (vector signed short,
20800 vector unsigned int);
20801 vector signed short vec_sll (vector signed short,
20802 vector unsigned short);
20803 vector signed short vec_sll (vector signed short,
20804 vector unsigned char);
20805 vector unsigned short vec_sll (vector unsigned short,
20806 vector unsigned int);
20807 vector unsigned short vec_sll (vector unsigned short,
20808 vector unsigned short);
20809 vector unsigned short vec_sll (vector unsigned short,
20810 vector unsigned char);
20811 vector bool short vec_sll (vector bool short, vector unsigned int);
20812 vector bool short vec_sll (vector bool short, vector unsigned short);
20813 vector bool short vec_sll (vector bool short, vector unsigned char);
20814 vector pixel vec_sll (vector pixel, vector unsigned int);
20815 vector pixel vec_sll (vector pixel, vector unsigned short);
20816 vector pixel vec_sll (vector pixel, vector unsigned char);
20817 vector signed char vec_sll (vector signed char, vector unsigned int);
20818 vector signed char vec_sll (vector signed char, vector unsigned short);
20819 vector signed char vec_sll (vector signed char, vector unsigned char);
20820 vector unsigned char vec_sll (vector unsigned char,
20821 vector unsigned int);
20822 vector unsigned char vec_sll (vector unsigned char,
20823 vector unsigned short);
20824 vector unsigned char vec_sll (vector unsigned char,
20825 vector unsigned char);
20826 vector bool char vec_sll (vector bool char, vector unsigned int);
20827 vector bool char vec_sll (vector bool char, vector unsigned short);
20828 vector bool char vec_sll (vector bool char, vector unsigned char);
20830 vector float vec_slo (vector float, vector signed char);
20831 vector float vec_slo (vector float, vector unsigned char);
20832 vector signed int vec_slo (vector signed int, vector signed char);
20833 vector signed int vec_slo (vector signed int, vector unsigned char);
20834 vector unsigned int vec_slo (vector unsigned int, vector signed char);
20835 vector unsigned int vec_slo (vector unsigned int, vector unsigned char);
20836 vector signed short vec_slo (vector signed short, vector signed char);
20837 vector signed short vec_slo (vector signed short, vector unsigned char);
20838 vector unsigned short vec_slo (vector unsigned short,
20839 vector signed char);
20840 vector unsigned short vec_slo (vector unsigned short,
20841 vector unsigned char);
20842 vector pixel vec_slo (vector pixel, vector signed char);
20843 vector pixel vec_slo (vector pixel, vector unsigned char);
20844 vector signed char vec_slo (vector signed char, vector signed char);
20845 vector signed char vec_slo (vector signed char, vector unsigned char);
20846 vector unsigned char vec_slo (vector unsigned char, vector signed char);
20847 vector unsigned char vec_slo (vector unsigned char,
20848 vector unsigned char);
20850 vector signed char vec_splat (vector signed char, const int);
20851 vector unsigned char vec_splat (vector unsigned char, const int);
20852 vector bool char vec_splat (vector bool char, const int);
20853 vector signed short vec_splat (vector signed short, const int);
20854 vector unsigned short vec_splat (vector unsigned short, const int);
20855 vector bool short vec_splat (vector bool short, const int);
20856 vector pixel vec_splat (vector pixel, const int);
20857 vector float vec_splat (vector float, const int);
20858 vector signed int vec_splat (vector signed int, const int);
20859 vector unsigned int vec_splat (vector unsigned int, const int);
20860 vector bool int vec_splat (vector bool int, const int);
20862 vector float vec_vspltw (vector float, const int);
20863 vector signed int vec_vspltw (vector signed int, const int);
20864 vector unsigned int vec_vspltw (vector unsigned int, const int);
20865 vector bool int vec_vspltw (vector bool int, const int);
20867 vector bool short vec_vsplth (vector bool short, const int);
20868 vector signed short vec_vsplth (vector signed short, const int);
20869 vector unsigned short vec_vsplth (vector unsigned short, const int);
20870 vector pixel vec_vsplth (vector pixel, const int);
20872 vector signed char vec_vspltb (vector signed char, const int);
20873 vector unsigned char vec_vspltb (vector unsigned char, const int);
20874 vector bool char vec_vspltb (vector bool char, const int);
20876 vector signed char vec_splat_s8 (const int);
20878 vector signed short vec_splat_s16 (const int);
20880 vector signed int vec_splat_s32 (const int);
20882 vector unsigned char vec_splat_u8 (const int);
20884 vector unsigned short vec_splat_u16 (const int);
20886 vector unsigned int vec_splat_u32 (const int);
20888 vector signed char vec_sr (vector signed char, vector unsigned char);
20889 vector unsigned char vec_sr (vector unsigned char,
20890 vector unsigned char);
20891 vector signed short vec_sr (vector signed short,
20892 vector unsigned short);
20893 vector unsigned short vec_sr (vector unsigned short,
20894 vector unsigned short);
20895 vector signed int vec_sr (vector signed int, vector unsigned int);
20896 vector unsigned int vec_sr (vector unsigned int, vector unsigned int);
20898 vector signed int vec_vsrw (vector signed int, vector unsigned int);
20899 vector unsigned int vec_vsrw (vector unsigned int, vector unsigned int);
20901 vector signed short vec_vsrh (vector signed short,
20902 vector unsigned short);
20903 vector unsigned short vec_vsrh (vector unsigned short,
20904 vector unsigned short);
20906 vector signed char vec_vsrb (vector signed char, vector unsigned char);
20907 vector unsigned char vec_vsrb (vector unsigned char,
20908 vector unsigned char);
20910 vector signed char vec_sra (vector signed char, vector unsigned char);
20911 vector unsigned char vec_sra (vector unsigned char,
20912 vector unsigned char);
20913 vector signed short vec_sra (vector signed short,
20914 vector unsigned short);
20915 vector unsigned short vec_sra (vector unsigned short,
20916 vector unsigned short);
20917 vector signed int vec_sra (vector signed int, vector unsigned int);
20918 vector unsigned int vec_sra (vector unsigned int, vector unsigned int);
20920 vector signed int vec_vsraw (vector signed int, vector unsigned int);
20921 vector unsigned int vec_vsraw (vector unsigned int,
20922 vector unsigned int);
20924 vector signed short vec_vsrah (vector signed short,
20925 vector unsigned short);
20926 vector unsigned short vec_vsrah (vector unsigned short,
20927 vector unsigned short);
20929 vector signed char vec_vsrab (vector signed char, vector unsigned char);
20930 vector unsigned char vec_vsrab (vector unsigned char,
20931 vector unsigned char);
20933 vector signed int vec_srl (vector signed int, vector unsigned int);
20934 vector signed int vec_srl (vector signed int, vector unsigned short);
20935 vector signed int vec_srl (vector signed int, vector unsigned char);
20936 vector unsigned int vec_srl (vector unsigned int, vector unsigned int);
20937 vector unsigned int vec_srl (vector unsigned int,
20938 vector unsigned short);
20939 vector unsigned int vec_srl (vector unsigned int, vector unsigned char);
20940 vector bool int vec_srl (vector bool int, vector unsigned int);
20941 vector bool int vec_srl (vector bool int, vector unsigned short);
20942 vector bool int vec_srl (vector bool int, vector unsigned char);
20943 vector signed short vec_srl (vector signed short, vector unsigned int);
20944 vector signed short vec_srl (vector signed short,
20945 vector unsigned short);
20946 vector signed short vec_srl (vector signed short, vector unsigned char);
20947 vector unsigned short vec_srl (vector unsigned short,
20948 vector unsigned int);
20949 vector unsigned short vec_srl (vector unsigned short,
20950 vector unsigned short);
20951 vector unsigned short vec_srl (vector unsigned short,
20952 vector unsigned char);
20953 vector bool short vec_srl (vector bool short, vector unsigned int);
20954 vector bool short vec_srl (vector bool short, vector unsigned short);
20955 vector bool short vec_srl (vector bool short, vector unsigned char);
20956 vector pixel vec_srl (vector pixel, vector unsigned int);
20957 vector pixel vec_srl (vector pixel, vector unsigned short);
20958 vector pixel vec_srl (vector pixel, vector unsigned char);
20959 vector signed char vec_srl (vector signed char, vector unsigned int);
20960 vector signed char vec_srl (vector signed char, vector unsigned short);
20961 vector signed char vec_srl (vector signed char, vector unsigned char);
20962 vector unsigned char vec_srl (vector unsigned char,
20963 vector unsigned int);
20964 vector unsigned char vec_srl (vector unsigned char,
20965 vector unsigned short);
20966 vector unsigned char vec_srl (vector unsigned char,
20967 vector unsigned char);
20968 vector bool char vec_srl (vector bool char, vector unsigned int);
20969 vector bool char vec_srl (vector bool char, vector unsigned short);
20970 vector bool char vec_srl (vector bool char, vector unsigned char);
20972 vector float vec_sro (vector float, vector signed char);
20973 vector float vec_sro (vector float, vector unsigned char);
20974 vector signed int vec_sro (vector signed int, vector signed char);
20975 vector signed int vec_sro (vector signed int, vector unsigned char);
20976 vector unsigned int vec_sro (vector unsigned int, vector signed char);
20977 vector unsigned int vec_sro (vector unsigned int, vector unsigned char);
20978 vector signed short vec_sro (vector signed short, vector signed char);
20979 vector signed short vec_sro (vector signed short, vector unsigned char);
20980 vector unsigned short vec_sro (vector unsigned short,
20981 vector signed char);
20982 vector unsigned short vec_sro (vector unsigned short,
20983 vector unsigned char);
20984 vector pixel vec_sro (vector pixel, vector signed char);
20985 vector pixel vec_sro (vector pixel, vector unsigned char);
20986 vector signed char vec_sro (vector signed char, vector signed char);
20987 vector signed char vec_sro (vector signed char, vector unsigned char);
20988 vector unsigned char vec_sro (vector unsigned char, vector signed char);
20989 vector unsigned char vec_sro (vector unsigned char,
20990 vector unsigned char);
20992 void vec_st (vector float, int, vector float *);
20993 void vec_st (vector float, int, float *);
20994 void vec_st (vector signed int, int, vector signed int *);
20995 void vec_st (vector signed int, int, int *);
20996 void vec_st (vector unsigned int, int, vector unsigned int *);
20997 void vec_st (vector unsigned int, int, unsigned int *);
20998 void vec_st (vector bool int, int, vector bool int *);
20999 void vec_st (vector bool int, int, unsigned int *);
21000 void vec_st (vector bool int, int, int *);
21001 void vec_st (vector signed short, int, vector signed short *);
21002 void vec_st (vector signed short, int, short *);
21003 void vec_st (vector unsigned short, int, vector unsigned short *);
21004 void vec_st (vector unsigned short, int, unsigned short *);
21005 void vec_st (vector bool short, int, vector bool short *);
21006 void vec_st (vector bool short, int, unsigned short *);
21007 void vec_st (vector pixel, int, vector pixel *);
21008 void vec_st (vector pixel, int, unsigned short *);
21009 void vec_st (vector pixel, int, short *);
21010 void vec_st (vector bool short, int, short *);
21011 void vec_st (vector signed char, int, vector signed char *);
21012 void vec_st (vector signed char, int, signed char *);
21013 void vec_st (vector unsigned char, int, vector unsigned char *);
21014 void vec_st (vector unsigned char, int, unsigned char *);
21015 void vec_st (vector bool char, int, vector bool char *);
21016 void vec_st (vector bool char, int, unsigned char *);
21017 void vec_st (vector bool char, int, signed char *);
21019 void vec_ste (vector signed char, int, signed char *);
21020 void vec_ste (vector unsigned char, int, unsigned char *);
21021 void vec_ste (vector bool char, int, signed char *);
21022 void vec_ste (vector bool char, int, unsigned char *);
21023 void vec_ste (vector signed short, int, short *);
21024 void vec_ste (vector unsigned short, int, unsigned short *);
21025 void vec_ste (vector bool short, int, short *);
21026 void vec_ste (vector bool short, int, unsigned short *);
21027 void vec_ste (vector pixel, int, short *);
21028 void vec_ste (vector pixel, int, unsigned short *);
21029 void vec_ste (vector float, int, float *);
21030 void vec_ste (vector signed int, int, int *);
21031 void vec_ste (vector unsigned int, int, unsigned int *);
21032 void vec_ste (vector bool int, int, int *);
21033 void vec_ste (vector bool int, int, unsigned int *);
21035 void vec_stvewx (vector float, int, float *);
21036 void vec_stvewx (vector signed int, int, int *);
21037 void vec_stvewx (vector unsigned int, int, unsigned int *);
21038 void vec_stvewx (vector bool int, int, int *);
21039 void vec_stvewx (vector bool int, int, unsigned int *);
21041 void vec_stvehx (vector signed short, int, short *);
21042 void vec_stvehx (vector unsigned short, int, unsigned short *);
21043 void vec_stvehx (vector bool short, int, short *);
21044 void vec_stvehx (vector bool short, int, unsigned short *);
21045 void vec_stvehx (vector pixel, int, short *);
21046 void vec_stvehx (vector pixel, int, unsigned short *);
21048 void vec_stvebx (vector signed char, int, signed char *);
21049 void vec_stvebx (vector unsigned char, int, unsigned char *);
21050 void vec_stvebx (vector bool char, int, signed char *);
21051 void vec_stvebx (vector bool char, int, unsigned char *);
21053 void vec_stl (vector float, int, vector float *);
21054 void vec_stl (vector float, int, float *);
21055 void vec_stl (vector signed int, int, vector signed int *);
21056 void vec_stl (vector signed int, int, int *);
21057 void vec_stl (vector unsigned int, int, vector unsigned int *);
21058 void vec_stl (vector unsigned int, int, unsigned int *);
21059 void vec_stl (vector bool int, int, vector bool int *);
21060 void vec_stl (vector bool int, int, unsigned int *);
21061 void vec_stl (vector bool int, int, int *);
21062 void vec_stl (vector signed short, int, vector signed short *);
21063 void vec_stl (vector signed short, int, short *);
21064 void vec_stl (vector unsigned short, int, vector unsigned short *);
21065 void vec_stl (vector unsigned short, int, unsigned short *);
21066 void vec_stl (vector bool short, int, vector bool short *);
21067 void vec_stl (vector bool short, int, unsigned short *);
21068 void vec_stl (vector bool short, int, short *);
21069 void vec_stl (vector pixel, int, vector pixel *);
21070 void vec_stl (vector pixel, int, unsigned short *);
21071 void vec_stl (vector pixel, int, short *);
21072 void vec_stl (vector signed char, int, vector signed char *);
21073 void vec_stl (vector signed char, int, signed char *);
21074 void vec_stl (vector unsigned char, int, vector unsigned char *);
21075 void vec_stl (vector unsigned char, int, unsigned char *);
21076 void vec_stl (vector bool char, int, vector bool char *);
21077 void vec_stl (vector bool char, int, unsigned char *);
21078 void vec_stl (vector bool char, int, signed char *);
21080 vector signed char vec_sub (vector bool char, vector signed char);
21081 vector signed char vec_sub (vector signed char, vector bool char);
21082 vector signed char vec_sub (vector signed char, vector signed char);
21083 vector unsigned char vec_sub (vector bool char, vector unsigned char);
21084 vector unsigned char vec_sub (vector unsigned char, vector bool char);
21085 vector unsigned char vec_sub (vector unsigned char,
21086 vector unsigned char);
21087 vector signed short vec_sub (vector bool short, vector signed short);
21088 vector signed short vec_sub (vector signed short, vector bool short);
21089 vector signed short vec_sub (vector signed short, vector signed short);
21090 vector unsigned short vec_sub (vector bool short,
21091 vector unsigned short);
21092 vector unsigned short vec_sub (vector unsigned short,
21093 vector bool short);
21094 vector unsigned short vec_sub (vector unsigned short,
21095 vector unsigned short);
21096 vector signed int vec_sub (vector bool int, vector signed int);
21097 vector signed int vec_sub (vector signed int, vector bool int);
21098 vector signed int vec_sub (vector signed int, vector signed int);
21099 vector unsigned int vec_sub (vector bool int, vector unsigned int);
21100 vector unsigned int vec_sub (vector unsigned int, vector bool int);
21101 vector unsigned int vec_sub (vector unsigned int, vector unsigned int);
21102 vector float vec_sub (vector float, vector float);
21104 vector float vec_vsubfp (vector float, vector float);
21106 vector signed int vec_vsubuwm (vector bool int, vector signed int);
21107 vector signed int vec_vsubuwm (vector signed int, vector bool int);
21108 vector signed int vec_vsubuwm (vector signed int, vector signed int);
21109 vector unsigned int vec_vsubuwm (vector bool int, vector unsigned int);
21110 vector unsigned int vec_vsubuwm (vector unsigned int, vector bool int);
21111 vector unsigned int vec_vsubuwm (vector unsigned int,
21112 vector unsigned int);
21114 vector signed short vec_vsubuhm (vector bool short,
21115 vector signed short);
21116 vector signed short vec_vsubuhm (vector signed short,
21117 vector bool short);
21118 vector signed short vec_vsubuhm (vector signed short,
21119 vector signed short);
21120 vector unsigned short vec_vsubuhm (vector bool short,
21121 vector unsigned short);
21122 vector unsigned short vec_vsubuhm (vector unsigned short,
21123 vector bool short);
21124 vector unsigned short vec_vsubuhm (vector unsigned short,
21125 vector unsigned short);
21127 vector signed char vec_vsububm (vector bool char, vector signed char);
21128 vector signed char vec_vsububm (vector signed char, vector bool char);
21129 vector signed char vec_vsububm (vector signed char, vector signed char);
21130 vector unsigned char vec_vsububm (vector bool char,
21131 vector unsigned char);
21132 vector unsigned char vec_vsububm (vector unsigned char,
21134 vector unsigned char vec_vsububm (vector unsigned char,
21135 vector unsigned char);
21137 vector unsigned int vec_subc (vector unsigned int, vector unsigned int);
21139 vector unsigned char vec_subs (vector bool char, vector unsigned char);
21140 vector unsigned char vec_subs (vector unsigned char, vector bool char);
21141 vector unsigned char vec_subs (vector unsigned char,
21142 vector unsigned char);
21143 vector signed char vec_subs (vector bool char, vector signed char);
21144 vector signed char vec_subs (vector signed char, vector bool char);
21145 vector signed char vec_subs (vector signed char, vector signed char);
21146 vector unsigned short vec_subs (vector bool short,
21147 vector unsigned short);
21148 vector unsigned short vec_subs (vector unsigned short,
21149 vector bool short);
21150 vector unsigned short vec_subs (vector unsigned short,
21151 vector unsigned short);
21152 vector signed short vec_subs (vector bool short, vector signed short);
21153 vector signed short vec_subs (vector signed short, vector bool short);
21154 vector signed short vec_subs (vector signed short, vector signed short);
21155 vector unsigned int vec_subs (vector bool int, vector unsigned int);
21156 vector unsigned int vec_subs (vector unsigned int, vector bool int);
21157 vector unsigned int vec_subs (vector unsigned int, vector unsigned int);
21158 vector signed int vec_subs (vector bool int, vector signed int);
21159 vector signed int vec_subs (vector signed int, vector bool int);
21160 vector signed int vec_subs (vector signed int, vector signed int);
21162 vector signed int vec_vsubsws (vector bool int, vector signed int);
21163 vector signed int vec_vsubsws (vector signed int, vector bool int);
21164 vector signed int vec_vsubsws (vector signed int, vector signed int);
21166 vector unsigned int vec_vsubuws (vector bool int, vector unsigned int);
21167 vector unsigned int vec_vsubuws (vector unsigned int, vector bool int);
21168 vector unsigned int vec_vsubuws (vector unsigned int,
21169 vector unsigned int);
21171 vector signed short vec_vsubshs (vector bool short,
21172 vector signed short);
21173 vector signed short vec_vsubshs (vector signed short,
21174 vector bool short);
21175 vector signed short vec_vsubshs (vector signed short,
21176 vector signed short);
21178 vector unsigned short vec_vsubuhs (vector bool short,
21179 vector unsigned short);
21180 vector unsigned short vec_vsubuhs (vector unsigned short,
21181 vector bool short);
21182 vector unsigned short vec_vsubuhs (vector unsigned short,
21183 vector unsigned short);
21185 vector signed char vec_vsubsbs (vector bool char, vector signed char);
21186 vector signed char vec_vsubsbs (vector signed char, vector bool char);
21187 vector signed char vec_vsubsbs (vector signed char, vector signed char);
21189 vector unsigned char vec_vsububs (vector bool char,
21190 vector unsigned char);
21191 vector unsigned char vec_vsububs (vector unsigned char,
21193 vector unsigned char vec_vsububs (vector unsigned char,
21194 vector unsigned char);
21196 vector unsigned int vec_sum4s (vector unsigned char,
21197 vector unsigned int);
21198 vector signed int vec_sum4s (vector signed char, vector signed int);
21199 vector signed int vec_sum4s (vector signed short, vector signed int);
21201 vector signed int vec_vsum4shs (vector signed short, vector signed int);
21203 vector signed int vec_vsum4sbs (vector signed char, vector signed int);
21205 vector unsigned int vec_vsum4ubs (vector unsigned char,
21206 vector unsigned int);
21208 vector signed int vec_sum2s (vector signed int, vector signed int);
21210 vector signed int vec_sums (vector signed int, vector signed int);
21212 vector float vec_trunc (vector float);
21214 vector signed short vec_unpackh (vector signed char);
21215 vector bool short vec_unpackh (vector bool char);
21216 vector signed int vec_unpackh (vector signed short);
21217 vector bool int vec_unpackh (vector bool short);
21218 vector unsigned int vec_unpackh (vector pixel);
21220 vector bool int vec_vupkhsh (vector bool short);
21221 vector signed int vec_vupkhsh (vector signed short);
21223 vector unsigned int vec_vupkhpx (vector pixel);
21225 vector bool short vec_vupkhsb (vector bool char);
21226 vector signed short vec_vupkhsb (vector signed char);
21228 vector signed short vec_unpackl (vector signed char);
21229 vector bool short vec_unpackl (vector bool char);
21230 vector unsigned int vec_unpackl (vector pixel);
21231 vector signed int vec_unpackl (vector signed short);
21232 vector bool int vec_unpackl (vector bool short);
21234 vector unsigned int vec_vupklpx (vector pixel);
21236 vector bool int vec_vupklsh (vector bool short);
21237 vector signed int vec_vupklsh (vector signed short);
21239 vector bool short vec_vupklsb (vector bool char);
21240 vector signed short vec_vupklsb (vector signed char);
21242 vector float vec_xor (vector float, vector float);
21243 vector float vec_xor (vector float, vector bool int);
21244 vector float vec_xor (vector bool int, vector float);
21245 vector bool int vec_xor (vector bool int, vector bool int);
21246 vector signed int vec_xor (vector bool int, vector signed int);
21247 vector signed int vec_xor (vector signed int, vector bool int);
21248 vector signed int vec_xor (vector signed int, vector signed int);
21249 vector unsigned int vec_xor (vector bool int, vector unsigned int);
21250 vector unsigned int vec_xor (vector unsigned int, vector bool int);
21251 vector unsigned int vec_xor (vector unsigned int, vector unsigned int);
21252 vector bool short vec_xor (vector bool short, vector bool short);
21253 vector signed short vec_xor (vector bool short, vector signed short);
21254 vector signed short vec_xor (vector signed short, vector bool short);
21255 vector signed short vec_xor (vector signed short, vector signed short);
21256 vector unsigned short vec_xor (vector bool short,
21257 vector unsigned short);
21258 vector unsigned short vec_xor (vector unsigned short,
21259 vector bool short);
21260 vector unsigned short vec_xor (vector unsigned short,
21261 vector unsigned short);
21262 vector signed char vec_xor (vector bool char, vector signed char);
21263 vector bool char vec_xor (vector bool char, vector bool char);
21264 vector signed char vec_xor (vector signed char, vector bool char);
21265 vector signed char vec_xor (vector signed char, vector signed char);
21266 vector unsigned char vec_xor (vector bool char, vector unsigned char);
21267 vector unsigned char vec_xor (vector unsigned char, vector bool char);
21268 vector unsigned char vec_xor (vector unsigned char,
21269 vector unsigned char);
21271 int vec_all_eq (vector signed char, vector bool char);
21272 int vec_all_eq (vector signed char, vector signed char);
21273 int vec_all_eq (vector unsigned char, vector bool char);
21274 int vec_all_eq (vector unsigned char, vector unsigned char);
21275 int vec_all_eq (vector bool char, vector bool char);
21276 int vec_all_eq (vector bool char, vector unsigned char);
21277 int vec_all_eq (vector bool char, vector signed char);
21278 int vec_all_eq (vector signed short, vector bool short);
21279 int vec_all_eq (vector signed short, vector signed short);
21280 int vec_all_eq (vector unsigned short, vector bool short);
21281 int vec_all_eq (vector unsigned short, vector unsigned short);
21282 int vec_all_eq (vector bool short, vector bool short);
21283 int vec_all_eq (vector bool short, vector unsigned short);
21284 int vec_all_eq (vector bool short, vector signed short);
21285 int vec_all_eq (vector pixel, vector pixel);
21286 int vec_all_eq (vector signed int, vector bool int);
21287 int vec_all_eq (vector signed int, vector signed int);
21288 int vec_all_eq (vector unsigned int, vector bool int);
21289 int vec_all_eq (vector unsigned int, vector unsigned int);
21290 int vec_all_eq (vector bool int, vector bool int);
21291 int vec_all_eq (vector bool int, vector unsigned int);
21292 int vec_all_eq (vector bool int, vector signed int);
21293 int vec_all_eq (vector float, vector float);
21295 int vec_all_ge (vector bool char, vector unsigned char);
21296 int vec_all_ge (vector unsigned char, vector bool char);
21297 int vec_all_ge (vector unsigned char, vector unsigned char);
21298 int vec_all_ge (vector bool char, vector signed char);
21299 int vec_all_ge (vector signed char, vector bool char);
21300 int vec_all_ge (vector signed char, vector signed char);
21301 int vec_all_ge (vector bool short, vector unsigned short);
21302 int vec_all_ge (vector unsigned short, vector bool short);
21303 int vec_all_ge (vector unsigned short, vector unsigned short);
21304 int vec_all_ge (vector signed short, vector signed short);
21305 int vec_all_ge (vector bool short, vector signed short);
21306 int vec_all_ge (vector signed short, vector bool short);
21307 int vec_all_ge (vector bool int, vector unsigned int);
21308 int vec_all_ge (vector unsigned int, vector bool int);
21309 int vec_all_ge (vector unsigned int, vector unsigned int);
21310 int vec_all_ge (vector bool int, vector signed int);
21311 int vec_all_ge (vector signed int, vector bool int);
21312 int vec_all_ge (vector signed int, vector signed int);
21313 int vec_all_ge (vector float, vector float);
21315 int vec_all_gt (vector bool char, vector unsigned char);
21316 int vec_all_gt (vector unsigned char, vector bool char);
21317 int vec_all_gt (vector unsigned char, vector unsigned char);
21318 int vec_all_gt (vector bool char, vector signed char);
21319 int vec_all_gt (vector signed char, vector bool char);
21320 int vec_all_gt (vector signed char, vector signed char);
21321 int vec_all_gt (vector bool short, vector unsigned short);
21322 int vec_all_gt (vector unsigned short, vector bool short);
21323 int vec_all_gt (vector unsigned short, vector unsigned short);
21324 int vec_all_gt (vector bool short, vector signed short);
21325 int vec_all_gt (vector signed short, vector bool short);
21326 int vec_all_gt (vector signed short, vector signed short);
21327 int vec_all_gt (vector bool int, vector unsigned int);
21328 int vec_all_gt (vector unsigned int, vector bool int);
21329 int vec_all_gt (vector unsigned int, vector unsigned int);
21330 int vec_all_gt (vector bool int, vector signed int);
21331 int vec_all_gt (vector signed int, vector bool int);
21332 int vec_all_gt (vector signed int, vector signed int);
21333 int vec_all_gt (vector float, vector float);
21335 int vec_all_in (vector float, vector float);
21337 int vec_all_le (vector bool char, vector unsigned char);
21338 int vec_all_le (vector unsigned char, vector bool char);
21339 int vec_all_le (vector unsigned char, vector unsigned char);
21340 int vec_all_le (vector bool char, vector signed char);
21341 int vec_all_le (vector signed char, vector bool char);
21342 int vec_all_le (vector signed char, vector signed char);
21343 int vec_all_le (vector bool short, vector unsigned short);
21344 int vec_all_le (vector unsigned short, vector bool short);
21345 int vec_all_le (vector unsigned short, vector unsigned short);
21346 int vec_all_le (vector bool short, vector signed short);
21347 int vec_all_le (vector signed short, vector bool short);
21348 int vec_all_le (vector signed short, vector signed short);
21349 int vec_all_le (vector bool int, vector unsigned int);
21350 int vec_all_le (vector unsigned int, vector bool int);
21351 int vec_all_le (vector unsigned int, vector unsigned int);
21352 int vec_all_le (vector bool int, vector signed int);
21353 int vec_all_le (vector signed int, vector bool int);
21354 int vec_all_le (vector signed int, vector signed int);
21355 int vec_all_le (vector float, vector float);
21357 int vec_all_lt (vector bool char, vector unsigned char);
21358 int vec_all_lt (vector unsigned char, vector bool char);
21359 int vec_all_lt (vector unsigned char, vector unsigned char);
21360 int vec_all_lt (vector bool char, vector signed char);
21361 int vec_all_lt (vector signed char, vector bool char);
21362 int vec_all_lt (vector signed char, vector signed char);
21363 int vec_all_lt (vector bool short, vector unsigned short);
21364 int vec_all_lt (vector unsigned short, vector bool short);
21365 int vec_all_lt (vector unsigned short, vector unsigned short);
21366 int vec_all_lt (vector bool short, vector signed short);
21367 int vec_all_lt (vector signed short, vector bool short);
21368 int vec_all_lt (vector signed short, vector signed short);
21369 int vec_all_lt (vector bool int, vector unsigned int);
21370 int vec_all_lt (vector unsigned int, vector bool int);
21371 int vec_all_lt (vector unsigned int, vector unsigned int);
21372 int vec_all_lt (vector bool int, vector signed int);
21373 int vec_all_lt (vector signed int, vector bool int);
21374 int vec_all_lt (vector signed int, vector signed int);
21375 int vec_all_lt (vector float, vector float);
21377 int vec_all_nan (vector float);
21379 int vec_all_ne (vector signed char, vector bool char);
21380 int vec_all_ne (vector signed char, vector signed char);
21381 int vec_all_ne (vector unsigned char, vector bool char);
21382 int vec_all_ne (vector unsigned char, vector unsigned char);
21383 int vec_all_ne (vector bool char, vector bool char);
21384 int vec_all_ne (vector bool char, vector unsigned char);
21385 int vec_all_ne (vector bool char, vector signed char);
21386 int vec_all_ne (vector signed short, vector bool short);
21387 int vec_all_ne (vector signed short, vector signed short);
21388 int vec_all_ne (vector unsigned short, vector bool short);
21389 int vec_all_ne (vector unsigned short, vector unsigned short);
21390 int vec_all_ne (vector bool short, vector bool short);
21391 int vec_all_ne (vector bool short, vector unsigned short);
21392 int vec_all_ne (vector bool short, vector signed short);
21393 int vec_all_ne (vector pixel, vector pixel);
21394 int vec_all_ne (vector signed int, vector bool int);
21395 int vec_all_ne (vector signed int, vector signed int);
21396 int vec_all_ne (vector unsigned int, vector bool int);
21397 int vec_all_ne (vector unsigned int, vector unsigned int);
21398 int vec_all_ne (vector bool int, vector bool int);
21399 int vec_all_ne (vector bool int, vector unsigned int);
21400 int vec_all_ne (vector bool int, vector signed int);
21401 int vec_all_ne (vector float, vector float);
21403 int vec_all_nge (vector float, vector float);
21405 int vec_all_ngt (vector float, vector float);
21407 int vec_all_nle (vector float, vector float);
21409 int vec_all_nlt (vector float, vector float);
21411 int vec_all_numeric (vector float);
21413 int vec_any_eq (vector signed char, vector bool char);
21414 int vec_any_eq (vector signed char, vector signed char);
21415 int vec_any_eq (vector unsigned char, vector bool char);
21416 int vec_any_eq (vector unsigned char, vector unsigned char);
21417 int vec_any_eq (vector bool char, vector bool char);
21418 int vec_any_eq (vector bool char, vector unsigned char);
21419 int vec_any_eq (vector bool char, vector signed char);
21420 int vec_any_eq (vector signed short, vector bool short);
21421 int vec_any_eq (vector signed short, vector signed short);
21422 int vec_any_eq (vector unsigned short, vector bool short);
21423 int vec_any_eq (vector unsigned short, vector unsigned short);
21424 int vec_any_eq (vector bool short, vector bool short);
21425 int vec_any_eq (vector bool short, vector unsigned short);
21426 int vec_any_eq (vector bool short, vector signed short);
21427 int vec_any_eq (vector pixel, vector pixel);
21428 int vec_any_eq (vector signed int, vector bool int);
21429 int vec_any_eq (vector signed int, vector signed int);
21430 int vec_any_eq (vector unsigned int, vector bool int);
21431 int vec_any_eq (vector unsigned int, vector unsigned int);
21432 int vec_any_eq (vector bool int, vector bool int);
21433 int vec_any_eq (vector bool int, vector unsigned int);
21434 int vec_any_eq (vector bool int, vector signed int);
21435 int vec_any_eq (vector float, vector float);
21437 int vec_any_ge (vector signed char, vector bool char);
21438 int vec_any_ge (vector unsigned char, vector bool char);
21439 int vec_any_ge (vector unsigned char, vector unsigned char);
21440 int vec_any_ge (vector signed char, vector signed char);
21441 int vec_any_ge (vector bool char, vector unsigned char);
21442 int vec_any_ge (vector bool char, vector signed char);
21443 int vec_any_ge (vector unsigned short, vector bool short);
21444 int vec_any_ge (vector unsigned short, vector unsigned short);
21445 int vec_any_ge (vector signed short, vector signed short);
21446 int vec_any_ge (vector signed short, vector bool short);
21447 int vec_any_ge (vector bool short, vector unsigned short);
21448 int vec_any_ge (vector bool short, vector signed short);
21449 int vec_any_ge (vector signed int, vector bool int);
21450 int vec_any_ge (vector unsigned int, vector bool int);
21451 int vec_any_ge (vector unsigned int, vector unsigned int);
21452 int vec_any_ge (vector signed int, vector signed int);
21453 int vec_any_ge (vector bool int, vector unsigned int);
21454 int vec_any_ge (vector bool int, vector signed int);
21455 int vec_any_ge (vector float, vector float);
21457 int vec_any_gt (vector bool char, vector unsigned char);
21458 int vec_any_gt (vector unsigned char, vector bool char);
21459 int vec_any_gt (vector unsigned char, vector unsigned char);
21460 int vec_any_gt (vector bool char, vector signed char);
21461 int vec_any_gt (vector signed char, vector bool char);
21462 int vec_any_gt (vector signed char, vector signed char);
21463 int vec_any_gt (vector bool short, vector unsigned short);
21464 int vec_any_gt (vector unsigned short, vector bool short);
21465 int vec_any_gt (vector unsigned short, vector unsigned short);
21466 int vec_any_gt (vector bool short, vector signed short);
21467 int vec_any_gt (vector signed short, vector bool short);
21468 int vec_any_gt (vector signed short, vector signed short);
21469 int vec_any_gt (vector bool int, vector unsigned int);
21470 int vec_any_gt (vector unsigned int, vector bool int);
21471 int vec_any_gt (vector unsigned int, vector unsigned int);
21472 int vec_any_gt (vector bool int, vector signed int);
21473 int vec_any_gt (vector signed int, vector bool int);
21474 int vec_any_gt (vector signed int, vector signed int);
21475 int vec_any_gt (vector float, vector float);
21477 int vec_any_le (vector bool char, vector unsigned char);
21478 int vec_any_le (vector unsigned char, vector bool char);
21479 int vec_any_le (vector unsigned char, vector unsigned char);
21480 int vec_any_le (vector bool char, vector signed char);
21481 int vec_any_le (vector signed char, vector bool char);
21482 int vec_any_le (vector signed char, vector signed char);
21483 int vec_any_le (vector bool short, vector unsigned short);
21484 int vec_any_le (vector unsigned short, vector bool short);
21485 int vec_any_le (vector unsigned short, vector unsigned short);
21486 int vec_any_le (vector bool short, vector signed short);
21487 int vec_any_le (vector signed short, vector bool short);
21488 int vec_any_le (vector signed short, vector signed short);
21489 int vec_any_le (vector bool int, vector unsigned int);
21490 int vec_any_le (vector unsigned int, vector bool int);
21491 int vec_any_le (vector unsigned int, vector unsigned int);
21492 int vec_any_le (vector bool int, vector signed int);
21493 int vec_any_le (vector signed int, vector bool int);
21494 int vec_any_le (vector signed int, vector signed int);
21495 int vec_any_le (vector float, vector float);
21497 int vec_any_lt (vector bool char, vector unsigned char);
21498 int vec_any_lt (vector unsigned char, vector bool char);
21499 int vec_any_lt (vector unsigned char, vector unsigned char);
21500 int vec_any_lt (vector bool char, vector signed char);
21501 int vec_any_lt (vector signed char, vector bool char);
21502 int vec_any_lt (vector signed char, vector signed char);
21503 int vec_any_lt (vector bool short, vector unsigned short);
21504 int vec_any_lt (vector unsigned short, vector bool short);
21505 int vec_any_lt (vector unsigned short, vector unsigned short);
21506 int vec_any_lt (vector bool short, vector signed short);
21507 int vec_any_lt (vector signed short, vector bool short);
21508 int vec_any_lt (vector signed short, vector signed short);
21509 int vec_any_lt (vector bool int, vector unsigned int);
21510 int vec_any_lt (vector unsigned int, vector bool int);
21511 int vec_any_lt (vector unsigned int, vector unsigned int);
21512 int vec_any_lt (vector bool int, vector signed int);
21513 int vec_any_lt (vector signed int, vector bool int);
21514 int vec_any_lt (vector signed int, vector signed int);
21515 int vec_any_lt (vector float, vector float);
21517 int vec_any_nan (vector float);
21519 int vec_any_ne (vector signed char, vector bool char);
21520 int vec_any_ne (vector signed char, vector signed char);
21521 int vec_any_ne (vector unsigned char, vector bool char);
21522 int vec_any_ne (vector unsigned char, vector unsigned char);
21523 int vec_any_ne (vector bool char, vector bool char);
21524 int vec_any_ne (vector bool char, vector unsigned char);
21525 int vec_any_ne (vector bool char, vector signed char);
21526 int vec_any_ne (vector signed short, vector bool short);
21527 int vec_any_ne (vector signed short, vector signed short);
21528 int vec_any_ne (vector unsigned short, vector bool short);
21529 int vec_any_ne (vector unsigned short, vector unsigned short);
21530 int vec_any_ne (vector bool short, vector bool short);
21531 int vec_any_ne (vector bool short, vector unsigned short);
21532 int vec_any_ne (vector bool short, vector signed short);
21533 int vec_any_ne (vector pixel, vector pixel);
21534 int vec_any_ne (vector signed int, vector bool int);
21535 int vec_any_ne (vector signed int, vector signed int);
21536 int vec_any_ne (vector unsigned int, vector bool int);
21537 int vec_any_ne (vector unsigned int, vector unsigned int);
21538 int vec_any_ne (vector bool int, vector bool int);
21539 int vec_any_ne (vector bool int, vector unsigned int);
21540 int vec_any_ne (vector bool int, vector signed int);
21541 int vec_any_ne (vector float, vector float);
21543 int vec_any_nge (vector float, vector float);
21545 int vec_any_ngt (vector float, vector float);
21547 int vec_any_nle (vector float, vector float);
21549 int vec_any_nlt (vector float, vector float);
21551 int vec_any_numeric (vector float);
21553 int vec_any_out (vector float, vector float);
21556 File: gcc.info, Node: SPARC VIS Built-in Functions, Prev: PowerPC AltiVec Built-in Functions, Up: Target Builtins
21558 5.45.7 SPARC VIS Built-in Functions
21559 -----------------------------------
21561 GCC supports SIMD operations on the SPARC using both the generic vector
21562 extensions (*note Vector Extensions::) as well as built-in functions for
21563 the SPARC Visual Instruction Set (VIS). When you use the `-mvis'
21564 switch, the VIS extension is exposed as the following built-in
21567 typedef int v2si __attribute__ ((vector_size (8)));
21568 typedef short v4hi __attribute__ ((vector_size (8)));
21569 typedef short v2hi __attribute__ ((vector_size (4)));
21570 typedef char v8qi __attribute__ ((vector_size (8)));
21571 typedef char v4qi __attribute__ ((vector_size (4)));
21573 void * __builtin_vis_alignaddr (void *, long);
21574 int64_t __builtin_vis_faligndatadi (int64_t, int64_t);
21575 v2si __builtin_vis_faligndatav2si (v2si, v2si);
21576 v4hi __builtin_vis_faligndatav4hi (v4si, v4si);
21577 v8qi __builtin_vis_faligndatav8qi (v8qi, v8qi);
21579 v4hi __builtin_vis_fexpand (v4qi);
21581 v4hi __builtin_vis_fmul8x16 (v4qi, v4hi);
21582 v4hi __builtin_vis_fmul8x16au (v4qi, v4hi);
21583 v4hi __builtin_vis_fmul8x16al (v4qi, v4hi);
21584 v4hi __builtin_vis_fmul8sux16 (v8qi, v4hi);
21585 v4hi __builtin_vis_fmul8ulx16 (v8qi, v4hi);
21586 v2si __builtin_vis_fmuld8sux16 (v4qi, v2hi);
21587 v2si __builtin_vis_fmuld8ulx16 (v4qi, v2hi);
21589 v4qi __builtin_vis_fpack16 (v4hi);
21590 v8qi __builtin_vis_fpack32 (v2si, v2si);
21591 v2hi __builtin_vis_fpackfix (v2si);
21592 v8qi __builtin_vis_fpmerge (v4qi, v4qi);
21594 int64_t __builtin_vis_pdist (v8qi, v8qi, int64_t);
21597 File: gcc.info, Node: Target Format Checks, Next: Pragmas, Prev: Target Builtins, Up: C Extensions
21599 5.46 Format Checks Specific to Particular Target Machines
21600 =========================================================
21602 For some target machines, GCC supports additional options to the format
21603 attribute (*note Declaring Attributes of Functions: Function
21608 * Solaris Format Checks::
21611 File: gcc.info, Node: Solaris Format Checks, Up: Target Format Checks
21613 5.46.1 Solaris Format Checks
21614 ----------------------------
21616 Solaris targets support the `cmn_err' (or `__cmn_err__') format check.
21617 `cmn_err' accepts a subset of the standard `printf' conversions, and
21618 the two-argument `%b' conversion for displaying bit-fields. See the
21619 Solaris man page for `cmn_err' for more information.
21622 File: gcc.info, Node: Pragmas, Next: Unnamed Fields, Prev: Target Format Checks, Up: C Extensions
21624 5.47 Pragmas Accepted by GCC
21625 ============================
21627 GCC supports several types of pragmas, primarily in order to compile
21628 code originally written for other compilers. Note that in general we
21629 do not recommend the use of pragmas; *Note Function Attributes::, for
21630 further explanation.
21635 * RS/6000 and PowerPC Pragmas::
21637 * Solaris Pragmas::
21638 * Symbol-Renaming Pragmas::
21639 * Structure-Packing Pragmas::
21643 File: gcc.info, Node: ARM Pragmas, Next: RS/6000 and PowerPC Pragmas, Up: Pragmas
21648 The ARM target defines pragmas for controlling the default addition of
21649 `long_call' and `short_call' attributes to functions. *Note Function
21650 Attributes::, for information about the effects of these attributes.
21653 Set all subsequent functions to have the `long_call' attribute.
21656 Set all subsequent functions to have the `short_call' attribute.
21659 Do not affect the `long_call' or `short_call' attributes of
21660 subsequent functions.
21663 File: gcc.info, Node: RS/6000 and PowerPC Pragmas, Next: Darwin Pragmas, Prev: ARM Pragmas, Up: Pragmas
21665 5.47.2 RS/6000 and PowerPC Pragmas
21666 ----------------------------------
21668 The RS/6000 and PowerPC targets define one pragma for controlling
21669 whether or not the `longcall' attribute is added to function
21670 declarations by default. This pragma overrides the `-mlongcall'
21671 option, but not the `longcall' and `shortcall' attributes. *Note
21672 RS/6000 and PowerPC Options::, for more information about when long
21673 calls are and are not necessary.
21676 Apply the `longcall' attribute to all subsequent function
21680 Do not apply the `longcall' attribute to subsequent function
21684 File: gcc.info, Node: Darwin Pragmas, Next: Solaris Pragmas, Prev: RS/6000 and PowerPC Pragmas, Up: Pragmas
21686 5.47.3 Darwin Pragmas
21687 ---------------------
21689 The following pragmas are available for all architectures running the
21690 Darwin operating system. These are useful for compatibility with other
21694 This pragma is accepted, but has no effect.
21696 `options align=ALIGNMENT'
21697 This pragma sets the alignment of fields in structures. The
21698 values of ALIGNMENT may be `mac68k', to emulate m68k alignment, or
21699 `power', to emulate PowerPC alignment. Uses of this pragma nest
21700 properly; to restore the previous setting, use `reset' for the
21703 `segment TOKENS...'
21704 This pragma is accepted, but has no effect.
21706 `unused (VAR [, VAR]...)'
21707 This pragma declares variables to be possibly unused. GCC will not
21708 produce warnings for the listed variables. The effect is similar
21709 to that of the `unused' attribute, except that this pragma may
21710 appear anywhere within the variables' scopes.
21713 File: gcc.info, Node: Solaris Pragmas, Next: Symbol-Renaming Pragmas, Prev: Darwin Pragmas, Up: Pragmas
21715 5.47.4 Solaris Pragmas
21716 ----------------------
21718 The Solaris target supports `#pragma redefine_extname' (*note
21719 Symbol-Renaming Pragmas::). It also supports additional `#pragma'
21720 directives for compatibility with the system compiler.
21722 `align ALIGNMENT (VARIABLE [, VARIABLE]...)'
21723 Increase the minimum alignment of each VARIABLE to ALIGNMENT.
21724 This is the same as GCC's `aligned' attribute *note Variable
21725 Attributes::). Macro expansion occurs on the arguments to this
21726 pragma when compiling C and Objective-C. It does not currently
21727 occur when compiling C++, but this is a bug which may be fixed in
21730 `fini (FUNCTION [, FUNCTION]...)'
21731 This pragma causes each listed FUNCTION to be called after main,
21732 or during shared module unloading, by adding a call to the `.fini'
21735 `init (FUNCTION [, FUNCTION]...)'
21736 This pragma causes each listed FUNCTION to be called during
21737 initialization (before `main') or during shared module loading, by
21738 adding a call to the `.init' section.
21742 File: gcc.info, Node: Symbol-Renaming Pragmas, Next: Structure-Packing Pragmas, Prev: Solaris Pragmas, Up: Pragmas
21744 5.47.5 Symbol-Renaming Pragmas
21745 ------------------------------
21747 For compatibility with the Solaris and Tru64 UNIX system headers, GCC
21748 supports two `#pragma' directives which change the name used in
21749 assembly for a given declaration. These pragmas are only available on
21750 platforms whose system headers need them. To get this effect on all
21751 platforms supported by GCC, use the asm labels extension (*note Asm
21754 `redefine_extname OLDNAME NEWNAME'
21755 This pragma gives the C function OLDNAME the assembly symbol
21756 NEWNAME. The preprocessor macro `__PRAGMA_REDEFINE_EXTNAME' will
21757 be defined if this pragma is available (currently only on Solaris).
21759 `extern_prefix STRING'
21760 This pragma causes all subsequent external function and variable
21761 declarations to have STRING prepended to their assembly symbols.
21762 This effect may be terminated with another `extern_prefix' pragma
21763 whose argument is an empty string. The preprocessor macro
21764 `__PRAGMA_EXTERN_PREFIX' will be defined if this pragma is
21765 available (currently only on Tru64 UNIX).
21767 These pragmas and the asm labels extension interact in a complicated
21768 manner. Here are some corner cases you may want to be aware of.
21770 1. Both pragmas silently apply only to declarations with external
21771 linkage. Asm labels do not have this restriction.
21773 2. In C++, both pragmas silently apply only to declarations with "C"
21774 linkage. Again, asm labels do not have this restriction.
21776 3. If any of the three ways of changing the assembly name of a
21777 declaration is applied to a declaration whose assembly name has
21778 already been determined (either by a previous use of one of these
21779 features, or because the compiler needed the assembly name in
21780 order to generate code), and the new name is different, a warning
21781 issues and the name does not change.
21783 4. The OLDNAME used by `#pragma redefine_extname' is always the
21786 5. If `#pragma extern_prefix' is in effect, and a declaration occurs
21787 with an asm label attached, the prefix is silently ignored for
21790 6. If `#pragma extern_prefix' and `#pragma redefine_extname' apply to
21791 the same declaration, whichever triggered first wins, and a
21792 warning issues if they contradict each other. (We would like to
21793 have `#pragma redefine_extname' always win, for consistency with
21794 asm labels, but if `#pragma extern_prefix' triggers first we have
21795 no way of knowing that that happened.)
21798 File: gcc.info, Node: Structure-Packing Pragmas, Next: Weak Pragmas, Prev: Symbol-Renaming Pragmas, Up: Pragmas
21800 5.47.6 Structure-Packing Pragmas
21801 --------------------------------
21803 For compatibility with Win32, GCC supports a set of `#pragma'
21804 directives which change the maximum alignment of members of structures,
21805 unions, and classes subsequently defined. The N value below always is
21806 required to be a small power of two and specifies the new alignment in
21809 1. `#pragma pack(N)' simply sets the new alignment.
21811 2. `#pragma pack()' sets the alignment to the one that was in effect
21812 when compilation started (see also command line option
21813 `-fpack-struct[=<n>]' *note Code Gen Options::).
21815 3. `#pragma pack(push[,N])' pushes the current alignment setting on
21816 an internal stack and then optionally sets the new alignment.
21818 4. `#pragma pack(pop)' restores the alignment setting to the one
21819 saved at the top of the internal stack (and removes that stack
21820 entry). Note that `#pragma pack([N])' does not influence this
21821 internal stack; thus it is possible to have `#pragma pack(push)'
21822 followed by multiple `#pragma pack(N)' instances and finalized by
21823 a single `#pragma pack(pop)'.
21826 File: gcc.info, Node: Weak Pragmas, Prev: Structure-Packing Pragmas, Up: Pragmas
21828 5.47.7 Weak Pragmas
21829 -------------------
21831 For compatibility with SVR4, GCC supports a set of `#pragma' directives
21832 for declaring symbols to be weak, and defining weak aliases.
21834 `#pragma weak SYMBOL'
21835 This pragma declares SYMBOL to be weak, as if the declaration had
21836 the attribute of the same name. The pragma may appear before or
21837 after the declaration of SYMBOL, but must appear before either its
21838 first use or its definition. It is not an error for SYMBOL to
21839 never be defined at all.
21841 `#pragma weak SYMBOL1 = SYMBOL2'
21842 This pragma declares SYMBOL1 to be a weak alias of SYMBOL2. It is
21843 an error if SYMBOL2 is not defined in the current translation unit.
21846 File: gcc.info, Node: Unnamed Fields, Next: Thread-Local, Prev: Pragmas, Up: C Extensions
21848 5.48 Unnamed struct/union fields within structs/unions
21849 ======================================================
21851 For compatibility with other compilers, GCC allows you to define a
21852 structure or union that contains, as fields, structures and unions
21853 without names. For example:
21864 In this example, the user would be able to access members of the
21865 unnamed union with code like `foo.b'. Note that only unnamed structs
21866 and unions are allowed, you may not have, for example, an unnamed `int'.
21868 You must never create such structures that cause ambiguous field
21869 definitions. For example, this structure:
21878 It is ambiguous which `a' is being referred to with `foo.a'. Such
21879 constructs are not supported and must be avoided. In the future, such
21880 constructs may be detected and treated as compilation errors.
21882 Unless `-fms-extensions' is used, the unnamed field must be a
21883 structure or union definition without a tag (for example, `struct { int
21884 a; };'). If `-fms-extensions' is used, the field may also be a
21885 definition with a tag such as `struct foo { int a; };', a reference to
21886 a previously defined structure or union such as `struct foo;', or a
21887 reference to a `typedef' name for a previously defined structure or
21891 File: gcc.info, Node: Thread-Local, Prev: Unnamed Fields, Up: C Extensions
21893 5.49 Thread-Local Storage
21894 =========================
21896 Thread-local storage (TLS) is a mechanism by which variables are
21897 allocated such that there is one instance of the variable per extant
21898 thread. The run-time model GCC uses to implement this originates in
21899 the IA-64 processor-specific ABI, but has since been migrated to other
21900 processors as well. It requires significant support from the linker
21901 (`ld'), dynamic linker (`ld.so'), and system libraries (`libc.so' and
21902 `libpthread.so'), so it is not available everywhere.
21904 At the user level, the extension is visible with a new storage class
21905 keyword: `__thread'. For example:
21908 extern __thread struct state s;
21909 static __thread char *p;
21911 The `__thread' specifier may be used alone, with the `extern' or
21912 `static' specifiers, but with no other storage class specifier. When
21913 used with `extern' or `static', `__thread' must appear immediately
21914 after the other storage class specifier.
21916 The `__thread' specifier may be applied to any global, file-scoped
21917 static, function-scoped static, or static data member of a class. It
21918 may not be applied to block-scoped automatic or non-static data member.
21920 When the address-of operator is applied to a thread-local variable, it
21921 is evaluated at run-time and returns the address of the current thread's
21922 instance of that variable. An address so obtained may be used by any
21923 thread. When a thread terminates, any pointers to thread-local
21924 variables in that thread become invalid.
21926 No static initialization may refer to the address of a thread-local
21929 In C++, if an initializer is present for a thread-local variable, it
21930 must be a CONSTANT-EXPRESSION, as defined in 5.19.2 of the ANSI/ISO C++
21933 See ELF Handling For Thread-Local Storage
21934 (http://people.redhat.com/drepper/tls.pdf) for a detailed explanation of
21935 the four thread-local storage addressing models, and how the run-time
21936 is expected to function.
21940 * C99 Thread-Local Edits::
21941 * C++98 Thread-Local Edits::
21944 File: gcc.info, Node: C99 Thread-Local Edits, Next: C++98 Thread-Local Edits, Up: Thread-Local
21946 5.49.1 ISO/IEC 9899:1999 Edits for Thread-Local Storage
21947 -------------------------------------------------------
21949 The following are a set of changes to ISO/IEC 9899:1999 (aka C99) that
21950 document the exact semantics of the language extension.
21952 * `5.1.2 Execution environments'
21954 Add new text after paragraph 1
21956 Within either execution environment, a "thread" is a flow of
21957 control within a program. It is implementation defined
21958 whether or not there may be more than one thread associated
21959 with a program. It is implementation defined how threads
21960 beyond the first are created, the name and type of the
21961 function called at thread startup, and how threads may be
21962 terminated. However, objects with thread storage duration
21963 shall be initialized before thread startup.
21965 * `6.2.4 Storage durations of objects'
21967 Add new text before paragraph 3
21969 An object whose identifier is declared with the storage-class
21970 specifier `__thread' has "thread storage duration". Its
21971 lifetime is the entire execution of the thread, and its
21972 stored value is initialized only once, prior to thread
21979 * `6.7.1 Storage-class specifiers'
21981 Add `__thread' to the list of storage class specifiers in
21984 Change paragraph 2 to
21986 With the exception of `__thread', at most one storage-class
21987 specifier may be given [...]. The `__thread' specifier may
21988 be used alone, or immediately following `extern' or `static'.
21990 Add new text after paragraph 6
21992 The declaration of an identifier for a variable that has
21993 block scope that specifies `__thread' shall also specify
21994 either `extern' or `static'.
21996 The `__thread' specifier shall be used only with variables.
21999 File: gcc.info, Node: C++98 Thread-Local Edits, Prev: C99 Thread-Local Edits, Up: Thread-Local
22001 5.49.2 ISO/IEC 14882:1998 Edits for Thread-Local Storage
22002 --------------------------------------------------------
22004 The following are a set of changes to ISO/IEC 14882:1998 (aka C++98)
22005 that document the exact semantics of the language extension.
22007 * [intro.execution]
22009 New text after paragraph 4
22011 A "thread" is a flow of control within the abstract machine.
22012 It is implementation defined whether or not there may be more
22015 New text after paragraph 7
22017 It is unspecified whether additional action must be taken to
22018 ensure when and whether side effects are visible to other
22025 * [basic.start.main]
22027 Add after paragraph 5
22029 The thread that begins execution at the `main' function is
22030 called the "main thread". It is implementation defined how
22031 functions beginning threads other than the main thread are
22032 designated or typed. A function so designated, as well as
22033 the `main' function, is called a "thread startup function".
22034 It is implementation defined what happens if a thread startup
22035 function returns. It is implementation defined what happens
22036 to other threads when any thread calls `exit'.
22038 * [basic.start.init]
22040 Add after paragraph 4
22042 The storage for an object of thread storage duration shall be
22043 statically initialized before the first statement of the
22044 thread startup function. An object of thread storage
22045 duration shall not require dynamic initialization.
22047 * [basic.start.term]
22049 Add after paragraph 3
22051 The type of an object with thread storage duration shall not
22052 have a non-trivial destructor, nor shall it be an array type
22053 whose elements (directly or indirectly) have non-trivial
22058 Add "thread storage duration" to the list in paragraph 1.
22062 Thread, static, and automatic storage durations are
22063 associated with objects introduced by declarations [...].
22065 Add `__thread' to the list of specifiers in paragraph 3.
22067 * [basic.stc.thread]
22069 New section before [basic.stc.static]
22071 The keyword `__thread' applied to a non-local object gives the
22072 object thread storage duration.
22074 A local variable or class data member declared both `static'
22075 and `__thread' gives the variable or member thread storage
22078 * [basic.stc.static]
22082 All objects which have neither thread storage duration,
22083 dynamic storage duration nor are local [...].
22087 Add `__thread' to the list in paragraph 1.
22091 With the exception of `__thread', at most one
22092 STORAGE-CLASS-SPECIFIER shall appear in a given
22093 DECL-SPECIFIER-SEQ. The `__thread' specifier may be used
22094 alone, or immediately following the `extern' or `static'
22097 Add after paragraph 5
22099 The `__thread' specifier can be applied only to the names of
22100 objects and to anonymous unions.
22104 Add after paragraph 6
22106 Non-`static' members shall not be `__thread'.
22109 File: gcc.info, Node: C++ Extensions, Next: Objective-C, Prev: C Extensions, Up: Top
22111 6 Extensions to the C++ Language
22112 ********************************
22114 The GNU compiler provides these extensions to the C++ language (and you
22115 can also use most of the C language extensions in your C++ programs).
22116 If you want to write code that checks whether these features are
22117 available, you can test for the GNU compiler the same way as for C
22118 programs: check for a predefined macro `__GNUC__'. You can also use
22119 `__GNUG__' to test specifically for GNU C++ (*note Predefined Macros:
22120 (cpp)Common Predefined Macros.).
22124 * Volatiles:: What constitutes an access to a volatile object.
22125 * Restricted Pointers:: C99 restricted pointers and references.
22126 * Vague Linkage:: Where G++ puts inlines, vtables and such.
22127 * C++ Interface:: You can use a single C++ header file for both
22128 declarations and definitions.
22129 * Template Instantiation:: Methods for ensuring that exactly one copy of
22130 each needed template instantiation is emitted.
22131 * Bound member functions:: You can extract a function pointer to the
22132 method denoted by a `->*' or `.*' expression.
22133 * C++ Attributes:: Variable, function, and type attributes for C++ only.
22134 * Strong Using:: Strong using-directives for namespace composition.
22135 * Java Exceptions:: Tweaking exception handling to work with Java.
22136 * Deprecated Features:: Things will disappear from g++.
22137 * Backwards Compatibility:: Compatibilities with earlier definitions of C++.
22140 File: gcc.info, Node: Volatiles, Next: Restricted Pointers, Up: C++ Extensions
22142 6.1 When is a Volatile Object Accessed?
22143 =======================================
22145 Both the C and C++ standard have the concept of volatile objects. These
22146 are normally accessed by pointers and used for accessing hardware. The
22147 standards encourage compilers to refrain from optimizations concerning
22148 accesses to volatile objects that it might perform on non-volatile
22149 objects. The C standard leaves it implementation defined as to what
22150 constitutes a volatile access. The C++ standard omits to specify this,
22151 except to say that C++ should behave in a similar manner to C with
22152 respect to volatiles, where possible. The minimum either standard
22153 specifies is that at a sequence point all previous accesses to volatile
22154 objects have stabilized and no subsequent accesses have occurred. Thus
22155 an implementation is free to reorder and combine volatile accesses
22156 which occur between sequence points, but cannot do so for accesses
22157 across a sequence point. The use of volatiles does not allow you to
22158 violate the restriction on updating objects multiple times within a
22161 In most expressions, it is intuitively obvious what is a read and what
22162 is a write. For instance
22164 volatile int *dst = SOMEVALUE;
22165 volatile int *src = SOMEOTHERVALUE;
22168 will cause a read of the volatile object pointed to by SRC and stores
22169 the value into the volatile object pointed to by DST. There is no
22170 guarantee that these reads and writes are atomic, especially for objects
22173 Less obvious expressions are where something which looks like an access
22174 is used in a void context. An example would be,
22176 volatile int *src = SOMEVALUE;
22179 With C, such expressions are rvalues, and as rvalues cause a read of
22180 the object, GCC interprets this as a read of the volatile being pointed
22181 to. The C++ standard specifies that such expressions do not undergo
22182 lvalue to rvalue conversion, and that the type of the dereferenced
22183 object may be incomplete. The C++ standard does not specify explicitly
22184 that it is this lvalue to rvalue conversion which is responsible for
22185 causing an access. However, there is reason to believe that it is,
22186 because otherwise certain simple expressions become undefined. However,
22187 because it would surprise most programmers, G++ treats dereferencing a
22188 pointer to volatile object of complete type in a void context as a read
22189 of the object. When the object has incomplete type, G++ issues a
22194 volatile S *ptr1 = SOMEVALUE;
22195 volatile T *ptr2 = SOMEVALUE;
22199 In this example, a warning is issued for `*ptr1', and `*ptr2' causes a
22200 read of the object pointed to. If you wish to force an error on the
22201 first case, you must force a conversion to rvalue with, for instance a
22202 static cast, `static_cast<S>(*ptr1)'.
22204 When using a reference to volatile, G++ does not treat equivalent
22205 expressions as accesses to volatiles, but instead issues a warning that
22206 no volatile is accessed. The rationale for this is that otherwise it
22207 becomes difficult to determine where volatile access occur, and not
22208 possible to ignore the return value from functions returning volatile
22209 references. Again, if you wish to force a read, cast the reference to
22213 File: gcc.info, Node: Restricted Pointers, Next: Vague Linkage, Prev: Volatiles, Up: C++ Extensions
22215 6.2 Restricting Pointer Aliasing
22216 ================================
22218 As with the C front end, G++ understands the C99 feature of restricted
22219 pointers, specified with the `__restrict__', or `__restrict' type
22220 qualifier. Because you cannot compile C++ by specifying the `-std=c99'
22221 language flag, `restrict' is not a keyword in C++.
22223 In addition to allowing restricted pointers, you can specify restricted
22224 references, which indicate that the reference is not aliased in the
22227 void fn (int *__restrict__ rptr, int &__restrict__ rref)
22232 In the body of `fn', RPTR points to an unaliased integer and RREF
22233 refers to a (different) unaliased integer.
22235 You may also specify whether a member function's THIS pointer is
22236 unaliased by using `__restrict__' as a member function qualifier.
22238 void T::fn () __restrict__
22243 Within the body of `T::fn', THIS will have the effective definition `T
22244 *__restrict__ const this'. Notice that the interpretation of a
22245 `__restrict__' member function qualifier is different to that of
22246 `const' or `volatile' qualifier, in that it is applied to the pointer
22247 rather than the object. This is consistent with other compilers which
22248 implement restricted pointers.
22250 As with all outermost parameter qualifiers, `__restrict__' is ignored
22251 in function definition matching. This means you only need to specify
22252 `__restrict__' in a function definition, rather than in a function
22256 File: gcc.info, Node: Vague Linkage, Next: C++ Interface, Prev: Restricted Pointers, Up: C++ Extensions
22261 There are several constructs in C++ which require space in the object
22262 file but are not clearly tied to a single translation unit. We say that
22263 these constructs have "vague linkage". Typically such constructs are
22264 emitted wherever they are needed, though sometimes we can be more
22268 Inline functions are typically defined in a header file which can
22269 be included in many different compilations. Hopefully they can
22270 usually be inlined, but sometimes an out-of-line copy is
22271 necessary, if the address of the function is taken or if inlining
22272 fails. In general, we emit an out-of-line copy in all translation
22273 units where one is needed. As an exception, we only emit inline
22274 virtual functions with the vtable, since it will always require a
22277 Local static variables and string constants used in an inline
22278 function are also considered to have vague linkage, since they
22279 must be shared between all inlined and out-of-line instances of
22283 C++ virtual functions are implemented in most compilers using a
22284 lookup table, known as a vtable. The vtable contains pointers to
22285 the virtual functions provided by a class, and each object of the
22286 class contains a pointer to its vtable (or vtables, in some
22287 multiple-inheritance situations). If the class declares any
22288 non-inline, non-pure virtual functions, the first one is chosen as
22289 the "key method" for the class, and the vtable is only emitted in
22290 the translation unit where the key method is defined.
22292 _Note:_ If the chosen key method is later defined as inline, the
22293 vtable will still be emitted in every translation unit which
22294 defines it. Make sure that any inline virtuals are declared
22295 inline in the class body, even if they are not defined there.
22298 C++ requires information about types to be written out in order to
22299 implement `dynamic_cast', `typeid' and exception handling. For
22300 polymorphic classes (classes with virtual functions), the type_info
22301 object is written out along with the vtable so that `dynamic_cast'
22302 can determine the dynamic type of a class object at runtime. For
22303 all other types, we write out the type_info object when it is
22304 used: when applying `typeid' to an expression, throwing an object,
22305 or referring to a type in a catch clause or exception
22308 Template Instantiations
22309 Most everything in this section also applies to template
22310 instantiations, but there are other options as well. *Note
22311 Where's the Template?: Template Instantiation.
22314 When used with GNU ld version 2.8 or later on an ELF system such as
22315 GNU/Linux or Solaris 2, or on Microsoft Windows, duplicate copies of
22316 these constructs will be discarded at link time. This is known as
22319 On targets that don't support COMDAT, but do support weak symbols, GCC
22320 will use them. This way one copy will override all the others, but the
22321 unused copies will still take up space in the executable.
22323 For targets which do not support either COMDAT or weak symbols, most
22324 entities with vague linkage will be emitted as local symbols to avoid
22325 duplicate definition errors from the linker. This will not happen for
22326 local statics in inlines, however, as having multiple copies will
22327 almost certainly break things.
22329 *Note Declarations and Definitions in One Header: C++ Interface, for
22330 another way to control placement of these constructs.
22333 File: gcc.info, Node: C++ Interface, Next: Template Instantiation, Prev: Vague Linkage, Up: C++ Extensions
22335 6.4 #pragma interface and implementation
22336 ========================================
22338 `#pragma interface' and `#pragma implementation' provide the user with
22339 a way of explicitly directing the compiler to emit entities with vague
22340 linkage (and debugging information) in a particular translation unit.
22342 _Note:_ As of GCC 2.7.2, these `#pragma's are not useful in most
22343 cases, because of COMDAT support and the "key method" heuristic
22344 mentioned in *Note Vague Linkage::. Using them can actually cause your
22345 program to grow due to unnecessary out-of-line copies of inline
22346 functions. Currently (3.4) the only benefit of these `#pragma's is
22347 reduced duplication of debugging information, and that should be
22348 addressed soon on DWARF 2 targets with the use of COMDAT groups.
22350 `#pragma interface'
22351 `#pragma interface "SUBDIR/OBJECTS.h"'
22352 Use this directive in _header files_ that define object classes,
22353 to save space in most of the object files that use those classes.
22354 Normally, local copies of certain information (backup copies of
22355 inline member functions, debugging information, and the internal
22356 tables that implement virtual functions) must be kept in each
22357 object file that includes class definitions. You can use this
22358 pragma to avoid such duplication. When a header file containing
22359 `#pragma interface' is included in a compilation, this auxiliary
22360 information will not be generated (unless the main input source
22361 file itself uses `#pragma implementation'). Instead, the object
22362 files will contain references to be resolved at link time.
22364 The second form of this directive is useful for the case where you
22365 have multiple headers with the same name in different directories.
22366 If you use this form, you must specify the same string to `#pragma
22369 `#pragma implementation'
22370 `#pragma implementation "OBJECTS.h"'
22371 Use this pragma in a _main input file_, when you want full output
22372 from included header files to be generated (and made globally
22373 visible). The included header file, in turn, should use `#pragma
22374 interface'. Backup copies of inline member functions, debugging
22375 information, and the internal tables used to implement virtual
22376 functions are all generated in implementation files.
22378 If you use `#pragma implementation' with no argument, it applies to
22379 an include file with the same basename(1) as your source file.
22380 For example, in `allclass.cc', giving just `#pragma implementation'
22381 by itself is equivalent to `#pragma implementation "allclass.h"'.
22383 In versions of GNU C++ prior to 2.6.0 `allclass.h' was treated as
22384 an implementation file whenever you would include it from
22385 `allclass.cc' even if you never specified `#pragma
22386 implementation'. This was deemed to be more trouble than it was
22387 worth, however, and disabled.
22389 Use the string argument if you want a single implementation file to
22390 include code from multiple header files. (You must also use
22391 `#include' to include the header file; `#pragma implementation'
22392 only specifies how to use the file--it doesn't actually include
22395 There is no way to split up the contents of a single header file
22396 into multiple implementation files.
22398 `#pragma implementation' and `#pragma interface' also have an effect
22399 on function inlining.
22401 If you define a class in a header file marked with `#pragma
22402 interface', the effect on an inline function defined in that class is
22403 similar to an explicit `extern' declaration--the compiler emits no code
22404 at all to define an independent version of the function. Its
22405 definition is used only for inlining with its callers.
22407 Conversely, when you include the same header file in a main source file
22408 that declares it as `#pragma implementation', the compiler emits code
22409 for the function itself; this defines a version of the function that
22410 can be found via pointers (or by callers compiled without inlining).
22411 If all calls to the function can be inlined, you can avoid emitting the
22412 function by compiling with `-fno-implement-inlines'. If any calls were
22413 not inlined, you will get linker errors.
22415 ---------- Footnotes ----------
22417 (1) A file's "basename" was the name stripped of all leading path
22418 information and of trailing suffixes, such as `.h' or `.C' or `.cc'.
22421 File: gcc.info, Node: Template Instantiation, Next: Bound member functions, Prev: C++ Interface, Up: C++ Extensions
22423 6.5 Where's the Template?
22424 =========================
22426 C++ templates are the first language feature to require more
22427 intelligence from the environment than one usually finds on a UNIX
22428 system. Somehow the compiler and linker have to make sure that each
22429 template instance occurs exactly once in the executable if it is needed,
22430 and not at all otherwise. There are two basic approaches to this
22431 problem, which are referred to as the Borland model and the Cfront
22435 Borland C++ solved the template instantiation problem by adding
22436 the code equivalent of common blocks to their linker; the compiler
22437 emits template instances in each translation unit that uses them,
22438 and the linker collapses them together. The advantage of this
22439 model is that the linker only has to consider the object files
22440 themselves; there is no external complexity to worry about. This
22441 disadvantage is that compilation time is increased because the
22442 template code is being compiled repeatedly. Code written for this
22443 model tends to include definitions of all templates in the header
22444 file, since they must be seen to be instantiated.
22447 The AT&T C++ translator, Cfront, solved the template instantiation
22448 problem by creating the notion of a template repository, an
22449 automatically maintained place where template instances are
22450 stored. A more modern version of the repository works as follows:
22451 As individual object files are built, the compiler places any
22452 template definitions and instantiations encountered in the
22453 repository. At link time, the link wrapper adds in the objects in
22454 the repository and compiles any needed instances that were not
22455 previously emitted. The advantages of this model are more optimal
22456 compilation speed and the ability to use the system linker; to
22457 implement the Borland model a compiler vendor also needs to
22458 replace the linker. The disadvantages are vastly increased
22459 complexity, and thus potential for error; for some code this can be
22460 just as transparent, but in practice it can been very difficult to
22461 build multiple programs in one directory and one program in
22462 multiple directories. Code written for this model tends to
22463 separate definitions of non-inline member templates into a
22464 separate file, which should be compiled separately.
22466 When used with GNU ld version 2.8 or later on an ELF system such as
22467 GNU/Linux or Solaris 2, or on Microsoft Windows, G++ supports the
22468 Borland model. On other systems, G++ implements neither automatic
22471 A future version of G++ will support a hybrid model whereby the
22472 compiler will emit any instantiations for which the template definition
22473 is included in the compile, and store template definitions and
22474 instantiation context information into the object file for the rest.
22475 The link wrapper will extract that information as necessary and invoke
22476 the compiler to produce the remaining instantiations. The linker will
22477 then combine duplicate instantiations.
22479 In the mean time, you have the following options for dealing with
22480 template instantiations:
22482 1. Compile your template-using code with `-frepo'. The compiler will
22483 generate files with the extension `.rpo' listing all of the
22484 template instantiations used in the corresponding object files
22485 which could be instantiated there; the link wrapper, `collect2',
22486 will then update the `.rpo' files to tell the compiler where to
22487 place those instantiations and rebuild any affected object files.
22488 The link-time overhead is negligible after the first pass, as the
22489 compiler will continue to place the instantiations in the same
22492 This is your best option for application code written for the
22493 Borland model, as it will just work. Code written for the Cfront
22494 model will need to be modified so that the template definitions
22495 are available at one or more points of instantiation; usually this
22496 is as simple as adding `#include <tmethods.cc>' to the end of each
22499 For library code, if you want the library to provide all of the
22500 template instantiations it needs, just try to link all of its
22501 object files together; the link will fail, but cause the
22502 instantiations to be generated as a side effect. Be warned,
22503 however, that this may cause conflicts if multiple libraries try
22504 to provide the same instantiations. For greater control, use
22505 explicit instantiation as described in the next option.
22507 2. Compile your code with `-fno-implicit-templates' to disable the
22508 implicit generation of template instances, and explicitly
22509 instantiate all the ones you use. This approach requires more
22510 knowledge of exactly which instances you need than do the others,
22511 but it's less mysterious and allows greater control. You can
22512 scatter the explicit instantiations throughout your program,
22513 perhaps putting them in the translation units where the instances
22514 are used or the translation units that define the templates
22515 themselves; you can put all of the explicit instantiations you
22516 need into one big file; or you can create small files like
22521 template class Foo<int>;
22522 template ostream& operator <<
22523 (ostream&, const Foo<int>&);
22525 for each of the instances you need, and create a template
22526 instantiation library from those.
22528 If you are using Cfront-model code, you can probably get away with
22529 not using `-fno-implicit-templates' when compiling files that don't
22530 `#include' the member template definitions.
22532 If you use one big file to do the instantiations, you may want to
22533 compile it without `-fno-implicit-templates' so you get all of the
22534 instances required by your explicit instantiations (but not by any
22535 other files) without having to specify them as well.
22537 G++ has extended the template instantiation syntax given in the ISO
22538 standard to allow forward declaration of explicit instantiations
22539 (with `extern'), instantiation of the compiler support data for a
22540 template class (i.e. the vtable) without instantiating any of its
22541 members (with `inline'), and instantiation of only the static data
22542 members of a template class, without the support data or member
22543 functions (with (`static'):
22545 extern template int max (int, int);
22546 inline template class Foo<int>;
22547 static template class Foo<int>;
22549 3. Do nothing. Pretend G++ does implement automatic instantiation
22550 management. Code written for the Borland model will work fine, but
22551 each translation unit will contain instances of each of the
22552 templates it uses. In a large program, this can lead to an
22553 unacceptable amount of code duplication.
22556 File: gcc.info, Node: Bound member functions, Next: C++ Attributes, Prev: Template Instantiation, Up: C++ Extensions
22558 6.6 Extracting the function pointer from a bound pointer to member function
22559 ===========================================================================
22561 In C++, pointer to member functions (PMFs) are implemented using a wide
22562 pointer of sorts to handle all the possible call mechanisms; the PMF
22563 needs to store information about how to adjust the `this' pointer, and
22564 if the function pointed to is virtual, where to find the vtable, and
22565 where in the vtable to look for the member function. If you are using
22566 PMFs in an inner loop, you should really reconsider that decision. If
22567 that is not an option, you can extract the pointer to the function that
22568 would be called for a given object/PMF pair and call it directly inside
22569 the inner loop, to save a bit of time.
22571 Note that you will still be paying the penalty for the call through a
22572 function pointer; on most modern architectures, such a call defeats the
22573 branch prediction features of the CPU. This is also true of normal
22574 virtual function calls.
22576 The syntax for this extension is
22579 extern int (A::*fp)();
22580 typedef int (*fptr)(A *);
22582 fptr p = (fptr)(a.*fp);
22584 For PMF constants (i.e. expressions of the form `&Klasse::Member'), no
22585 object is needed to obtain the address of the function. They can be
22586 converted to function pointers directly:
22588 fptr p1 = (fptr)(&A::foo);
22590 You must specify `-Wno-pmf-conversions' to use this extension.
22593 File: gcc.info, Node: C++ Attributes, Next: Strong Using, Prev: Bound member functions, Up: C++ Extensions
22595 6.7 C++-Specific Variable, Function, and Type Attributes
22596 ========================================================
22598 Some attributes only make sense for C++ programs.
22600 `init_priority (PRIORITY)'
22601 In Standard C++, objects defined at namespace scope are guaranteed
22602 to be initialized in an order in strict accordance with that of
22603 their definitions _in a given translation unit_. No guarantee is
22604 made for initializations across translation units. However, GNU
22605 C++ allows users to control the order of initialization of objects
22606 defined at namespace scope with the `init_priority' attribute by
22607 specifying a relative PRIORITY, a constant integral expression
22608 currently bounded between 101 and 65535 inclusive. Lower numbers
22609 indicate a higher priority.
22611 In the following example, `A' would normally be created before
22612 `B', but the `init_priority' attribute has reversed that order:
22614 Some_Class A __attribute__ ((init_priority (2000)));
22615 Some_Class B __attribute__ ((init_priority (543)));
22617 Note that the particular values of PRIORITY do not matter; only
22618 their relative ordering.
22621 This type attribute informs C++ that the class is a Java
22622 interface. It may only be applied to classes declared within an
22623 `extern "Java"' block. Calls to methods declared in this
22624 interface will be dispatched using GCJ's interface table
22625 mechanism, instead of regular virtual table dispatch.
22628 See also *Note Strong Using::.
22631 File: gcc.info, Node: Strong Using, Next: Java Exceptions, Prev: C++ Attributes, Up: C++ Extensions
22636 *Caution:* The semantics of this extension are not fully defined.
22637 Users should refrain from using this extension as its semantics may
22638 change subtly over time. It is possible that this extension wil be
22639 removed in future versions of G++.
22641 A using-directive with `__attribute ((strong))' is stronger than a
22642 normal using-directive in two ways:
22644 * Templates from the used namespace can be specialized as though
22645 they were members of the using namespace.
22647 * The using namespace is considered an associated namespace of all
22648 templates in the used namespace for purposes of argument-dependent
22651 This is useful for composing a namespace transparently from
22652 implementation namespaces. For example:
22656 template <class T> struct A { };
22658 using namespace debug __attribute ((__strong__));
22659 template <> struct A<int> { }; // ok to specialize
22661 template <class T> void f (A<T>);
22666 f (std::A<float>()); // lookup finds std::f
22671 File: gcc.info, Node: Java Exceptions, Next: Deprecated Features, Prev: Strong Using, Up: C++ Extensions
22673 6.9 Java Exceptions
22674 ===================
22676 The Java language uses a slightly different exception handling model
22677 from C++. Normally, GNU C++ will automatically detect when you are
22678 writing C++ code that uses Java exceptions, and handle them
22679 appropriately. However, if C++ code only needs to execute destructors
22680 when Java exceptions are thrown through it, GCC will guess incorrectly.
22681 Sample problematic code is:
22683 struct S { ~S(); };
22684 extern void bar(); // is written in Java, and may throw exceptions
22691 The usual effect of an incorrect guess is a link failure, complaining of
22692 a missing routine called `__gxx_personality_v0'.
22694 You can inform the compiler that Java exceptions are to be used in a
22695 translation unit, irrespective of what it might think, by writing
22696 `#pragma GCC java_exceptions' at the head of the file. This `#pragma'
22697 must appear before any functions that throw or catch exceptions, or run
22698 destructors when exceptions are thrown through them.
22700 You cannot mix Java and C++ exceptions in the same translation unit.
22701 It is believed to be safe to throw a C++ exception from one file through
22702 another file compiled for the Java exception model, or vice versa, but
22703 there may be bugs in this area.
22706 File: gcc.info, Node: Deprecated Features, Next: Backwards Compatibility, Prev: Java Exceptions, Up: C++ Extensions
22708 6.10 Deprecated Features
22709 ========================
22711 In the past, the GNU C++ compiler was extended to experiment with new
22712 features, at a time when the C++ language was still evolving. Now that
22713 the C++ standard is complete, some of those features are superseded by
22714 superior alternatives. Using the old features might cause a warning in
22715 some cases that the feature will be dropped in the future. In other
22716 cases, the feature might be gone already.
22718 While the list below is not exhaustive, it documents some of the
22719 options that are now deprecated:
22721 `-fexternal-templates'
22722 `-falt-external-templates'
22723 These are two of the many ways for G++ to implement template
22724 instantiation. *Note Template Instantiation::. The C++ standard
22725 clearly defines how template definitions have to be organized
22726 across implementation units. G++ has an implicit instantiation
22727 mechanism that should work just fine for standard-conforming code.
22729 `-fstrict-prototype'
22730 `-fno-strict-prototype'
22731 Previously it was possible to use an empty prototype parameter
22732 list to indicate an unspecified number of parameters (like C),
22733 rather than no parameters, as C++ demands. This feature has been
22734 removed, except where it is required for backwards compatibility
22735 *Note Backwards Compatibility::.
22737 G++ allows a virtual function returning `void *' to be overridden by
22738 one returning a different pointer type. This extension to the
22739 covariant return type rules is now deprecated and will be removed from a
22742 The G++ minimum and maximum operators (`<?' and `>?') and their
22743 compound forms (`<?=') and `>?=') have been deprecated and will be
22744 removed in a future version. Code using these operators should be
22745 modified to use `std::min' and `std::max' instead.
22747 The named return value extension has been deprecated, and is now
22750 The use of initializer lists with new expressions has been deprecated,
22751 and is now removed from G++.
22753 Floating and complex non-type template parameters have been deprecated,
22754 and are now removed from G++.
22756 The implicit typename extension has been deprecated and is now removed
22759 The use of default arguments in function pointers, function typedefs
22760 and and other places where they are not permitted by the standard is
22761 deprecated and will be removed from a future version of G++.
22763 G++ allows floating-point literals to appear in integral constant
22764 expressions, e.g. ` enum E { e = int(2.2 * 3.7) } ' This extension is
22765 deprecated and will be removed from a future version.
22767 G++ allows static data members of const floating-point type to be
22768 declared with an initializer in a class definition. The standard only
22769 allows initializers for static members of const integral types and const
22770 enumeration types so this extension has been deprecated and will be
22771 removed from a future version.
22774 File: gcc.info, Node: Backwards Compatibility, Prev: Deprecated Features, Up: C++ Extensions
22776 6.11 Backwards Compatibility
22777 ============================
22779 Now that there is a definitive ISO standard C++, G++ has a specification
22780 to adhere to. The C++ language evolved over time, and features that
22781 used to be acceptable in previous drafts of the standard, such as the
22782 ARM [Annotated C++ Reference Manual], are no longer accepted. In order
22783 to allow compilation of C++ written to such drafts, G++ contains some
22784 backwards compatibilities. _All such backwards compatibility features
22785 are liable to disappear in future versions of G++._ They should be
22786 considered deprecated *Note Deprecated Features::.
22789 If a variable is declared at for scope, it used to remain in scope
22790 until the end of the scope which contained the for statement
22791 (rather than just within the for scope). G++ retains this, but
22792 issues a warning, if such a variable is accessed outside the for
22795 `Implicit C language'
22796 Old C system header files did not contain an `extern "C" {...}'
22797 scope to set the language. On such systems, all header files are
22798 implicitly scoped inside a C language scope. Also, an empty
22799 prototype `()' will be treated as an unspecified number of
22800 arguments, rather than no arguments, as C++ demands.
22803 File: gcc.info, Node: Objective-C, Next: Compatibility, Prev: C++ Extensions, Up: Top
22805 7 GNU Objective-C runtime features
22806 **********************************
22808 This document is meant to describe some of the GNU Objective-C runtime
22809 features. It is not intended to teach you Objective-C, there are
22810 several resources on the Internet that present the language. Questions
22811 and comments about this document to Ovidiu Predescu <ovidiu@cup.hp.com>.
22815 * Executing code before main::
22817 * Garbage Collection::
22818 * Constant string objects::
22819 * compatibility_alias::
22822 File: gcc.info, Node: Executing code before main, Next: Type encoding, Prev: Objective-C, Up: Objective-C
22824 7.1 `+load': Executing code before main
22825 =======================================
22827 The GNU Objective-C runtime provides a way that allows you to execute
22828 code before the execution of the program enters the `main' function.
22829 The code is executed on a per-class and a per-category basis, through a
22830 special class method `+load'.
22832 This facility is very useful if you want to initialize global variables
22833 which can be accessed by the program directly, without sending a message
22834 to the class first. The usual way to initialize global variables, in
22835 the `+initialize' method, might not be useful because `+initialize' is
22836 only called when the first message is sent to a class object, which in
22837 some cases could be too late.
22839 Suppose for example you have a `FileStream' class that declares
22840 `Stdin', `Stdout' and `Stderr' as global variables, like below:
22843 FileStream *Stdin = nil;
22844 FileStream *Stdout = nil;
22845 FileStream *Stderr = nil;
22847 @implementation FileStream
22851 Stdin = [[FileStream new] initWithFd:0];
22852 Stdout = [[FileStream new] initWithFd:1];
22853 Stderr = [[FileStream new] initWithFd:2];
22856 /* Other methods here */
22859 In this example, the initialization of `Stdin', `Stdout' and `Stderr'
22860 in `+initialize' occurs too late. The programmer can send a message to
22861 one of these objects before the variables are actually initialized,
22862 thus sending messages to the `nil' object. The `+initialize' method
22863 which actually initializes the global variables is not invoked until
22864 the first message is sent to the class object. The solution would
22865 require these variables to be initialized just before entering `main'.
22867 The correct solution of the above problem is to use the `+load' method
22868 instead of `+initialize':
22871 @implementation FileStream
22875 Stdin = [[FileStream new] initWithFd:0];
22876 Stdout = [[FileStream new] initWithFd:1];
22877 Stderr = [[FileStream new] initWithFd:2];
22880 /* Other methods here */
22883 The `+load' is a method that is not overridden by categories. If a
22884 class and a category of it both implement `+load', both methods are
22885 invoked. This allows some additional initializations to be performed in
22888 This mechanism is not intended to be a replacement for `+initialize'.
22889 You should be aware of its limitations when you decide to use it
22890 instead of `+initialize'.
22894 * What you can and what you cannot do in +load::
22897 File: gcc.info, Node: What you can and what you cannot do in +load, Prev: Executing code before main, Up: Executing code before main
22899 7.1.1 What you can and what you cannot do in `+load'
22900 ----------------------------------------------------
22902 The `+load' implementation in the GNU runtime guarantees you the
22905 * you can write whatever C code you like;
22907 * you can send messages to Objective-C constant strings (`@"this is a
22908 constant string"');
22910 * you can allocate and send messages to objects whose class is
22911 implemented in the same file;
22913 * the `+load' implementation of all super classes of a class are
22914 executed before the `+load' of that class is executed;
22916 * the `+load' implementation of a class is executed before the
22917 `+load' implementation of any category.
22920 In particular, the following things, even if they can work in a
22921 particular case, are not guaranteed:
22923 * allocation of or sending messages to arbitrary objects;
22925 * allocation of or sending messages to objects whose classes have a
22926 category implemented in the same file;
22929 You should make no assumptions about receiving `+load' in sibling
22930 classes when you write `+load' of a class. The order in which sibling
22931 classes receive `+load' is not guaranteed.
22933 The order in which `+load' and `+initialize' are called could be
22934 problematic if this matters. If you don't allocate objects inside
22935 `+load', it is guaranteed that `+load' is called before `+initialize'.
22936 If you create an object inside `+load' the `+initialize' method of
22937 object's class is invoked even if `+load' was not invoked. Note if you
22938 explicitly call `+load' on a class, `+initialize' will be called first.
22939 To avoid possible problems try to implement only one of these methods.
22941 The `+load' method is also invoked when a bundle is dynamically loaded
22942 into your running program. This happens automatically without any
22943 intervening operation from you. When you write bundles and you need to
22944 write `+load' you can safely create and send messages to objects whose
22945 classes already exist in the running program. The same restrictions as
22946 above apply to classes defined in bundle.
22949 File: gcc.info, Node: Type encoding, Next: Garbage Collection, Prev: Executing code before main, Up: Objective-C
22954 The Objective-C compiler generates type encodings for all the types.
22955 These type encodings are used at runtime to find out information about
22956 selectors and methods and about objects and classes.
22958 The types are encoded in the following way:
22961 `unsigned char' `C'
22963 `unsigned short' `S'
22967 `unsigned long' `L'
22979 bit-fields `b' followed by the starting position of the
22980 bit-field, the type of the bit-field and the size of
22981 the bit-field (the bit-fields encoding was changed
22982 from the NeXT's compiler encoding, see below)
22984 The encoding of bit-fields has changed to allow bit-fields to be
22985 properly handled by the runtime functions that compute sizes and
22986 alignments of types that contain bit-fields. The previous encoding
22987 contained only the size of the bit-field. Using only this information
22988 it is not possible to reliably compute the size occupied by the
22989 bit-field. This is very important in the presence of the Boehm's
22990 garbage collector because the objects are allocated using the typed
22991 memory facility available in this collector. The typed memory
22992 allocation requires information about where the pointers are located
22995 The position in the bit-field is the position, counting in bits, of the
22996 bit closest to the beginning of the structure.
22998 The non-atomic types are encoded as follows:
23000 pointers `^' followed by the pointed type.
23001 arrays `[' followed by the number of elements in the array
23002 followed by the type of the elements followed by `]'
23003 structures `{' followed by the name of the structure (or `?' if the
23004 structure is unnamed), the `=' sign, the type of the
23006 unions `(' followed by the name of the structure (or `?' if the
23007 union is unnamed), the `=' sign, the type of the members
23010 Here are some types and their encodings, as they are generated by the
23011 compiler on an i386 machine:
23014 Objective-C type Compiler encoding
23016 struct { `{?=i[3f]b128i3b131i2c}'
23025 In addition to the types the compiler also encodes the type
23026 specifiers. The table below describes the encoding of the current
23027 Objective-C type specifiers:
23039 The type specifiers are encoded just before the type. Unlike types
23040 however, the type specifiers are only encoded when they appear in method
23044 File: gcc.info, Node: Garbage Collection, Next: Constant string objects, Prev: Type encoding, Up: Objective-C
23046 7.3 Garbage Collection
23047 ======================
23049 Support for a new memory management policy has been added by using a
23050 powerful conservative garbage collector, known as the
23051 Boehm-Demers-Weiser conservative garbage collector. It is available
23052 from `http://www.hpl.hp.com/personal/Hans_Boehm/gc/'.
23054 To enable the support for it you have to configure the compiler using
23055 an additional argument, `--enable-objc-gc'. You need to have garbage
23056 collector installed before building the compiler. This will build an
23057 additional runtime library which has several enhancements to support
23058 the garbage collector. The new library has a new name, `libobjc_gc.a'
23059 to not conflict with the non-garbage-collected library.
23061 When the garbage collector is used, the objects are allocated using the
23062 so-called typed memory allocation mechanism available in the
23063 Boehm-Demers-Weiser collector. This mode requires precise information
23064 on where pointers are located inside objects. This information is
23065 computed once per class, immediately after the class has been
23068 There is a new runtime function `class_ivar_set_gcinvisible()' which
23069 can be used to declare a so-called "weak pointer" reference. Such a
23070 pointer is basically hidden for the garbage collector; this can be
23071 useful in certain situations, especially when you want to keep track of
23072 the allocated objects, yet allow them to be collected. This kind of
23073 pointers can only be members of objects, you cannot declare a global
23074 pointer as a weak reference. Every type which is a pointer type can be
23075 declared a weak pointer, including `id', `Class' and `SEL'.
23077 Here is an example of how to use this feature. Suppose you want to
23078 implement a class whose instances hold a weak pointer reference; the
23079 following class does this:
23082 @interface WeakPointer : Object
23084 const void* weakPointer;
23087 - initWithPointer:(const void*)p;
23088 - (const void*)weakPointer;
23092 @implementation WeakPointer
23096 class_ivar_set_gcinvisible (self, "weakPointer", YES);
23099 - initWithPointer:(const void*)p
23105 - (const void*)weakPointer
23107 return weakPointer;
23112 Weak pointers are supported through a new type character specifier
23113 represented by the `!' character. The `class_ivar_set_gcinvisible()'
23114 function adds or removes this specifier to the string type description
23115 of the instance variable named as argument.
23118 File: gcc.info, Node: Constant string objects, Next: compatibility_alias, Prev: Garbage Collection, Up: Objective-C
23120 7.4 Constant string objects
23121 ===========================
23123 GNU Objective-C provides constant string objects that are generated
23124 directly by the compiler. You declare a constant string object by
23125 prefixing a C constant string with the character `@':
23127 id myString = @"this is a constant string object";
23129 The constant string objects are by default instances of the
23130 `NXConstantString' class which is provided by the GNU Objective-C
23131 runtime. To get the definition of this class you must include the
23132 `objc/NXConstStr.h' header file.
23134 User defined libraries may want to implement their own constant string
23135 class. To be able to support them, the GNU Objective-C compiler
23136 provides a new command line options
23137 `-fconstant-string-class=CLASS-NAME'. The provided class should adhere
23138 to a strict structure, the same as `NXConstantString''s structure:
23141 @interface MyConstantStringClass
23149 `NXConstantString' inherits from `Object'; user class libraries may
23150 choose to inherit the customized constant string class from a different
23151 class than `Object'. There is no requirement in the methods the
23152 constant string class has to implement, but the final ivar layout of
23153 the class must be the compatible with the given structure.
23155 When the compiler creates the statically allocated constant string
23156 object, the `c_string' field will be filled by the compiler with the
23157 string; the `length' field will be filled by the compiler with the
23158 string length; the `isa' pointer will be filled with `NULL' by the
23159 compiler, and it will later be fixed up automatically at runtime by the
23160 GNU Objective-C runtime library to point to the class which was set by
23161 the `-fconstant-string-class' option when the object file is loaded (if
23162 you wonder how it works behind the scenes, the name of the class to
23163 use, and the list of static objects to fixup, are stored by the
23164 compiler in the object file in a place where the GNU runtime library
23165 will find them at runtime).
23167 As a result, when a file is compiled with the
23168 `-fconstant-string-class' option, all the constant string objects will
23169 be instances of the class specified as argument to this option. It is
23170 possible to have multiple compilation units referring to different
23171 constant string classes, neither the compiler nor the linker impose any
23172 restrictions in doing this.
23175 File: gcc.info, Node: compatibility_alias, Prev: Constant string objects, Up: Objective-C
23177 7.5 compatibility_alias
23178 =======================
23180 This is a feature of the Objective-C compiler rather than of the
23181 runtime, anyway since it is documented nowhere and its existence was
23182 forgotten, we are documenting it here.
23184 The keyword `@compatibility_alias' allows you to define a class name
23185 as equivalent to another class name. For example:
23187 @compatibility_alias WOApplication GSWApplication;
23189 tells the compiler that each time it encounters `WOApplication' as a
23190 class name, it should replace it with `GSWApplication' (that is,
23191 `WOApplication' is just an alias for `GSWApplication').
23193 There are some constraints on how this can be used--
23195 * `WOApplication' (the alias) must not be an existing class;
23197 * `GSWApplication' (the real class) must be an existing class.
23201 File: gcc.info, Node: Compatibility, Next: Gcov, Prev: Objective-C, Up: Top
23203 8 Binary Compatibility
23204 **********************
23206 Binary compatibility encompasses several related concepts:
23208 "application binary interface (ABI)"
23209 The set of runtime conventions followed by all of the tools that
23210 deal with binary representations of a program, including
23211 compilers, assemblers, linkers, and language runtime support.
23212 Some ABIs are formal with a written specification, possibly
23213 designed by multiple interested parties. Others are simply the
23214 way things are actually done by a particular set of tools.
23217 A compiler conforms to an ABI if it generates code that follows
23218 all of the specifications enumerated by that ABI. A library
23219 conforms to an ABI if it is implemented according to that ABI. An
23220 application conforms to an ABI if it is built using tools that
23221 conform to that ABI and does not contain source code that
23222 specifically changes behavior specified by the ABI.
23224 "calling conventions"
23225 Calling conventions are a subset of an ABI that specify of how
23226 arguments are passed and function results are returned.
23229 Different sets of tools are interoperable if they generate files
23230 that can be used in the same program. The set of tools includes
23231 compilers, assemblers, linkers, libraries, header files, startup
23232 files, and debuggers. Binaries produced by different sets of
23233 tools are not interoperable unless they implement the same ABI.
23234 This applies to different versions of the same tools as well as
23235 tools from different vendors.
23238 Whether a function in a binary built by one set of tools can call a
23239 function in a binary built by a different set of tools is a subset
23240 of interoperability.
23242 "implementation-defined features"
23243 Language standards include lists of implementation-defined
23244 features whose behavior can vary from one implementation to
23245 another. Some of these features are normally covered by a
23246 platform's ABI and others are not. The features that are not
23247 covered by an ABI generally affect how a program behaves, but not
23251 Conformance to the same ABI and the same behavior of
23252 implementation-defined features are both relevant for
23255 The application binary interface implemented by a C or C++ compiler
23256 affects code generation and runtime support for:
23258 * size and alignment of data types
23260 * layout of structured types
23262 * calling conventions
23264 * register usage conventions
23266 * interfaces for runtime arithmetic support
23268 * object file formats
23270 In addition, the application binary interface implemented by a C++
23271 compiler affects code generation and runtime support for:
23274 * exception handling
23276 * invoking constructors and destructors
23278 * layout, alignment, and padding of classes
23280 * layout and alignment of virtual tables
23282 Some GCC compilation options cause the compiler to generate code that
23283 does not conform to the platform's default ABI. Other options cause
23284 different program behavior for implementation-defined features that are
23285 not covered by an ABI. These options are provided for consistency with
23286 other compilers that do not follow the platform's default ABI or the
23287 usual behavior of implementation-defined features for the platform. Be
23288 very careful about using such options.
23290 Most platforms have a well-defined ABI that covers C code, but ABIs
23291 that cover C++ functionality are not yet common.
23293 Starting with GCC 3.2, GCC binary conventions for C++ are based on a
23294 written, vendor-neutral C++ ABI that was designed to be specific to
23295 64-bit Itanium but also includes generic specifications that apply to
23296 any platform. This C++ ABI is also implemented by other compiler
23297 vendors on some platforms, notably GNU/Linux and BSD systems. We have
23298 tried hard to provide a stable ABI that will be compatible with future
23299 GCC releases, but it is possible that we will encounter problems that
23300 make this difficult. Such problems could include different
23301 interpretations of the C++ ABI by different vendors, bugs in the ABI, or
23302 bugs in the implementation of the ABI in different compilers. GCC's
23303 `-Wabi' switch warns when G++ generates code that is probably not
23304 compatible with the C++ ABI.
23306 The C++ library used with a C++ compiler includes the Standard C++
23307 Library, with functionality defined in the C++ Standard, plus language
23308 runtime support. The runtime support is included in a C++ ABI, but
23309 there is no formal ABI for the Standard C++ Library. Two
23310 implementations of that library are interoperable if one follows the
23311 de-facto ABI of the other and if they are both built with the same
23312 compiler, or with compilers that conform to the same ABI for C++
23313 compiler and runtime support.
23315 When G++ and another C++ compiler conform to the same C++ ABI, but the
23316 implementations of the Standard C++ Library that they normally use do
23317 not follow the same ABI for the Standard C++ Library, object files
23318 built with those compilers can be used in the same program only if they
23319 use the same C++ library. This requires specifying the location of the
23320 C++ library header files when invoking the compiler whose usual library
23321 is not being used. The location of GCC's C++ header files depends on
23322 how the GCC build was configured, but can be seen by using the G++ `-v'
23323 option. With default configuration options for G++ 3.3 the compile
23324 line for a different C++ compiler needs to include
23326 -IGCC_INSTALL_DIRECTORY/include/c++/3.3
23328 Similarly, compiling code with G++ that must use a C++ library other
23329 than the GNU C++ library requires specifying the location of the header
23330 files for that other library.
23332 The most straightforward way to link a program to use a particular C++
23333 library is to use a C++ driver that specifies that C++ library by
23334 default. The `g++' driver, for example, tells the linker where to find
23335 GCC's C++ library (`libstdc++') plus the other libraries and startup
23336 files it needs, in the proper order.
23338 If a program must use a different C++ library and it's not possible to
23339 do the final link using a C++ driver that uses that library by default,
23340 it is necessary to tell `g++' the location and name of that library.
23341 It might also be necessary to specify different startup files and other
23342 runtime support libraries, and to suppress the use of GCC's support
23343 libraries with one or more of the options `-nostdlib', `-nostartfiles',
23344 and `-nodefaultlibs'.
23347 File: gcc.info, Node: Gcov, Next: Trouble, Prev: Compatibility, Up: Top
23349 9 `gcov'--a Test Coverage Program
23350 *********************************
23352 `gcov' is a tool you can use in conjunction with GCC to test code
23353 coverage in your programs.
23357 * Gcov Intro:: Introduction to gcov.
23358 * Invoking Gcov:: How to use gcov.
23359 * Gcov and Optimization:: Using gcov with GCC optimization.
23360 * Gcov Data Files:: The files used by gcov.
23363 File: gcc.info, Node: Gcov Intro, Next: Invoking Gcov, Up: Gcov
23365 9.1 Introduction to `gcov'
23366 ==========================
23368 `gcov' is a test coverage program. Use it in concert with GCC to
23369 analyze your programs to help create more efficient, faster running
23370 code and to discover untested parts of your program. You can use
23371 `gcov' as a profiling tool to help discover where your optimization
23372 efforts will best affect your code. You can also use `gcov' along with
23373 the other profiling tool, `gprof', to assess which parts of your code
23374 use the greatest amount of computing time.
23376 Profiling tools help you analyze your code's performance. Using a
23377 profiler such as `gcov' or `gprof', you can find out some basic
23378 performance statistics, such as:
23380 * how often each line of code executes
23382 * what lines of code are actually executed
23384 * how much computing time each section of code uses
23386 Once you know these things about how your code works when compiled, you
23387 can look at each module to see which modules should be optimized.
23388 `gcov' helps you determine where to work on optimization.
23390 Software developers also use coverage testing in concert with
23391 testsuites, to make sure software is actually good enough for a release.
23392 Testsuites can verify that a program works as expected; a coverage
23393 program tests to see how much of the program is exercised by the
23394 testsuite. Developers can then determine what kinds of test cases need
23395 to be added to the testsuites to create both better testing and a better
23398 You should compile your code without optimization if you plan to use
23399 `gcov' because the optimization, by combining some lines of code into
23400 one function, may not give you as much information as you need to look
23401 for `hot spots' where the code is using a great deal of computer time.
23402 Likewise, because `gcov' accumulates statistics by line (at the lowest
23403 resolution), it works best with a programming style that places only
23404 one statement on each line. If you use complicated macros that expand
23405 to loops or to other control structures, the statistics are less
23406 helpful--they only report on the line where the macro call appears. If
23407 your complex macros behave like functions, you can replace them with
23408 inline functions to solve this problem.
23410 `gcov' creates a logfile called `SOURCEFILE.gcov' which indicates how
23411 many times each line of a source file `SOURCEFILE.c' has executed. You
23412 can use these logfiles along with `gprof' to aid in fine-tuning the
23413 performance of your programs. `gprof' gives timing information you can
23414 use along with the information you get from `gcov'.
23416 `gcov' works only on code compiled with GCC. It is not compatible
23417 with any other profiling or test coverage mechanism.
23420 File: gcc.info, Node: Invoking Gcov, Next: Gcov and Optimization, Prev: Gcov Intro, Up: Gcov
23425 gcov [OPTIONS] SOURCEFILE
23427 `gcov' accepts the following options:
23431 Display help about using `gcov' (on the standard output), and exit
23432 without doing any further processing.
23436 Display the `gcov' version number (on the standard output), and
23437 exit without doing any further processing.
23441 Write individual execution counts for every basic block. Normally
23442 gcov outputs execution counts only for the main blocks of a line.
23443 With this option you can determine if blocks within a single line
23444 are not being executed.
23447 `--branch-probabilities'
23448 Write branch frequencies to the output file, and write branch
23449 summary info to the standard output. This option allows you to
23450 see how often each branch in your program was taken.
23451 Unconditional branches will not be shown, unless the `-u' option
23456 Write branch frequencies as the number of branches taken, rather
23457 than the percentage of branches taken.
23461 Do not create the `gcov' output file.
23464 `--long-file-names'
23465 Create long file names for included source files. For example, if
23466 the header file `x.h' contains code, and was included in the file
23467 `a.c', then running `gcov' on the file `a.c' will produce an
23468 output file called `a.c##x.h.gcov' instead of `x.h.gcov'. This
23469 can be useful if `x.h' is included in multiple source files. If
23470 you use the `-p' option, both the including and included file
23471 names will be complete path names.
23475 Preserve complete path information in the names of generated
23476 `.gcov' files. Without this option, just the filename component is
23477 used. With this option, all directories are used, with `/'
23478 characters translated to `#' characters, `.' directory components
23479 removed and `..' components renamed to `^'. This is useful if
23480 sourcefiles are in several different directories. It also affects
23484 `--function-summaries'
23485 Output summaries for each function in addition to the file level
23488 `-o DIRECTORY|FILE'
23489 `--object-directory DIRECTORY'
23490 `--object-file FILE'
23491 Specify either the directory containing the gcov data files, or the
23492 object path name. The `.gcno', and `.gcda' data files are
23493 searched for using this option. If a directory is specified, the
23494 data files are in that directory and named after the source file
23495 name, without its extension. If a file is specified here, the
23496 data files are named after that file, without its extension. If
23497 this option is not supplied, it defaults to the current directory.
23500 `--unconditional-branches'
23501 When branch probabilities are given, include those of
23502 unconditional branches. Unconditional branches are normally not
23506 `gcov' should be run with the current directory the same as that when
23507 you invoked the compiler. Otherwise it will not be able to locate the
23508 source files. `gcov' produces files called `MANGLEDNAME.gcov' in the
23509 current directory. These contain the coverage information of the
23510 source file they correspond to. One `.gcov' file is produced for each
23511 source file containing code, which was compiled to produce the data
23512 files. The MANGLEDNAME part of the output file name is usually simply
23513 the source file name, but can be something more complicated if the `-l'
23514 or `-p' options are given. Refer to those options for details.
23516 The `.gcov' files contain the `:' separated fields along with program
23517 source code. The format is
23519 EXECUTION_COUNT:LINE_NUMBER:SOURCE LINE TEXT
23521 Additional block information may succeed each line, when requested by
23522 command line option. The EXECUTION_COUNT is `-' for lines containing
23523 no code and `#####' for lines which were never executed. Some lines of
23524 information at the start have LINE_NUMBER of zero.
23526 The preamble lines are of the form
23530 The ordering and number of these preamble lines will be augmented as
23531 `gcov' development progresses -- do not rely on them remaining
23532 unchanged. Use TAG to locate a particular preamble line.
23534 The additional block information is of the form
23538 The INFORMATION is human readable, but designed to be simple enough
23539 for machine parsing too.
23541 When printing percentages, 0% and 100% are only printed when the values
23542 are _exactly_ 0% and 100% respectively. Other values which would
23543 conventionally be rounded to 0% or 100% are instead printed as the
23544 nearest non-boundary value.
23546 When using `gcov', you must first compile your program with two
23547 special GCC options: `-fprofile-arcs -ftest-coverage'. This tells the
23548 compiler to generate additional information needed by gcov (basically a
23549 flow graph of the program) and also includes additional code in the
23550 object files for generating the extra profiling information needed by
23551 gcov. These additional files are placed in the directory where the
23552 object file is located.
23554 Running the program will cause profile output to be generated. For
23555 each source file compiled with `-fprofile-arcs', an accompanying
23556 `.gcda' file will be placed in the object file directory.
23558 Running `gcov' with your program's source file names as arguments will
23559 now produce a listing of the code along with frequency of execution for
23560 each line. For example, if your program is called `tmp.c', this is
23561 what you see when you use the basic `gcov' facility:
23563 $ gcc -fprofile-arcs -ftest-coverage tmp.c
23566 90.00% of 10 source lines executed in file tmp.c
23567 Creating tmp.c.gcov.
23569 The file `tmp.c.gcov' contains output from `gcov'. Here is a sample:
23572 -: 0:Graph:tmp.gcno
23576 -: 1:#include <stdio.h>
23578 -: 3:int main (void)
23580 1: 5: int i, total;
23584 11: 9: for (i = 0; i < 10; i++)
23585 10: 10: total += i;
23587 1: 12: if (total != 45)
23588 #####: 13: printf ("Failure\n");
23590 1: 15: printf ("Success\n");
23594 When you use the `-a' option, you will get individual block counts,
23595 and the output looks like this:
23598 -: 0:Graph:tmp.gcno
23602 -: 1:#include <stdio.h>
23604 -: 3:int main (void)
23607 1: 5: int i, total;
23611 11: 9: for (i = 0; i < 10; i++)
23613 10: 10: total += i;
23616 1: 12: if (total != 45)
23618 #####: 13: printf ("Failure\n");
23621 1: 15: printf ("Success\n");
23627 In this mode, each basic block is only shown on one line - the last
23628 line of the block. A multi-line block will only contribute to the
23629 execution count of that last line, and other lines will not be shown to
23630 contain code, unless previous blocks end on those lines. The total
23631 execution count of a line is shown and subsequent lines show the
23632 execution counts for individual blocks that end on that line. After
23633 each block, the branch and call counts of the block will be shown, if
23634 the `-b' option is given.
23636 Because of the way GCC instruments calls, a call count can be shown
23637 after a line with no individual blocks. As you can see, line 13
23638 contains a basic block that was not executed.
23640 When you use the `-b' option, your output looks like this:
23643 90.00% of 10 source lines executed in file tmp.c
23644 80.00% of 5 branches executed in file tmp.c
23645 80.00% of 5 branches taken at least once in file tmp.c
23646 50.00% of 2 calls executed in file tmp.c
23647 Creating tmp.c.gcov.
23649 Here is a sample of a resulting `tmp.c.gcov' file:
23652 -: 0:Graph:tmp.gcno
23656 -: 1:#include <stdio.h>
23658 -: 3:int main (void)
23659 function main called 1 returned 1 blocks executed 75%
23661 1: 5: int i, total;
23665 11: 9: for (i = 0; i < 10; i++)
23666 branch 0 taken 91% (fallthrough)
23668 10: 10: total += i;
23670 1: 12: if (total != 45)
23671 branch 0 taken 0% (fallthrough)
23672 branch 1 taken 100%
23673 #####: 13: printf ("Failure\n");
23674 call 0 never executed
23676 1: 15: printf ("Success\n");
23677 call 0 called 1 returned 100%
23681 For each function, a line is printed showing how many times the
23682 function is called, how many times it returns and what percentage of the
23683 function's blocks were executed.
23685 For each basic block, a line is printed after the last line of the
23686 basic block describing the branch or call that ends the basic block.
23687 There can be multiple branches and calls listed for a single source
23688 line if there are multiple basic blocks that end on that line. In this
23689 case, the branches and calls are each given a number. There is no
23690 simple way to map these branches and calls back to source constructs.
23691 In general, though, the lowest numbered branch or call will correspond
23692 to the leftmost construct on the source line.
23694 For a branch, if it was executed at least once, then a percentage
23695 indicating the number of times the branch was taken divided by the
23696 number of times the branch was executed will be printed. Otherwise, the
23697 message "never executed" is printed.
23699 For a call, if it was executed at least once, then a percentage
23700 indicating the number of times the call returned divided by the number
23701 of times the call was executed will be printed. This will usually be
23702 100%, but may be less for functions call `exit' or `longjmp', and thus
23703 may not return every time they are called.
23705 The execution counts are cumulative. If the example program were
23706 executed again without removing the `.gcda' file, the count for the
23707 number of times each line in the source was executed would be added to
23708 the results of the previous run(s). This is potentially useful in
23709 several ways. For example, it could be used to accumulate data over a
23710 number of program runs as part of a test verification suite, or to
23711 provide more accurate long-term information over a large number of
23714 The data in the `.gcda' files is saved immediately before the program
23715 exits. For each source file compiled with `-fprofile-arcs', the
23716 profiling code first attempts to read in an existing `.gcda' file; if
23717 the file doesn't match the executable (differing number of basic block
23718 counts) it will ignore the contents of the file. It then adds in the
23719 new execution counts and finally writes the data to the file.
23722 File: gcc.info, Node: Gcov and Optimization, Next: Gcov Data Files, Prev: Invoking Gcov, Up: Gcov
23724 9.3 Using `gcov' with GCC Optimization
23725 ======================================
23727 If you plan to use `gcov' to help optimize your code, you must first
23728 compile your program with two special GCC options: `-fprofile-arcs
23729 -ftest-coverage'. Aside from that, you can use any other GCC options;
23730 but if you want to prove that every single line in your program was
23731 executed, you should not compile with optimization at the same time.
23732 On some machines the optimizer can eliminate some simple code lines by
23733 combining them with other lines. For example, code like this:
23740 can be compiled into one instruction on some machines. In this case,
23741 there is no way for `gcov' to calculate separate execution counts for
23742 each line because there isn't separate code for each line. Hence the
23743 `gcov' output looks like this if you compiled the program with
23746 100: 12:if (a != b)
23751 The output shows that this block of code, combined by optimization,
23752 executed 100 times. In one sense this result is correct, because there
23753 was only one instruction representing all four of these lines. However,
23754 the output does not indicate how many times the result was 0 and how
23755 many times the result was 1.
23757 Inlineable functions can create unexpected line counts. Line counts
23758 are shown for the source code of the inlineable function, but what is
23759 shown depends on where the function is inlined, or if it is not inlined
23762 If the function is not inlined, the compiler must emit an out of line
23763 copy of the function, in any object file that needs it. If `fileA.o'
23764 and `fileB.o' both contain out of line bodies of a particular
23765 inlineable function, they will also both contain coverage counts for
23766 that function. When `fileA.o' and `fileB.o' are linked together, the
23767 linker will, on many systems, select one of those out of line bodies
23768 for all calls to that function, and remove or ignore the other.
23769 Unfortunately, it will not remove the coverage counters for the unused
23770 function body. Hence when instrumented, all but one use of that
23771 function will show zero counts.
23773 If the function is inlined in several places, the block structure in
23774 each location might not be the same. For instance, a condition might
23775 now be calculable at compile time in some instances. Because the
23776 coverage of all the uses of the inline function will be shown for the
23777 same source lines, the line counts themselves might seem inconsistent.
23780 File: gcc.info, Node: Gcov Data Files, Prev: Gcov and Optimization, Up: Gcov
23782 9.4 Brief description of `gcov' data files
23783 ==========================================
23785 `gcov' uses two files for profiling. The names of these files are
23786 derived from the original _object_ file by substituting the file suffix
23787 with either `.gcno', or `.gcda'. All of these files are placed in the
23788 same directory as the object file, and contain data stored in a
23789 platform-independent format.
23791 The `.gcno' file is generated when the source file is compiled with
23792 the GCC `-ftest-coverage' option. It contains information to
23793 reconstruct the basic block graphs and assign source line numbers to
23796 The `.gcda' file is generated when a program containing object files
23797 built with the GCC `-fprofile-arcs' option is executed. A separate
23798 `.gcda' file is created for each object file compiled with this option.
23799 It contains arc transition counts, and some summary information.
23801 The full details of the file format is specified in `gcov-io.h', and
23802 functions provided in that header file should be used to access the
23806 File: gcc.info, Node: Trouble, Next: Bugs, Prev: Gcov, Up: Top
23808 10 Known Causes of Trouble with GCC
23809 ***********************************
23811 This section describes known problems that affect users of GCC. Most
23812 of these are not GCC bugs per se--if they were, we would fix them. But
23813 the result for a user may be like the result of a bug.
23815 Some of these problems are due to bugs in other software, some are
23816 missing features that are too much work to add, and some are places
23817 where people's opinions differ as to what is best.
23821 * Actual Bugs:: Bugs we will fix later.
23822 * Cross-Compiler Problems:: Common problems of cross compiling with GCC.
23823 * Interoperation:: Problems using GCC with other compilers,
23824 and with certain linkers, assemblers and debuggers.
23825 * Incompatibilities:: GCC is incompatible with traditional C.
23826 * Fixed Headers:: GCC uses corrected versions of system header files.
23827 This is necessary, but doesn't always work smoothly.
23828 * Standard Libraries:: GCC uses the system C library, which might not be
23829 compliant with the ISO C standard.
23830 * Disappointments:: Regrettable things we can't change, but not quite bugs.
23831 * C++ Misunderstandings:: Common misunderstandings with GNU C++.
23832 * Protoize Caveats:: Things to watch out for when using `protoize'.
23833 * Non-bugs:: Things we think are right, but some others disagree.
23834 * Warnings and Errors:: Which problems in your code get warnings,
23835 and which get errors.
23838 File: gcc.info, Node: Actual Bugs, Next: Cross-Compiler Problems, Up: Trouble
23840 10.1 Actual Bugs We Haven't Fixed Yet
23841 =====================================
23843 * The `fixincludes' script interacts badly with automounters; if the
23844 directory of system header files is automounted, it tends to be
23845 unmounted while `fixincludes' is running. This would seem to be a
23846 bug in the automounter. We don't know any good way to work around
23849 * The `fixproto' script will sometimes add prototypes for the
23850 `sigsetjmp' and `siglongjmp' functions that reference the
23851 `jmp_buf' type before that type is defined. To work around this,
23852 edit the offending file and place the typedef in front of the
23856 File: gcc.info, Node: Cross-Compiler Problems, Next: Interoperation, Prev: Actual Bugs, Up: Trouble
23858 10.2 Cross-Compiler Problems
23859 ============================
23861 You may run into problems with cross compilation on certain machines,
23862 for several reasons.
23864 * At present, the program `mips-tfile' which adds debug support to
23865 object files on MIPS systems does not work in a cross compile
23869 File: gcc.info, Node: Interoperation, Next: Incompatibilities, Prev: Cross-Compiler Problems, Up: Trouble
23871 10.3 Interoperation
23872 ===================
23874 This section lists various difficulties encountered in using GCC
23875 together with other compilers or with the assemblers, linkers,
23876 libraries and debuggers on certain systems.
23878 * On many platforms, GCC supports a different ABI for C++ than do
23879 other compilers, so the object files compiled by GCC cannot be
23880 used with object files generated by another C++ compiler.
23882 An area where the difference is most apparent is name mangling.
23883 The use of different name mangling is intentional, to protect you
23884 from more subtle problems. Compilers differ as to many internal
23885 details of C++ implementation, including: how class instances are
23886 laid out, how multiple inheritance is implemented, and how virtual
23887 function calls are handled. If the name encoding were made the
23888 same, your programs would link against libraries provided from
23889 other compilers--but the programs would then crash when run.
23890 Incompatible libraries are then detected at link time, rather than
23893 * On some BSD systems, including some versions of Ultrix, use of
23894 profiling causes static variable destructors (currently used only
23895 in C++) not to be run.
23897 * On some SGI systems, when you use `-lgl_s' as an option, it gets
23898 translated magically to `-lgl_s -lX11_s -lc_s'. Naturally, this
23899 does not happen when you use GCC. You must specify all three
23900 options explicitly.
23902 * On a SPARC, GCC aligns all values of type `double' on an 8-byte
23903 boundary, and it expects every `double' to be so aligned. The Sun
23904 compiler usually gives `double' values 8-byte alignment, with one
23905 exception: function arguments of type `double' may not be aligned.
23907 As a result, if a function compiled with Sun CC takes the address
23908 of an argument of type `double' and passes this pointer of type
23909 `double *' to a function compiled with GCC, dereferencing the
23910 pointer may cause a fatal signal.
23912 One way to solve this problem is to compile your entire program
23913 with GCC. Another solution is to modify the function that is
23914 compiled with Sun CC to copy the argument into a local variable;
23915 local variables are always properly aligned. A third solution is
23916 to modify the function that uses the pointer to dereference it via
23917 the following function `access_double' instead of directly with
23921 access_double (double *unaligned_ptr)
23923 union d2i { double d; int i[2]; };
23925 union d2i *p = (union d2i *) unaligned_ptr;
23934 Storing into the pointer can be done likewise with the same union.
23936 * On Solaris, the `malloc' function in the `libmalloc.a' library may
23937 allocate memory that is only 4 byte aligned. Since GCC on the
23938 SPARC assumes that doubles are 8 byte aligned, this may result in a
23939 fatal signal if doubles are stored in memory allocated by the
23940 `libmalloc.a' library.
23942 The solution is to not use the `libmalloc.a' library. Use instead
23943 `malloc' and related functions from `libc.a'; they do not have
23946 * On the HP PA machine, ADB sometimes fails to work on functions
23947 compiled with GCC. Specifically, it fails to work on functions
23948 that use `alloca' or variable-size arrays. This is because GCC
23949 doesn't generate HP-UX unwind descriptors for such functions. It
23950 may even be impossible to generate them.
23952 * Debugging (`-g') is not supported on the HP PA machine, unless you
23953 use the preliminary GNU tools.
23955 * Taking the address of a label may generate errors from the HP-UX
23956 PA assembler. GAS for the PA does not have this problem.
23958 * Using floating point parameters for indirect calls to static
23959 functions will not work when using the HP assembler. There simply
23960 is no way for GCC to specify what registers hold arguments for
23961 static functions when using the HP assembler. GAS for the PA does
23962 not have this problem.
23964 * In extremely rare cases involving some very large functions you may
23965 receive errors from the HP linker complaining about an out of
23966 bounds unconditional branch offset. This used to occur more often
23967 in previous versions of GCC, but is now exceptionally rare. If
23968 you should run into it, you can work around by making your
23971 * GCC compiled code sometimes emits warnings from the HP-UX
23972 assembler of the form:
23974 (warning) Use of GR3 when
23975 frame >= 8192 may cause conflict.
23977 These warnings are harmless and can be safely ignored.
23979 * In extremely rare cases involving some very large functions you may
23980 receive errors from the AIX Assembler complaining about a
23981 displacement that is too large. If you should run into it, you
23982 can work around by making your function smaller.
23984 * The `libstdc++.a' library in GCC relies on the SVR4 dynamic linker
23985 semantics which merges global symbols between libraries and
23986 applications, especially necessary for C++ streams functionality.
23987 This is not the default behavior of AIX shared libraries and
23988 dynamic linking. `libstdc++.a' is built on AIX with
23989 "runtime-linking" enabled so that symbol merging can occur. To
23990 utilize this feature, the application linked with `libstdc++.a'
23991 must include the `-Wl,-brtl' flag on the link line. G++ cannot
23992 impose this because this option may interfere with the semantics
23993 of the user program and users may not always use `g++' to link his
23994 or her application. Applications are not required to use the
23995 `-Wl,-brtl' flag on the link line--the rest of the `libstdc++.a'
23996 library which is not dependent on the symbol merging semantics
23997 will continue to function correctly.
23999 * An application can interpose its own definition of functions for
24000 functions invoked by `libstdc++.a' with "runtime-linking" enabled
24001 on AIX. To accomplish this the application must be linked with
24002 "runtime-linking" option and the functions explicitly must be
24003 exported by the application (`-Wl,-brtl,-bE:exportfile').
24005 * AIX on the RS/6000 provides support (NLS) for environments outside
24006 of the United States. Compilers and assemblers use NLS to support
24007 locale-specific representations of various objects including
24008 floating-point numbers (`.' vs `,' for separating decimal
24009 fractions). There have been problems reported where the library
24010 linked with GCC does not produce the same floating-point formats
24011 that the assembler accepts. If you have this problem, set the
24012 `LANG' environment variable to `C' or `En_US'.
24014 * Even if you specify `-fdollars-in-identifiers', you cannot
24015 successfully use `$' in identifiers on the RS/6000 due to a
24016 restriction in the IBM assembler. GAS supports these identifiers.
24018 * On Ultrix, the Fortran compiler expects registers 2 through 5 to
24019 be saved by function calls. However, the C compiler uses
24020 conventions compatible with BSD Unix: registers 2 through 5 may be
24021 clobbered by function calls.
24023 GCC uses the same convention as the Ultrix C compiler. You can use
24024 these options to produce code compatible with the Fortran compiler:
24026 -fcall-saved-r2 -fcall-saved-r3 -fcall-saved-r4 -fcall-saved-r5
24029 File: gcc.info, Node: Incompatibilities, Next: Fixed Headers, Prev: Interoperation, Up: Trouble
24031 10.4 Incompatibilities of GCC
24032 =============================
24034 There are several noteworthy incompatibilities between GNU C and K&R
24035 (non-ISO) versions of C.
24037 * GCC normally makes string constants read-only. If several
24038 identical-looking string constants are used, GCC stores only one
24039 copy of the string.
24041 One consequence is that you cannot call `mktemp' with a string
24042 constant argument. The function `mktemp' always alters the string
24043 its argument points to.
24045 Another consequence is that `sscanf' does not work on some very
24046 old systems when passed a string constant as its format control
24047 string or input. This is because `sscanf' incorrectly tries to
24048 write into the string constant. Likewise `fscanf' and `scanf'.
24050 The solution to these problems is to change the program to use
24051 `char'-array variables with initialization strings for these
24052 purposes instead of string constants.
24054 * `-2147483648' is positive.
24056 This is because 2147483648 cannot fit in the type `int', so
24057 (following the ISO C rules) its data type is `unsigned long int'.
24058 Negating this value yields 2147483648 again.
24060 * GCC does not substitute macro arguments when they appear inside of
24061 string constants. For example, the following macro in GCC
24065 will produce output `"a"' regardless of what the argument A is.
24067 * When you use `setjmp' and `longjmp', the only automatic variables
24068 guaranteed to remain valid are those declared `volatile'. This is
24069 a consequence of automatic register allocation. Consider this
24083 /* `longjmp (j)' may occur in `fun3'. */
24084 return a + fun3 ();
24087 Here `a' may or may not be restored to its first value when the
24088 `longjmp' occurs. If `a' is allocated in a register, then its
24089 first value is restored; otherwise, it keeps the last value stored
24092 If you use the `-W' option with the `-O' option, you will get a
24093 warning when GCC thinks such a problem might be possible.
24095 * Programs that use preprocessing directives in the middle of macro
24096 arguments do not work with GCC. For example, a program like this
24103 ISO C does not permit such a construct.
24105 * K&R compilers allow comments to cross over an inclusion boundary
24106 (i.e. started in an include file and ended in the including file).
24108 * Declarations of external variables and functions within a block
24109 apply only to the block containing the declaration. In other
24110 words, they have the same scope as any other declaration in the
24113 In some other C compilers, a `extern' declaration affects all the
24114 rest of the file even if it happens within a block.
24116 * In traditional C, you can combine `long', etc., with a typedef
24117 name, as shown here:
24120 typedef long foo bar;
24122 In ISO C, this is not allowed: `long' and other type modifiers
24123 require an explicit `int'.
24125 * PCC allows typedef names to be used as function parameters.
24127 * Traditional C allows the following erroneous pair of declarations
24128 to appear together in a given scope:
24133 * GCC treats all characters of identifiers as significant.
24134 According to K&R-1 (2.2), "No more than the first eight characters
24135 are significant, although more may be used.". Also according to
24136 K&R-1 (2.2), "An identifier is a sequence of letters and digits;
24137 the first character must be a letter. The underscore _ counts as
24138 a letter.", but GCC also allows dollar signs in identifiers.
24140 * PCC allows whitespace in the middle of compound assignment
24141 operators such as `+='. GCC, following the ISO standard, does not
24144 * GCC complains about unterminated character constants inside of
24145 preprocessing conditionals that fail. Some programs have English
24146 comments enclosed in conditionals that are guaranteed to fail; if
24147 these comments contain apostrophes, GCC will probably report an
24148 error. For example, this code would produce an error:
24151 You can't expect this to work.
24154 The best solution to such a problem is to put the text into an
24155 actual C comment delimited by `/*...*/'.
24157 * Many user programs contain the declaration `long time ();'. In the
24158 past, the system header files on many systems did not actually
24159 declare `time', so it did not matter what type your program
24160 declared it to return. But in systems with ISO C headers, `time'
24161 is declared to return `time_t', and if that is not the same as
24162 `long', then `long time ();' is erroneous.
24164 The solution is to change your program to use appropriate system
24165 headers (`<time.h>' on systems with ISO C headers) and not to
24166 declare `time' if the system header files declare it, or failing
24167 that to use `time_t' as the return type of `time'.
24169 * When compiling functions that return `float', PCC converts it to a
24170 double. GCC actually returns a `float'. If you are concerned
24171 with PCC compatibility, you should declare your functions to return
24172 `double'; you might as well say what you mean.
24174 * When compiling functions that return structures or unions, GCC
24175 output code normally uses a method different from that used on most
24176 versions of Unix. As a result, code compiled with GCC cannot call
24177 a structure-returning function compiled with PCC, and vice versa.
24179 The method used by GCC is as follows: a structure or union which is
24180 1, 2, 4 or 8 bytes long is returned like a scalar. A structure or
24181 union with any other size is stored into an address supplied by
24182 the caller (usually in a special, fixed register, but on some
24183 machines it is passed on the stack). The target hook
24184 `TARGET_STRUCT_VALUE_RTX' tells GCC where to pass this address.
24186 By contrast, PCC on most target machines returns structures and
24187 unions of any size by copying the data into an area of static
24188 storage, and then returning the address of that storage as if it
24189 were a pointer value. The caller must copy the data from that
24190 memory area to the place where the value is wanted. GCC does not
24191 use this method because it is slower and nonreentrant.
24193 On some newer machines, PCC uses a reentrant convention for all
24194 structure and union returning. GCC on most of these machines uses
24195 a compatible convention when returning structures and unions in
24196 memory, but still returns small structures and unions in registers.
24198 You can tell GCC to use a compatible convention for all structure
24199 and union returning with the option `-fpcc-struct-return'.
24201 * GCC complains about program fragments such as `0x74ae-0x4000'
24202 which appear to be two hexadecimal constants separated by the minus
24203 operator. Actually, this string is a single "preprocessing token".
24204 Each such token must correspond to one token in C. Since this
24205 does not, GCC prints an error message. Although it may appear
24206 obvious that what is meant is an operator and two values, the ISO
24207 C standard specifically requires that this be treated as erroneous.
24209 A "preprocessing token" is a "preprocessing number" if it begins
24210 with a digit and is followed by letters, underscores, digits,
24211 periods and `e+', `e-', `E+', `E-', `p+', `p-', `P+', or `P-'
24212 character sequences. (In strict C89 mode, the sequences `p+',
24213 `p-', `P+' and `P-' cannot appear in preprocessing numbers.)
24215 To make the above program fragment valid, place whitespace in
24216 front of the minus sign. This whitespace will end the
24217 preprocessing number.
24220 File: gcc.info, Node: Fixed Headers, Next: Standard Libraries, Prev: Incompatibilities, Up: Trouble
24222 10.5 Fixed Header Files
24223 =======================
24225 GCC needs to install corrected versions of some system header files.
24226 This is because most target systems have some header files that won't
24227 work with GCC unless they are changed. Some have bugs, some are
24228 incompatible with ISO C, and some depend on special features of other
24231 Installing GCC automatically creates and installs the fixed header
24232 files, by running a program called `fixincludes'. Normally, you don't
24233 need to pay attention to this. But there are cases where it doesn't do
24234 the right thing automatically.
24236 * If you update the system's header files, such as by installing a
24237 new system version, the fixed header files of GCC are not
24238 automatically updated. They can be updated using the `mkheaders'
24239 script installed in `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
24241 * On some systems, header file directories contain machine-specific
24242 symbolic links in certain places. This makes it possible to share
24243 most of the header files among hosts running the same version of
24244 the system on different machine models.
24246 The programs that fix the header files do not understand this
24247 special way of using symbolic links; therefore, the directory of
24248 fixed header files is good only for the machine model used to
24251 It is possible to make separate sets of fixed header files for the
24252 different machine models, and arrange a structure of symbolic
24253 links so as to use the proper set, but you'll have to do this by
24257 File: gcc.info, Node: Standard Libraries, Next: Disappointments, Prev: Fixed Headers, Up: Trouble
24259 10.6 Standard Libraries
24260 =======================
24262 GCC by itself attempts to be a conforming freestanding implementation.
24263 *Note Language Standards Supported by GCC: Standards, for details of
24264 what this means. Beyond the library facilities required of such an
24265 implementation, the rest of the C library is supplied by the vendor of
24266 the operating system. If that C library doesn't conform to the C
24267 standards, then your programs might get warnings (especially when using
24268 `-Wall') that you don't expect.
24270 For example, the `sprintf' function on SunOS 4.1.3 returns `char *'
24271 while the C standard says that `sprintf' returns an `int'. The
24272 `fixincludes' program could make the prototype for this function match
24273 the Standard, but that would be wrong, since the function will still
24276 If you need a Standard compliant library, then you need to find one, as
24277 GCC does not provide one. The GNU C library (called `glibc') provides
24278 ISO C, POSIX, BSD, SystemV and X/Open compatibility for GNU/Linux and
24279 HURD-based GNU systems; no recent version of it supports other systems,
24280 though some very old versions did. Version 2.2 of the GNU C library
24281 includes nearly complete C99 support. You could also ask your
24282 operating system vendor if newer libraries are available.
24285 File: gcc.info, Node: Disappointments, Next: C++ Misunderstandings, Prev: Standard Libraries, Up: Trouble
24287 10.7 Disappointments and Misunderstandings
24288 ==========================================
24290 These problems are perhaps regrettable, but we don't know any practical
24293 * Certain local variables aren't recognized by debuggers when you
24294 compile with optimization.
24296 This occurs because sometimes GCC optimizes the variable out of
24297 existence. There is no way to tell the debugger how to compute the
24298 value such a variable "would have had", and it is not clear that
24299 would be desirable anyway. So GCC simply does not mention the
24300 eliminated variable when it writes debugging information.
24302 You have to expect a certain amount of disagreement between the
24303 executable and your source code, when you use optimization.
24305 * Users often think it is a bug when GCC reports an error for code
24308 int foo (struct mumble *);
24310 struct mumble { ... };
24312 int foo (struct mumble *x)
24315 This code really is erroneous, because the scope of `struct
24316 mumble' in the prototype is limited to the argument list
24317 containing it. It does not refer to the `struct mumble' defined
24318 with file scope immediately below--they are two unrelated types
24319 with similar names in different scopes.
24321 But in the definition of `foo', the file-scope type is used
24322 because that is available to be inherited. Thus, the definition
24323 and the prototype do not match, and you get an error.
24325 This behavior may seem silly, but it's what the ISO standard
24326 specifies. It is easy enough for you to make your code work by
24327 moving the definition of `struct mumble' above the prototype.
24328 It's not worth being incompatible with ISO C just to avoid an
24329 error for the example shown above.
24331 * Accesses to bit-fields even in volatile objects works by accessing
24332 larger objects, such as a byte or a word. You cannot rely on what
24333 size of object is accessed in order to read or write the
24334 bit-field; it may even vary for a given bit-field according to the
24337 If you care about controlling the amount of memory that is
24338 accessed, use volatile but do not use bit-fields.
24340 * GCC comes with shell scripts to fix certain known problems in
24341 system header files. They install corrected copies of various
24342 header files in a special directory where only GCC will normally
24343 look for them. The scripts adapt to various systems by searching
24344 all the system header files for the problem cases that we know
24347 If new system header files are installed, nothing automatically
24348 arranges to update the corrected header files. They can be
24349 updated using the `mkheaders' script installed in
24350 `LIBEXECDIR/gcc/TARGET/VERSION/install-tools/'.
24352 * On 68000 and x86 systems, for instance, you can get paradoxical
24353 results if you test the precise values of floating point numbers.
24354 For example, you can find that a floating point value which is not
24355 a NaN is not equal to itself. This results from the fact that the
24356 floating point registers hold a few more bits of precision than
24357 fit in a `double' in memory. Compiled code moves values between
24358 memory and floating point registers at its convenience, and moving
24359 them into memory truncates them.
24361 You can partially avoid this problem by using the `-ffloat-store'
24362 option (*note Optimize Options::).
24364 * On AIX and other platforms without weak symbol support, templates
24365 need to be instantiated explicitly and symbols for static members
24366 of templates will not be generated.
24368 * On AIX, GCC scans object files and library archives for static
24369 constructors and destructors when linking an application before the
24370 linker prunes unreferenced symbols. This is necessary to prevent
24371 the AIX linker from mistakenly assuming that static constructor or
24372 destructor are unused and removing them before the scanning can
24373 occur. All static constructors and destructors found will be
24374 referenced even though the modules in which they occur may not be
24375 used by the program. This may lead to both increased executable
24376 size and unexpected symbol references.
24379 File: gcc.info, Node: C++ Misunderstandings, Next: Protoize Caveats, Prev: Disappointments, Up: Trouble
24381 10.8 Common Misunderstandings with GNU C++
24382 ==========================================
24384 C++ is a complex language and an evolving one, and its standard
24385 definition (the ISO C++ standard) was only recently completed. As a
24386 result, your C++ compiler may occasionally surprise you, even when its
24387 behavior is correct. This section discusses some areas that frequently
24388 give rise to questions of this sort.
24392 * Static Definitions:: Static member declarations are not definitions
24393 * Name lookup:: Name lookup, templates, and accessing members of base classes
24394 * Temporaries:: Temporaries may vanish before you expect
24395 * Copy Assignment:: Copy Assignment operators copy virtual bases twice
24398 File: gcc.info, Node: Static Definitions, Next: Name lookup, Up: C++ Misunderstandings
24400 10.8.1 Declare _and_ Define Static Members
24401 ------------------------------------------
24403 When a class has static data members, it is not enough to _declare_ the
24404 static member; you must also _define_ it. For example:
24413 This declaration only establishes that the class `Foo' has an `int'
24414 named `Foo::bar', and a member function named `Foo::method'. But you
24415 still need to define _both_ `method' and `bar' elsewhere. According to
24416 the ISO standard, you must supply an initializer in one (and only one)
24417 source file, such as:
24421 Other C++ compilers may not correctly implement the standard behavior.
24422 As a result, when you switch to `g++' from one of these compilers, you
24423 may discover that a program that appeared to work correctly in fact
24424 does not conform to the standard: `g++' reports as undefined symbols
24425 any static data members that lack definitions.
24428 File: gcc.info, Node: Name lookup, Next: Temporaries, Prev: Static Definitions, Up: C++ Misunderstandings
24430 10.8.2 Name lookup, templates, and accessing members of base classes
24431 --------------------------------------------------------------------
24433 The C++ standard prescribes that all names that are not dependent on
24434 template parameters are bound to their present definitions when parsing
24435 a template function or class.(1) Only names that are dependent are
24436 looked up at the point of instantiation. For example, consider
24441 template <typename T>
24450 static const int N;
24453 Here, the names `foo' and `N' appear in a context that does not depend
24454 on the type of `T'. The compiler will thus require that they are
24455 defined in the context of use in the template, not only before the
24456 point of instantiation, and will here use `::foo(double)' and `A::N',
24457 respectively. In particular, it will convert the integer value to a
24458 `double' when passing it to `::foo(double)'.
24460 Conversely, `bar' and the call to `foo' in the fourth marked line are
24461 used in contexts that do depend on the type of `T', so they are only
24462 looked up at the point of instantiation, and you can provide
24463 declarations for them after declaring the template, but before
24464 instantiating it. In particular, if you instantiate `A::f<int>', the
24465 last line will call an overloaded `::foo(int)' if one was provided,
24466 even if after the declaration of `struct A'.
24468 This distinction between lookup of dependent and non-dependent names is
24469 called two-stage (or dependent) name lookup. G++ implements it since
24472 Two-stage name lookup sometimes leads to situations with behavior
24473 different from non-template codes. The most common is probably this:
24475 template <typename T> struct Base {
24479 template <typename T> struct Derived : public Base<T> {
24480 int get_i() { return i; }
24483 In `get_i()', `i' is not used in a dependent context, so the compiler
24484 will look for a name declared at the enclosing namespace scope (which
24485 is the global scope here). It will not look into the base class, since
24486 that is dependent and you may declare specializations of `Base' even
24487 after declaring `Derived', so the compiler can't really know what `i'
24488 would refer to. If there is no global variable `i', then you will get
24491 In order to make it clear that you want the member of the base class,
24492 you need to defer lookup until instantiation time, at which the base
24493 class is known. For this, you need to access `i' in a dependent
24494 context, by either using `this->i' (remember that `this' is of type
24495 `Derived<T>*', so is obviously dependent), or using `Base<T>::i'.
24496 Alternatively, `Base<T>::i' might be brought into scope by a
24497 `using'-declaration.
24499 Another, similar example involves calling member functions of a base
24502 template <typename T> struct Base {
24506 template <typename T> struct Derived : Base<T> {
24507 int g() { return f(); };
24510 Again, the call to `f()' is not dependent on template arguments (there
24511 are no arguments that depend on the type `T', and it is also not
24512 otherwise specified that the call should be in a dependent context).
24513 Thus a global declaration of such a function must be available, since
24514 the one in the base class is not visible until instantiation time. The
24515 compiler will consequently produce the following error message:
24517 x.cc: In member function `int Derived<T>::g()':
24518 x.cc:6: error: there are no arguments to `f' that depend on a template
24519 parameter, so a declaration of `f' must be available
24520 x.cc:6: error: (if you use `-fpermissive', G++ will accept your code, but
24521 allowing the use of an undeclared name is deprecated)
24523 To make the code valid either use `this->f()', or `Base<T>::f()'.
24524 Using the `-fpermissive' flag will also let the compiler accept the
24525 code, by marking all function calls for which no declaration is visible
24526 at the time of definition of the template for later lookup at
24527 instantiation time, as if it were a dependent call. We do not
24528 recommend using `-fpermissive' to work around invalid code, and it will
24529 also only catch cases where functions in base classes are called, not
24530 where variables in base classes are used (as in the example above).
24532 Note that some compilers (including G++ versions prior to 3.4) get
24533 these examples wrong and accept above code without an error. Those
24534 compilers do not implement two-stage name lookup correctly.
24536 ---------- Footnotes ----------
24538 (1) The C++ standard just uses the term "dependent" for names that
24539 depend on the type or value of template parameters. This shorter term
24540 will also be used in the rest of this section.
24543 File: gcc.info, Node: Temporaries, Next: Copy Assignment, Prev: Name lookup, Up: C++ Misunderstandings
24545 10.8.3 Temporaries May Vanish Before You Expect
24546 -----------------------------------------------
24548 It is dangerous to use pointers or references to _portions_ of a
24549 temporary object. The compiler may very well delete the object before
24550 you expect it to, leaving a pointer to garbage. The most common place
24551 where this problem crops up is in classes like string classes,
24552 especially ones that define a conversion function to type `char *' or
24553 `const char *'--which is one reason why the standard `string' class
24554 requires you to call the `c_str' member function. However, any class
24555 that returns a pointer to some internal structure is potentially
24556 subject to this problem.
24558 For example, a program may use a function `strfunc' that returns
24559 `string' objects, and another function `charfunc' that operates on
24560 pointers to `char':
24563 void charfunc (const char *);
24568 const char *p = strfunc().c_str();
24575 In this situation, it may seem reasonable to save a pointer to the C
24576 string returned by the `c_str' member function and use that rather than
24577 call `c_str' repeatedly. However, the temporary string created by the
24578 call to `strfunc' is destroyed after `p' is initialized, at which point
24579 `p' is left pointing to freed memory.
24581 Code like this may run successfully under some other compilers,
24582 particularly obsolete cfront-based compilers that delete temporaries
24583 along with normal local variables. However, the GNU C++ behavior is
24584 standard-conforming, so if your program depends on late destruction of
24585 temporaries it is not portable.
24587 The safe way to write such code is to give the temporary a name, which
24588 forces it to remain until the end of the scope of the name. For
24591 const string& tmp = strfunc ();
24592 charfunc (tmp.c_str ());
24595 File: gcc.info, Node: Copy Assignment, Prev: Temporaries, Up: C++ Misunderstandings
24597 10.8.4 Implicit Copy-Assignment for Virtual Bases
24598 -------------------------------------------------
24600 When a base class is virtual, only one subobject of the base class
24601 belongs to each full object. Also, the constructors and destructors are
24602 invoked only once, and called from the most-derived class. However,
24603 such objects behave unspecified when being assigned. For example:
24607 Base(char *n) : name(strdup(n)){}
24608 Base& operator= (const Base& other){
24610 name = strdup (other.name);
24614 struct A:virtual Base{
24619 struct B:virtual Base{
24624 struct Derived:public A, public B{
24625 Derived():Base("Derived"){}
24628 void func(Derived &d1, Derived &d2)
24633 The C++ standard specifies that `Base::Base' is only called once when
24634 constructing or copy-constructing a Derived object. It is unspecified
24635 whether `Base::operator=' is called more than once when the implicit
24636 copy-assignment for Derived objects is invoked (as it is inside `func'
24639 G++ implements the "intuitive" algorithm for copy-assignment: assign
24640 all direct bases, then assign all members. In that algorithm, the
24641 virtual base subobject can be encountered more than once. In the
24642 example, copying proceeds in the following order: `val', `name' (via
24643 `strdup'), `bval', and `name' again.
24645 If application code relies on copy-assignment, a user-defined
24646 copy-assignment operator removes any uncertainties. With such an
24647 operator, the application can define whether and how the virtual base
24648 subobject is assigned.
24651 File: gcc.info, Node: Protoize Caveats, Next: Non-bugs, Prev: C++ Misunderstandings, Up: Trouble
24653 10.9 Caveats of using `protoize'
24654 ================================
24656 The conversion programs `protoize' and `unprotoize' can sometimes
24657 change a source file in a way that won't work unless you rearrange it.
24659 * `protoize' can insert references to a type name or type tag before
24660 the definition, or in a file where they are not defined.
24662 If this happens, compiler error messages should show you where the
24663 new references are, so fixing the file by hand is straightforward.
24665 * There are some C constructs which `protoize' cannot figure out.
24666 For example, it can't determine argument types for declaring a
24667 pointer-to-function variable; this you must do by hand. `protoize'
24668 inserts a comment containing `???' each time it finds such a
24669 variable; so you can find all such variables by searching for this
24670 string. ISO C does not require declaring the argument types of
24671 pointer-to-function types.
24673 * Using `unprotoize' can easily introduce bugs. If the program
24674 relied on prototypes to bring about conversion of arguments, these
24675 conversions will not take place in the program without prototypes.
24676 One case in which you can be sure `unprotoize' is safe is when you
24677 are removing prototypes that were made with `protoize'; if the
24678 program worked before without any prototypes, it will work again
24681 You can find all the places where this problem might occur by
24682 compiling the program with the `-Wconversion' option. It prints a
24683 warning whenever an argument is converted.
24685 * Both conversion programs can be confused if there are macro calls
24686 in and around the text to be converted. In other words, the
24687 standard syntax for a declaration or definition must not result
24688 from expanding a macro. This problem is inherent in the design of
24689 C and cannot be fixed. If only a few functions have confusing
24690 macro calls, you can easily convert them manually.
24692 * `protoize' cannot get the argument types for a function whose
24693 definition was not actually compiled due to preprocessing
24694 conditionals. When this happens, `protoize' changes nothing in
24695 regard to such a function. `protoize' tries to detect such
24696 instances and warn about them.
24698 You can generally work around this problem by using `protoize' step
24699 by step, each time specifying a different set of `-D' options for
24700 compilation, until all of the functions have been converted.
24701 There is no automatic way to verify that you have got them all,
24704 * Confusion may result if there is an occasion to convert a function
24705 declaration or definition in a region of source code where there
24706 is more than one formal parameter list present. Thus, attempts to
24707 convert code containing multiple (conditionally compiled) versions
24708 of a single function header (in the same vicinity) may not produce
24709 the desired (or expected) results.
24711 If you plan on converting source files which contain such code, it
24712 is recommended that you first make sure that each conditionally
24713 compiled region of source code which contains an alternative
24714 function header also contains at least one additional follower
24715 token (past the final right parenthesis of the function header).
24716 This should circumvent the problem.
24718 * `unprotoize' can become confused when trying to convert a function
24719 definition or declaration which contains a declaration for a
24720 pointer-to-function formal argument which has the same name as the
24721 function being defined or declared. We recommend you avoid such
24722 choices of formal parameter names.
24724 * You might also want to correct some of the indentation by hand and
24725 break long lines. (The conversion programs don't write lines
24726 longer than eighty characters in any case.)
24729 File: gcc.info, Node: Non-bugs, Next: Warnings and Errors, Prev: Protoize Caveats, Up: Trouble
24731 10.10 Certain Changes We Don't Want to Make
24732 ===========================================
24734 This section lists changes that people frequently request, but which we
24735 do not make because we think GCC is better without them.
24737 * Checking the number and type of arguments to a function which has
24738 an old-fashioned definition and no prototype.
24740 Such a feature would work only occasionally--only for calls that
24741 appear in the same file as the called function, following the
24742 definition. The only way to check all calls reliably is to add a
24743 prototype for the function. But adding a prototype eliminates the
24744 motivation for this feature. So the feature is not worthwhile.
24746 * Warning about using an expression whose type is signed as a shift
24749 Shift count operands are probably signed more often than unsigned.
24750 Warning about this would cause far more annoyance than good.
24752 * Warning about assigning a signed value to an unsigned variable.
24754 Such assignments must be very common; warning about them would
24755 cause more annoyance than good.
24757 * Warning when a non-void function value is ignored.
24759 C contains many standard functions that return a value that most
24760 programs choose to ignore. One obvious example is `printf'.
24761 Warning about this practice only leads the defensive programmer to
24762 clutter programs with dozens of casts to `void'. Such casts are
24763 required so frequently that they become visual noise. Writing
24764 those casts becomes so automatic that they no longer convey useful
24765 information about the intentions of the programmer. For functions
24766 where the return value should never be ignored, use the
24767 `warn_unused_result' function attribute (*note Function
24770 * Making `-fshort-enums' the default.
24772 This would cause storage layout to be incompatible with most other
24773 C compilers. And it doesn't seem very important, given that you
24774 can get the same result in other ways. The case where it matters
24775 most is when the enumeration-valued object is inside a structure,
24776 and in that case you can specify a field width explicitly.
24778 * Making bit-fields unsigned by default on particular machines where
24779 "the ABI standard" says to do so.
24781 The ISO C standard leaves it up to the implementation whether a
24782 bit-field declared plain `int' is signed or not. This in effect
24783 creates two alternative dialects of C.
24785 The GNU C compiler supports both dialects; you can specify the
24786 signed dialect with `-fsigned-bitfields' and the unsigned dialect
24787 with `-funsigned-bitfields'. However, this leaves open the
24788 question of which dialect to use by default.
24790 Currently, the preferred dialect makes plain bit-fields signed,
24791 because this is simplest. Since `int' is the same as `signed int'
24792 in every other context, it is cleanest for them to be the same in
24793 bit-fields as well.
24795 Some computer manufacturers have published Application Binary
24796 Interface standards which specify that plain bit-fields should be
24797 unsigned. It is a mistake, however, to say anything about this
24798 issue in an ABI. This is because the handling of plain bit-fields
24799 distinguishes two dialects of C. Both dialects are meaningful on
24800 every type of machine. Whether a particular object file was
24801 compiled using signed bit-fields or unsigned is of no concern to
24802 other object files, even if they access the same bit-fields in the
24803 same data structures.
24805 A given program is written in one or the other of these two
24806 dialects. The program stands a chance to work on most any machine
24807 if it is compiled with the proper dialect. It is unlikely to work
24808 at all if compiled with the wrong dialect.
24810 Many users appreciate the GNU C compiler because it provides an
24811 environment that is uniform across machines. These users would be
24812 inconvenienced if the compiler treated plain bit-fields
24813 differently on certain machines.
24815 Occasionally users write programs intended only for a particular
24816 machine type. On these occasions, the users would benefit if the
24817 GNU C compiler were to support by default the same dialect as the
24818 other compilers on that machine. But such applications are rare.
24819 And users writing a program to run on more than one type of
24820 machine cannot possibly benefit from this kind of compatibility.
24822 This is why GCC does and will treat plain bit-fields in the same
24823 fashion on all types of machines (by default).
24825 There are some arguments for making bit-fields unsigned by default
24826 on all machines. If, for example, this becomes a universal de
24827 facto standard, it would make sense for GCC to go along with it.
24828 This is something to be considered in the future.
24830 (Of course, users strongly concerned about portability should
24831 indicate explicitly in each bit-field whether it is signed or not.
24832 In this way, they write programs which have the same meaning in
24835 * Undefining `__STDC__' when `-ansi' is not used.
24837 Currently, GCC defines `__STDC__' unconditionally. This provides
24838 good results in practice.
24840 Programmers normally use conditionals on `__STDC__' to ask whether
24841 it is safe to use certain features of ISO C, such as function
24842 prototypes or ISO token concatenation. Since plain `gcc' supports
24843 all the features of ISO C, the correct answer to these questions is
24846 Some users try to use `__STDC__' to check for the availability of
24847 certain library facilities. This is actually incorrect usage in
24848 an ISO C program, because the ISO C standard says that a conforming
24849 freestanding implementation should define `__STDC__' even though it
24850 does not have the library facilities. `gcc -ansi -pedantic' is a
24851 conforming freestanding implementation, and it is therefore
24852 required to define `__STDC__', even though it does not come with
24855 Sometimes people say that defining `__STDC__' in a compiler that
24856 does not completely conform to the ISO C standard somehow violates
24857 the standard. This is illogical. The standard is a standard for
24858 compilers that claim to support ISO C, such as `gcc -ansi'--not
24859 for other compilers such as plain `gcc'. Whatever the ISO C
24860 standard says is relevant to the design of plain `gcc' without
24861 `-ansi' only for pragmatic reasons, not as a requirement.
24863 GCC normally defines `__STDC__' to be 1, and in addition defines
24864 `__STRICT_ANSI__' if you specify the `-ansi' option, or a `-std'
24865 option for strict conformance to some version of ISO C. On some
24866 hosts, system include files use a different convention, where
24867 `__STDC__' is normally 0, but is 1 if the user specifies strict
24868 conformance to the C Standard. GCC follows the host convention
24869 when processing system include files, but when processing user
24870 files it follows the usual GNU C convention.
24872 * Undefining `__STDC__' in C++.
24874 Programs written to compile with C++-to-C translators get the
24875 value of `__STDC__' that goes with the C compiler that is
24876 subsequently used. These programs must test `__STDC__' to
24877 determine what kind of C preprocessor that compiler uses: whether
24878 they should concatenate tokens in the ISO C fashion or in the
24879 traditional fashion.
24881 These programs work properly with GNU C++ if `__STDC__' is defined.
24882 They would not work otherwise.
24884 In addition, many header files are written to provide prototypes
24885 in ISO C but not in traditional C. Many of these header files can
24886 work without change in C++ provided `__STDC__' is defined. If
24887 `__STDC__' is not defined, they will all fail, and will all need
24888 to be changed to test explicitly for C++ as well.
24890 * Deleting "empty" loops.
24892 Historically, GCC has not deleted "empty" loops under the
24893 assumption that the most likely reason you would put one in a
24894 program is to have a delay, so deleting them will not make real
24895 programs run any faster.
24897 However, the rationale here is that optimization of a nonempty loop
24898 cannot produce an empty one. This held for carefully written C
24899 compiled with less powerful optimizers but is not always the case
24900 for carefully written C++ or with more powerful optimizers.
24902 Thus GCC will remove operations from loops whenever it can
24903 determine those operations are not externally visible (apart from
24904 the time taken to execute them, of course). As GCC improves, it
24905 will remove the loop itself. Indeed, with `-funroll-loops' small
24906 loops can already be removed, so leaving an empty non-unrolled
24907 loop is both sub-optimal and inconsistent.
24909 Be aware of this when performing timing tests, for instance the
24910 following loop can be completely removed, provided
24911 `some_expression' can provably not change any global state.
24917 for (ix = 0; ix != 10000; ix++)
24918 sum += some_expression;
24921 Even though `sum' is accumulated in the loop, no use is made of
24922 that summation, so the accumulation can be removed.
24924 * Making side effects happen in the same order as in some other
24927 It is never safe to depend on the order of evaluation of side
24928 effects. For example, a function call like this may very well
24929 behave differently from one compiler to another:
24931 void func (int, int);
24936 There is no guarantee (in either the C or the C++ standard language
24937 definitions) that the increments will be evaluated in any
24938 particular order. Either increment might happen first. `func'
24939 might get the arguments `2, 3', or it might get `3, 2', or even
24942 * Making certain warnings into errors by default.
24944 Some ISO C testsuites report failure when the compiler does not
24945 produce an error message for a certain program.
24947 ISO C requires a "diagnostic" message for certain kinds of invalid
24948 programs, but a warning is defined by GCC to count as a
24949 diagnostic. If GCC produces a warning but not an error, that is
24950 correct ISO C support. If testsuites call this "failure", they
24951 should be run with the GCC option `-pedantic-errors', which will
24952 turn these warnings into errors.
24956 File: gcc.info, Node: Warnings and Errors, Prev: Non-bugs, Up: Trouble
24958 10.11 Warning Messages and Error Messages
24959 =========================================
24961 The GNU compiler can produce two kinds of diagnostics: errors and
24962 warnings. Each kind has a different purpose:
24964 "Errors" report problems that make it impossible to compile your
24965 program. GCC reports errors with the source file name and line
24966 number where the problem is apparent.
24968 "Warnings" report other unusual conditions in your code that _may_
24969 indicate a problem, although compilation can (and does) proceed.
24970 Warning messages also report the source file name and line number,
24971 but include the text `warning:' to distinguish them from error
24974 Warnings may indicate danger points where you should check to make sure
24975 that your program really does what you intend; or the use of obsolete
24976 features; or the use of nonstandard features of GNU C or C++. Many
24977 warnings are issued only if you ask for them, with one of the `-W'
24978 options (for instance, `-Wall' requests a variety of useful warnings).
24980 GCC always tries to compile your program if possible; it never
24981 gratuitously rejects a program whose meaning is clear merely because
24982 (for instance) it fails to conform to a standard. In some cases,
24983 however, the C and C++ standards specify that certain extensions are
24984 forbidden, and a diagnostic _must_ be issued by a conforming compiler.
24985 The `-pedantic' option tells GCC to issue warnings in such cases;
24986 `-pedantic-errors' says to make them errors instead. This does not
24987 mean that _all_ non-ISO constructs get warnings or errors.
24989 *Note Options to Request or Suppress Warnings: Warning Options, for
24990 more detail on these and related command-line options.
24993 File: gcc.info, Node: Bugs, Next: Service, Prev: Trouble, Up: Top
24998 Your bug reports play an essential role in making GCC reliable.
25000 When you encounter a problem, the first thing to do is to see if it is
25001 already known. *Note Trouble::. If it isn't known, then you should
25002 report the problem.
25006 * Criteria: Bug Criteria. Have you really found a bug?
25007 * Reporting: Bug Reporting. How to report a bug effectively.
25008 * Known: Trouble. Known problems.
25009 * Help: Service. Where to ask for help.
25012 File: gcc.info, Node: Bug Criteria, Next: Bug Reporting, Up: Bugs
25014 11.1 Have You Found a Bug?
25015 ==========================
25017 If you are not sure whether you have found a bug, here are some
25020 * If the compiler gets a fatal signal, for any input whatever, that
25021 is a compiler bug. Reliable compilers never crash.
25023 * If the compiler produces invalid assembly code, for any input
25024 whatever (except an `asm' statement), that is a compiler bug,
25025 unless the compiler reports errors (not just warnings) which would
25026 ordinarily prevent the assembler from being run.
25028 * If the compiler produces valid assembly code that does not
25029 correctly execute the input source code, that is a compiler bug.
25031 However, you must double-check to make sure, because you may have a
25032 program whose behavior is undefined, which happened by chance to
25033 give the desired results with another C or C++ compiler.
25035 For example, in many nonoptimizing compilers, you can write `x;'
25036 at the end of a function instead of `return x;', with the same
25037 results. But the value of the function is undefined if `return'
25038 is omitted; it is not a bug when GCC produces different results.
25040 Problems often result from expressions with two increment
25041 operators, as in `f (*p++, *p++)'. Your previous compiler might
25042 have interpreted that expression the way you intended; GCC might
25043 interpret it another way. Neither compiler is wrong. The bug is
25046 After you have localized the error to a single source line, it
25047 should be easy to check for these things. If your program is
25048 correct and well defined, you have found a compiler bug.
25050 * If the compiler produces an error message for valid input, that is
25053 * If the compiler does not produce an error message for invalid
25054 input, that is a compiler bug. However, you should note that your
25055 idea of "invalid input" might be someone else's idea of "an
25056 extension" or "support for traditional practice".
25058 * If you are an experienced user of one of the languages GCC
25059 supports, your suggestions for improvement of GCC are welcome in
25063 File: gcc.info, Node: Bug Reporting, Prev: Bug Criteria, Up: Bugs
25065 11.2 How and where to Report Bugs
25066 =================================
25068 Bugs should be reported to the GCC bug database. Please refer to
25069 `http://gcc.gnu.org/bugs.html' for up-to-date instructions how to
25070 submit bug reports. Copies of this file in HTML (`bugs.html') and
25071 plain text (`BUGS') are also part of GCC releases.
25074 File: gcc.info, Node: Service, Next: Contributing, Prev: Bugs, Up: Top
25076 12 How To Get Help with GCC
25077 ***************************
25079 If you need help installing, using or changing GCC, there are two ways
25082 * Send a message to a suitable network mailing list. First try
25083 <gcc-help@gcc.gnu.org> (for help installing or using GCC), and if
25084 that brings no response, try <gcc@gcc.gnu.org>. For help changing
25085 GCC, ask <gcc@gcc.gnu.org>. If you think you have found a bug in
25086 GCC, please report it following the instructions at *note Bug
25089 * Look in the service directory for someone who might help you for a
25090 fee. The service directory is found at
25091 `http://www.gnu.org/prep/service.html'.
25093 For further information, see `http://gcc.gnu.org/faq.html#support'.
25096 File: gcc.info, Node: Contributing, Next: Funding, Prev: Service, Up: Top
25098 13 Contributing to GCC Development
25099 **********************************
25101 If you would like to help pretest GCC releases to assure they work well,
25102 current development sources are available by CVS (see
25103 `http://gcc.gnu.org/cvs.html'). Source and binary snapshots are also
25104 available for FTP; see `http://gcc.gnu.org/snapshots.html'.
25106 If you would like to work on improvements to GCC, please read the
25107 advice at these URLs:
25109 `http://gcc.gnu.org/contribute.html'
25110 `http://gcc.gnu.org/contributewhy.html'
25112 for information on how to make useful contributions and avoid
25113 duplication of effort. Suggested projects are listed at
25114 `http://gcc.gnu.org/projects/'.
25117 File: gcc.info, Node: Funding, Next: GNU Project, Prev: Contributing, Up: Top
25119 Funding Free Software
25120 *********************
25122 If you want to have more free software a few years from now, it makes
25123 sense for you to help encourage people to contribute funds for its
25124 development. The most effective approach known is to encourage
25125 commercial redistributors to donate.
25127 Users of free software systems can boost the pace of development by
25128 encouraging for-a-fee distributors to donate part of their selling price
25129 to free software developers--the Free Software Foundation, and others.
25131 The way to convince distributors to do this is to demand it and expect
25132 it from them. So when you compare distributors, judge them partly by
25133 how much they give to free software development. Show distributors
25134 they must compete to be the one who gives the most.
25136 To make this approach work, you must insist on numbers that you can
25137 compare, such as, "We will donate ten dollars to the Frobnitz project
25138 for each disk sold." Don't be satisfied with a vague promise, such as
25139 "A portion of the profits are donated," since it doesn't give a basis
25142 Even a precise fraction "of the profits from this disk" is not very
25143 meaningful, since creative accounting and unrelated business decisions
25144 can greatly alter what fraction of the sales price counts as profit.
25145 If the price you pay is $50, ten percent of the profit is probably less
25146 than a dollar; it might be a few cents, or nothing at all.
25148 Some redistributors do development work themselves. This is useful
25149 too; but to keep everyone honest, you need to inquire how much they do,
25150 and what kind. Some kinds of development make much more long-term
25151 difference than others. For example, maintaining a separate version of
25152 a program contributes very little; maintaining the standard version of a
25153 program for the whole community contributes much. Easy new ports
25154 contribute little, since someone else would surely do them; difficult
25155 ports such as adding a new CPU to the GNU Compiler Collection
25156 contribute more; major new features or packages contribute the most.
25158 By establishing the idea that supporting further development is "the
25159 proper thing to do" when distributing free software for a fee, we can
25160 assure a steady flow of resources into making more free software.
25162 Copyright (C) 1994 Free Software Foundation, Inc.
25163 Verbatim copying and redistribution of this section is permitted
25164 without royalty; alteration is not permitted.
25167 File: gcc.info, Node: GNU Project, Next: Copying, Prev: Funding, Up: Top
25169 The GNU Project and GNU/Linux
25170 *****************************
25172 The GNU Project was launched in 1984 to develop a complete Unix-like
25173 operating system which is free software: the GNU system. (GNU is a
25174 recursive acronym for "GNU's Not Unix"; it is pronounced "guh-NEW".)
25175 Variants of the GNU operating system, which use the kernel Linux, are
25176 now widely used; though these systems are often referred to as "Linux",
25177 they are more accurately called GNU/Linux systems.
25179 For more information, see:
25180 `http://www.gnu.org/'
25181 `http://www.gnu.org/gnu/linux-and-gnu.html'
25184 File: gcc.info, Node: Copying, Next: GNU Free Documentation License, Prev: GNU Project, Up: Top
25186 GNU GENERAL PUBLIC LICENSE
25187 **************************
25189 Version 2, June 1991
25191 Copyright (C) 1989, 1991 Free Software Foundation, Inc.
25192 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
25194 Everyone is permitted to copy and distribute verbatim copies
25195 of this license document, but changing it is not allowed.
25200 The licenses for most software are designed to take away your freedom
25201 to share and change it. By contrast, the GNU General Public License is
25202 intended to guarantee your freedom to share and change free
25203 software--to make sure the software is free for all its users. This
25204 General Public License applies to most of the Free Software
25205 Foundation's software and to any other program whose authors commit to
25206 using it. (Some other Free Software Foundation software is covered by
25207 the GNU Library General Public License instead.) You can apply it to
25208 your programs, too.
25210 When we speak of free software, we are referring to freedom, not
25211 price. Our General Public Licenses are designed to make sure that you
25212 have the freedom to distribute copies of free software (and charge for
25213 this service if you wish), that you receive source code or can get it
25214 if you want it, that you can change the software or use pieces of it in
25215 new free programs; and that you know you can do these things.
25217 To protect your rights, we need to make restrictions that forbid
25218 anyone to deny you these rights or to ask you to surrender the rights.
25219 These restrictions translate to certain responsibilities for you if you
25220 distribute copies of the software, or if you modify it.
25222 For example, if you distribute copies of such a program, whether
25223 gratis or for a fee, you must give the recipients all the rights that
25224 you have. You must make sure that they, too, receive or can get the
25225 source code. And you must show them these terms so they know their
25228 We protect your rights with two steps: (1) copyright the software, and
25229 (2) offer you this license which gives you legal permission to copy,
25230 distribute and/or modify the software.
25232 Also, for each author's protection and ours, we want to make certain
25233 that everyone understands that there is no warranty for this free
25234 software. If the software is modified by someone else and passed on, we
25235 want its recipients to know that what they have is not the original, so
25236 that any problems introduced by others will not reflect on the original
25237 authors' reputations.
25239 Finally, any free program is threatened constantly by software
25240 patents. We wish to avoid the danger that redistributors of a free
25241 program will individually obtain patent licenses, in effect making the
25242 program proprietary. To prevent this, we have made it clear that any
25243 patent must be licensed for everyone's free use or not licensed at all.
25245 The precise terms and conditions for copying, distribution and
25246 modification follow.
25248 TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
25249 0. This License applies to any program or other work which contains a
25250 notice placed by the copyright holder saying it may be distributed
25251 under the terms of this General Public License. The "Program",
25252 below, refers to any such program or work, and a "work based on
25253 the Program" means either the Program or any derivative work under
25254 copyright law: that is to say, a work containing the Program or a
25255 portion of it, either verbatim or with modifications and/or
25256 translated into another language. (Hereinafter, translation is
25257 included without limitation in the term "modification".) Each
25258 licensee is addressed as "you".
25260 Activities other than copying, distribution and modification are
25261 not covered by this License; they are outside its scope. The act
25262 of running the Program is not restricted, and the output from the
25263 Program is covered only if its contents constitute a work based on
25264 the Program (independent of having been made by running the
25265 Program). Whether that is true depends on what the Program does.
25267 1. You may copy and distribute verbatim copies of the Program's
25268 source code as you receive it, in any medium, provided that you
25269 conspicuously and appropriately publish on each copy an appropriate
25270 copyright notice and disclaimer of warranty; keep intact all the
25271 notices that refer to this License and to the absence of any
25272 warranty; and give any other recipients of the Program a copy of
25273 this License along with the Program.
25275 You may charge a fee for the physical act of transferring a copy,
25276 and you may at your option offer warranty protection in exchange
25279 2. You may modify your copy or copies of the Program or any portion
25280 of it, thus forming a work based on the Program, and copy and
25281 distribute such modifications or work under the terms of Section 1
25282 above, provided that you also meet all of these conditions:
25284 a. You must cause the modified files to carry prominent notices
25285 stating that you changed the files and the date of any change.
25287 b. You must cause any work that you distribute or publish, that
25288 in whole or in part contains or is derived from the Program
25289 or any part thereof, to be licensed as a whole at no charge
25290 to all third parties under the terms of this License.
25292 c. If the modified program normally reads commands interactively
25293 when run, you must cause it, when started running for such
25294 interactive use in the most ordinary way, to print or display
25295 an announcement including an appropriate copyright notice and
25296 a notice that there is no warranty (or else, saying that you
25297 provide a warranty) and that users may redistribute the
25298 program under these conditions, and telling the user how to
25299 view a copy of this License. (Exception: if the Program
25300 itself is interactive but does not normally print such an
25301 announcement, your work based on the Program is not required
25302 to print an announcement.)
25304 These requirements apply to the modified work as a whole. If
25305 identifiable sections of that work are not derived from the
25306 Program, and can be reasonably considered independent and separate
25307 works in themselves, then this License, and its terms, do not
25308 apply to those sections when you distribute them as separate
25309 works. But when you distribute the same sections as part of a
25310 whole which is a work based on the Program, the distribution of
25311 the whole must be on the terms of this License, whose permissions
25312 for other licensees extend to the entire whole, and thus to each
25313 and every part regardless of who wrote it.
25315 Thus, it is not the intent of this section to claim rights or
25316 contest your rights to work written entirely by you; rather, the
25317 intent is to exercise the right to control the distribution of
25318 derivative or collective works based on the Program.
25320 In addition, mere aggregation of another work not based on the
25321 Program with the Program (or with a work based on the Program) on
25322 a volume of a storage or distribution medium does not bring the
25323 other work under the scope of this License.
25325 3. You may copy and distribute the Program (or a work based on it,
25326 under Section 2) in object code or executable form under the terms
25327 of Sections 1 and 2 above provided that you also do one of the
25330 a. Accompany it with the complete corresponding machine-readable
25331 source code, which must be distributed under the terms of
25332 Sections 1 and 2 above on a medium customarily used for
25333 software interchange; or,
25335 b. Accompany it with a written offer, valid for at least three
25336 years, to give any third party, for a charge no more than your
25337 cost of physically performing source distribution, a complete
25338 machine-readable copy of the corresponding source code, to be
25339 distributed under the terms of Sections 1 and 2 above on a
25340 medium customarily used for software interchange; or,
25342 c. Accompany it with the information you received as to the offer
25343 to distribute corresponding source code. (This alternative is
25344 allowed only for noncommercial distribution and only if you
25345 received the program in object code or executable form with
25346 such an offer, in accord with Subsection b above.)
25348 The source code for a work means the preferred form of the work for
25349 making modifications to it. For an executable work, complete
25350 source code means all the source code for all modules it contains,
25351 plus any associated interface definition files, plus the scripts
25352 used to control compilation and installation of the executable.
25353 However, as a special exception, the source code distributed need
25354 not include anything that is normally distributed (in either
25355 source or binary form) with the major components (compiler,
25356 kernel, and so on) of the operating system on which the executable
25357 runs, unless that component itself accompanies the executable.
25359 If distribution of executable or object code is made by offering
25360 access to copy from a designated place, then offering equivalent
25361 access to copy the source code from the same place counts as
25362 distribution of the source code, even though third parties are not
25363 compelled to copy the source along with the object code.
25365 4. You may not copy, modify, sublicense, or distribute the Program
25366 except as expressly provided under this License. Any attempt
25367 otherwise to copy, modify, sublicense or distribute the Program is
25368 void, and will automatically terminate your rights under this
25369 License. However, parties who have received copies, or rights,
25370 from you under this License will not have their licenses
25371 terminated so long as such parties remain in full compliance.
25373 5. You are not required to accept this License, since you have not
25374 signed it. However, nothing else grants you permission to modify
25375 or distribute the Program or its derivative works. These actions
25376 are prohibited by law if you do not accept this License.
25377 Therefore, by modifying or distributing the Program (or any work
25378 based on the Program), you indicate your acceptance of this
25379 License to do so, and all its terms and conditions for copying,
25380 distributing or modifying the Program or works based on it.
25382 6. Each time you redistribute the Program (or any work based on the
25383 Program), the recipient automatically receives a license from the
25384 original licensor to copy, distribute or modify the Program
25385 subject to these terms and conditions. You may not impose any
25386 further restrictions on the recipients' exercise of the rights
25387 granted herein. You are not responsible for enforcing compliance
25388 by third parties to this License.
25390 7. If, as a consequence of a court judgment or allegation of patent
25391 infringement or for any other reason (not limited to patent
25392 issues), conditions are imposed on you (whether by court order,
25393 agreement or otherwise) that contradict the conditions of this
25394 License, they do not excuse you from the conditions of this
25395 License. If you cannot distribute so as to satisfy simultaneously
25396 your obligations under this License and any other pertinent
25397 obligations, then as a consequence you may not distribute the
25398 Program at all. For example, if a patent license would not permit
25399 royalty-free redistribution of the Program by all those who
25400 receive copies directly or indirectly through you, then the only
25401 way you could satisfy both it and this License would be to refrain
25402 entirely from distribution of the Program.
25404 If any portion of this section is held invalid or unenforceable
25405 under any particular circumstance, the balance of the section is
25406 intended to apply and the section as a whole is intended to apply
25407 in other circumstances.
25409 It is not the purpose of this section to induce you to infringe any
25410 patents or other property right claims or to contest validity of
25411 any such claims; this section has the sole purpose of protecting
25412 the integrity of the free software distribution system, which is
25413 implemented by public license practices. Many people have made
25414 generous contributions to the wide range of software distributed
25415 through that system in reliance on consistent application of that
25416 system; it is up to the author/donor to decide if he or she is
25417 willing to distribute software through any other system and a
25418 licensee cannot impose that choice.
25420 This section is intended to make thoroughly clear what is believed
25421 to be a consequence of the rest of this License.
25423 8. If the distribution and/or use of the Program is restricted in
25424 certain countries either by patents or by copyrighted interfaces,
25425 the original copyright holder who places the Program under this
25426 License may add an explicit geographical distribution limitation
25427 excluding those countries, so that distribution is permitted only
25428 in or among countries not thus excluded. In such case, this
25429 License incorporates the limitation as if written in the body of
25432 9. The Free Software Foundation may publish revised and/or new
25433 versions of the General Public License from time to time. Such
25434 new versions will be similar in spirit to the present version, but
25435 may differ in detail to address new problems or concerns.
25437 Each version is given a distinguishing version number. If the
25438 Program specifies a version number of this License which applies
25439 to it and "any later version", you have the option of following
25440 the terms and conditions either of that version or of any later
25441 version published by the Free Software Foundation. If the Program
25442 does not specify a version number of this License, you may choose
25443 any version ever published by the Free Software Foundation.
25445 10. If you wish to incorporate parts of the Program into other free
25446 programs whose distribution conditions are different, write to the
25447 author to ask for permission. For software which is copyrighted
25448 by the Free Software Foundation, write to the Free Software
25449 Foundation; we sometimes make exceptions for this. Our decision
25450 will be guided by the two goals of preserving the free status of
25451 all derivatives of our free software and of promoting the sharing
25452 and reuse of software generally.
25455 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO
25456 WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE
25457 LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
25458 HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT
25459 WARRANTY OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT
25460 NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
25461 FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE
25462 QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE
25463 PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY
25464 SERVICING, REPAIR OR CORRECTION.
25466 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
25467 WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY
25468 MODIFY AND/OR REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE
25469 LIABLE TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL,
25470 INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR
25471 INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
25472 DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU
25473 OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY
25474 OTHER PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN
25475 ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
25477 END OF TERMS AND CONDITIONS
25478 How to Apply These Terms to Your New Programs
25479 =============================================
25481 If you develop a new program, and you want it to be of the greatest
25482 possible use to the public, the best way to achieve this is to make it
25483 free software which everyone can redistribute and change under these
25486 To do so, attach the following notices to the program. It is safest
25487 to attach them to the start of each source file to most effectively
25488 convey the exclusion of warranty; and each file should have at least
25489 the "copyright" line and a pointer to where the full notice is found.
25491 ONE LINE TO GIVE THE PROGRAM'S NAME AND A BRIEF IDEA OF WHAT IT DOES.
25492 Copyright (C) YEAR NAME OF AUTHOR
25494 This program is free software; you can redistribute it and/or modify
25495 it under the terms of the GNU General Public License as published by
25496 the Free Software Foundation; either version 2 of the License, or
25497 (at your option) any later version.
25499 This program is distributed in the hope that it will be useful,
25500 but WITHOUT ANY WARRANTY; without even the implied warranty of
25501 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
25502 GNU General Public License for more details.
25504 You should have received a copy of the GNU General Public License
25505 along with this program; if not, write to the Free Software Foundation,
25506 Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
25508 Also add information on how to contact you by electronic and paper
25511 If the program is interactive, make it output a short notice like this
25512 when it starts in an interactive mode:
25514 Gnomovision version 69, Copyright (C) YEAR NAME OF AUTHOR
25515 Gnomovision comes with ABSOLUTELY NO WARRANTY; for details
25517 This is free software, and you are welcome to redistribute it
25518 under certain conditions; type `show c' for details.
25520 The hypothetical commands `show w' and `show c' should show the
25521 appropriate parts of the General Public License. Of course, the
25522 commands you use may be called something other than `show w' and `show
25523 c'; they could even be mouse-clicks or menu items--whatever suits your
25526 You should also get your employer (if you work as a programmer) or your
25527 school, if any, to sign a "copyright disclaimer" for the program, if
25528 necessary. Here is a sample; alter the names:
25530 Yoyodyne, Inc., hereby disclaims all copyright interest in the program
25531 `Gnomovision' (which makes passes at compilers) written by James Hacker.
25533 SIGNATURE OF TY COON, 1 April 1989
25534 Ty Coon, President of Vice
25536 This General Public License does not permit incorporating your program
25537 into proprietary programs. If your program is a subroutine library,
25538 you may consider it more useful to permit linking proprietary
25539 applications with the library. If this is what you want to do, use the
25540 GNU Library General Public License instead of this License.
25543 File: gcc.info, Node: GNU Free Documentation License, Next: Contributors, Prev: Copying, Up: Top
25545 GNU Free Documentation License
25546 ******************************
25548 Version 1.2, November 2002
25550 Copyright (C) 2000,2001,2002 Free Software Foundation, Inc.
25551 59 Temple Place, Suite 330, Boston, MA 02111-1307, USA
25553 Everyone is permitted to copy and distribute verbatim copies
25554 of this license document, but changing it is not allowed.
25558 The purpose of this License is to make a manual, textbook, or other
25559 functional and useful document "free" in the sense of freedom: to
25560 assure everyone the effective freedom to copy and redistribute it,
25561 with or without modifying it, either commercially or
25562 noncommercially. Secondarily, this License preserves for the
25563 author and publisher a way to get credit for their work, while not
25564 being considered responsible for modifications made by others.
25566 This License is a kind of "copyleft", which means that derivative
25567 works of the document must themselves be free in the same sense.
25568 It complements the GNU General Public License, which is a copyleft
25569 license designed for free software.
25571 We have designed this License in order to use it for manuals for
25572 free software, because free software needs free documentation: a
25573 free program should come with manuals providing the same freedoms
25574 that the software does. But this License is not limited to
25575 software manuals; it can be used for any textual work, regardless
25576 of subject matter or whether it is published as a printed book.
25577 We recommend this License principally for works whose purpose is
25578 instruction or reference.
25580 1. APPLICABILITY AND DEFINITIONS
25582 This License applies to any manual or other work, in any medium,
25583 that contains a notice placed by the copyright holder saying it
25584 can be distributed under the terms of this License. Such a notice
25585 grants a world-wide, royalty-free license, unlimited in duration,
25586 to use that work under the conditions stated herein. The
25587 "Document", below, refers to any such manual or work. Any member
25588 of the public is a licensee, and is addressed as "you". You
25589 accept the license if you copy, modify or distribute the work in a
25590 way requiring permission under copyright law.
25592 A "Modified Version" of the Document means any work containing the
25593 Document or a portion of it, either copied verbatim, or with
25594 modifications and/or translated into another language.
25596 A "Secondary Section" is a named appendix or a front-matter section
25597 of the Document that deals exclusively with the relationship of the
25598 publishers or authors of the Document to the Document's overall
25599 subject (or to related matters) and contains nothing that could
25600 fall directly within that overall subject. (Thus, if the Document
25601 is in part a textbook of mathematics, a Secondary Section may not
25602 explain any mathematics.) The relationship could be a matter of
25603 historical connection with the subject or with related matters, or
25604 of legal, commercial, philosophical, ethical or political position
25607 The "Invariant Sections" are certain Secondary Sections whose
25608 titles are designated, as being those of Invariant Sections, in
25609 the notice that says that the Document is released under this
25610 License. If a section does not fit the above definition of
25611 Secondary then it is not allowed to be designated as Invariant.
25612 The Document may contain zero Invariant Sections. If the Document
25613 does not identify any Invariant Sections then there are none.
25615 The "Cover Texts" are certain short passages of text that are
25616 listed, as Front-Cover Texts or Back-Cover Texts, in the notice
25617 that says that the Document is released under this License. A
25618 Front-Cover Text may be at most 5 words, and a Back-Cover Text may
25619 be at most 25 words.
25621 A "Transparent" copy of the Document means a machine-readable copy,
25622 represented in a format whose specification is available to the
25623 general public, that is suitable for revising the document
25624 straightforwardly with generic text editors or (for images
25625 composed of pixels) generic paint programs or (for drawings) some
25626 widely available drawing editor, and that is suitable for input to
25627 text formatters or for automatic translation to a variety of
25628 formats suitable for input to text formatters. A copy made in an
25629 otherwise Transparent file format whose markup, or absence of
25630 markup, has been arranged to thwart or discourage subsequent
25631 modification by readers is not Transparent. An image format is
25632 not Transparent if used for any substantial amount of text. A
25633 copy that is not "Transparent" is called "Opaque".
25635 Examples of suitable formats for Transparent copies include plain
25636 ASCII without markup, Texinfo input format, LaTeX input format,
25637 SGML or XML using a publicly available DTD, and
25638 standard-conforming simple HTML, PostScript or PDF designed for
25639 human modification. Examples of transparent image formats include
25640 PNG, XCF and JPG. Opaque formats include proprietary formats that
25641 can be read and edited only by proprietary word processors, SGML or
25642 XML for which the DTD and/or processing tools are not generally
25643 available, and the machine-generated HTML, PostScript or PDF
25644 produced by some word processors for output purposes only.
25646 The "Title Page" means, for a printed book, the title page itself,
25647 plus such following pages as are needed to hold, legibly, the
25648 material this License requires to appear in the title page. For
25649 works in formats which do not have any title page as such, "Title
25650 Page" means the text near the most prominent appearance of the
25651 work's title, preceding the beginning of the body of the text.
25653 A section "Entitled XYZ" means a named subunit of the Document
25654 whose title either is precisely XYZ or contains XYZ in parentheses
25655 following text that translates XYZ in another language. (Here XYZ
25656 stands for a specific section name mentioned below, such as
25657 "Acknowledgements", "Dedications", "Endorsements", or "History".)
25658 To "Preserve the Title" of such a section when you modify the
25659 Document means that it remains a section "Entitled XYZ" according
25660 to this definition.
25662 The Document may include Warranty Disclaimers next to the notice
25663 which states that this License applies to the Document. These
25664 Warranty Disclaimers are considered to be included by reference in
25665 this License, but only as regards disclaiming warranties: any other
25666 implication that these Warranty Disclaimers may have is void and
25667 has no effect on the meaning of this License.
25669 2. VERBATIM COPYING
25671 You may copy and distribute the Document in any medium, either
25672 commercially or noncommercially, provided that this License, the
25673 copyright notices, and the license notice saying this License
25674 applies to the Document are reproduced in all copies, and that you
25675 add no other conditions whatsoever to those of this License. You
25676 may not use technical measures to obstruct or control the reading
25677 or further copying of the copies you make or distribute. However,
25678 you may accept compensation in exchange for copies. If you
25679 distribute a large enough number of copies you must also follow
25680 the conditions in section 3.
25682 You may also lend copies, under the same conditions stated above,
25683 and you may publicly display copies.
25685 3. COPYING IN QUANTITY
25687 If you publish printed copies (or copies in media that commonly
25688 have printed covers) of the Document, numbering more than 100, and
25689 the Document's license notice requires Cover Texts, you must
25690 enclose the copies in covers that carry, clearly and legibly, all
25691 these Cover Texts: Front-Cover Texts on the front cover, and
25692 Back-Cover Texts on the back cover. Both covers must also clearly
25693 and legibly identify you as the publisher of these copies. The
25694 front cover must present the full title with all words of the
25695 title equally prominent and visible. You may add other material
25696 on the covers in addition. Copying with changes limited to the
25697 covers, as long as they preserve the title of the Document and
25698 satisfy these conditions, can be treated as verbatim copying in
25701 If the required texts for either cover are too voluminous to fit
25702 legibly, you should put the first ones listed (as many as fit
25703 reasonably) on the actual cover, and continue the rest onto
25706 If you publish or distribute Opaque copies of the Document
25707 numbering more than 100, you must either include a
25708 machine-readable Transparent copy along with each Opaque copy, or
25709 state in or with each Opaque copy a computer-network location from
25710 which the general network-using public has access to download
25711 using public-standard network protocols a complete Transparent
25712 copy of the Document, free of added material. If you use the
25713 latter option, you must take reasonably prudent steps, when you
25714 begin distribution of Opaque copies in quantity, to ensure that
25715 this Transparent copy will remain thus accessible at the stated
25716 location until at least one year after the last time you
25717 distribute an Opaque copy (directly or through your agents or
25718 retailers) of that edition to the public.
25720 It is requested, but not required, that you contact the authors of
25721 the Document well before redistributing any large number of
25722 copies, to give them a chance to provide you with an updated
25723 version of the Document.
25727 You may copy and distribute a Modified Version of the Document
25728 under the conditions of sections 2 and 3 above, provided that you
25729 release the Modified Version under precisely this License, with
25730 the Modified Version filling the role of the Document, thus
25731 licensing distribution and modification of the Modified Version to
25732 whoever possesses a copy of it. In addition, you must do these
25733 things in the Modified Version:
25735 A. Use in the Title Page (and on the covers, if any) a title
25736 distinct from that of the Document, and from those of
25737 previous versions (which should, if there were any, be listed
25738 in the History section of the Document). You may use the
25739 same title as a previous version if the original publisher of
25740 that version gives permission.
25742 B. List on the Title Page, as authors, one or more persons or
25743 entities responsible for authorship of the modifications in
25744 the Modified Version, together with at least five of the
25745 principal authors of the Document (all of its principal
25746 authors, if it has fewer than five), unless they release you
25747 from this requirement.
25749 C. State on the Title page the name of the publisher of the
25750 Modified Version, as the publisher.
25752 D. Preserve all the copyright notices of the Document.
25754 E. Add an appropriate copyright notice for your modifications
25755 adjacent to the other copyright notices.
25757 F. Include, immediately after the copyright notices, a license
25758 notice giving the public permission to use the Modified
25759 Version under the terms of this License, in the form shown in
25760 the Addendum below.
25762 G. Preserve in that license notice the full lists of Invariant
25763 Sections and required Cover Texts given in the Document's
25766 H. Include an unaltered copy of this License.
25768 I. Preserve the section Entitled "History", Preserve its Title,
25769 and add to it an item stating at least the title, year, new
25770 authors, and publisher of the Modified Version as given on
25771 the Title Page. If there is no section Entitled "History" in
25772 the Document, create one stating the title, year, authors,
25773 and publisher of the Document as given on its Title Page,
25774 then add an item describing the Modified Version as stated in
25775 the previous sentence.
25777 J. Preserve the network location, if any, given in the Document
25778 for public access to a Transparent copy of the Document, and
25779 likewise the network locations given in the Document for
25780 previous versions it was based on. These may be placed in
25781 the "History" section. You may omit a network location for a
25782 work that was published at least four years before the
25783 Document itself, or if the original publisher of the version
25784 it refers to gives permission.
25786 K. For any section Entitled "Acknowledgements" or "Dedications",
25787 Preserve the Title of the section, and preserve in the
25788 section all the substance and tone of each of the contributor
25789 acknowledgements and/or dedications given therein.
25791 L. Preserve all the Invariant Sections of the Document,
25792 unaltered in their text and in their titles. Section numbers
25793 or the equivalent are not considered part of the section
25796 M. Delete any section Entitled "Endorsements". Such a section
25797 may not be included in the Modified Version.
25799 N. Do not retitle any existing section to be Entitled
25800 "Endorsements" or to conflict in title with any Invariant
25803 O. Preserve any Warranty Disclaimers.
25805 If the Modified Version includes new front-matter sections or
25806 appendices that qualify as Secondary Sections and contain no
25807 material copied from the Document, you may at your option
25808 designate some or all of these sections as invariant. To do this,
25809 add their titles to the list of Invariant Sections in the Modified
25810 Version's license notice. These titles must be distinct from any
25811 other section titles.
25813 You may add a section Entitled "Endorsements", provided it contains
25814 nothing but endorsements of your Modified Version by various
25815 parties--for example, statements of peer review or that the text
25816 has been approved by an organization as the authoritative
25817 definition of a standard.
25819 You may add a passage of up to five words as a Front-Cover Text,
25820 and a passage of up to 25 words as a Back-Cover Text, to the end
25821 of the list of Cover Texts in the Modified Version. Only one
25822 passage of Front-Cover Text and one of Back-Cover Text may be
25823 added by (or through arrangements made by) any one entity. If the
25824 Document already includes a cover text for the same cover,
25825 previously added by you or by arrangement made by the same entity
25826 you are acting on behalf of, you may not add another; but you may
25827 replace the old one, on explicit permission from the previous
25828 publisher that added the old one.
25830 The author(s) and publisher(s) of the Document do not by this
25831 License give permission to use their names for publicity for or to
25832 assert or imply endorsement of any Modified Version.
25834 5. COMBINING DOCUMENTS
25836 You may combine the Document with other documents released under
25837 this License, under the terms defined in section 4 above for
25838 modified versions, provided that you include in the combination
25839 all of the Invariant Sections of all of the original documents,
25840 unmodified, and list them all as Invariant Sections of your
25841 combined work in its license notice, and that you preserve all
25842 their Warranty Disclaimers.
25844 The combined work need only contain one copy of this License, and
25845 multiple identical Invariant Sections may be replaced with a single
25846 copy. If there are multiple Invariant Sections with the same name
25847 but different contents, make the title of each such section unique
25848 by adding at the end of it, in parentheses, the name of the
25849 original author or publisher of that section if known, or else a
25850 unique number. Make the same adjustment to the section titles in
25851 the list of Invariant Sections in the license notice of the
25854 In the combination, you must combine any sections Entitled
25855 "History" in the various original documents, forming one section
25856 Entitled "History"; likewise combine any sections Entitled
25857 "Acknowledgements", and any sections Entitled "Dedications". You
25858 must delete all sections Entitled "Endorsements."
25860 6. COLLECTIONS OF DOCUMENTS
25862 You may make a collection consisting of the Document and other
25863 documents released under this License, and replace the individual
25864 copies of this License in the various documents with a single copy
25865 that is included in the collection, provided that you follow the
25866 rules of this License for verbatim copying of each of the
25867 documents in all other respects.
25869 You may extract a single document from such a collection, and
25870 distribute it individually under this License, provided you insert
25871 a copy of this License into the extracted document, and follow
25872 this License in all other respects regarding verbatim copying of
25875 7. AGGREGATION WITH INDEPENDENT WORKS
25877 A compilation of the Document or its derivatives with other
25878 separate and independent documents or works, in or on a volume of
25879 a storage or distribution medium, is called an "aggregate" if the
25880 copyright resulting from the compilation is not used to limit the
25881 legal rights of the compilation's users beyond what the individual
25882 works permit. When the Document is included an aggregate, this
25883 License does not apply to the other works in the aggregate which
25884 are not themselves derivative works of the Document.
25886 If the Cover Text requirement of section 3 is applicable to these
25887 copies of the Document, then if the Document is less than one half
25888 of the entire aggregate, the Document's Cover Texts may be placed
25889 on covers that bracket the Document within the aggregate, or the
25890 electronic equivalent of covers if the Document is in electronic
25891 form. Otherwise they must appear on printed covers that bracket
25892 the whole aggregate.
25896 Translation is considered a kind of modification, so you may
25897 distribute translations of the Document under the terms of section
25898 4. Replacing Invariant Sections with translations requires special
25899 permission from their copyright holders, but you may include
25900 translations of some or all Invariant Sections in addition to the
25901 original versions of these Invariant Sections. You may include a
25902 translation of this License, and all the license notices in the
25903 Document, and any Warrany Disclaimers, provided that you also
25904 include the original English version of this License and the
25905 original versions of those notices and disclaimers. In case of a
25906 disagreement between the translation and the original version of
25907 this License or a notice or disclaimer, the original version will
25910 If a section in the Document is Entitled "Acknowledgements",
25911 "Dedications", or "History", the requirement (section 4) to
25912 Preserve its Title (section 1) will typically require changing the
25917 You may not copy, modify, sublicense, or distribute the Document
25918 except as expressly provided for under this License. Any other
25919 attempt to copy, modify, sublicense or distribute the Document is
25920 void, and will automatically terminate your rights under this
25921 License. However, parties who have received copies, or rights,
25922 from you under this License will not have their licenses
25923 terminated so long as such parties remain in full compliance.
25925 10. FUTURE REVISIONS OF THIS LICENSE
25927 The Free Software Foundation may publish new, revised versions of
25928 the GNU Free Documentation License from time to time. Such new
25929 versions will be similar in spirit to the present version, but may
25930 differ in detail to address new problems or concerns. See
25931 `http://www.gnu.org/copyleft/'.
25933 Each version of the License is given a distinguishing version
25934 number. If the Document specifies that a particular numbered
25935 version of this License "or any later version" applies to it, you
25936 have the option of following the terms and conditions either of
25937 that specified version or of any later version that has been
25938 published (not as a draft) by the Free Software Foundation. If
25939 the Document does not specify a version number of this License,
25940 you may choose any version ever published (not as a draft) by the
25941 Free Software Foundation.
25943 ADDENDUM: How to use this License for your documents
25944 ====================================================
25946 To use this License in a document you have written, include a copy of
25947 the License in the document and put the following copyright and license
25948 notices just after the title page:
25950 Copyright (C) YEAR YOUR NAME.
25951 Permission is granted to copy, distribute and/or modify this document
25952 under the terms of the GNU Free Documentation License, Version 1.2
25953 or any later version published by the Free Software Foundation;
25954 with no Invariant Sections, no Front-Cover Texts, and no Back-Cover Texts.
25955 A copy of the license is included in the section entitled ``GNU
25956 Free Documentation License''.
25958 If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts,
25959 replace the "with...Texts." line with this:
25961 with the Invariant Sections being LIST THEIR TITLES, with
25962 the Front-Cover Texts being LIST, and with the Back-Cover Texts
25965 If you have Invariant Sections without Cover Texts, or some other
25966 combination of the three, merge those two alternatives to suit the
25969 If your document contains nontrivial examples of program code, we
25970 recommend releasing these examples in parallel under your choice of
25971 free software license, such as the GNU General Public License, to
25972 permit their use in free software.
25975 File: gcc.info, Node: Contributors, Next: Option Index, Prev: GNU Free Documentation License, Up: Top
25977 Contributors to GCC
25978 *******************
25980 The GCC project would like to thank its many contributors. Without
25981 them the project would not have been nearly as successful as it has
25982 been. Any omissions in this list are accidental. Feel free to contact
25983 <law@redhat.com> or <gerald@pfeifer.com> if you have been left out or
25984 some of your contributions are not listed. Please keep this list in
25985 alphabetical order.
25987 * Analog Devices helped implement the support for complex data types
25990 * John David Anglin for threading-related fixes and improvements to
25991 libstdc++-v3, and the HP-UX port.
25993 * James van Artsdalen wrote the code that makes efficient use of the
25994 Intel 80387 register stack.
25996 * Abramo and Roberto Bagnara for the SysV68 Motorola 3300 Delta
25999 * Alasdair Baird for various bug fixes.
26001 * Giovanni Bajo for analyzing lots of complicated C++ problem
26004 * Peter Barada for his work to improve code generation for new
26007 * Gerald Baumgartner added the signature extension to the C++ front
26010 * Godmar Back for his Java improvements and encouragement.
26012 * Scott Bambrough for help porting the Java compiler.
26014 * Wolfgang Bangerth for processing tons of bug reports.
26016 * Jon Beniston for his Microsoft Windows port of Java.
26018 * Daniel Berlin for better DWARF2 support, faster/better
26019 optimizations, improved alias analysis, plus migrating GCC to
26022 * Geoff Berry for his Java object serialization work and various
26025 * Eric Blake for helping to make GCJ and libgcj conform to the
26028 * Janne Blomqvist for contributions to gfortran.
26030 * Segher Boessenkool for various fixes.
26032 * Hans-J. Boehm for his garbage collector, IA-64 libffi port, and
26035 * Neil Booth for work on cpplib, lang hooks, debug hooks and other
26036 miscellaneous clean-ups.
26038 * Steven Bosscher for integrating the gfortran front end into GCC
26039 and for contributing to the tree-ssa branch.
26041 * Eric Botcazou for fixing middle- and backend bugs left and right.
26043 * Per Bothner for his direction via the steering committee and
26044 various improvements to the infrastructure for supporting new
26045 languages. Chill front end implementation. Initial
26046 implementations of cpplib, fix-header, config.guess, libio, and
26047 past C++ library (libg++) maintainer. Dreaming up, designing and
26048 implementing much of GCJ.
26050 * Devon Bowen helped port GCC to the Tahoe.
26052 * Don Bowman for mips-vxworks contributions.
26054 * Dave Brolley for work on cpplib and Chill.
26056 * Paul Brook for work on the ARM architecture and maintaining
26059 * Robert Brown implemented the support for Encore 32000 systems.
26061 * Christian Bruel for improvements to local store elimination.
26063 * Herman A.J. ten Brugge for various fixes.
26065 * Joerg Brunsmann for Java compiler hacking and help with the GCJ
26068 * Joe Buck for his direction via the steering committee.
26070 * Craig Burley for leadership of the G77 Fortran effort.
26072 * Stephan Buys for contributing Doxygen notes for libstdc++.
26074 * Paolo Carlini for libstdc++ work: lots of efficiency improvements
26075 to the C++ strings, streambufs and formatted I/O, hard detective
26076 work on the frustrating localization issues, and keeping up with
26077 the problem reports.
26079 * John Carr for his alias work, SPARC hacking, infrastructure
26080 improvements, previous contributions to the steering committee,
26081 loop optimizations, etc.
26083 * Stephane Carrez for 68HC11 and 68HC12 ports.
26085 * Steve Chamberlain for support for the Renesas SH and H8 processors
26086 and the PicoJava processor, and for GCJ config fixes.
26088 * Glenn Chambers for help with the GCJ FAQ.
26090 * John-Marc Chandonia for various libgcj patches.
26092 * Scott Christley for his Objective-C contributions.
26094 * Eric Christopher for his Java porting help and clean-ups.
26096 * Branko Cibej for more warning contributions.
26098 * The GNU Classpath project for all of their merged runtime code.
26100 * Nick Clifton for arm, mcore, fr30, v850, m32r work, `--help', and
26101 other random hacking.
26103 * Michael Cook for libstdc++ cleanup patches to reduce warnings.
26105 * R. Kelley Cook for making GCC buildable from a read-only directory
26106 as well as other miscellaneous build process and documentation
26109 * Ralf Corsepius for SH testing and minor bugfixing.
26111 * Stan Cox for care and feeding of the x86 port and lots of behind
26112 the scenes hacking.
26114 * Alex Crain provided changes for the 3b1.
26116 * Ian Dall for major improvements to the NS32k port.
26118 * Paul Dale for his work to add uClinux platform support to the m68k
26121 * Dario Dariol contributed the four varieties of sample programs
26122 that print a copy of their source.
26124 * Russell Davidson for fstream and stringstream fixes in libstdc++.
26126 * Bud Davis for work on the G77 and gfortran compilers.
26128 * Mo DeJong for GCJ and libgcj bug fixes.
26130 * DJ Delorie for the DJGPP port, build and libiberty maintenance, and
26133 * Arnaud Desitter for helping to debug gfortran.
26135 * Gabriel Dos Reis for contributions to G++, contributions and
26136 maintenance of GCC diagnostics infrastructure, libstdc++-v3,
26137 including `valarray<>', `complex<>', maintaining the numerics
26138 library (including that pesky `<limits>' :-) and keeping
26139 up-to-date anything to do with numbers.
26141 * Ulrich Drepper for his work on glibc, testing of GCC using glibc,
26142 ISO C99 support, CFG dumping support, etc., plus support of the
26143 C++ runtime libraries including for all kinds of C interface
26144 issues, contributing and maintaining `complex<>', sanity checking
26145 and disbursement, configuration architecture, libio maintenance,
26146 and early math work.
26148 * Zdenek Dvorak for a new loop unroller and various fixes.
26150 * Richard Earnshaw for his ongoing work with the ARM.
26152 * David Edelsohn for his direction via the steering committee,
26153 ongoing work with the RS6000/PowerPC port, help cleaning up Haifa
26154 loop changes, doing the entire AIX port of libstdc++ with his bare
26155 hands, and for ensuring GCC properly keeps working on AIX.
26157 * Kevin Ediger for the floating point formatting of num_put::do_put
26160 * Phil Edwards for libstdc++ work including configuration hackery,
26161 documentation maintainer, chief breaker of the web pages, the
26162 occasional iostream bug fix, and work on shared library symbol
26165 * Paul Eggert for random hacking all over GCC.
26167 * Mark Elbrecht for various DJGPP improvements, and for libstdc++
26168 configuration support for locales and fstream-related fixes.
26170 * Vadim Egorov for libstdc++ fixes in strings, streambufs, and
26173 * Christian Ehrhardt for dealing with bug reports.
26175 * Ben Elliston for his work to move the Objective-C runtime into its
26176 own subdirectory and for his work on autoconf.
26178 * Marc Espie for OpenBSD support.
26180 * Doug Evans for much of the global optimization framework, arc,
26181 m32r, and SPARC work.
26183 * Christopher Faylor for his work on the Cygwin port and for caring
26184 and feeding the gcc.gnu.org box and saving its users tons of spam.
26186 * Fred Fish for BeOS support and Ada fixes.
26188 * Ivan Fontes Garcia for the Portuguese translation of the GCJ FAQ.
26190 * Peter Gerwinski for various bug fixes and the Pascal front end.
26192 * Kaveh Ghazi for his direction via the steering committee, amazing
26193 work to make `-W -Wall' useful, and continuously testing GCC on a
26194 plethora of platforms.
26196 * John Gilmore for a donation to the FSF earmarked improving GNU
26199 * Judy Goldberg for c++ contributions.
26201 * Torbjorn Granlund for various fixes and the c-torture testsuite,
26202 multiply- and divide-by-constant optimization, improved long long
26203 support, improved leaf function register allocation, and his
26204 direction via the steering committee.
26206 * Anthony Green for his `-Os' contributions and Java front end work.
26208 * Stu Grossman for gdb hacking, allowing GCJ developers to debug
26211 * Michael K. Gschwind contributed the port to the PDP-11.
26213 * Ron Guilmette implemented the `protoize' and `unprotoize' tools,
26214 the support for Dwarf symbolic debugging information, and much of
26215 the support for System V Release 4. He has also worked heavily on
26216 the Intel 386 and 860 support.
26218 * Mostafa Hagog for Swing Modulo Scheduling (SMS) and post reload
26221 * Bruno Haible for improvements in the runtime overhead for EH, new
26222 warnings and assorted bug fixes.
26224 * Andrew Haley for his amazing Java compiler and library efforts.
26226 * Chris Hanson assisted in making GCC work on HP-UX for the 9000
26229 * Michael Hayes for various thankless work he's done trying to get
26230 the c30/c40 ports functional. Lots of loop and unroll
26231 improvements and fixes.
26233 * Dara Hazeghi for wading through myriads of target-specific bug
26236 * Kate Hedstrom for staking the G77 folks with an initial testsuite.
26238 * Richard Henderson for his ongoing SPARC, alpha, ia32, and ia64
26239 work, loop opts, and generally fixing lots of old problems we've
26240 ignored for years, flow rewrite and lots of further stuff,
26241 including reviewing tons of patches.
26243 * Aldy Hernandez for working on the PowerPC port, SIMD support, and
26246 * Nobuyuki Hikichi of Software Research Associates, Tokyo,
26247 contributed the support for the Sony NEWS machine.
26249 * Kazu Hirata for caring and feeding the Renesas H8/300 port and
26252 * Katherine Holcomb for work on gfortran.
26254 * Manfred Hollstein for his ongoing work to keep the m88k alive, lots
26255 of testing and bug fixing, particularly of GCC configury code.
26257 * Steve Holmgren for MachTen patches.
26259 * Jan Hubicka for his x86 port improvements.
26261 * Falk Hueffner for working on C and optimization bug reports.
26263 * Bernardo Innocenti for his m68k work, including merging of
26264 ColdFire improvements and uClinux support.
26266 * Christian Iseli for various bug fixes.
26268 * Kamil Iskra for general m68k hacking.
26270 * Lee Iverson for random fixes and MIPS testing.
26272 * Andreas Jaeger for testing and benchmarking of GCC and various bug
26275 * Jakub Jelinek for his SPARC work and sibling call optimizations as
26276 well as lots of bug fixes and test cases, and for improving the
26279 * Janis Johnson for ia64 testing and fixes, her quality improvement
26280 sidetracks, and web page maintenance.
26282 * Kean Johnston for SCO OpenServer support and various fixes.
26284 * Tim Josling for the sample language treelang based originally on
26285 Richard Kenner's "toy" language.
26287 * Nicolai Josuttis for additional libstdc++ documentation.
26289 * Klaus Kaempf for his ongoing work to make alpha-vms a viable
26292 * Steven G. Kargl for work on gfortran.
26294 * David Kashtan of SRI adapted GCC to VMS.
26296 * Ryszard Kabatek for many, many libstdc++ bug fixes and
26297 optimizations of strings, especially member functions, and for
26300 * Geoffrey Keating for his ongoing work to make the PPC work for
26301 GNU/Linux and his automatic regression tester.
26303 * Brendan Kehoe for his ongoing work with G++ and for a lot of early
26304 work in just about every part of libstdc++.
26306 * Oliver M. Kellogg of Deutsche Aerospace contributed the port to the
26309 * Richard Kenner of the New York University Ultracomputer Research
26310 Laboratory wrote the machine descriptions for the AMD 29000, the
26311 DEC Alpha, the IBM RT PC, and the IBM RS/6000 as well as the
26312 support for instruction attributes. He also made changes to
26313 better support RISC processors including changes to common
26314 subexpression elimination, strength reduction, function calling
26315 sequence handling, and condition code support, in addition to
26316 generalizing the code for frame pointer elimination and delay slot
26317 scheduling. Richard Kenner was also the head maintainer of GCC
26320 * Mumit Khan for various contributions to the Cygwin and Mingw32
26321 ports and maintaining binary releases for Microsoft Windows hosts,
26322 and for massive libstdc++ porting work to Cygwin/Mingw32.
26324 * Robin Kirkham for cpu32 support.
26326 * Mark Klein for PA improvements.
26328 * Thomas Koenig for various bug fixes.
26330 * Bruce Korb for the new and improved fixincludes code.
26332 * Benjamin Kosnik for his G++ work and for leading the libstdc++-v3
26335 * Charles LaBrec contributed the support for the Integrated Solutions
26338 * Jeff Law for his direction via the steering committee,
26339 coordinating the entire egcs project and GCC 2.95, rolling out
26340 snapshots and releases, handling merges from GCC2, reviewing tons
26341 of patches that might have fallen through the cracks else, and
26342 random but extensive hacking.
26344 * Marc Lehmann for his direction via the steering committee and
26345 helping with analysis and improvements of x86 performance.
26347 * Victor Leikehman for work on gfortran.
26349 * Ted Lemon wrote parts of the RTL reader and printer.
26351 * Kriang Lerdsuwanakij for C++ improvements including template as
26352 template parameter support, and many C++ fixes.
26354 * Warren Levy for tremendous work on libgcj (Java Runtime Library)
26355 and random work on the Java front end.
26357 * Alain Lichnewsky ported GCC to the MIPS CPU.
26359 * Oskar Liljeblad for hacking on AWT and his many Java bug reports
26362 * Robert Lipe for OpenServer support, new testsuites, testing, etc.
26364 * Weiwen Liu for testing and various bug fixes.
26366 * Dave Love for his ongoing work with the Fortran front end and
26369 * Martin von Lo"wis for internal consistency checking infrastructure,
26370 various C++ improvements including namespace support, and tons of
26371 assistance with libstdc++/compiler merges.
26373 * H.J. Lu for his previous contributions to the steering committee,
26374 many x86 bug reports, prototype patches, and keeping the GNU/Linux
26377 * Greg McGary for random fixes and (someday) bounded pointers.
26379 * Andrew MacLeod for his ongoing work in building a real EH system,
26380 various code generation improvements, work on the global
26383 * Vladimir Makarov for hacking some ugly i960 problems, PowerPC
26384 hacking improvements to compile-time performance, overall
26385 knowledge and direction in the area of instruction scheduling, and
26386 design and implementation of the automaton based instruction
26389 * Bob Manson for his behind the scenes work on dejagnu.
26391 * Philip Martin for lots of libstdc++ string and vector iterator
26392 fixes and improvements, and string clean up and testsuites.
26394 * All of the Mauve project contributors, for Java test code.
26396 * Bryce McKinlay for numerous GCJ and libgcj fixes and improvements.
26398 * Adam Megacz for his work on the Microsoft Windows port of GCJ.
26400 * Michael Meissner for LRS framework, ia32, m32r, v850, m88k, MIPS,
26401 powerpc, haifa, ECOFF debug support, and other assorted hacking.
26403 * Jason Merrill for his direction via the steering committee and
26404 leading the G++ effort.
26406 * David Miller for his direction via the steering committee, lots of
26407 SPARC work, improvements in jump.c and interfacing with the Linux
26410 * Gary Miller ported GCC to Charles River Data Systems machines.
26412 * Alfred Minarik for libstdc++ string and ios bug fixes, and turning
26413 the entire libstdc++ testsuite namespace-compatible.
26415 * Mark Mitchell for his direction via the steering committee,
26416 mountains of C++ work, load/store hoisting out of loops, alias
26417 analysis improvements, ISO C `restrict' support, and serving as
26418 release manager for GCC 3.x.
26420 * Alan Modra for various GNU/Linux bits and testing.
26422 * Toon Moene for his direction via the steering committee, Fortran
26423 maintenance, and his ongoing work to make us make Fortran run fast.
26425 * Jason Molenda for major help in the care and feeding of all the
26426 services on the gcc.gnu.org (formerly egcs.cygnus.com)
26427 machine--mail, web services, ftp services, etc etc. Doing all
26428 this work on scrap paper and the backs of envelopes would have
26431 * Catherine Moore for fixing various ugly problems we have sent her
26432 way, including the haifa bug which was killing the Alpha & PowerPC
26435 * Mike Moreton for his various Java patches.
26437 * David Mosberger-Tang for various Alpha improvements, and for the
26438 initial IA-64 port.
26440 * Stephen Moshier contributed the floating point emulator that
26441 assists in cross-compilation and permits support for floating
26442 point numbers wider than 64 bits and for ISO C99 support.
26444 * Bill Moyer for his behind the scenes work on various issues.
26446 * Philippe De Muyter for his work on the m68k port.
26448 * Joseph S. Myers for his work on the PDP-11 port, format checking
26449 and ISO C99 support, and continuous emphasis on (and contributions
26452 * Nathan Myers for his work on libstdc++-v3: architecture and
26453 authorship through the first three snapshots, including
26454 implementation of locale infrastructure, string, shadow C headers,
26455 and the initial project documentation (DESIGN, CHECKLIST, and so
26456 forth). Later, more work on MT-safe string and shadow headers.
26458 * Felix Natter for documentation on porting libstdc++.
26460 * Nathanael Nerode for cleaning up the configuration/build process.
26462 * NeXT, Inc. donated the front end that supports the Objective-C
26465 * Hans-Peter Nilsson for the CRIS and MMIX ports, improvements to
26466 the search engine setup, various documentation fixes and other
26469 * Geoff Noer for his work on getting cygwin native builds working.
26471 * Diego Novillo for his SPEC performance tracking web pages and
26472 assorted fixes in the middle end and various back ends.
26474 * David O'Brien for the FreeBSD/alpha, FreeBSD/AMD x86-64,
26475 FreeBSD/ARM, FreeBSD/PowerPC, and FreeBSD/SPARC64 ports and
26476 related infrastructure improvements.
26478 * Alexandre Oliva for various build infrastructure improvements,
26479 scripts and amazing testing work, including keeping libtool issues
26482 * Stefan Olsson for work on mt_alloc.
26484 * Melissa O'Neill for various NeXT fixes.
26486 * Rainer Orth for random MIPS work, including improvements to GCC's
26487 o32 ABI support, improvements to dejagnu's MIPS support, Java
26488 configuration clean-ups and porting work, etc.
26490 * Hartmut Penner for work on the s390 port.
26492 * Paul Petersen wrote the machine description for the Alliant FX/8.
26494 * Alexandre Petit-Bianco for implementing much of the Java compiler
26495 and continued Java maintainership.
26497 * Matthias Pfaller for major improvements to the NS32k port.
26499 * Gerald Pfeifer for his direction via the steering committee,
26500 pointing out lots of problems we need to solve, maintenance of the
26501 web pages, and taking care of documentation maintenance in general.
26503 * Andrew Pinski for processing bug reports by the dozen.
26505 * Ovidiu Predescu for his work on the Objective-C front end and
26508 * Jerry Quinn for major performance improvements in C++ formatted
26511 * Ken Raeburn for various improvements to checker, MIPS ports and
26512 various cleanups in the compiler.
26514 * Rolf W. Rasmussen for hacking on AWT.
26516 * David Reese of Sun Microsystems contributed to the Solaris on
26519 * Volker Reichelt for keeping up with the problem reports.
26521 * Joern Rennecke for maintaining the sh port, loop, regmove & reload
26524 * Loren J. Rittle for improvements to libstdc++-v3 including the
26525 FreeBSD port, threading fixes, thread-related configury changes,
26526 critical threading documentation, and solutions to really tricky
26527 I/O problems, as well as keeping GCC properly working on FreeBSD
26528 and continuous testing.
26530 * Craig Rodrigues for processing tons of bug reports.
26532 * Ola Ro"nnerup for work on mt_alloc.
26534 * Gavin Romig-Koch for lots of behind the scenes MIPS work.
26536 * David Ronis inspired and encouraged Craig to rewrite the G77
26537 documentation in texinfo format by contributing a first pass at a
26538 translation of the old `g77-0.5.16/f/DOC' file.
26540 * Ken Rose for fixes to GCC's delay slot filling code.
26542 * Paul Rubin wrote most of the preprocessor.
26544 * Pe'tur Runo'lfsson for major performance improvements in C++
26545 formatted I/O and large file support in C++ filebuf.
26547 * Chip Salzenberg for libstdc++ patches and improvements to locales,
26548 traits, Makefiles, libio, libtool hackery, and "long long" support.
26550 * Juha Sarlin for improvements to the H8 code generator.
26552 * Greg Satz assisted in making GCC work on HP-UX for the 9000 series
26555 * Roger Sayle for improvements to constant folding and GCC's RTL
26556 optimizers as well as for fixing numerous bugs.
26558 * Bradley Schatz for his work on the GCJ FAQ.
26560 * Peter Schauer wrote the code to allow debugging to work on the
26563 * William Schelter did most of the work on the Intel 80386 support.
26565 * Tobias Schlu"ter for work on gfortran.
26567 * Bernd Schmidt for various code generation improvements and major
26568 work in the reload pass as well a serving as release manager for
26571 * Peter Schmid for constant testing of libstdc++--especially
26572 application testing, going above and beyond what was requested for
26573 the release criteria--and libstdc++ header file tweaks.
26575 * Jason Schroeder for jcf-dump patches.
26577 * Andreas Schwab for his work on the m68k port.
26579 * Lars Segerlund for work on gfortran.
26581 * Joel Sherrill for his direction via the steering committee, RTEMS
26582 contributions and RTEMS testing.
26584 * Nathan Sidwell for many C++ fixes/improvements.
26586 * Jeffrey Siegal for helping RMS with the original design of GCC,
26587 some code which handles the parse tree and RTL data structures,
26588 constant folding and help with the original VAX & m68k ports.
26590 * Kenny Simpson for prompting libstdc++ fixes due to defect reports
26591 from the LWG (thereby keeping GCC in line with updates from the
26594 * Franz Sirl for his ongoing work with making the PPC port stable
26597 * Andrey Slepuhin for assorted AIX hacking.
26599 * Christopher Smith did the port for Convex machines.
26601 * Danny Smith for his major efforts on the Mingw (and Cygwin) ports.
26603 * Randy Smith finished the Sun FPA support.
26605 * Scott Snyder for queue, iterator, istream, and string fixes and
26606 libstdc++ testsuite entries. Also for providing the patch to G77
26607 to add rudimentary support for `INTEGER*1', `INTEGER*2', and
26610 * Brad Spencer for contributions to the GLIBCPP_FORCE_NEW technique.
26612 * Richard Stallman, for writing the original GCC and launching the
26615 * Jan Stein of the Chalmers Computer Society provided support for
26616 Genix, as well as part of the 32000 machine description.
26618 * Nigel Stephens for various mips16 related fixes/improvements.
26620 * Jonathan Stone wrote the machine description for the Pyramid
26623 * Graham Stott for various infrastructure improvements.
26625 * John Stracke for his Java HTTP protocol fixes.
26627 * Mike Stump for his Elxsi port, G++ contributions over the years
26628 and more recently his vxworks contributions
26630 * Jeff Sturm for Java porting help, bug fixes, and encouragement.
26632 * Shigeya Suzuki for this fixes for the bsdi platforms.
26634 * Ian Lance Taylor for his mips16 work, general configury hacking,
26637 * Holger Teutsch provided the support for the Clipper CPU.
26639 * Gary Thomas for his ongoing work to make the PPC work for
26642 * Philipp Thomas for random bug fixes throughout the compiler
26644 * Jason Thorpe for thread support in libstdc++ on NetBSD.
26646 * Kresten Krab Thorup wrote the run time support for the Objective-C
26647 language and the fantastic Java bytecode interpreter.
26649 * Michael Tiemann for random bug fixes, the first instruction
26650 scheduler, initial C++ support, function integration, NS32k, SPARC
26651 and M88k machine description work, delay slot scheduling.
26653 * Andreas Tobler for his work porting libgcj to Darwin.
26655 * Teemu Torma for thread safe exception handling support.
26657 * Leonard Tower wrote parts of the parser, RTL generator, and RTL
26658 definitions, and of the VAX machine description.
26660 * Tom Tromey for internationalization support and for his many Java
26661 contributions and libgcj maintainership.
26663 * Lassi Tuura for improvements to config.guess to determine HP
26666 * Petter Urkedal for libstdc++ CXXFLAGS, math, and algorithms fixes.
26668 * Andy Vaught for the design and initial implementation of the
26669 gfortran front end.
26671 * Brent Verner for work with the libstdc++ cshadow files and their
26672 associated configure steps.
26674 * Todd Vierling for contributions for NetBSD ports.
26676 * Jonathan Wakely for contributing libstdc++ Doxygen notes and XHTML
26679 * Dean Wakerley for converting the install documentation from HTML
26680 to texinfo in time for GCC 3.0.
26682 * Krister Walfridsson for random bug fixes.
26684 * Feng Wang for contributions to gfortran.
26686 * Stephen M. Webb for time and effort on making libstdc++ shadow
26687 files work with the tricky Solaris 8+ headers, and for pushing the
26688 build-time header tree.
26690 * John Wehle for various improvements for the x86 code generator,
26691 related infrastructure improvements to help x86 code generation,
26692 value range propagation and other work, WE32k port.
26694 * Ulrich Weigand for work on the s390 port.
26696 * Zack Weinberg for major work on cpplib and various other bug fixes.
26698 * Matt Welsh for help with Linux Threads support in GCJ.
26700 * Urban Widmark for help fixing java.io.
26702 * Mark Wielaard for new Java library code and his work integrating
26705 * Dale Wiles helped port GCC to the Tahoe.
26707 * Bob Wilson from Tensilica, Inc. for the Xtensa port.
26709 * Jim Wilson for his direction via the steering committee, tackling
26710 hard problems in various places that nobody else wanted to work
26711 on, strength reduction and other loop optimizations.
26713 * Carlo Wood for various fixes.
26715 * Tom Wood for work on the m88k port.
26717 * Canqun Yang for work on gfortran.
26719 * Masanobu Yuhara of Fujitsu Laboratories implemented the machine
26720 description for the Tron architecture (specifically, the Gmicro).
26722 * Kevin Zachmann helped port GCC to the Tahoe.
26724 * Ayal Zaks for Swing Modulo Scheduling (SMS).
26726 * Xiaoqiang Zhang for work on gfortran.
26728 * Gilles Zunino for help porting Java to Irix.
26731 The following people are recognized for their contributions to GNAT,
26732 the Ada front end of GCC:
26735 * Romain Berrendonner
26785 * Hristian Kirtchev
26828 In addition to the above, all of which also contributed time and
26829 energy in testing GCC, we would like to thank the following for their
26830 contributions to testing:
26832 * Michael Abd-El-Malek
26842 * David Billinghurst
26846 * Stephane Bortzmeyer
26856 * Bradford Castalia
26876 * Charles-Antoine Gauthier
26898 * Kevin B. Hendricks
26902 * Christian Joensson
26910 * Anand Krishnaswamy
26912 * A. O. V. Le Blanc
26976 * Pedro A. M. Vazquez
26986 And finally we'd like to thank everyone who uses the compiler, submits
26987 bug reports and generally reminds us why we're doing this work in the
26991 File: gcc.info, Node: Option Index, Next: Keyword Index, Prev: Contributors, Up: Top
26996 GCC's command line options are indexed here without any initial `-' or
26997 `--'. Where an option has both positive and negative forms (such as
26998 `-fOPTION' and `-fno-OPTION'), relevant entries in the manual are
26999 indexed under the most appropriate form; it may sometimes be useful to
27000 look up both forms.
27005 * ###: Overall Options. (line 184)
27006 * -dynamiclib: Darwin Options. (line 108)
27007 * -force_cpusubtype_ALL: Darwin Options. (line 113)
27008 * -fsplit-ivs-in-unroller: Optimize Options. (line 695)
27009 * -fvariable-expansion-in-unroller: Optimize Options. (line 710)
27010 * -gfull: Darwin Options. (line 64)
27011 * -gused: Darwin Options. (line 59)
27012 * -mone-byte-bool: Darwin Options. (line 67)
27013 * A: Preprocessor Options.
27015 * all_load: Darwin Options. (line 87)
27016 * allowable_client: Darwin Options. (line 174)
27017 * ansi <1>: Non-bugs. (line 107)
27018 * ansi <2>: Other Builtins. (line 22)
27019 * ansi <3>: Preprocessor Options.
27021 * ansi <4>: C Dialect Options. (line 11)
27022 * ansi: Standards. (line 13)
27023 * arch_errors_fatal: Darwin Options. (line 91)
27024 * aux-info: C Dialect Options. (line 98)
27025 * b: Target Options. (line 13)
27026 * B: Directory Options. (line 41)
27027 * bcopy-builtin: PDP-11 Options. (line 32)
27028 * bind_at_load: Darwin Options. (line 95)
27029 * bundle: Darwin Options. (line 100)
27030 * bundle_loader: Darwin Options. (line 104)
27031 * c: Link Options. (line 20)
27032 * C: Preprocessor Options.
27034 * c: Overall Options. (line 139)
27035 * client_name: Darwin Options. (line 174)
27036 * combine: Overall Options. (line 195)
27037 * compatibility_version: Darwin Options. (line 174)
27038 * crossjumping: Optimize Options. (line 436)
27039 * current_version: Darwin Options. (line 174)
27040 * D: Preprocessor Options.
27042 * d: Debugging Options. (line 216)
27043 * da: Debugging Options. (line 387)
27044 * dA: Debugging Options. (line 229)
27045 * dB: Debugging Options. (line 238)
27046 * db: Debugging Options. (line 234)
27047 * dC: Debugging Options. (line 248)
27048 * dc: Debugging Options. (line 242)
27049 * dD <1>: Preprocessor Options.
27051 * dD: Debugging Options. (line 262)
27052 * dd: Debugging Options. (line 256)
27053 * dE: Debugging Options. (line 267)
27054 * dead_strip: Darwin Options. (line 174)
27055 * dependency-file: Darwin Options. (line 174)
27056 * df: Debugging Options. (line 272)
27057 * dG: Debugging Options. (line 284)
27058 * dg: Debugging Options. (line 279)
27059 * dH: Debugging Options. (line 390)
27060 * dh: Debugging Options. (line 291)
27061 * dI: Preprocessor Options.
27063 * di: Debugging Options. (line 295)
27064 * dj: Debugging Options. (line 299)
27065 * dk: Debugging Options. (line 303)
27066 * dL: Debugging Options. (line 313)
27067 * dl: Debugging Options. (line 308)
27068 * dM: Preprocessor Options.
27070 * dm: Debugging Options. (line 393)
27071 * dM: Debugging Options. (line 324)
27072 * dm: Debugging Options. (line 320)
27073 * dN <1>: Preprocessor Options.
27075 * dN: Debugging Options. (line 333)
27076 * dn: Debugging Options. (line 329)
27077 * do: Debugging Options. (line 337)
27078 * dP: Debugging Options. (line 402)
27079 * dp: Debugging Options. (line 397)
27080 * dR: Debugging Options. (line 345)
27081 * dr: Debugging Options. (line 341)
27082 * dS: Debugging Options. (line 354)
27083 * ds: Debugging Options. (line 349)
27084 * dT: Debugging Options. (line 363)
27085 * dt: Debugging Options. (line 358)
27086 * dumpmachine: Debugging Options. (line 719)
27087 * dumpspecs: Debugging Options. (line 727)
27088 * dumpversion: Debugging Options. (line 723)
27089 * dv: Debugging Options. (line 406)
27090 * dV: Debugging Options. (line 368)
27091 * dw: Debugging Options. (line 375)
27092 * dx: Debugging Options. (line 411)
27093 * dy: Debugging Options. (line 415)
27094 * dylib_file: Darwin Options. (line 174)
27095 * dylinker_install_name: Darwin Options. (line 174)
27096 * dynamic: Darwin Options. (line 174)
27097 * dZ: Debugging Options. (line 383)
27098 * dz: Debugging Options. (line 379)
27099 * E <1>: Link Options. (line 20)
27100 * E: Overall Options. (line 160)
27101 * EB <1>: MIPS Options. (line 7)
27102 * EB: ARC Options. (line 12)
27103 * EL <1>: MIPS Options. (line 10)
27104 * EL: ARC Options. (line 9)
27105 * exported_symbols_list: Darwin Options. (line 174)
27106 * F: Darwin Options. (line 32)
27107 * fabi-version: C++ Dialect Options.
27109 * falign-functions: Optimize Options. (line 823)
27110 * falign-jumps: Optimize Options. (line 873)
27111 * falign-labels: Optimize Options. (line 841)
27112 * falign-loops: Optimize Options. (line 859)
27113 * fargument-alias: Code Gen Options. (line 328)
27114 * fargument-noalias: Code Gen Options. (line 328)
27115 * fargument-noalias-global: Code Gen Options. (line 328)
27116 * fasynchronous-unwind-tables: Code Gen Options. (line 64)
27117 * fbounds-check <1>: Code Gen Options. (line 15)
27118 * fbounds-check: Optimize Options. (line 312)
27119 * fbranch-probabilities: Optimize Options. (line 1089)
27120 * fbranch-target-load-optimize: Optimize Options. (line 1221)
27121 * fbranch-target-load-optimize2: Optimize Options. (line 1227)
27122 * fbtr-bb-exclusive: Optimize Options. (line 1231)
27123 * fcall-saved <1>: Interoperation. (line 150)
27124 * fcall-saved: Code Gen Options. (line 230)
27125 * fcall-used: Code Gen Options. (line 216)
27126 * fcaller-saves: Optimize Options. (line 576)
27127 * fcheck-new: C++ Dialect Options.
27129 * fcommon: Variable Attributes.
27131 * fcond-mismatch: C Dialect Options. (line 208)
27132 * fconserve-space: C++ Dialect Options.
27134 * fconstant-string-class: Objective-C and Objective-C++ Dialect Options.
27136 * fcse-follow-jumps: Optimize Options. (line 355)
27137 * fcse-skip-blocks: Optimize Options. (line 364)
27138 * fcx-limited-range: Optimize Options. (line 1075)
27139 * fdata-sections: Optimize Options. (line 1202)
27140 * fdelayed-branch: Optimize Options. (line 489)
27141 * fdelete-null-pointer-checks: Optimize Options. (line 458)
27142 * fdiagnostics-show-location: Language Independent Options.
27144 * fdollars-in-identifiers <1>: Interoperation. (line 146)
27145 * fdollars-in-identifiers: Preprocessor Options.
27147 * fdump-class-hierarchy: Debugging Options. (line 434)
27148 * fdump-ipa: Debugging Options. (line 441)
27149 * fdump-rtl-all: Debugging Options. (line 387)
27150 * fdump-rtl-bbro: Debugging Options. (line 238)
27151 * fdump-rtl-bp: Debugging Options. (line 234)
27152 * fdump-rtl-btl: Debugging Options. (line 256)
27153 * fdump-rtl-bypass: Debugging Options. (line 284)
27154 * fdump-rtl-ce1: Debugging Options. (line 248)
27155 * fdump-rtl-ce2: Debugging Options. (line 248)
27156 * fdump-rtl-ce3: Debugging Options. (line 267)
27157 * fdump-rtl-cfg: Debugging Options. (line 272)
27158 * fdump-rtl-combine: Debugging Options. (line 242)
27159 * fdump-rtl-cse: Debugging Options. (line 349)
27160 * fdump-rtl-cse2: Debugging Options. (line 358)
27161 * fdump-rtl-dbr: Debugging Options. (line 256)
27162 * fdump-rtl-eh: Debugging Options. (line 291)
27163 * fdump-rtl-expand: Debugging Options. (line 341)
27164 * fdump-rtl-flow2: Debugging Options. (line 375)
27165 * fdump-rtl-gcse: Debugging Options. (line 284)
27166 * fdump-rtl-greg: Debugging Options. (line 279)
27167 * fdump-rtl-jump: Debugging Options. (line 299)
27168 * fdump-rtl-life: Debugging Options. (line 272)
27169 * fdump-rtl-loop: Debugging Options. (line 313)
27170 * fdump-rtl-loop2: Debugging Options. (line 313)
27171 * fdump-rtl-lreg: Debugging Options. (line 308)
27172 * fdump-rtl-mach: Debugging Options. (line 324)
27173 * fdump-rtl-peephole2: Debugging Options. (line 379)
27174 * fdump-rtl-postreload: Debugging Options. (line 337)
27175 * fdump-rtl-regmove: Debugging Options. (line 333)
27176 * fdump-rtl-rnreg: Debugging Options. (line 329)
27177 * fdump-rtl-sched: Debugging Options. (line 354)
27178 * fdump-rtl-sched2: Debugging Options. (line 345)
27179 * fdump-rtl-sibling: Debugging Options. (line 295)
27180 * fdump-rtl-sms: Debugging Options. (line 320)
27181 * fdump-rtl-stack: Debugging Options. (line 303)
27182 * fdump-rtl-tracer: Debugging Options. (line 363)
27183 * fdump-rtl-vartrack: Debugging Options. (line 368)
27184 * fdump-rtl-vpt: Debugging Options. (line 368)
27185 * fdump-rtl-web: Debugging Options. (line 383)
27186 * fdump-translation-unit: Debugging Options. (line 426)
27187 * fdump-tree: Debugging Options. (line 456)
27188 * fdump-tree-alias: Debugging Options. (line 539)
27189 * fdump-tree-all: Debugging Options. (line 604)
27190 * fdump-tree-ccp: Debugging Options. (line 543)
27191 * fdump-tree-cfg: Debugging Options. (line 519)
27192 * fdump-tree-ch: Debugging Options. (line 531)
27193 * fdump-tree-copyrename: Debugging Options. (line 589)
27194 * fdump-tree-dce: Debugging Options. (line 555)
27195 * fdump-tree-dom: Debugging Options. (line 569)
27196 * fdump-tree-dse: Debugging Options. (line 574)
27197 * fdump-tree-forwprop: Debugging Options. (line 584)
27198 * fdump-tree-fre: Debugging Options. (line 551)
27199 * fdump-tree-gimple: Debugging Options. (line 514)
27200 * fdump-tree-mudflap: Debugging Options. (line 559)
27201 * fdump-tree-nrv: Debugging Options. (line 594)
27202 * fdump-tree-phiopt: Debugging Options. (line 579)
27203 * fdump-tree-pre: Debugging Options. (line 547)
27204 * fdump-tree-sra: Debugging Options. (line 564)
27205 * fdump-tree-ssa: Debugging Options. (line 535)
27206 * fdump-tree-vcg: Debugging Options. (line 523)
27207 * fdump-tree-vect: Debugging Options. (line 599)
27208 * fdump-unnumbered: Debugging Options. (line 418)
27209 * feliminate-dwarf2-dups: Debugging Options. (line 117)
27210 * feliminate-unused-debug-symbols: Debugging Options. (line 52)
27211 * feliminate-unused-debug-types: Debugging Options. (line 731)
27212 * fexceptions: Code Gen Options. (line 34)
27213 * fexec-charset: Preprocessor Options.
27215 * fexpensive-optimizations: Optimize Options. (line 471)
27216 * ffast-math: Optimize Options. (line 970)
27217 * ffinite-math-only: Optimize Options. (line 1010)
27218 * ffix-and-continue: Darwin Options. (line 81)
27219 * ffixed: Code Gen Options. (line 204)
27220 * ffloat-store <1>: Disappointments. (line 77)
27221 * ffloat-store: Optimize Options. (line 956)
27222 * ffor-scope: C++ Dialect Options.
27224 * fforce-addr: Optimize Options. (line 156)
27225 * fforce-mem: Optimize Options. (line 147)
27226 * ffreestanding <1>: Function Attributes.
27228 * ffreestanding <2>: Warning Options. (line 94)
27229 * ffreestanding <3>: C Dialect Options. (line 169)
27230 * ffreestanding: Standards. (line 81)
27231 * ffunction-sections: Optimize Options. (line 1202)
27232 * fgcse: Optimize Options. (line 383)
27233 * fgcse-after-reload: Optimize Options. (line 419)
27234 * fgcse-las: Optimize Options. (line 412)
27235 * fgcse-lm: Optimize Options. (line 394)
27236 * fgcse-sm: Optimize Options. (line 403)
27237 * fgnu-runtime: Objective-C and Objective-C++ Dialect Options.
27239 * fhosted: C Dialect Options. (line 162)
27240 * filelist: Darwin Options. (line 174)
27241 * findirect-data: Darwin Options. (line 81)
27242 * finhibit-size-directive: Code Gen Options. (line 154)
27243 * finline-functions: Optimize Options. (line 188)
27244 * finline-limit: Optimize Options. (line 199)
27245 * finput-charset: Preprocessor Options.
27247 * finstrument-functions <1>: Function Attributes.
27249 * finstrument-functions: Code Gen Options. (line 260)
27250 * fkeep-inline-functions <1>: Inline. (line 51)
27251 * fkeep-inline-functions: Optimize Options. (line 237)
27252 * fkeep-static-consts: Optimize Options. (line 244)
27253 * flat_namespace: Darwin Options. (line 174)
27254 * fleading-underscore: Code Gen Options. (line 343)
27255 * floop-optimize: Optimize Options. (line 424)
27256 * floop-optimize2: Optimize Options. (line 431)
27257 * fmem-report: Debugging Options. (line 142)
27258 * fmessage-length: Language Independent Options.
27260 * fmodulo-sched: Optimize Options. (line 273)
27261 * fmove-loop-invariants: Optimize Options. (line 1185)
27262 * fms-extensions <1>: Unnamed Fields. (line 37)
27263 * fms-extensions <2>: C++ Dialect Options.
27265 * fms-extensions: C Dialect Options. (line 179)
27266 * fmudflap: Optimize Options. (line 319)
27267 * fmudflapir: Optimize Options. (line 319)
27268 * fmudflapth: Optimize Options. (line 319)
27269 * fnext-runtime: Objective-C and Objective-C++ Dialect Options.
27271 * fno-access-control: C++ Dialect Options.
27273 * fno-asm: C Dialect Options. (line 114)
27274 * fno-branch-count-reg: Optimize Options. (line 278)
27275 * fno-builtin <1>: Other Builtins. (line 14)
27276 * fno-builtin <2>: Function Attributes.
27278 * fno-builtin <3>: Warning Options. (line 94)
27279 * fno-builtin: C Dialect Options. (line 128)
27280 * fno-common <1>: Variable Attributes.
27282 * fno-common: Code Gen Options. (line 142)
27283 * fno-const-strings: C++ Dialect Options.
27285 * fno-cprop-registers: Optimize Options. (line 928)
27286 * fno-cx-limited-range: Optimize Options. (line 1075)
27287 * fno-default-inline <1>: Inline. (line 46)
27288 * fno-default-inline <2>: Optimize Options. (line 132)
27289 * fno-default-inline: C++ Dialect Options.
27291 * fno-defer-pop: Optimize Options. (line 139)
27292 * fno-elide-constructors: C++ Dialect Options.
27294 * fno-enforce-eh-specs: C++ Dialect Options.
27296 * fno-for-scope: C++ Dialect Options.
27298 * fno-function-cse: Optimize Options. (line 289)
27299 * fno-gnu-keywords: C++ Dialect Options.
27301 * fno-guess-branch-probability: Optimize Options. (line 733)
27302 * fno-ident: Code Gen Options. (line 151)
27303 * fno-implement-inlines <1>: C++ Interface. (line 75)
27304 * fno-implement-inlines: C++ Dialect Options.
27306 * fno-implicit-inline-templates: C++ Dialect Options.
27308 * fno-implicit-templates <1>: Template Instantiation.
27310 * fno-implicit-templates: C++ Dialect Options.
27312 * fno-inline: Optimize Options. (line 182)
27313 * fno-math-errno: Optimize Options. (line 983)
27314 * fno-nil-receivers: Objective-C and Objective-C++ Dialect Options.
27316 * fno-nonansi-builtins: C++ Dialect Options.
27318 * fno-operator-names: C++ Dialect Options.
27320 * fno-optional-diags: C++ Dialect Options.
27322 * fno-peephole: Optimize Options. (line 724)
27323 * fno-peephole2: Optimize Options. (line 724)
27324 * fno-rtti: C++ Dialect Options.
27326 * fno-sched-interblock: Optimize Options. (line 515)
27327 * fno-sched-spec: Optimize Options. (line 520)
27328 * fno-show-column: Preprocessor Options.
27330 * fno-signed-bitfields: C Dialect Options. (line 241)
27331 * fno-stack-limit: Code Gen Options. (line 312)
27332 * fno-threadsafe-statics: C++ Dialect Options.
27334 * fno-trapping-math: Optimize Options. (line 1020)
27335 * fno-unsigned-bitfields: C Dialect Options. (line 241)
27336 * fno-weak: C++ Dialect Options.
27338 * fno-working-directory: Preprocessor Options.
27340 * fno-zero-initialized-in-bss: Optimize Options. (line 300)
27341 * fnon-call-exceptions: Code Gen Options. (line 48)
27342 * fobjc-exceptions: Objective-C and Objective-C++ Dialect Options.
27344 * fomit-frame-pointer: Optimize Options. (line 161)
27345 * foptimize-register-move: Optimize Options. (line 478)
27346 * foptimize-sibling-calls: Optimize Options. (line 177)
27347 * force_flat_namespace: Darwin Options. (line 174)
27348 * fpack-struct: Code Gen Options. (line 247)
27349 * fpcc-struct-return <1>: Incompatibilities. (line 170)
27350 * fpcc-struct-return: Code Gen Options. (line 70)
27351 * fpch-deps: Preprocessor Options.
27353 * fpch-preprocess: Preprocessor Options.
27355 * fpeel-loops: Optimize Options. (line 1177)
27356 * fpermissive: C++ Dialect Options.
27358 * fPIC: Code Gen Options. (line 188)
27359 * fpic: Code Gen Options. (line 170)
27360 * fPIE: Code Gen Options. (line 198)
27361 * fpie: Code Gen Options. (line 198)
27362 * fprefetch-loop-arrays: Optimize Options. (line 715)
27363 * fpreprocessed: Preprocessor Options.
27365 * fprofile-arcs <1>: Other Builtins. (line 236)
27366 * fprofile-arcs: Debugging Options. (line 146)
27367 * fprofile-generate: Optimize Options. (line 935)
27368 * fprofile-use: Optimize Options. (line 944)
27369 * fprofile-values: Optimize Options. (line 1108)
27370 * frandom-string: Debugging Options. (line 614)
27371 * freg-struct-return: Code Gen Options. (line 88)
27372 * fregmove: Optimize Options. (line 478)
27373 * frename-registers: Optimize Options. (line 1144)
27374 * freorder-blocks: Optimize Options. (line 750)
27375 * freorder-blocks-and-partition: Optimize Options. (line 756)
27376 * freorder-functions: Optimize Options. (line 767)
27377 * freplace-objc-classes: Objective-C and Objective-C++ Dialect Options.
27379 * frepo <1>: Template Instantiation.
27381 * frepo: C++ Dialect Options.
27383 * frerun-cse-after-loop: Optimize Options. (line 372)
27384 * frerun-loop-opt: Optimize Options. (line 378)
27385 * frounding-math: Optimize Options. (line 1035)
27386 * fsched-spec-load: Optimize Options. (line 525)
27387 * fsched-spec-load-dangerous: Optimize Options. (line 530)
27388 * fsched-stalled-insns: Optimize Options. (line 535)
27389 * fsched-stalled-insns-dep: Optimize Options. (line 540)
27390 * fsched-verbose: Debugging Options. (line 624)
27391 * fsched2-use-superblocks: Optimize Options. (line 547)
27392 * fsched2-use-traces: Optimize Options. (line 558)
27393 * fschedule-insns: Optimize Options. (line 496)
27394 * fschedule-insns2: Optimize Options. (line 506)
27395 * fscheduling-in-modulo-scheduled-loops: Optimize Options. (line 570)
27396 * fshared-data: Code Gen Options. (line 135)
27397 * fshort-double: Code Gen Options. (line 117)
27398 * fshort-enums <1>: Non-bugs. (line 42)
27399 * fshort-enums <2>: Type Attributes. (line 112)
27400 * fshort-enums <3>: Structures unions enumerations and bit-fields implementation.
27402 * fshort-enums: Code Gen Options. (line 106)
27403 * fshort-wchar: Code Gen Options. (line 125)
27404 * fsignaling-nans: Optimize Options. (line 1055)
27405 * fsigned-bitfields <1>: Non-bugs. (line 57)
27406 * fsigned-bitfields: C Dialect Options. (line 241)
27407 * fsigned-char <1>: Characters implementation.
27409 * fsigned-char: C Dialect Options. (line 231)
27410 * fsingle-precision-constant: Optimize Options. (line 1070)
27411 * fspeculative-prefetching: Optimize Options. (line 1127)
27412 * fstack-check: Code Gen Options. (line 297)
27413 * fstack-limit-register: Code Gen Options. (line 312)
27414 * fstack-limit-symbol: Code Gen Options. (line 312)
27415 * fstats: C++ Dialect Options.
27417 * fstrength-reduce: Optimize Options. (line 340)
27418 * fstrict-aliasing: Optimize Options. (line 780)
27419 * fsyntax-only: Warning Options. (line 23)
27420 * ftabstop: Preprocessor Options.
27422 * ftemplate-depth: C++ Dialect Options.
27424 * ftest-coverage: Debugging Options. (line 205)
27425 * fthread-jumps: Optimize Options. (line 346)
27426 * ftime-report: Debugging Options. (line 138)
27427 * ftracer: Optimize Options. (line 678)
27428 * ftrapv: Code Gen Options. (line 22)
27429 * ftree-based-profiling: Debugging Options. (line 195)
27430 * ftree-vectorizer-verbose: Debugging Options. (line 608)
27431 * funit-at-a-time: Optimize Options. (line 886)
27432 * funroll-all-loops: Optimize Options. (line 689)
27433 * funroll-loops <1>: Non-bugs. (line 174)
27434 * funroll-loops: Optimize Options. (line 683)
27435 * funsafe-math-optimizations: Optimize Options. (line 996)
27436 * funsigned-bitfields <1>: Non-bugs. (line 57)
27437 * funsigned-bitfields <2>: Structures unions enumerations and bit-fields implementation.
27439 * funsigned-bitfields: C Dialect Options. (line 241)
27440 * funsigned-char <1>: Characters implementation.
27442 * funsigned-char: C Dialect Options. (line 213)
27443 * funswitch-loops: Optimize Options. (line 1189)
27444 * funwind-tables: Code Gen Options. (line 57)
27445 * fuse-cxa-atexit: C++ Dialect Options.
27447 * fvar-tracking: Debugging Options. (line 667)
27448 * fverbose-asm: Code Gen Options. (line 161)
27449 * fvisibility: Code Gen Options. (line 362)
27450 * fvisibility-inlines-hidden: C++ Dialect Options.
27452 * fvpt: Optimize Options. (line 1118)
27453 * fweb: Optimize Options. (line 915)
27454 * fwide-exec-charset: Preprocessor Options.
27456 * fworking-directory: Preprocessor Options.
27458 * fwrapv: Code Gen Options. (line 26)
27459 * fzero-link: Objective-C and Objective-C++ Dialect Options.
27461 * G <1>: System V Options. (line 10)
27462 * G <2>: RS/6000 and PowerPC Options.
27464 * G <3>: MIPS Options. (line 195)
27465 * G: M32R/D Options. (line 57)
27466 * g: Debugging Options. (line 10)
27467 * gcoff: Debugging Options. (line 62)
27468 * gdwarf-2: Debugging Options. (line 80)
27469 * gen-decls: Objective-C and Objective-C++ Dialect Options.
27471 * ggdb: Debugging Options. (line 38)
27472 * gnu-ld: HPPA Options. (line 113)
27473 * gstabs: Debugging Options. (line 44)
27474 * gstabs+: Debugging Options. (line 56)
27475 * gvms: Debugging Options. (line 87)
27476 * gxcoff: Debugging Options. (line 67)
27477 * gxcoff+: Debugging Options. (line 72)
27478 * H: Preprocessor Options.
27480 * headerpad_max_install_names: Darwin Options. (line 174)
27481 * help <1>: Preprocessor Options.
27483 * help: Overall Options. (line 211)
27484 * hp-ld: HPPA Options. (line 123)
27485 * I <1>: Directory Options. (line 10)
27486 * I: Preprocessor Options.
27488 * I- <1>: Directory Options. (line 92)
27489 * I-: Preprocessor Options.
27491 * idirafter: Preprocessor Options.
27493 * if-conversion: Optimize Options. (line 443)
27494 * if-conversion2: Optimize Options. (line 452)
27495 * imacros: Preprocessor Options.
27497 * image_base: Darwin Options. (line 174)
27498 * include: Preprocessor Options.
27500 * init: Darwin Options. (line 174)
27501 * install_name: Darwin Options. (line 174)
27502 * iprefix: Preprocessor Options.
27504 * iquote <1>: Directory Options. (line 31)
27505 * iquote: Preprocessor Options.
27507 * isystem: Preprocessor Options.
27509 * iwithprefix: Preprocessor Options.
27511 * iwithprefixbefore: Preprocessor Options.
27513 * keep_private_externs: Darwin Options. (line 174)
27514 * L: Directory Options. (line 37)
27515 * l: Link Options. (line 26)
27516 * lobjc: Link Options. (line 53)
27517 * M: Preprocessor Options.
27519 * m1: SH Options. (line 9)
27520 * m10: PDP-11 Options. (line 29)
27521 * m128bit-long-double: i386 and x86-64 Options.
27523 * m16-bit: CRIS Options. (line 69)
27524 * m2: SH Options. (line 12)
27525 * m210: MCore Options. (line 43)
27526 * m3: SH Options. (line 18)
27527 * m31: S/390 and zSeries Options.
27529 * m32 <1>: SPARC Options. (line 187)
27530 * m32 <2>: RS/6000 and PowerPC Options.
27532 * m32: i386 and x86-64 Options.
27534 * m32-bit: CRIS Options. (line 69)
27535 * m32032: NS32K Options. (line 13)
27536 * m32081: NS32K Options. (line 27)
27537 * m32332: NS32K Options. (line 18)
27538 * m32381: NS32K Options. (line 31)
27539 * m32532: NS32K Options. (line 23)
27540 * m32r: M32R/D Options. (line 15)
27541 * m32r2: M32R/D Options. (line 9)
27542 * m32rx: M32R/D Options. (line 12)
27543 * m340: MCore Options. (line 43)
27544 * m386: i386 and x86-64 Options.
27546 * m3dnow: i386 and x86-64 Options.
27548 * m3e: SH Options. (line 21)
27549 * m4: SH Options. (line 35)
27550 * m4-nofpu: SH Options. (line 24)
27551 * m4-single: SH Options. (line 31)
27552 * m4-single-only: SH Options. (line 27)
27553 * m40: PDP-11 Options. (line 23)
27554 * m45: PDP-11 Options. (line 26)
27555 * m486: i386 and x86-64 Options.
27557 * m4a: SH Options. (line 50)
27558 * m4a-nofpu: SH Options. (line 38)
27559 * m4a-single: SH Options. (line 46)
27560 * m4a-single-only: SH Options. (line 42)
27561 * m4al: SH Options. (line 53)
27562 * m4byte-functions: MCore Options. (line 27)
27563 * m5200: M680x0 Options. (line 59)
27564 * m64 <1>: SPARC Options. (line 187)
27565 * m64 <2>: S/390 and zSeries Options.
27567 * m64 <3>: RS/6000 and PowerPC Options.
27569 * m64: i386 and x86-64 Options.
27571 * m68000: M680x0 Options. (line 13)
27572 * m68020: M680x0 Options. (line 21)
27573 * m68020-40: M680x0 Options. (line 66)
27574 * m68020-60: M680x0 Options. (line 73)
27575 * m68030: M680x0 Options. (line 30)
27576 * m68040: M680x0 Options. (line 34)
27577 * m68060: M680x0 Options. (line 42)
27578 * m6811: M68hc1x Options. (line 13)
27579 * m6812: M68hc1x Options. (line 18)
27580 * m68881: M680x0 Options. (line 25)
27581 * m68hc11: M68hc1x Options. (line 13)
27582 * m68hc12: M68hc1x Options. (line 18)
27583 * m68hcs12: M68hc1x Options. (line 23)
27584 * m68S12: M68hc1x Options. (line 23)
27585 * m8-bit: CRIS Options. (line 69)
27586 * m96bit-long-double: i386 and x86-64 Options.
27588 * mabi: ARM Options. (line 10)
27589 * mabi-mmixware: MMIX Options. (line 20)
27590 * mabi=32: MIPS Options. (line 87)
27591 * mabi=64: MIPS Options. (line 87)
27592 * mabi=altivec: RS/6000 and PowerPC Options.
27594 * mabi=eabi: MIPS Options. (line 87)
27595 * mabi=gnu: MMIX Options. (line 20)
27596 * mabi=n32: MIPS Options. (line 87)
27597 * mabi=no-altivec: RS/6000 and PowerPC Options.
27599 * mabi=no-spe: RS/6000 and PowerPC Options.
27601 * mabi=o64: MIPS Options. (line 87)
27602 * mabi=spe: RS/6000 and PowerPC Options.
27604 * mabicalls: MIPS Options. (line 98)
27605 * mabort-on-noreturn: ARM Options. (line 144)
27606 * mabshi: PDP-11 Options. (line 55)
27607 * mac0: PDP-11 Options. (line 16)
27608 * macc-4: FRV Options. (line 113)
27609 * macc-8: FRV Options. (line 116)
27610 * maccumulate-outgoing-args: i386 and x86-64 Options.
27612 * mads: RS/6000 and PowerPC Options.
27614 * maix-struct-return: RS/6000 and PowerPC Options.
27616 * maix32: RS/6000 and PowerPC Options.
27618 * maix64: RS/6000 and PowerPC Options.
27620 * malign-300: H8/300 Options. (line 31)
27621 * malign-double: i386 and x86-64 Options.
27623 * malign-int: M680x0 Options. (line 128)
27624 * malign-labels: FRV Options. (line 104)
27625 * malign-loops: M32R/D Options. (line 73)
27626 * malign-natural: RS/6000 and PowerPC Options.
27628 * malign-power: RS/6000 and PowerPC Options.
27630 * malloc-cc: FRV Options. (line 25)
27631 * malpha-as: DEC Alpha Options. (line 159)
27632 * maltivec: RS/6000 and PowerPC Options.
27634 * mam33: MN10300 Options. (line 17)
27635 * maout: CRIS Options. (line 92)
27636 * mapcs: ARM Options. (line 22)
27637 * mapcs-frame: ARM Options. (line 14)
27638 * mapp-regs <1>: V850 Options. (line 57)
27639 * mapp-regs: SPARC Options. (line 10)
27640 * march <1>: S/390 and zSeries Options.
27642 * march <2>: MIPS Options. (line 14)
27643 * march <3>: i386 and x86-64 Options.
27645 * march <4>: HPPA Options. (line 9)
27646 * march <5>: CRIS Options. (line 10)
27647 * march: ARM Options. (line 109)
27648 * masm=DIALECT: i386 and x86-64 Options.
27650 * mauto-incdec: M68hc1x Options. (line 26)
27651 * mauto-pic: IA-64 Options. (line 50)
27652 * mb: SH Options. (line 58)
27653 * mbackchain: S/390 and zSeries Options.
27655 * mbase-addresses: MMIX Options. (line 54)
27656 * mbcopy: PDP-11 Options. (line 36)
27657 * mbig <1>: TMS320C3x/C4x Options.
27659 * mbig: RS/6000 and PowerPC Options.
27661 * mbig-endian <1>: RS/6000 and PowerPC Options.
27663 * mbig-endian <2>: MCore Options. (line 39)
27664 * mbig-endian <3>: IA-64 Options. (line 9)
27665 * mbig-endian: ARM Options. (line 72)
27666 * mbig-memory: TMS320C3x/C4x Options.
27668 * mbig-switch <1>: V850 Options. (line 52)
27669 * mbig-switch: HPPA Options. (line 23)
27670 * mbigtable: SH Options. (line 74)
27671 * mbit-align: RS/6000 and PowerPC Options.
27673 * mbitfield <1>: NS32K Options. (line 66)
27674 * mbitfield: M680x0 Options. (line 100)
27675 * mbk: TMS320C3x/C4x Options.
27677 * mbranch-cheap: PDP-11 Options. (line 65)
27678 * mbranch-cost=NUMBER: M32R/D Options. (line 82)
27679 * mbranch-expensive: PDP-11 Options. (line 61)
27680 * mbranch-likely: MIPS Options. (line 346)
27681 * mbranch-predict: MMIX Options. (line 49)
27682 * mbuild-constants: DEC Alpha Options. (line 142)
27683 * mbwx: DEC Alpha Options. (line 171)
27684 * mc68000: M680x0 Options. (line 13)
27685 * mc68020: M680x0 Options. (line 21)
27686 * mcall-gnu: RS/6000 and PowerPC Options.
27688 * mcall-linux: RS/6000 and PowerPC Options.
27690 * mcall-netbsd: RS/6000 and PowerPC Options.
27692 * mcall-prologues: AVR Options. (line 43)
27693 * mcall-solaris: RS/6000 and PowerPC Options.
27695 * mcall-sysv: RS/6000 and PowerPC Options.
27697 * mcall-sysv-eabi: RS/6000 and PowerPC Options.
27699 * mcall-sysv-noeabi: RS/6000 and PowerPC Options.
27701 * mcallee-super-interworking: ARM Options. (line 234)
27702 * mcaller-super-interworking: ARM Options. (line 240)
27703 * mcallgraph-data: MCore Options. (line 31)
27704 * mcc-init: CRIS Options. (line 46)
27705 * mcheck-zero-division: MIPS Options. (line 233)
27706 * mcirrus-fix-invalid-insns: ARM Options. (line 187)
27707 * mcix: DEC Alpha Options. (line 171)
27708 * mcmodel=embmedany: SPARC Options. (line 209)
27709 * mcmodel=kernel: i386 and x86-64 Options.
27711 * mcmodel=large: i386 and x86-64 Options.
27713 * mcmodel=medany: SPARC Options. (line 203)
27714 * mcmodel=medium: i386 and x86-64 Options.
27716 * mcmodel=medlow: SPARC Options. (line 192)
27717 * mcmodel=medmid: SPARC Options. (line 197)
27718 * mcmodel=small: i386 and x86-64 Options.
27720 * mcond-exec: FRV Options. (line 152)
27721 * mcond-move: FRV Options. (line 128)
27722 * mconst-align: CRIS Options. (line 60)
27723 * mconst16: Xtensa Options. (line 10)
27724 * mconstant-gp: IA-64 Options. (line 46)
27725 * mcpu <1>: TMS320C3x/C4x Options.
27727 * mcpu <2>: SPARC Options. (line 96)
27728 * mcpu <3>: RS/6000 and PowerPC Options.
27730 * mcpu <4>: i386 and x86-64 Options.
27732 * mcpu <5>: FRV Options. (line 204)
27733 * mcpu <6>: DEC Alpha Options. (line 223)
27734 * mcpu <7>: CRIS Options. (line 10)
27735 * mcpu <8>: ARM Options. (line 84)
27736 * mcpu: ARC Options. (line 23)
27737 * mcpu32: M680x0 Options. (line 51)
27738 * mcsync: Blackfin Options. (line 14)
27739 * MD: Preprocessor Options.
27741 * mdalign: SH Options. (line 64)
27742 * mdata: ARC Options. (line 30)
27743 * mdata-align: CRIS Options. (line 60)
27744 * mdb: TMS320C3x/C4x Options.
27746 * mdebug <1>: S/390 and zSeries Options.
27748 * mdebug: M32R/D Options. (line 69)
27749 * mdec-asm: PDP-11 Options. (line 78)
27750 * mdisable-callt: V850 Options. (line 80)
27751 * mdisable-fpregs: HPPA Options. (line 33)
27752 * mdisable-indexing: HPPA Options. (line 40)
27753 * mdiv: MCore Options. (line 15)
27754 * mdivide-breaks: MIPS Options. (line 238)
27755 * mdivide-traps: MIPS Options. (line 238)
27756 * mdouble: FRV Options. (line 38)
27757 * mdouble-float: MIPS Options. (line 149)
27758 * mdp-isr-reload: TMS320C3x/C4x Options.
27760 * mdwarf2-asm: IA-64 Options. (line 79)
27761 * mdword: FRV Options. (line 32)
27762 * mdynamic-no-pic: RS/6000 and PowerPC Options.
27764 * meabi: RS/6000 and PowerPC Options.
27766 * mearly-stop-bits: IA-64 Options. (line 85)
27767 * melf <1>: MMIX Options. (line 44)
27768 * melf: CRIS Options. (line 95)
27769 * melinux: CRIS Options. (line 99)
27770 * melinux-stacksize: CRIS Options. (line 25)
27771 * memb: RS/6000 and PowerPC Options.
27773 * membedded-data: MIPS Options. (line 204)
27774 * mep: V850 Options. (line 16)
27775 * mepsilon: MMIX Options. (line 15)
27776 * mesa: S/390 and zSeries Options.
27778 * metrax100: CRIS Options. (line 31)
27779 * metrax4: CRIS Options. (line 31)
27780 * mexplicit-relocs <1>: MIPS Options. (line 224)
27781 * mexplicit-relocs: DEC Alpha Options. (line 184)
27782 * MF: Preprocessor Options.
27784 * mfast-fix: TMS320C3x/C4x Options.
27786 * mfast-indirect-calls: HPPA Options. (line 52)
27787 * mfaster-structs: SPARC Options. (line 71)
27788 * mfdpic: FRV Options. (line 56)
27789 * mfix: DEC Alpha Options. (line 171)
27790 * mfix-and-continue: Darwin Options. (line 81)
27791 * mfix-r4000: MIPS Options. (line 288)
27792 * mfix-r4400: MIPS Options. (line 302)
27793 * mfix-sb1: MIPS Options. (line 330)
27794 * mfix-vr4120: MIPS Options. (line 309)
27795 * mfix-vr4130: MIPS Options. (line 323)
27796 * mfixed-cc: FRV Options. (line 28)
27797 * mfixed-range <1>: IA-64 Options. (line 90)
27798 * mfixed-range: HPPA Options. (line 59)
27799 * mfloat-abi: ARM Options. (line 59)
27800 * mfloat-gprs: RS/6000 and PowerPC Options.
27802 * mfloat-ieee: DEC Alpha Options. (line 179)
27803 * mfloat-vax: DEC Alpha Options. (line 179)
27804 * mfloat32: PDP-11 Options. (line 52)
27805 * mfloat64: PDP-11 Options. (line 48)
27806 * mflush-func: MIPS Options. (line 336)
27807 * mflush-func=NAME: M32R/D Options. (line 94)
27808 * mflush-trap=NUMBER: M32R/D Options. (line 87)
27809 * mfmovd: SH Options. (line 78)
27810 * mfp: ARM Options. (line 119)
27811 * mfp-exceptions: MIPS Options. (line 357)
27812 * mfp-reg: DEC Alpha Options. (line 25)
27813 * mfp-rounding-mode: DEC Alpha Options. (line 85)
27814 * mfp-trap-mode: DEC Alpha Options. (line 63)
27815 * mfp32: MIPS Options. (line 132)
27816 * mfp64: MIPS Options. (line 135)
27817 * mfpe: ARM Options. (line 119)
27818 * mfpr-32: FRV Options. (line 13)
27819 * mfpr-64: FRV Options. (line 16)
27820 * mfpu <1>: SPARC Options. (line 20)
27821 * mfpu <2>: PDP-11 Options. (line 9)
27822 * mfpu: ARM Options. (line 119)
27823 * mfull-toc: RS/6000 and PowerPC Options.
27825 * mfused-madd <1>: Xtensa Options. (line 19)
27826 * mfused-madd <2>: S/390 and zSeries Options.
27828 * mfused-madd <3>: RS/6000 and PowerPC Options.
27830 * mfused-madd: MIPS Options. (line 273)
27831 * mg: VAX Options. (line 17)
27832 * MG: Preprocessor Options.
27834 * mgas <1>: HPPA Options. (line 75)
27835 * mgas: DEC Alpha Options. (line 159)
27836 * mgnu: VAX Options. (line 13)
27837 * mgnu-as: IA-64 Options. (line 18)
27838 * mgnu-ld: IA-64 Options. (line 23)
27839 * mgotplt: CRIS Options. (line 86)
27840 * mgp32: MIPS Options. (line 126)
27841 * mgp64: MIPS Options. (line 129)
27842 * mgpr-32: FRV Options. (line 7)
27843 * mgpr-64: FRV Options. (line 10)
27844 * mgprel-ro: FRV Options. (line 79)
27845 * mh: H8/300 Options. (line 14)
27846 * mhard-float <1>: SPARC Options. (line 20)
27847 * mhard-float <2>: S/390 and zSeries Options.
27849 * mhard-float <3>: RS/6000 and PowerPC Options.
27851 * mhard-float <4>: MIPS Options. (line 138)
27852 * mhard-float <5>: FRV Options. (line 19)
27853 * mhard-float: ARM Options. (line 41)
27854 * mhard-quad-float: SPARC Options. (line 41)
27855 * mhardlit: MCore Options. (line 10)
27856 * mhimem: NS32K Options. (line 111)
27857 * mhitachi: SH Options. (line 81)
27858 * mid-shared-library: Blackfin Options. (line 30)
27859 * mieee <1>: SH Options. (line 96)
27860 * mieee: DEC Alpha Options. (line 39)
27861 * mieee-compare: NS32K Options. (line 55)
27862 * mieee-conformant: DEC Alpha Options. (line 134)
27863 * mieee-fp: i386 and x86-64 Options.
27865 * mieee-with-inexact: DEC Alpha Options. (line 52)
27866 * milp32: IA-64 Options. (line 114)
27867 * mimpure-text: SPARC Options. (line 81)
27868 * minit-stack: AVR Options. (line 35)
27869 * minline-all-stringops: i386 and x86-64 Options.
27871 * minline-float-divide-max-throughput: IA-64 Options. (line 58)
27872 * minline-float-divide-min-latency: IA-64 Options. (line 54)
27873 * minline-int-divide-max-throughput: IA-64 Options. (line 66)
27874 * minline-int-divide-min-latency: IA-64 Options. (line 62)
27875 * minline-plt: FRV Options. (line 64)
27876 * minline-sqrt-max-throughput: IA-64 Options. (line 74)
27877 * minline-sqrt-min-latency: IA-64 Options. (line 70)
27878 * minmax: M68hc1x Options. (line 31)
27879 * minsert-sched-nops: RS/6000 and PowerPC Options.
27881 * mint16: PDP-11 Options. (line 40)
27882 * mint32 <1>: PDP-11 Options. (line 44)
27883 * mint32: H8/300 Options. (line 28)
27884 * mint64: MIPS Options. (line 165)
27885 * mint8: AVR Options. (line 53)
27886 * mips1: MIPS Options. (line 58)
27887 * mips16: MIPS Options. (line 80)
27888 * mips2: MIPS Options. (line 61)
27889 * mips3: MIPS Options. (line 64)
27890 * mips32: MIPS Options. (line 70)
27891 * mips32r2: MIPS Options. (line 73)
27892 * mips3d: MIPS Options. (line 161)
27893 * mips4: MIPS Options. (line 67)
27894 * mips64: MIPS Options. (line 76)
27895 * misel: RS/6000 and PowerPC Options.
27897 * misize: SH Options. (line 99)
27898 * missue-rate=NUMBER: M32R/D Options. (line 79)
27899 * mjump-in-delay: HPPA Options. (line 28)
27900 * mknuthdiv: MMIX Options. (line 33)
27901 * ml: SH Options. (line 61)
27902 * mlarge-data: DEC Alpha Options. (line 195)
27903 * mlarge-text: DEC Alpha Options. (line 213)
27904 * mlibfuncs: MMIX Options. (line 10)
27905 * mlibrary-pic: FRV Options. (line 110)
27906 * mlinked-fp: FRV Options. (line 94)
27907 * mlinker-opt: HPPA Options. (line 85)
27908 * mlinux: CRIS Options. (line 104)
27909 * mlittle: RS/6000 and PowerPC Options.
27911 * mlittle-endian <1>: SPARC Options. (line 181)
27912 * mlittle-endian <2>: RS/6000 and PowerPC Options.
27914 * mlittle-endian <3>: MCore Options. (line 39)
27915 * mlittle-endian <4>: IA-64 Options. (line 13)
27916 * mlittle-endian: ARM Options. (line 68)
27917 * mlong-calls <1>: V850 Options. (line 10)
27918 * mlong-calls <2>: MIPS Options. (line 259)
27919 * mlong-calls <3>: M68hc1x Options. (line 35)
27920 * mlong-calls <4>: FRV Options. (line 99)
27921 * mlong-calls: ARM Options. (line 149)
27922 * mlong-load-store: HPPA Options. (line 66)
27923 * mlong32: MIPS Options. (line 178)
27924 * mlong64: MIPS Options. (line 173)
27925 * mlongcall: RS/6000 and PowerPC Options.
27927 * mlongcalls: Xtensa Options. (line 60)
27928 * mloop-unsigned: TMS320C3x/C4x Options.
27930 * mlow-64k: Blackfin Options. (line 23)
27931 * mlp64: IA-64 Options. (line 114)
27932 * MM: Preprocessor Options.
27934 * mmad: MIPS Options. (line 268)
27935 * mmangle-cpu: ARC Options. (line 15)
27936 * mmax: DEC Alpha Options. (line 171)
27937 * mmax-stack-frame: CRIS Options. (line 22)
27938 * mmcu: AVR Options. (line 9)
27939 * MMD: Preprocessor Options.
27941 * mmedia: FRV Options. (line 44)
27942 * mmemcpy: MIPS Options. (line 253)
27943 * mmemory-latency: DEC Alpha Options. (line 266)
27944 * mmemparm: TMS320C3x/C4x Options.
27946 * mminimal-toc: RS/6000 and PowerPC Options.
27948 * mmmx: i386 and x86-64 Options.
27950 * mmodel=large: M32R/D Options. (line 33)
27951 * mmodel=medium: M32R/D Options. (line 27)
27952 * mmodel=small: M32R/D Options. (line 18)
27953 * mmpyi: TMS320C3x/C4x Options.
27955 * mmul-bug-workaround: CRIS Options. (line 36)
27956 * mmuladd: FRV Options. (line 50)
27957 * mmult-bug: MN10300 Options. (line 9)
27958 * mmulti-add: NS32K Options. (line 37)
27959 * mmulti-cond-exec: FRV Options. (line 176)
27960 * mmultiple: RS/6000 and PowerPC Options.
27962 * mmvcle: S/390 and zSeries Options.
27964 * mmvme: RS/6000 and PowerPC Options.
27966 * mn: H8/300 Options. (line 20)
27967 * mnested-cond-exec: FRV Options. (line 189)
27968 * mnew-mnemonics: RS/6000 and PowerPC Options.
27970 * mno-3dnow: i386 and x86-64 Options.
27972 * mno-4byte-functions: MCore Options. (line 27)
27973 * mno-abicalls: MIPS Options. (line 98)
27974 * mno-abshi: PDP-11 Options. (line 58)
27975 * mno-ac0: PDP-11 Options. (line 20)
27976 * mno-align-double: i386 and x86-64 Options.
27978 * mno-align-int: M680x0 Options. (line 128)
27979 * mno-align-loops: M32R/D Options. (line 76)
27980 * mno-align-stringops: i386 and x86-64 Options.
27982 * mno-altivec: RS/6000 and PowerPC Options.
27984 * mno-am33: MN10300 Options. (line 20)
27985 * mno-app-regs <1>: V850 Options. (line 61)
27986 * mno-app-regs: SPARC Options. (line 10)
27987 * mno-backchain: S/390 and zSeries Options.
27989 * mno-base-addresses: MMIX Options. (line 54)
27990 * mno-bit-align: RS/6000 and PowerPC Options.
27992 * mno-bk: TMS320C3x/C4x Options.
27994 * mno-branch-likely: MIPS Options. (line 346)
27995 * mno-branch-predict: MMIX Options. (line 49)
27996 * mno-bwx: DEC Alpha Options. (line 171)
27997 * mno-callgraph-data: MCore Options. (line 31)
27998 * mno-check-zero-division: MIPS Options. (line 233)
27999 * mno-cirrus-fix-invalid-insns: ARM Options. (line 187)
28000 * mno-cix: DEC Alpha Options. (line 171)
28001 * mno-cond-exec: FRV Options. (line 158)
28002 * mno-cond-move: FRV Options. (line 134)
28003 * mno-const-align: CRIS Options. (line 60)
28004 * mno-const16: Xtensa Options. (line 10)
28005 * mno-crt0: MN10300 Options. (line 24)
28006 * mno-csync: Blackfin Options. (line 19)
28007 * mno-data-align: CRIS Options. (line 60)
28008 * mno-db: TMS320C3x/C4x Options.
28010 * mno-debug: S/390 and zSeries Options.
28012 * mno-div: MCore Options. (line 15)
28013 * mno-double: FRV Options. (line 41)
28014 * mno-dwarf2-asm: IA-64 Options. (line 79)
28015 * mno-dword: FRV Options. (line 35)
28016 * mno-eabi: RS/6000 and PowerPC Options.
28018 * mno-early-stop-bits: IA-64 Options. (line 85)
28019 * mno-eflags: FRV Options. (line 125)
28020 * mno-embedded-data: MIPS Options. (line 204)
28021 * mno-ep: V850 Options. (line 16)
28022 * mno-epsilon: MMIX Options. (line 15)
28023 * mno-explicit-relocs <1>: MIPS Options. (line 224)
28024 * mno-explicit-relocs: DEC Alpha Options. (line 184)
28025 * mno-fancy-math-387: i386 and x86-64 Options.
28027 * mno-fast-fix: TMS320C3x/C4x Options.
28029 * mno-faster-structs: SPARC Options. (line 71)
28030 * mno-fix: DEC Alpha Options. (line 171)
28031 * mno-fix-r4000: MIPS Options. (line 288)
28032 * mno-fix-r4400: MIPS Options. (line 302)
28033 * mno-float32: PDP-11 Options. (line 48)
28034 * mno-float64: PDP-11 Options. (line 52)
28035 * mno-flush-func: M32R/D Options. (line 99)
28036 * mno-flush-trap: M32R/D Options. (line 91)
28037 * mno-fp-in-toc: RS/6000 and PowerPC Options.
28039 * mno-fp-regs: DEC Alpha Options. (line 25)
28040 * mno-fp-ret-in-387: i386 and x86-64 Options.
28042 * mno-fpu: SPARC Options. (line 25)
28043 * mno-fused-madd <1>: Xtensa Options. (line 19)
28044 * mno-fused-madd <2>: S/390 and zSeries Options.
28046 * mno-fused-madd <3>: RS/6000 and PowerPC Options.
28048 * mno-fused-madd: MIPS Options. (line 273)
28049 * mno-gnu-as: IA-64 Options. (line 18)
28050 * mno-gnu-ld: IA-64 Options. (line 23)
28051 * mno-gotplt: CRIS Options. (line 86)
28052 * mno-hardlit: MCore Options. (line 10)
28053 * mno-id-shared-library: Blackfin Options. (line 36)
28054 * mno-ieee-compare: NS32K Options. (line 55)
28055 * mno-ieee-fp: i386 and x86-64 Options.
28057 * mno-int16: PDP-11 Options. (line 44)
28058 * mno-int32: PDP-11 Options. (line 40)
28059 * mno-interrupts: AVR Options. (line 39)
28060 * mno-knuthdiv: MMIX Options. (line 33)
28061 * mno-libfuncs: MMIX Options. (line 10)
28062 * mno-long-calls <1>: V850 Options. (line 10)
28063 * mno-long-calls <2>: MIPS Options. (line 259)
28064 * mno-long-calls <3>: M68hc1x Options. (line 35)
28065 * mno-long-calls <4>: HPPA Options. (line 134)
28066 * mno-long-calls: ARM Options. (line 149)
28067 * mno-longcall: RS/6000 and PowerPC Options.
28069 * mno-longcalls: Xtensa Options. (line 60)
28070 * mno-loop-unsigned: TMS320C3x/C4x Options.
28072 * mno-low-64k: Blackfin Options. (line 27)
28073 * mno-mad: MIPS Options. (line 268)
28074 * mno-max: DEC Alpha Options. (line 171)
28075 * mno-media: FRV Options. (line 47)
28076 * mno-memcpy: MIPS Options. (line 253)
28077 * mno-mips16: MIPS Options. (line 80)
28078 * mno-mips3d: MIPS Options. (line 161)
28079 * mno-mmx: i386 and x86-64 Options.
28081 * mno-mpyi: TMS320C3x/C4x Options.
28083 * mno-mul-bug-workaround: CRIS Options. (line 36)
28084 * mno-muladd: FRV Options. (line 53)
28085 * mno-mult-bug: MN10300 Options. (line 13)
28086 * mno-multi-cond-exec: FRV Options. (line 183)
28087 * mno-multiple: RS/6000 and PowerPC Options.
28089 * mno-mvcle: S/390 and zSeries Options.
28091 * mno-nested-cond-exec: FRV Options. (line 195)
28092 * mno-pack: FRV Options. (line 122)
28093 * mno-packed-stack: S/390 and zSeries Options.
28095 * mno-paired-single: MIPS Options. (line 154)
28096 * mno-parallel-insns: TMS320C3x/C4x Options.
28098 * mno-parallel-mpy: TMS320C3x/C4x Options.
28100 * mno-pic: IA-64 Options. (line 26)
28101 * mno-power: RS/6000 and PowerPC Options.
28103 * mno-power2: RS/6000 and PowerPC Options.
28105 * mno-powerpc: RS/6000 and PowerPC Options.
28107 * mno-powerpc-gfxopt: RS/6000 and PowerPC Options.
28109 * mno-powerpc-gpopt: RS/6000 and PowerPC Options.
28111 * mno-powerpc64: RS/6000 and PowerPC Options.
28113 * mno-prolog-function: V850 Options. (line 23)
28114 * mno-prologue-epilogue: CRIS Options. (line 76)
28115 * mno-prototype: RS/6000 and PowerPC Options.
28117 * mno-push-args: i386 and x86-64 Options.
28119 * mno-register-names: IA-64 Options. (line 37)
28120 * mno-regnames: RS/6000 and PowerPC Options.
28122 * mno-relax-immediate: MCore Options. (line 19)
28123 * mno-relocatable: RS/6000 and PowerPC Options.
28125 * mno-relocatable-lib: RS/6000 and PowerPC Options.
28127 * mno-rptb: TMS320C3x/C4x Options.
28129 * mno-rpts: TMS320C3x/C4x Options.
28131 * mno-scc: FRV Options. (line 146)
28132 * mno-sched-prolog: ARM Options. (line 32)
28133 * mno-sdata <1>: RS/6000 and PowerPC Options.
28135 * mno-sdata: IA-64 Options. (line 42)
28136 * mno-side-effects: CRIS Options. (line 51)
28137 * mno-single-exit: MMIX Options. (line 66)
28138 * mno-slow-bytes: MCore Options. (line 35)
28139 * mno-small-exec: S/390 and zSeries Options.
28141 * mno-soft-float: DEC Alpha Options. (line 10)
28142 * mno-space-regs: HPPA Options. (line 45)
28143 * mno-split: PDP-11 Options. (line 71)
28144 * mno-split-addresses: MIPS Options. (line 218)
28145 * mno-sse: i386 and x86-64 Options.
28147 * mno-stack-align: CRIS Options. (line 60)
28148 * mno-stack-bias: SPARC Options. (line 218)
28149 * mno-strict-align <1>: RS/6000 and PowerPC Options.
28151 * mno-strict-align: M680x0 Options. (line 148)
28152 * mno-string: RS/6000 and PowerPC Options.
28154 * mno-sum-in-toc: RS/6000 and PowerPC Options.
28156 * mno-svr3-shlib: i386 and x86-64 Options.
28158 * mno-sym32: MIPS Options. (line 188)
28159 * mno-tablejump: AVR Options. (line 47)
28160 * mno-target-align: Xtensa Options. (line 47)
28161 * mno-text-section-literals: Xtensa Options. (line 35)
28162 * mno-toc: RS/6000 and PowerPC Options.
28164 * mno-toplevel-symbols: MMIX Options. (line 40)
28165 * mno-tpf-trace: S/390 and zSeries Options.
28167 * mno-unaligned-doubles: SPARC Options. (line 59)
28168 * mno-uninit-const-in-rodata: MIPS Options. (line 212)
28169 * mno-update: RS/6000 and PowerPC Options.
28171 * mno-v8plus: SPARC Options. (line 166)
28172 * mno-vis: SPARC Options. (line 173)
28173 * mno-vliw-branch: FRV Options. (line 170)
28174 * mno-volatile-asm-stop: IA-64 Options. (line 32)
28175 * mno-wide-bitfields: MCore Options. (line 23)
28176 * mno-xgot: MIPS Options. (line 103)
28177 * mno-xl-compat: RS/6000 and PowerPC Options.
28179 * mno-zero-extend: MMIX Options. (line 27)
28180 * mnobitfield <1>: NS32K Options. (line 61)
28181 * mnobitfield: M680x0 Options. (line 96)
28182 * mnohimem: NS32K Options. (line 118)
28183 * mnomacsave: SH Options. (line 92)
28184 * mnominmax: M68hc1x Options. (line 31)
28185 * mnomulti-add: NS32K Options. (line 46)
28186 * mnop-fun-dllimport: ARM Options. (line 174)
28187 * mnoregparam: NS32K Options. (line 97)
28188 * mnosb: NS32K Options. (line 105)
28189 * mold-mnemonics: RS/6000 and PowerPC Options.
28191 * momit-leaf-frame-pointer <1>: i386 and x86-64 Options.
28193 * momit-leaf-frame-pointer: Blackfin Options. (line 7)
28194 * MP: Preprocessor Options.
28196 * mpa-risc-1-0: HPPA Options. (line 19)
28197 * mpa-risc-1-1: HPPA Options. (line 19)
28198 * mpa-risc-2-0: HPPA Options. (line 19)
28199 * mpack: FRV Options. (line 119)
28200 * mpacked-stack: S/390 and zSeries Options.
28202 * mpadstruct: SH Options. (line 102)
28203 * mpaired-single: MIPS Options. (line 154)
28204 * mparallel-insns: TMS320C3x/C4x Options.
28206 * mparallel-mpy: TMS320C3x/C4x Options.
28208 * mparanoid: TMS320C3x/C4x Options.
28210 * mpcrel: M680x0 Options. (line 140)
28211 * mpdebug: CRIS Options. (line 40)
28212 * mpe: RS/6000 and PowerPC Options.
28214 * mpentium: i386 and x86-64 Options.
28216 * mpentiumpro: i386 and x86-64 Options.
28218 * mpic-register: ARM Options. (line 183)
28219 * mpoke-function-name: ARM Options. (line 197)
28220 * mportable-runtime: HPPA Options. (line 71)
28221 * mpower: RS/6000 and PowerPC Options.
28223 * mpower2: RS/6000 and PowerPC Options.
28225 * mpowerpc: RS/6000 and PowerPC Options.
28227 * mpowerpc-gfxopt: RS/6000 and PowerPC Options.
28229 * mpowerpc-gpopt: RS/6000 and PowerPC Options.
28231 * mpowerpc64: RS/6000 and PowerPC Options.
28233 * mprefergot: SH Options. (line 109)
28234 * mpreferred-stack-boundary: i386 and x86-64 Options.
28236 * mprioritize-restricted-insns: RS/6000 and PowerPC Options.
28238 * mprolog-function: V850 Options. (line 23)
28239 * mprologue-epilogue: CRIS Options. (line 76)
28240 * mprototype: RS/6000 and PowerPC Options.
28242 * mpush-args: i386 and x86-64 Options.
28244 * MQ: Preprocessor Options.
28246 * mregister-names: IA-64 Options. (line 37)
28247 * mregnames: RS/6000 and PowerPC Options.
28249 * mregparam: NS32K Options. (line 89)
28250 * mregparm <1>: TMS320C3x/C4x Options.
28252 * mregparm: i386 and x86-64 Options.
28254 * mrelax <1>: SH Options. (line 70)
28255 * mrelax <2>: MN10300 Options. (line 27)
28256 * mrelax: H8/300 Options. (line 9)
28257 * mrelax-immediate: MCore Options. (line 19)
28258 * mrelocatable: RS/6000 and PowerPC Options.
28260 * mrelocatable-lib: RS/6000 and PowerPC Options.
28262 * mrodata: ARC Options. (line 30)
28263 * mrptb: TMS320C3x/C4x Options.
28265 * mrpts: TMS320C3x/C4x Options.
28267 * mrtd <1>: Function Attributes.
28269 * mrtd <2>: NS32K Options. (line 70)
28270 * mrtd <3>: M680x0 Options. (line 105)
28271 * mrtd: i386 and x86-64 Options.
28273 * ms: H8/300 Options. (line 17)
28274 * ms2600: H8/300 Options. (line 24)
28275 * msb: NS32K Options. (line 101)
28276 * mscc: FRV Options. (line 140)
28277 * msched-costly-dep: RS/6000 and PowerPC Options.
28279 * mschedule: HPPA Options. (line 78)
28280 * msda: V850 Options. (line 40)
28281 * msdata <1>: RS/6000 and PowerPC Options.
28283 * msdata: IA-64 Options. (line 42)
28284 * msdata-data: RS/6000 and PowerPC Options.
28286 * msdata=default: RS/6000 and PowerPC Options.
28288 * msdata=eabi: RS/6000 and PowerPC Options.
28290 * msdata=none <1>: RS/6000 and PowerPC Options.
28292 * msdata=none: M32R/D Options. (line 40)
28293 * msdata=sdata: M32R/D Options. (line 49)
28294 * msdata=sysv: RS/6000 and PowerPC Options.
28296 * msdata=use: M32R/D Options. (line 53)
28297 * mshared-library-id: Blackfin Options. (line 40)
28298 * mshort <1>: M68hc1x Options. (line 40)
28299 * mshort: M680x0 Options. (line 90)
28300 * msim <1>: Xstormy16 Options. (line 9)
28301 * msim: RS/6000 and PowerPC Options.
28303 * msingle-exit: MMIX Options. (line 66)
28304 * msingle-float: MIPS Options. (line 145)
28305 * msingle-pic-base: ARM Options. (line 177)
28306 * msio: HPPA Options. (line 107)
28307 * msize: AVR Options. (line 32)
28308 * mslow-bytes: MCore Options. (line 35)
28309 * msmall: TMS320C3x/C4x Options.
28311 * msmall-data: DEC Alpha Options. (line 195)
28312 * msmall-exec: S/390 and zSeries Options.
28314 * msmall-memory: TMS320C3x/C4x Options.
28316 * msmall-text: DEC Alpha Options. (line 213)
28317 * msoft-float <1>: SPARC Options. (line 25)
28318 * msoft-float <2>: S/390 and zSeries Options.
28320 * msoft-float <3>: RS/6000 and PowerPC Options.
28322 * msoft-float <4>: PDP-11 Options. (line 13)
28323 * msoft-float <5>: NS32K Options. (line 50)
28324 * msoft-float <6>: MIPS Options. (line 141)
28325 * msoft-float <7>: M680x0 Options. (line 80)
28326 * msoft-float <8>: i386 and x86-64 Options.
28328 * msoft-float <9>: HPPA Options. (line 91)
28329 * msoft-float <10>: FRV Options. (line 22)
28330 * msoft-float <11>: DEC Alpha Options. (line 10)
28331 * msoft-float: ARM Options. (line 45)
28332 * msoft-quad-float: SPARC Options. (line 45)
28333 * msoft-reg-count: M68hc1x Options. (line 43)
28334 * mspace <1>: V850 Options. (line 30)
28335 * mspace: SH Options. (line 106)
28336 * mspe: RS/6000 and PowerPC Options.
28338 * msplit: PDP-11 Options. (line 68)
28339 * msplit-addresses: MIPS Options. (line 218)
28340 * msse: i386 and x86-64 Options.
28342 * mstack-align: CRIS Options. (line 60)
28343 * mstack-bias: SPARC Options. (line 218)
28344 * mstack-guard: S/390 and zSeries Options.
28346 * mstack-size: S/390 and zSeries Options.
28348 * mstrict-align <1>: RS/6000 and PowerPC Options.
28350 * mstrict-align: M680x0 Options. (line 148)
28351 * mstring: RS/6000 and PowerPC Options.
28353 * mstructure-size-boundary: ARM Options. (line 129)
28354 * msvr3-shlib: i386 and x86-64 Options.
28356 * msvr4-struct-return: RS/6000 and PowerPC Options.
28358 * msym32: MIPS Options. (line 188)
28359 * mt: IA-64 Options. (line 106)
28360 * MT: Preprocessor Options.
28362 * mtarget-align: Xtensa Options. (line 47)
28363 * mtda: V850 Options. (line 34)
28364 * mtext: ARC Options. (line 30)
28365 * mtext-section-literals: Xtensa Options. (line 35)
28366 * mthreads: i386 and x86-64 Options.
28368 * mthumb: ARM Options. (line 218)
28369 * mthumb-interwork: ARM Options. (line 25)
28370 * mti: TMS320C3x/C4x Options.
28372 * mtiny-stack: AVR Options. (line 50)
28373 * mtls-direct-seg-refs: i386 and x86-64 Options.
28375 * mtls-size: IA-64 Options. (line 97)
28376 * mtoc: RS/6000 and PowerPC Options.
28378 * mtomcat-stats: FRV Options. (line 201)
28379 * mtoplevel-symbols: MMIX Options. (line 40)
28380 * mtpcs-frame: ARM Options. (line 222)
28381 * mtpcs-leaf-frame: ARM Options. (line 228)
28382 * mtpf-trace: S/390 and zSeries Options.
28384 * mtrap-precision: DEC Alpha Options. (line 109)
28385 * mtune <1>: SPARC Options. (line 154)
28386 * mtune <2>: S/390 and zSeries Options.
28388 * mtune <3>: RS/6000 and PowerPC Options.
28390 * mtune <4>: MIPS Options. (line 43)
28391 * mtune <5>: i386 and x86-64 Options.
28393 * mtune <6>: DEC Alpha Options. (line 262)
28394 * mtune <7>: CRIS Options. (line 16)
28395 * mtune: ARM Options. (line 99)
28396 * mtune-arch: IA-64 Options. (line 101)
28397 * multi_module: Darwin Options. (line 174)
28398 * multilib-library-pic: FRV Options. (line 89)
28399 * multiply_defined: Darwin Options. (line 174)
28400 * multiply_defined_unused: Darwin Options. (line 174)
28401 * munaligned-doubles: SPARC Options. (line 59)
28402 * muninit-const-in-rodata: MIPS Options. (line 212)
28403 * munix: VAX Options. (line 9)
28404 * munix-asm: PDP-11 Options. (line 74)
28405 * mupdate: RS/6000 and PowerPC Options.
28407 * musermode: SH Options. (line 114)
28408 * mv850: V850 Options. (line 49)
28409 * mv850e: V850 Options. (line 69)
28410 * mv850e1: V850 Options. (line 64)
28411 * mv8plus: SPARC Options. (line 166)
28412 * mvis: SPARC Options. (line 173)
28413 * mvliw-branch: FRV Options. (line 164)
28414 * mvms-return-codes: DEC Alpha/VMS Options.
28416 * mvolatile-asm-stop: IA-64 Options. (line 32)
28417 * mvr4130-align: MIPS Options. (line 367)
28418 * mvxworks: RS/6000 and PowerPC Options.
28420 * mwarn-dynamicstack: S/390 and zSeries Options.
28422 * mwarn-framesize: S/390 and zSeries Options.
28424 * mwide-bitfields: MCore Options. (line 23)
28425 * mwindiss: RS/6000 and PowerPC Options.
28427 * mwords-little-endian: ARM Options. (line 76)
28428 * mxgot: MIPS Options. (line 103)
28429 * mxl-compat: RS/6000 and PowerPC Options.
28431 * myellowknife: RS/6000 and PowerPC Options.
28433 * mzarch: S/390 and zSeries Options.
28435 * mzda: V850 Options. (line 45)
28436 * mzero-extend: MMIX Options. (line 27)
28437 * no-integrated-cpp: C Dialect Options. (line 190)
28438 * no-red-zone: i386 and x86-64 Options.
28440 * no_dead_strip_inits_and_terms: Darwin Options. (line 174)
28441 * noall_load: Darwin Options. (line 174)
28442 * nocpp: MIPS Options. (line 283)
28443 * nodefaultlibs: Link Options. (line 62)
28444 * nofixprebinding: Darwin Options. (line 174)
28445 * nolibdld: HPPA Options. (line 186)
28446 * nomultidefs: Darwin Options. (line 174)
28447 * noprebind: Darwin Options. (line 174)
28448 * noseglinkedit: Darwin Options. (line 174)
28449 * nostartfiles: Link Options. (line 57)
28450 * nostdinc: Preprocessor Options.
28452 * nostdinc++ <1>: Preprocessor Options.
28454 * nostdinc++: C++ Dialect Options.
28456 * nostdlib: Link Options. (line 71)
28457 * o: Preprocessor Options.
28459 * O: Optimize Options. (line 32)
28460 * o: Overall Options. (line 167)
28461 * O0: Optimize Options. (line 106)
28462 * O1: Optimize Options. (line 32)
28463 * O2: Optimize Options. (line 67)
28464 * O3: Optimize Options. (line 101)
28465 * Os: Optimize Options. (line 109)
28466 * P: Preprocessor Options.
28468 * p: Debugging Options. (line 122)
28469 * pagezero_size: Darwin Options. (line 174)
28470 * param: Optimize Options. (line 1235)
28471 * pass-exit-codes: Overall Options. (line 126)
28472 * pedantic <1>: Warnings and Errors.
28474 * pedantic <2>: Alternate Keywords. (line 29)
28475 * pedantic <3>: C Extensions. (line 6)
28476 * pedantic <4>: Preprocessor Options.
28478 * pedantic <5>: Warning Options. (line 27)
28479 * pedantic: Standards. (line 13)
28480 * pedantic-errors <1>: Warnings and Errors.
28482 * pedantic-errors <2>: Non-bugs. (line 219)
28483 * pedantic-errors <3>: Preprocessor Options.
28485 * pedantic-errors <4>: Warning Options. (line 69)
28486 * pedantic-errors: Standards. (line 13)
28487 * pg: Debugging Options. (line 128)
28488 * pie: Link Options. (line 92)
28489 * pipe: Overall Options. (line 189)
28490 * prebind: Darwin Options. (line 174)
28491 * prebind_all_twolevel_modules: Darwin Options. (line 174)
28492 * preprocessor: Preprocessor Options.
28494 * print-file-name: Debugging Options. (line 677)
28495 * print-libgcc-file-name: Debugging Options. (line 698)
28496 * print-multi-directory: Debugging Options. (line 683)
28497 * print-multi-lib: Debugging Options. (line 688)
28498 * print-objc-runtime-info: Objective-C and Objective-C++ Dialect Options.
28500 * print-prog-name: Debugging Options. (line 695)
28501 * print-search-dirs: Debugging Options. (line 706)
28502 * private_bundle: Darwin Options. (line 174)
28503 * pthread <1>: RS/6000 and PowerPC Options.
28505 * pthread: IA-64 Options. (line 106)
28506 * pthreads: SPARC Options. (line 232)
28507 * Q: Debugging Options. (line 134)
28508 * Qn: System V Options. (line 18)
28509 * Qy: System V Options. (line 14)
28510 * read_only_relocs: Darwin Options. (line 174)
28511 * remap: Preprocessor Options.
28513 * s: Link Options. (line 98)
28514 * S <1>: Link Options. (line 20)
28515 * S: Overall Options. (line 150)
28516 * save-temps: Debugging Options. (line 639)
28517 * sectalign: Darwin Options. (line 174)
28518 * sectcreate: Darwin Options. (line 174)
28519 * sectobjectsymbols: Darwin Options. (line 174)
28520 * sectorder: Darwin Options. (line 174)
28521 * seg1addr: Darwin Options. (line 174)
28522 * seg_addr_table: Darwin Options. (line 174)
28523 * seg_addr_table_filename: Darwin Options. (line 174)
28524 * segaddr: Darwin Options. (line 174)
28525 * seglinkedit: Darwin Options. (line 174)
28526 * segprot: Darwin Options. (line 174)
28527 * segs_read_only_addr: Darwin Options. (line 174)
28528 * segs_read_write_addr: Darwin Options. (line 174)
28529 * shared: Link Options. (line 107)
28530 * shared-libgcc: Link Options. (line 115)
28531 * sim: CRIS Options. (line 108)
28532 * sim2: CRIS Options. (line 114)
28533 * single_module: Darwin Options. (line 174)
28534 * specs: Directory Options. (line 84)
28535 * static <1>: HPPA Options. (line 190)
28536 * static <2>: Darwin Options. (line 174)
28537 * static: Link Options. (line 102)
28538 * static-libgcc: Link Options. (line 115)
28539 * std <1>: Non-bugs. (line 107)
28540 * std <2>: Other Builtins. (line 22)
28541 * std <3>: C Dialect Options. (line 47)
28542 * std: Standards. (line 13)
28543 * std=: Preprocessor Options.
28545 * sub_library: Darwin Options. (line 174)
28546 * sub_umbrella: Darwin Options. (line 174)
28547 * symbolic: Link Options. (line 150)
28548 * target-help <1>: Preprocessor Options.
28550 * target-help: Overall Options. (line 220)
28551 * threads <1>: SPARC Options. (line 226)
28552 * threads: HPPA Options. (line 203)
28553 * time: Debugging Options. (line 653)
28554 * tls: FRV Options. (line 75)
28555 * TLS: FRV Options. (line 72)
28556 * traditional <1>: Incompatibilities. (line 6)
28557 * traditional: C Dialect Options. (line 202)
28558 * traditional-cpp <1>: Preprocessor Options.
28560 * traditional-cpp: C Dialect Options. (line 202)
28561 * trigraphs <1>: Preprocessor Options.
28563 * trigraphs: C Dialect Options. (line 186)
28564 * twolevel_namespace: Darwin Options. (line 174)
28565 * u: Link Options. (line 172)
28566 * U: Preprocessor Options.
28568 * umbrella: Darwin Options. (line 174)
28569 * undef: Preprocessor Options.
28571 * undefined: Darwin Options. (line 174)
28572 * unexported_symbols_list: Darwin Options. (line 174)
28573 * V: Target Options. (line 22)
28574 * v <1>: Preprocessor Options.
28576 * v: Overall Options. (line 178)
28577 * version <1>: Preprocessor Options.
28579 * version: Overall Options. (line 224)
28580 * W: Incompatibilities. (line 64)
28581 * w: Preprocessor Options.
28583 * W: Warning Options. (line 513)
28584 * w: Warning Options. (line 73)
28585 * Wa: Assembler Options. (line 9)
28586 * Wabi: C++ Dialect Options.
28588 * Waggregate-return: Warning Options. (line 767)
28589 * Wall <1>: Standard Libraries. (line 6)
28590 * Wall <2>: Preprocessor Options.
28592 * Wall: Warning Options. (line 497)
28593 * Wbad-function-cast: Warning Options. (line 721)
28594 * Wcast-align: Warning Options. (line 730)
28595 * Wcast-qual: Warning Options. (line 725)
28596 * Wchar-subscripts: Warning Options. (line 79)
28597 * Wcomment <1>: Preprocessor Options.
28599 * Wcomment: Warning Options. (line 84)
28600 * Wcomments: Preprocessor Options.
28602 * Wconversion <1>: Protoize Caveats. (line 31)
28603 * Wconversion: Warning Options. (line 747)
28604 * Wctor-dtor-privacy: C++ Dialect Options.
28606 * Wdeclaration-after-statement: Warning Options. (line 693)
28607 * Wdisabled-optimization: Warning Options. (line 939)
28608 * Wdiv-by-zero: Warning Options. (line 585)
28609 * weak_reference_mismatches: Darwin Options. (line 174)
28610 * Weffc++: C++ Dialect Options.
28612 * Wendif-labels <1>: Preprocessor Options.
28614 * Wendif-labels: Warning Options. (line 703)
28615 * Werror <1>: Preprocessor Options.
28617 * Werror: Warning Options. (line 954)
28618 * Werror-implicit-function-declaration: Warning Options. (line 198)
28619 * Wextra: Warning Options. (line 513)
28620 * Wfatal-errors: Warning Options. (line 89)
28621 * Wfloat-equal: Warning Options. (line 601)
28622 * Wformat <1>: Function Attributes.
28624 * Wformat: Warning Options. (line 94)
28625 * Wformat-nonliteral <1>: Function Attributes.
28627 * Wformat-nonliteral: Warning Options. (line 151)
28628 * Wformat-security: Warning Options. (line 156)
28629 * Wformat-y2k: Warning Options. (line 129)
28630 * Wformat=2: Warning Options. (line 167)
28631 * whatsloaded: Darwin Options. (line 174)
28632 * whyload: Darwin Options. (line 174)
28633 * Wimplicit: Warning Options. (line 204)
28634 * Wimplicit-function-declaration: Warning Options. (line 198)
28635 * Wimplicit-int: Warning Options. (line 193)
28636 * Wimport: Preprocessor Options.
28638 * Winit-self: Warning Options. (line 179)
28639 * Winline <1>: Inline. (line 35)
28640 * Winline: Warning Options. (line 896)
28641 * Winvalid-pch: Warning Options. (line 923)
28642 * Wl: Link Options. (line 168)
28643 * Wlarger-than: Warning Options. (line 712)
28644 * Wlong-long: Warning Options. (line 927)
28645 * Wmain: Warning Options. (line 208)
28646 * Wmissing-braces: Warning Options. (line 214)
28647 * Wmissing-declarations: Warning Options. (line 788)
28648 * Wmissing-field-initializers: Warning Options. (line 794)
28649 * Wmissing-format-attribute: Warning Options. (line 820)
28650 * Wmissing-include-dirs: Warning Options. (line 224)
28651 * Wmissing-noreturn: Warning Options. (line 812)
28652 * Wmissing-prototypes: Warning Options. (line 782)
28653 * Wmultichar: Warning Options. (line 830)
28654 * Wnested-externs: Warning Options. (line 871)
28655 * Wno-deprecated: C++ Dialect Options.
28657 * Wno-deprecated-declarations: Warning Options. (line 836)
28658 * Wno-div-by-zero: Warning Options. (line 585)
28659 * Wno-endif-labels: Warning Options. (line 703)
28660 * Wno-format-extra-args: Warning Options. (line 133)
28661 * Wno-format-zero-length: Warning Options. (line 147)
28662 * Wno-import: Warning Options. (line 76)
28663 * Wno-invalid-offsetof: Warning Options. (line 909)
28664 * Wno-long-long: Warning Options. (line 927)
28665 * Wno-multichar: Warning Options. (line 830)
28666 * Wno-non-template-friend: C++ Dialect Options.
28668 * Wno-pmf-conversions <1>: Bound member functions.
28670 * Wno-pmf-conversions: C++ Dialect Options.
28672 * Wno-pointer-sign: Warning Options. (line 948)
28673 * Wno-protocol: Objective-C and Objective-C++ Dialect Options.
28675 * Wno-variadic-macros: Warning Options. (line 933)
28676 * Wnon-virtual-dtor: C++ Dialect Options.
28678 * Wnonnull: Warning Options. (line 172)
28679 * Wold-style-cast: C++ Dialect Options.
28681 * Wold-style-definition: Warning Options. (line 778)
28682 * Woverloaded-virtual: C++ Dialect Options.
28684 * Wp: Preprocessor Options.
28686 * Wpacked: Warning Options. (line 842)
28687 * Wpadded: Warning Options. (line 859)
28688 * Wparentheses: Warning Options. (line 227)
28689 * Wpointer-arith <1>: Pointer Arith. (line 13)
28690 * Wpointer-arith: Warning Options. (line 715)
28691 * Wredundant-decls: Warning Options. (line 866)
28692 * Wreorder: C++ Dialect Options.
28694 * Wreturn-type: Warning Options. (line 322)
28695 * Wselector: Objective-C and Objective-C++ Dialect Options.
28697 * Wsequence-point: Warning Options. (line 273)
28698 * Wshadow: Warning Options. (line 707)
28699 * Wsign-compare: Warning Options. (line 760)
28700 * Wsign-promo: C++ Dialect Options.
28702 * Wstrict-aliasing: Warning Options. (line 483)
28703 * Wstrict-aliasing=2: Warning Options. (line 490)
28704 * Wstrict-null-sentinel: C++ Dialect Options.
28706 * Wstrict-prototypes: Warning Options. (line 772)
28707 * Wswitch: Warning Options. (line 341)
28708 * Wswitch-enum: Warning Options. (line 352)
28709 * Wswitch-switch: Warning Options. (line 349)
28710 * Wsystem-headers <1>: Preprocessor Options.
28712 * Wsystem-headers: Warning Options. (line 590)
28713 * Wtraditional <1>: Preprocessor Options.
28715 * Wtraditional: Warning Options. (line 616)
28716 * Wtrigraphs <1>: Preprocessor Options.
28718 * Wtrigraphs: Warning Options. (line 358)
28719 * Wundeclared-selector: Objective-C and Objective-C++ Dialect Options.
28721 * Wundef <1>: Preprocessor Options.
28723 * Wundef: Warning Options. (line 700)
28724 * Wuninitialized: Warning Options. (line 403)
28725 * Wunknown-pragmas: Warning Options. (line 476)
28726 * Wunreachable-code: Warning Options. (line 874)
28727 * Wunused: Warning Options. (line 396)
28728 * Wunused-function: Warning Options. (line 363)
28729 * Wunused-label: Warning Options. (line 368)
28730 * Wunused-macros: Preprocessor Options.
28732 * Wunused-parameter: Warning Options. (line 375)
28733 * Wunused-value: Warning Options. (line 390)
28734 * Wunused-variable: Warning Options. (line 382)
28735 * Wvariadic-macros: Warning Options. (line 933)
28736 * Wwrite-strings: Warning Options. (line 736)
28737 * x <1>: Preprocessor Options.
28739 * x: Overall Options. (line 101)
28740 * Xassembler: Assembler Options. (line 13)
28741 * Xlinker: Link Options. (line 156)
28742 * Ym: System V Options. (line 26)
28743 * YP: System V Options. (line 22)
28746 File: gcc.info, Node: Keyword Index, Prev: Option Index, Up: Top
28754 * ! in constraint: Multi-Alternative. (line 33)
28755 * # in constraint: Modifiers. (line 54)
28756 * #pragma: Pragmas. (line 6)
28757 * #pragma implementation: C++ Interface. (line 39)
28758 * #pragma implementation, implied: C++ Interface. (line 46)
28759 * #pragma interface: C++ Interface. (line 20)
28760 * #pragma, reason for not using: Function Attributes. (line 736)
28761 * $: Dollar Signs. (line 6)
28762 * % in constraint: Modifiers. (line 45)
28763 * %include: Spec Files. (line 27)
28764 * %include_noerr: Spec Files. (line 31)
28765 * %rename: Spec Files. (line 35)
28766 * & in constraint: Modifiers. (line 25)
28767 * ': Incompatibilities. (line 116)
28768 * * in constraint: Modifiers. (line 59)
28769 * + in constraint: Modifiers. (line 12)
28770 * -lgcc, use with -nodefaultlibs: Link Options. (line 79)
28771 * -lgcc, use with -nostdlib: Link Options. (line 79)
28772 * -nodefaultlibs and unresolved references: Link Options. (line 79)
28773 * -nostdlib and unresolved references: Link Options. (line 79)
28774 * .sdata/.sdata2 references (PowerPC): RS/6000 and PowerPC Options.
28776 * //: C++ Comments. (line 6)
28777 * 0 in constraint: Simple Constraints. (line 115)
28778 * < in constraint: Simple Constraints. (line 46)
28779 * = in constraint: Modifiers. (line 8)
28780 * > in constraint: Simple Constraints. (line 50)
28781 * ? in constraint: Multi-Alternative. (line 27)
28782 * ?: extensions: Conditionals. (line 6)
28783 * ?: side effect: Conditionals. (line 20)
28784 * _ in variables in macros: Typeof. (line 42)
28785 * __builtin_apply: Constructing Calls. (line 31)
28786 * __builtin_apply_args: Constructing Calls. (line 20)
28787 * __builtin_choose_expr: Other Builtins. (line 150)
28788 * __builtin_clz: Other Builtins. (line 362)
28789 * __builtin_clzl: Other Builtins. (line 380)
28790 * __builtin_clzll: Other Builtins. (line 400)
28791 * __builtin_constant_p: Other Builtins. (line 190)
28792 * __builtin_ctz: Other Builtins. (line 366)
28793 * __builtin_ctzl: Other Builtins. (line 384)
28794 * __builtin_ctzll: Other Builtins. (line 404)
28795 * __builtin_expect: Other Builtins. (line 236)
28796 * __builtin_ffs: Other Builtins. (line 358)
28797 * __builtin_ffsl: Other Builtins. (line 376)
28798 * __builtin_ffsll: Other Builtins. (line 396)
28799 * __builtin_frame_address: Return Address. (line 34)
28800 * __builtin_huge_val: Other Builtins. (line 300)
28801 * __builtin_huge_valf: Other Builtins. (line 305)
28802 * __builtin_huge_vall: Other Builtins. (line 308)
28803 * __builtin_inf: Other Builtins. (line 312)
28804 * __builtin_inff: Other Builtins. (line 316)
28805 * __builtin_infl: Other Builtins. (line 321)
28806 * __builtin_isgreater: Other Builtins. (line 6)
28807 * __builtin_isgreaterequal: Other Builtins. (line 6)
28808 * __builtin_isless: Other Builtins. (line 6)
28809 * __builtin_islessequal: Other Builtins. (line 6)
28810 * __builtin_islessgreater: Other Builtins. (line 6)
28811 * __builtin_isunordered: Other Builtins. (line 6)
28812 * __builtin_nan: Other Builtins. (line 325)
28813 * __builtin_nanf: Other Builtins. (line 340)
28814 * __builtin_nanl: Other Builtins. (line 343)
28815 * __builtin_nans: Other Builtins. (line 347)
28816 * __builtin_nansf: Other Builtins. (line 351)
28817 * __builtin_nansl: Other Builtins. (line 354)
28818 * __builtin_offsetof: Offsetof. (line 6)
28819 * __builtin_parity: Other Builtins. (line 373)
28820 * __builtin_parityl: Other Builtins. (line 392)
28821 * __builtin_parityll: Other Builtins. (line 412)
28822 * __builtin_popcount: Other Builtins. (line 370)
28823 * __builtin_popcountl: Other Builtins. (line 388)
28824 * __builtin_popcountll: Other Builtins. (line 408)
28825 * __builtin_powi: Other Builtins. (line 6)
28826 * __builtin_powif: Other Builtins. (line 6)
28827 * __builtin_powil: Other Builtins. (line 6)
28828 * __builtin_prefetch: Other Builtins. (line 261)
28829 * __builtin_return: Constructing Calls. (line 48)
28830 * __builtin_return_address: Return Address. (line 11)
28831 * __builtin_types_compatible_p: Other Builtins. (line 104)
28832 * __complex__ keyword: Complex. (line 6)
28833 * __declspec(dllexport): Function Attributes. (line 110)
28834 * __declspec(dllimport): Function Attributes. (line 142)
28835 * __extension__: Alternate Keywords. (line 29)
28836 * __func__ identifier: Function Names. (line 6)
28837 * __FUNCTION__ identifier: Function Names. (line 6)
28838 * __imag__ keyword: Complex. (line 27)
28839 * __PRETTY_FUNCTION__ identifier: Function Names. (line 6)
28840 * __real__ keyword: Complex. (line 27)
28841 * __STDC_HOSTED__: Standards. (line 6)
28842 * __thread: Thread-Local. (line 6)
28843 * _Complex keyword: Complex. (line 6)
28844 * _exit: Other Builtins. (line 6)
28845 * _Exit: Other Builtins. (line 6)
28846 * ABI: Compatibility. (line 6)
28847 * abort: Other Builtins. (line 6)
28848 * abs: Other Builtins. (line 6)
28849 * accessing volatiles: Volatiles. (line 6)
28850 * acos: Other Builtins. (line 6)
28851 * acosf: Other Builtins. (line 6)
28852 * acosh: Other Builtins. (line 6)
28853 * acoshf: Other Builtins. (line 6)
28854 * acoshl: Other Builtins. (line 6)
28855 * acosl: Other Builtins. (line 6)
28856 * Ada: G++ and GCC. (line 6)
28857 * address constraints: Simple Constraints. (line 142)
28858 * address of a label: Labels as Values. (line 6)
28859 * address_operand: Simple Constraints. (line 146)
28860 * alias attribute: Function Attributes. (line 31)
28861 * aliasing of parameters: Code Gen Options. (line 325)
28862 * aligned attribute <1>: Type Attributes. (line 30)
28863 * aligned attribute: Variable Attributes. (line 23)
28864 * alignment: Alignment. (line 6)
28865 * alloca: Other Builtins. (line 6)
28866 * alloca vs variable-length arrays: Variable Length. (line 27)
28867 * Allow nesting in an interrupt handler on the Blackfin processor.: Function Attributes.
28869 * alternate keywords: Alternate Keywords. (line 6)
28870 * always_inline function attribute: Function Attributes. (line 44)
28871 * AMD x86-64 Options: i386 and x86-64 Options.
28873 * AMD1: Standards. (line 6)
28874 * ANSI C: Standards. (line 6)
28875 * ANSI C standard: Standards. (line 6)
28876 * ANSI C89: Standards. (line 6)
28877 * ANSI support: C Dialect Options. (line 10)
28878 * ANSI X3.159-1989: Standards. (line 6)
28879 * apostrophes: Incompatibilities. (line 116)
28880 * application binary interface: Compatibility. (line 6)
28881 * ARC Options: ARC Options. (line 6)
28882 * ARM [Annotated C++ Reference Manual]: Backwards Compatibility.
28884 * ARM options: ARM Options. (line 6)
28885 * arrays of length zero: Zero Length. (line 6)
28886 * arrays of variable length: Variable Length. (line 6)
28887 * arrays, non-lvalue: Subscripting. (line 6)
28888 * asin: Other Builtins. (line 6)
28889 * asinf: Other Builtins. (line 6)
28890 * asinh: Other Builtins. (line 6)
28891 * asinhf: Other Builtins. (line 6)
28892 * asinhl: Other Builtins. (line 6)
28893 * asinl: Other Builtins. (line 6)
28894 * asm constraints: Constraints. (line 6)
28895 * asm expressions: Extended Asm. (line 6)
28896 * assembler instructions: Extended Asm. (line 6)
28897 * assembler names for identifiers: Asm Labels. (line 6)
28898 * assembly code, invalid: Bug Criteria. (line 12)
28899 * atan: Other Builtins. (line 6)
28900 * atan2: Other Builtins. (line 6)
28901 * atan2f: Other Builtins. (line 6)
28902 * atan2l: Other Builtins. (line 6)
28903 * atanf: Other Builtins. (line 6)
28904 * atanh: Other Builtins. (line 6)
28905 * atanhf: Other Builtins. (line 6)
28906 * atanhl: Other Builtins. (line 6)
28907 * atanl: Other Builtins. (line 6)
28908 * attribute of types: Type Attributes. (line 6)
28909 * attribute of variables: Variable Attributes. (line 6)
28910 * attribute syntax: Attribute Syntax. (line 6)
28911 * autoincrement/decrement addressing: Simple Constraints. (line 28)
28912 * automatic inline for C++ member fns: Inline. (line 46)
28913 * AVR Options: AVR Options. (line 6)
28914 * Backwards Compatibility: Backwards Compatibility.
28916 * base class members: Name lookup. (line 6)
28917 * bcmp: Other Builtins. (line 6)
28918 * below100 attribute: Variable Attributes. (line 313)
28919 * binary compatibility: Compatibility. (line 6)
28920 * Blackfin Options: Blackfin Options. (line 6)
28921 * bound pointer to member function: Bound member functions.
28923 * bounds checking: Optimize Options. (line 319)
28924 * bug criteria: Bug Criteria. (line 6)
28925 * bugs: Bugs. (line 6)
28926 * bugs, known: Trouble. (line 6)
28927 * built-in functions <1>: Other Builtins. (line 6)
28928 * built-in functions: C Dialect Options. (line 128)
28929 * bzero: Other Builtins. (line 6)
28930 * C compilation options: Invoking GCC. (line 17)
28931 * C intermediate output, nonexistent: G++ and GCC. (line 35)
28932 * C language extensions: C Extensions. (line 6)
28933 * C language, traditional: C Dialect Options. (line 200)
28934 * C standard: Standards. (line 6)
28935 * C standards: Standards. (line 6)
28936 * c++: Invoking G++. (line 13)
28937 * C++: G++ and GCC. (line 30)
28938 * C++ comments: C++ Comments. (line 6)
28939 * C++ compilation options: Invoking GCC. (line 23)
28940 * C++ interface and implementation headers: C++ Interface. (line 6)
28941 * C++ language extensions: C++ Extensions. (line 6)
28942 * C++ member fns, automatically inline: Inline. (line 46)
28943 * C++ misunderstandings: C++ Misunderstandings.
28945 * C++ options, command line: C++ Dialect Options. (line 6)
28946 * C++ pragmas, effect on inlining: C++ Interface. (line 66)
28947 * C++ source file suffixes: Invoking G++. (line 6)
28948 * C++ static data, declaring and defining: Static Definitions.
28950 * C89: Standards. (line 6)
28951 * C90: Standards. (line 6)
28952 * C94: Standards. (line 6)
28953 * C95: Standards. (line 6)
28954 * C99: Standards. (line 6)
28955 * C9X: Standards. (line 6)
28956 * C_INCLUDE_PATH: Environment Variables.
28958 * cabs: Other Builtins. (line 6)
28959 * cabsf: Other Builtins. (line 6)
28960 * cabsl: Other Builtins. (line 6)
28961 * cacos: Other Builtins. (line 6)
28962 * cacosf: Other Builtins. (line 6)
28963 * cacosh: Other Builtins. (line 6)
28964 * cacoshf: Other Builtins. (line 6)
28965 * cacoshl: Other Builtins. (line 6)
28966 * cacosl: Other Builtins. (line 6)
28967 * calling functions through the function vector on the H8/300 processors: Function Attributes.
28969 * calloc: Other Builtins. (line 6)
28970 * carg: Other Builtins. (line 6)
28971 * cargf: Other Builtins. (line 6)
28972 * cargl: Other Builtins. (line 6)
28973 * case labels in initializers: Designated Inits. (line 6)
28974 * case ranges: Case Ranges. (line 6)
28975 * casin: Other Builtins. (line 6)
28976 * casinf: Other Builtins. (line 6)
28977 * casinh: Other Builtins. (line 6)
28978 * casinhf: Other Builtins. (line 6)
28979 * casinhl: Other Builtins. (line 6)
28980 * casinl: Other Builtins. (line 6)
28981 * cast to a union: Cast to Union. (line 6)
28982 * catan: Other Builtins. (line 6)
28983 * catanf: Other Builtins. (line 6)
28984 * catanh: Other Builtins. (line 6)
28985 * catanhf: Other Builtins. (line 6)
28986 * catanhl: Other Builtins. (line 6)
28987 * catanl: Other Builtins. (line 6)
28988 * cbrt: Other Builtins. (line 6)
28989 * cbrtf: Other Builtins. (line 6)
28990 * cbrtl: Other Builtins. (line 6)
28991 * ccos: Other Builtins. (line 6)
28992 * ccosf: Other Builtins. (line 6)
28993 * ccosh: Other Builtins. (line 6)
28994 * ccoshf: Other Builtins. (line 6)
28995 * ccoshl: Other Builtins. (line 6)
28996 * ccosl: Other Builtins. (line 6)
28997 * ceil: Other Builtins. (line 6)
28998 * ceilf: Other Builtins. (line 6)
28999 * ceill: Other Builtins. (line 6)
29000 * cexp: Other Builtins. (line 6)
29001 * cexpf: Other Builtins. (line 6)
29002 * cexpl: Other Builtins. (line 6)
29003 * character set, execution: Preprocessor Options.
29005 * character set, input: Preprocessor Options.
29007 * character set, wide execution: Preprocessor Options.
29009 * cimag: Other Builtins. (line 6)
29010 * cimagf: Other Builtins. (line 6)
29011 * cimagl: Other Builtins. (line 6)
29012 * cleanup attribute: Variable Attributes. (line 76)
29013 * COBOL: G++ and GCC. (line 23)
29014 * code generation conventions: Code Gen Options. (line 6)
29015 * code, mixed with declarations: Mixed Declarations. (line 6)
29016 * command options: Invoking GCC. (line 6)
29017 * comments, C++ style: C++ Comments. (line 6)
29018 * common attribute: Variable Attributes. (line 92)
29019 * comparison of signed and unsigned values, warning: Warning Options.
29021 * compiler bugs, reporting: Bug Reporting. (line 6)
29022 * compiler compared to C++ preprocessor: G++ and GCC. (line 35)
29023 * compiler options, C++: C++ Dialect Options. (line 6)
29024 * compiler options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
29026 * compiler version, specifying: Target Options. (line 6)
29027 * COMPILER_PATH: Environment Variables.
29029 * complex conjugation: Complex. (line 34)
29030 * complex numbers: Complex. (line 6)
29031 * compound literals: Compound Literals. (line 6)
29032 * computed gotos: Labels as Values. (line 6)
29033 * conditional expressions, extensions: Conditionals. (line 6)
29034 * conflicting types: Disappointments. (line 21)
29035 * conj: Other Builtins. (line 6)
29036 * conjf: Other Builtins. (line 6)
29037 * conjl: Other Builtins. (line 6)
29038 * const applied to function: Function Attributes. (line 6)
29039 * const function attribute: Function Attributes. (line 55)
29040 * constants in constraints: Simple Constraints. (line 58)
29041 * constraint modifier characters: Modifiers. (line 6)
29042 * constraint, matching: Simple Constraints. (line 127)
29043 * constraints, asm: Constraints. (line 6)
29044 * constraints, machine specific: Machine Constraints. (line 6)
29045 * constructing calls: Constructing Calls. (line 6)
29046 * constructor expressions: Compound Literals. (line 6)
29047 * constructor function attribute: Function Attributes. (line 81)
29048 * contributors: Contributors. (line 6)
29049 * copysign: Other Builtins. (line 6)
29050 * copysignf: Other Builtins. (line 6)
29051 * copysignl: Other Builtins. (line 6)
29052 * core dump: Bug Criteria. (line 9)
29053 * cos: Other Builtins. (line 6)
29054 * cosf: Other Builtins. (line 6)
29055 * cosh: Other Builtins. (line 6)
29056 * coshf: Other Builtins. (line 6)
29057 * coshl: Other Builtins. (line 6)
29058 * cosl: Other Builtins. (line 6)
29059 * CPATH: Environment Variables.
29061 * CPLUS_INCLUDE_PATH: Environment Variables.
29063 * cpow: Other Builtins. (line 6)
29064 * cpowf: Other Builtins. (line 6)
29065 * cpowl: Other Builtins. (line 6)
29066 * cproj: Other Builtins. (line 6)
29067 * cprojf: Other Builtins. (line 6)
29068 * cprojl: Other Builtins. (line 6)
29069 * creal: Other Builtins. (line 6)
29070 * crealf: Other Builtins. (line 6)
29071 * creall: Other Builtins. (line 6)
29072 * CRIS Options: CRIS Options. (line 6)
29073 * cross compiling: Target Options. (line 6)
29074 * csin: Other Builtins. (line 6)
29075 * csinf: Other Builtins. (line 6)
29076 * csinh: Other Builtins. (line 6)
29077 * csinhf: Other Builtins. (line 6)
29078 * csinhl: Other Builtins. (line 6)
29079 * csinl: Other Builtins. (line 6)
29080 * csqrt: Other Builtins. (line 6)
29081 * csqrtf: Other Builtins. (line 6)
29082 * csqrtl: Other Builtins. (line 6)
29083 * ctan: Other Builtins. (line 6)
29084 * ctanf: Other Builtins. (line 6)
29085 * ctanh: Other Builtins. (line 6)
29086 * ctanhf: Other Builtins. (line 6)
29087 * ctanhl: Other Builtins. (line 6)
29088 * ctanl: Other Builtins. (line 6)
29089 * Darwin options: Darwin Options. (line 6)
29090 * dcgettext: Other Builtins. (line 6)
29091 * deallocating variable length arrays: Variable Length. (line 23)
29092 * debugging information options: Debugging Options. (line 6)
29093 * declaration scope: Incompatibilities. (line 80)
29094 * declarations inside expressions: Statement Exprs. (line 6)
29095 * declarations, mixed with code: Mixed Declarations. (line 6)
29096 * declaring attributes of functions: Function Attributes. (line 6)
29097 * declaring static data in C++: Static Definitions. (line 6)
29098 * defining static data in C++: Static Definitions. (line 6)
29099 * dependencies for make as output: Environment Variables.
29101 * dependencies, make: Preprocessor Options.
29103 * DEPENDENCIES_OUTPUT: Environment Variables.
29105 * dependent name lookup: Name lookup. (line 6)
29106 * deprecated attribute: Variable Attributes. (line 100)
29107 * deprecated attribute.: Function Attributes. (line 92)
29108 * designated initializers: Designated Inits. (line 6)
29109 * designator lists: Designated Inits. (line 94)
29110 * designators: Designated Inits. (line 61)
29111 * destructor function attribute: Function Attributes. (line 81)
29112 * dgettext: Other Builtins. (line 6)
29113 * diagnostic messages: Language Independent Options.
29115 * dialect options: C Dialect Options. (line 6)
29116 * digits in constraint: Simple Constraints. (line 115)
29117 * directory options: Directory Options. (line 6)
29118 * dollar signs in identifier names: Dollar Signs. (line 6)
29119 * double-word arithmetic: Long Long. (line 6)
29120 * downward funargs: Nested Functions. (line 6)
29121 * drem: Other Builtins. (line 6)
29122 * dremf: Other Builtins. (line 6)
29123 * dreml: Other Builtins. (line 6)
29124 * E in constraint: Simple Constraints. (line 77)
29125 * earlyclobber operand: Modifiers. (line 25)
29126 * eight bit data on the H8/300, H8/300H, and H8S: Function Attributes.
29128 * empty structures: Empty Structures. (line 6)
29129 * environment variables: Environment Variables.
29131 * erf: Other Builtins. (line 6)
29132 * erfc: Other Builtins. (line 6)
29133 * erfcf: Other Builtins. (line 6)
29134 * erfcl: Other Builtins. (line 6)
29135 * erff: Other Builtins. (line 6)
29136 * erfl: Other Builtins. (line 6)
29137 * error messages: Warnings and Errors. (line 6)
29138 * escaped newlines: Escaped Newlines. (line 6)
29139 * exception handler functions on the Blackfin processor: Function Attributes.
29141 * exclamation point: Multi-Alternative. (line 33)
29142 * exit: Other Builtins. (line 6)
29143 * exp: Other Builtins. (line 6)
29144 * exp10: Other Builtins. (line 6)
29145 * exp10f: Other Builtins. (line 6)
29146 * exp10l: Other Builtins. (line 6)
29147 * exp2: Other Builtins. (line 6)
29148 * exp2f: Other Builtins. (line 6)
29149 * exp2l: Other Builtins. (line 6)
29150 * expf: Other Builtins. (line 6)
29151 * expl: Other Builtins. (line 6)
29152 * explicit register variables: Explicit Reg Vars. (line 6)
29153 * expm1: Other Builtins. (line 6)
29154 * expm1f: Other Builtins. (line 6)
29155 * expm1l: Other Builtins. (line 6)
29156 * expressions containing statements: Statement Exprs. (line 6)
29157 * expressions, constructor: Compound Literals. (line 6)
29158 * extended asm: Extended Asm. (line 6)
29159 * extensible constraints: Simple Constraints. (line 151)
29160 * extensions, ?:: Conditionals. (line 6)
29161 * extensions, C language: C Extensions. (line 6)
29162 * extensions, C++ language: C++ Extensions. (line 6)
29163 * external declaration scope: Incompatibilities. (line 80)
29164 * F in constraint: Simple Constraints. (line 82)
29165 * fabs: Other Builtins. (line 6)
29166 * fabsf: Other Builtins. (line 6)
29167 * fabsl: Other Builtins. (line 6)
29168 * fatal signal: Bug Criteria. (line 9)
29169 * fdim: Other Builtins. (line 6)
29170 * fdimf: Other Builtins. (line 6)
29171 * fdiml: Other Builtins. (line 6)
29172 * FDL, GNU Free Documentation License: GNU Free Documentation License.
29174 * ffs: Other Builtins. (line 6)
29175 * file name suffix: Overall Options. (line 14)
29176 * file names: Link Options. (line 10)
29177 * flexible array members: Zero Length. (line 6)
29178 * float as function value type: Incompatibilities. (line 141)
29179 * floating point precision <1>: Disappointments. (line 68)
29180 * floating point precision: Optimize Options. (line 960)
29181 * floor: Other Builtins. (line 6)
29182 * floorf: Other Builtins. (line 6)
29183 * floorl: Other Builtins. (line 6)
29184 * fma: Other Builtins. (line 6)
29185 * fmaf: Other Builtins. (line 6)
29186 * fmal: Other Builtins. (line 6)
29187 * fmax: Other Builtins. (line 6)
29188 * fmaxf: Other Builtins. (line 6)
29189 * fmaxl: Other Builtins. (line 6)
29190 * fmin: Other Builtins. (line 6)
29191 * fminf: Other Builtins. (line 6)
29192 * fminl: Other Builtins. (line 6)
29193 * fmod: Other Builtins. (line 6)
29194 * fmodf: Other Builtins. (line 6)
29195 * fmodl: Other Builtins. (line 6)
29196 * format function attribute: Function Attributes. (line 228)
29197 * format_arg function attribute: Function Attributes. (line 281)
29198 * Fortran: G++ and GCC. (line 6)
29199 * forwarding calls: Constructing Calls. (line 6)
29200 * fprintf: Other Builtins. (line 6)
29201 * fprintf_unlocked: Other Builtins. (line 6)
29202 * fputs: Other Builtins. (line 6)
29203 * fputs_unlocked: Other Builtins. (line 6)
29204 * freestanding environment: Standards. (line 6)
29205 * freestanding implementation: Standards. (line 6)
29206 * frexp: Other Builtins. (line 6)
29207 * frexpf: Other Builtins. (line 6)
29208 * frexpl: Other Builtins. (line 6)
29209 * FRV Options: FRV Options. (line 6)
29210 * fscanf: Other Builtins. (line 6)
29211 * fscanf, and constant strings: Incompatibilities. (line 17)
29212 * function addressability on the M32R/D: Function Attributes. (line 396)
29213 * function attributes: Function Attributes. (line 6)
29214 * function pointers, arithmetic: Pointer Arith. (line 6)
29215 * function prototype declarations: Function Prototypes. (line 6)
29216 * function without a prologue/epilogue code: Function Attributes.
29218 * function, size of pointer to: Pointer Arith. (line 6)
29219 * functions called via pointer on the RS/6000 and PowerPC: Function Attributes.
29221 * functions in arbitrary sections: Function Attributes. (line 6)
29222 * functions that are passed arguments in registers on the 386: Function Attributes.
29224 * functions that behave like malloc: Function Attributes. (line 6)
29225 * functions that do not pop the argument stack on the 386: Function Attributes.
29227 * functions that do pop the argument stack on the 386: Function Attributes.
29229 * functions that have no side effects: Function Attributes. (line 6)
29230 * functions that never return: Function Attributes. (line 6)
29231 * functions that pop the argument stack on the 386: Function Attributes.
29233 * functions which do not handle memory bank switching on 68HC11/68HC12: Function Attributes.
29235 * functions which handle memory bank switching: Function Attributes.
29237 * functions with non-null pointer arguments: Function Attributes.
29239 * functions with printf, scanf, strftime or strfmon style arguments: Function Attributes.
29241 * g in constraint: Simple Constraints. (line 108)
29242 * G in constraint: Simple Constraints. (line 86)
29243 * g++: Invoking G++. (line 13)
29244 * G++: G++ and GCC. (line 30)
29245 * gamma: Other Builtins. (line 6)
29246 * gammaf: Other Builtins. (line 6)
29247 * gammal: Other Builtins. (line 6)
29248 * GCC: G++ and GCC. (line 6)
29249 * GCC command options: Invoking GCC. (line 6)
29250 * GCC_EXEC_PREFIX: Environment Variables.
29252 * gcc_struct: Type Attributes. (line 288)
29253 * gcc_struct attribute: Variable Attributes. (line 296)
29254 * gettext: Other Builtins. (line 6)
29255 * global offset table: Code Gen Options. (line 170)
29256 * global register after longjmp: Global Reg Vars. (line 66)
29257 * global register variables: Global Reg Vars. (line 6)
29258 * GNAT: G++ and GCC. (line 30)
29259 * GNU C Compiler: G++ and GCC. (line 6)
29260 * GNU Compiler Collection: G++ and GCC. (line 6)
29261 * goto with computed label: Labels as Values. (line 6)
29262 * gp-relative references (MIPS): MIPS Options. (line 195)
29263 * gprof: Debugging Options. (line 127)
29264 * grouping options: Invoking GCC. (line 26)
29265 * H in constraint: Simple Constraints. (line 86)
29266 * hardware models and configurations, specifying: Submodel Options.
29268 * hex floats: Hex Floats. (line 6)
29269 * hosted environment <1>: C Dialect Options. (line 162)
29270 * hosted environment: Standards. (line 6)
29271 * hosted implementation: Standards. (line 6)
29272 * HPPA Options: HPPA Options. (line 6)
29273 * hypot: Other Builtins. (line 6)
29274 * hypotf: Other Builtins. (line 6)
29275 * hypotl: Other Builtins. (line 6)
29276 * I in constraint: Simple Constraints. (line 69)
29277 * i in constraint: Simple Constraints. (line 58)
29278 * i386 Options: i386 and x86-64 Options.
29280 * IA-64 Options: IA-64 Options. (line 6)
29281 * IBM RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
29283 * identifier names, dollar signs in: Dollar Signs. (line 6)
29284 * identifiers, names in assembler code: Asm Labels. (line 6)
29285 * ilogb: Other Builtins. (line 6)
29286 * ilogbf: Other Builtins. (line 6)
29287 * ilogbl: Other Builtins. (line 6)
29288 * imaxabs: Other Builtins. (line 6)
29289 * implementation-defined behavior, C language: C Implementation.
29291 * implied #pragma implementation: C++ Interface. (line 46)
29292 * incompatibilities of GCC: Incompatibilities. (line 6)
29293 * increment operators: Bug Criteria. (line 17)
29294 * index: Other Builtins. (line 6)
29295 * indirect calls on ARM: Function Attributes. (line 365)
29296 * init_priority attribute: C++ Attributes. (line 9)
29297 * initializations in expressions: Compound Literals. (line 6)
29298 * initializers with labeled elements: Designated Inits. (line 6)
29299 * initializers, non-constant: Initializers. (line 6)
29300 * inline automatic for C++ member fns: Inline. (line 46)
29301 * inline functions: Inline. (line 6)
29302 * inline functions, omission of: Inline. (line 51)
29303 * inlining and C++ pragmas: C++ Interface. (line 66)
29304 * installation trouble: Trouble. (line 6)
29305 * integrating function code: Inline. (line 6)
29306 * Intel 386 Options: i386 and x86-64 Options.
29308 * interface and implementation headers, C++: C++ Interface. (line 6)
29309 * intermediate C version, nonexistent: G++ and GCC. (line 35)
29310 * interrupt handler functions: Function Attributes. (line 331)
29311 * interrupt handler functions on the Blackfin, m68k, H8/300 and SH processors: Function Attributes.
29313 * introduction: Top. (line 6)
29314 * invalid assembly code: Bug Criteria. (line 12)
29315 * invalid input: Bug Criteria. (line 42)
29316 * invoking g++: Invoking G++. (line 23)
29317 * isalnum: Other Builtins. (line 6)
29318 * isalpha: Other Builtins. (line 6)
29319 * isascii: Other Builtins. (line 6)
29320 * isblank: Other Builtins. (line 6)
29321 * iscntrl: Other Builtins. (line 6)
29322 * isdigit: Other Builtins. (line 6)
29323 * isgraph: Other Builtins. (line 6)
29324 * islower: Other Builtins. (line 6)
29325 * ISO 9899: Standards. (line 6)
29326 * ISO C: Standards. (line 6)
29327 * ISO C standard: Standards. (line 6)
29328 * ISO C90: Standards. (line 6)
29329 * ISO C94: Standards. (line 6)
29330 * ISO C95: Standards. (line 6)
29331 * ISO C99: Standards. (line 6)
29332 * ISO C9X: Standards. (line 6)
29333 * ISO support: C Dialect Options. (line 10)
29334 * ISO/IEC 9899: Standards. (line 6)
29335 * isprint: Other Builtins. (line 6)
29336 * ispunct: Other Builtins. (line 6)
29337 * isspace: Other Builtins. (line 6)
29338 * isupper: Other Builtins. (line 6)
29339 * iswalnum: Other Builtins. (line 6)
29340 * iswalpha: Other Builtins. (line 6)
29341 * iswblank: Other Builtins. (line 6)
29342 * iswcntrl: Other Builtins. (line 6)
29343 * iswdigit: Other Builtins. (line 6)
29344 * iswgraph: Other Builtins. (line 6)
29345 * iswlower: Other Builtins. (line 6)
29346 * iswprint: Other Builtins. (line 6)
29347 * iswpunct: Other Builtins. (line 6)
29348 * iswspace: Other Builtins. (line 6)
29349 * iswupper: Other Builtins. (line 6)
29350 * iswxdigit: Other Builtins. (line 6)
29351 * isxdigit: Other Builtins. (line 6)
29352 * j0: Other Builtins. (line 6)
29353 * j0f: Other Builtins. (line 6)
29354 * j0l: Other Builtins. (line 6)
29355 * j1: Other Builtins. (line 6)
29356 * j1f: Other Builtins. (line 6)
29357 * j1l: Other Builtins. (line 6)
29358 * Java: G++ and GCC. (line 6)
29359 * java_interface attribute: C++ Attributes. (line 29)
29360 * jn: Other Builtins. (line 6)
29361 * jnf: Other Builtins. (line 6)
29362 * jnl: Other Builtins. (line 6)
29363 * keywords, alternate: Alternate Keywords. (line 6)
29364 * known causes of trouble: Trouble. (line 6)
29365 * labeled elements in initializers: Designated Inits. (line 6)
29366 * labels as values: Labels as Values. (line 6)
29367 * labs: Other Builtins. (line 6)
29368 * LANG: Environment Variables.
29370 * language dialect options: C Dialect Options. (line 6)
29371 * LC_ALL: Environment Variables.
29373 * LC_CTYPE: Environment Variables.
29375 * LC_MESSAGES: Environment Variables.
29377 * ldexp: Other Builtins. (line 6)
29378 * ldexpf: Other Builtins. (line 6)
29379 * ldexpl: Other Builtins. (line 6)
29380 * length-zero arrays: Zero Length. (line 6)
29381 * lgamma: Other Builtins. (line 6)
29382 * lgammaf: Other Builtins. (line 6)
29383 * lgammal: Other Builtins. (line 6)
29384 * Libraries: Link Options. (line 24)
29385 * LIBRARY_PATH: Environment Variables.
29387 * link options: Link Options. (line 6)
29388 * LL integer suffix: Long Long. (line 6)
29389 * llabs: Other Builtins. (line 6)
29390 * llrint: Other Builtins. (line 6)
29391 * llrintf: Other Builtins. (line 6)
29392 * llrintl: Other Builtins. (line 6)
29393 * llround: Other Builtins. (line 6)
29394 * llroundf: Other Builtins. (line 6)
29395 * llroundl: Other Builtins. (line 6)
29396 * load address instruction: Simple Constraints. (line 142)
29397 * local labels: Local Labels. (line 6)
29398 * local variables in macros: Typeof. (line 42)
29399 * local variables, specifying registers: Local Reg Vars. (line 6)
29400 * locale: Environment Variables.
29402 * locale definition: Environment Variables.
29404 * log: Other Builtins. (line 6)
29405 * log10: Other Builtins. (line 6)
29406 * log10f: Other Builtins. (line 6)
29407 * log10l: Other Builtins. (line 6)
29408 * log1p: Other Builtins. (line 6)
29409 * log1pf: Other Builtins. (line 6)
29410 * log1pl: Other Builtins. (line 6)
29411 * log2: Other Builtins. (line 6)
29412 * log2f: Other Builtins. (line 6)
29413 * log2l: Other Builtins. (line 6)
29414 * logb: Other Builtins. (line 6)
29415 * logbf: Other Builtins. (line 6)
29416 * logbl: Other Builtins. (line 6)
29417 * logf: Other Builtins. (line 6)
29418 * logl: Other Builtins. (line 6)
29419 * long long data types: Long Long. (line 6)
29420 * longjmp: Global Reg Vars. (line 66)
29421 * longjmp incompatibilities: Incompatibilities. (line 39)
29422 * longjmp warnings: Warning Options. (line 459)
29423 * lrint: Other Builtins. (line 6)
29424 * lrintf: Other Builtins. (line 6)
29425 * lrintl: Other Builtins. (line 6)
29426 * lround: Other Builtins. (line 6)
29427 * lroundf: Other Builtins. (line 6)
29428 * lroundl: Other Builtins. (line 6)
29429 * m in constraint: Simple Constraints. (line 17)
29430 * M32R/D options: M32R/D Options. (line 6)
29431 * M680x0 options: M680x0 Options. (line 6)
29432 * M68hc1x options: M68hc1x Options. (line 6)
29433 * machine dependent options: Submodel Options. (line 6)
29434 * machine specific constraints: Machine Constraints. (line 6)
29435 * macro with variable arguments: Variadic Macros. (line 6)
29436 * macros containing asm: Extended Asm. (line 239)
29437 * macros, inline alternative: Inline. (line 6)
29438 * macros, local labels: Local Labels. (line 6)
29439 * macros, local variables in: Typeof. (line 42)
29440 * macros, statements in expressions: Statement Exprs. (line 6)
29441 * macros, types of arguments: Typeof. (line 6)
29442 * make: Preprocessor Options.
29444 * malloc: Other Builtins. (line 6)
29445 * malloc attribute: Function Attributes. (line 386)
29446 * matching constraint: Simple Constraints. (line 127)
29447 * MCore options: MCore Options. (line 6)
29448 * member fns, automatically inline: Inline. (line 46)
29449 * memcmp: Other Builtins. (line 6)
29450 * memcpy: Other Builtins. (line 6)
29451 * memory references in constraints: Simple Constraints. (line 17)
29452 * mempcpy: Other Builtins. (line 6)
29453 * memset: Other Builtins. (line 6)
29454 * Mercury: G++ and GCC. (line 23)
29455 * message formatting: Language Independent Options.
29457 * messages, warning: Warning Options. (line 6)
29458 * messages, warning and error: Warnings and Errors. (line 6)
29459 * middle-operands, omitted: Conditionals. (line 6)
29460 * MIPS options: MIPS Options. (line 6)
29461 * misunderstandings in C++: C++ Misunderstandings.
29463 * mixed declarations and code: Mixed Declarations. (line 6)
29464 * mktemp, and constant strings: Incompatibilities. (line 13)
29465 * MMIX Options: MMIX Options. (line 6)
29466 * MN10300 options: MN10300 Options. (line 6)
29467 * mode attribute: Variable Attributes. (line 118)
29468 * modf: Other Builtins. (line 6)
29469 * modff: Other Builtins. (line 6)
29470 * modfl: Other Builtins. (line 6)
29471 * modifiers in constraints: Modifiers. (line 6)
29472 * ms_struct: Type Attributes. (line 288)
29473 * ms_struct attribute: Variable Attributes. (line 296)
29474 * mudflap: Optimize Options. (line 319)
29475 * multiple alternative constraints: Multi-Alternative. (line 6)
29476 * multiprecision arithmetic: Long Long. (line 6)
29477 * n in constraint: Simple Constraints. (line 63)
29478 * names used in assembler code: Asm Labels. (line 6)
29479 * naming convention, implementation headers: C++ Interface. (line 46)
29480 * nearbyint: Other Builtins. (line 6)
29481 * nearbyintf: Other Builtins. (line 6)
29482 * nearbyintl: Other Builtins. (line 6)
29483 * nested functions: Nested Functions. (line 6)
29484 * newlines (escaped): Escaped Newlines. (line 6)
29485 * nextafter: Other Builtins. (line 6)
29486 * nextafterf: Other Builtins. (line 6)
29487 * nextafterl: Other Builtins. (line 6)
29488 * nexttoward: Other Builtins. (line 6)
29489 * nexttowardf: Other Builtins. (line 6)
29490 * nexttowardl: Other Builtins. (line 6)
29491 * NMI handler functions on the Blackfin processor: Function Attributes.
29493 * no_instrument_function function attribute: Function Attributes.
29495 * nocommon attribute: Variable Attributes. (line 92)
29496 * noinline function attribute: Function Attributes. (line 452)
29497 * non-constant initializers: Initializers. (line 6)
29498 * non-static inline function: Inline. (line 63)
29499 * nonnull function attribute: Function Attributes. (line 456)
29500 * noreturn function attribute: Function Attributes. (line 479)
29501 * nothrow function attribute: Function Attributes. (line 521)
29502 * NS32K options: NS32K Options. (line 6)
29503 * o in constraint: Simple Constraints. (line 21)
29504 * OBJC_INCLUDE_PATH: Environment Variables.
29506 * Objective-C <1>: Standards. (line 110)
29507 * Objective-C: G++ and GCC. (line 6)
29508 * Objective-C and Objective-C++ options, command line: Objective-C and Objective-C++ Dialect Options.
29510 * Objective-C++ <1>: Standards. (line 110)
29511 * Objective-C++: G++ and GCC. (line 6)
29512 * offsettable address: Simple Constraints. (line 21)
29513 * old-style function definitions: Function Prototypes. (line 6)
29514 * omitted middle-operands: Conditionals. (line 6)
29515 * open coding: Inline. (line 6)
29516 * operand constraints, asm: Constraints. (line 6)
29517 * optimize options: Optimize Options. (line 6)
29518 * options to control diagnostics formatting: Language Independent Options.
29520 * options to control warnings: Warning Options. (line 6)
29521 * options, C++: C++ Dialect Options. (line 6)
29522 * options, code generation: Code Gen Options. (line 6)
29523 * options, debugging: Debugging Options. (line 6)
29524 * options, dialect: C Dialect Options. (line 6)
29525 * options, directory search: Directory Options. (line 6)
29526 * options, GCC command: Invoking GCC. (line 6)
29527 * options, grouping: Invoking GCC. (line 26)
29528 * options, linking: Link Options. (line 6)
29529 * options, Objective-C and Objective-C++: Objective-C and Objective-C++ Dialect Options.
29531 * options, optimization: Optimize Options. (line 6)
29532 * options, order: Invoking GCC. (line 30)
29533 * options, preprocessor: Preprocessor Options.
29535 * order of evaluation, side effects: Non-bugs. (line 199)
29536 * order of options: Invoking GCC. (line 30)
29537 * other register constraints: Simple Constraints. (line 151)
29538 * output file option: Overall Options. (line 166)
29539 * overloaded virtual fn, warning: C++ Dialect Options. (line 379)
29540 * p in constraint: Simple Constraints. (line 142)
29541 * packed attribute: Variable Attributes. (line 129)
29542 * parameter forward declaration: Variable Length. (line 60)
29543 * parameters, aliased: Code Gen Options. (line 325)
29544 * Pascal: G++ and GCC. (line 23)
29545 * PDP-11 Options: PDP-11 Options. (line 6)
29546 * PIC: Code Gen Options. (line 170)
29547 * pmf: Bound member functions.
29549 * pointer arguments: Function Attributes. (line 60)
29550 * pointer to member function: Bound member functions.
29552 * portions of temporary objects, pointers to: Temporaries. (line 6)
29553 * pow: Other Builtins. (line 6)
29554 * pow10: Other Builtins. (line 6)
29555 * pow10f: Other Builtins. (line 6)
29556 * pow10l: Other Builtins. (line 6)
29557 * PowerPC options: PowerPC Options. (line 6)
29558 * powf: Other Builtins. (line 6)
29559 * powl: Other Builtins. (line 6)
29560 * pragma, align: Solaris Pragmas. (line 11)
29561 * pragma, extern_prefix: Symbol-Renaming Pragmas.
29563 * pragma, fini: Solaris Pragmas. (line 19)
29564 * pragma, init: Solaris Pragmas. (line 24)
29565 * pragma, long_calls: ARM Pragmas. (line 11)
29566 * pragma, long_calls_off: ARM Pragmas. (line 17)
29567 * pragma, longcall: RS/6000 and PowerPC Pragmas.
29569 * pragma, mark: Darwin Pragmas. (line 11)
29570 * pragma, no_long_calls: ARM Pragmas. (line 14)
29571 * pragma, options align: Darwin Pragmas. (line 14)
29572 * pragma, reason for not using: Function Attributes. (line 736)
29573 * pragma, redefine_extname: Symbol-Renaming Pragmas.
29575 * pragma, segment: Darwin Pragmas. (line 21)
29576 * pragma, unused: Darwin Pragmas. (line 24)
29577 * pragma, weak: Weak Pragmas. (line 10)
29578 * pragmas: Pragmas. (line 6)
29579 * pragmas in C++, effect on inlining: C++ Interface. (line 66)
29580 * pragmas, interface and implementation: C++ Interface. (line 6)
29581 * pragmas, warning of unknown: Warning Options. (line 476)
29582 * precompiled headers: Precompiled Headers. (line 6)
29583 * preprocessing numbers: Incompatibilities. (line 173)
29584 * preprocessing tokens: Incompatibilities. (line 173)
29585 * preprocessor options: Preprocessor Options.
29587 * printf: Other Builtins. (line 6)
29588 * printf_unlocked: Other Builtins. (line 6)
29589 * prof: Debugging Options. (line 121)
29590 * promotion of formal parameters: Function Prototypes. (line 6)
29591 * pure function attribute: Function Attributes. (line 529)
29592 * push address instruction: Simple Constraints. (line 142)
29593 * putchar: Other Builtins. (line 6)
29594 * puts: Other Builtins. (line 6)
29595 * qsort, and global register variables: Global Reg Vars. (line 42)
29596 * question mark: Multi-Alternative. (line 27)
29597 * r in constraint: Simple Constraints. (line 54)
29598 * ranges in case statements: Case Ranges. (line 6)
29599 * read-only strings: Incompatibilities. (line 9)
29600 * register variable after longjmp: Global Reg Vars. (line 66)
29601 * registers: Extended Asm. (line 6)
29602 * registers for local variables: Local Reg Vars. (line 6)
29603 * registers in constraints: Simple Constraints. (line 54)
29604 * registers, global allocation: Explicit Reg Vars. (line 6)
29605 * registers, global variables in: Global Reg Vars. (line 6)
29606 * regparm attribute: Function Attributes. (line 551)
29607 * relocation truncated to fit (MIPS): MIPS Options. (line 111)
29608 * remainder: Other Builtins. (line 6)
29609 * remainderf: Other Builtins. (line 6)
29610 * remainderl: Other Builtins. (line 6)
29611 * remquo: Other Builtins. (line 6)
29612 * remquof: Other Builtins. (line 6)
29613 * remquol: Other Builtins. (line 6)
29614 * reordering, warning: C++ Dialect Options. (line 300)
29615 * reporting bugs: Bugs. (line 6)
29616 * rest argument (in macro): Variadic Macros. (line 6)
29617 * restricted pointers: Restricted Pointers. (line 6)
29618 * restricted references: Restricted Pointers. (line 6)
29619 * restricted this pointer: Restricted Pointers. (line 6)
29620 * rindex: Other Builtins. (line 6)
29621 * rint: Other Builtins. (line 6)
29622 * rintf: Other Builtins. (line 6)
29623 * rintl: Other Builtins. (line 6)
29624 * round: Other Builtins. (line 6)
29625 * roundf: Other Builtins. (line 6)
29626 * roundl: Other Builtins. (line 6)
29627 * RS/6000 and PowerPC Options: RS/6000 and PowerPC Options.
29629 * RTTI: Vague Linkage. (line 43)
29630 * run-time options: Code Gen Options. (line 6)
29631 * s in constraint: Simple Constraints. (line 90)
29632 * S/390 and zSeries Options: S/390 and zSeries Options.
29634 * save all registers on the Blackfin, H8/300, H8/300H, and H8S: Function Attributes.
29636 * scalb: Other Builtins. (line 6)
29637 * scalbf: Other Builtins. (line 6)
29638 * scalbl: Other Builtins. (line 6)
29639 * scalbln: Other Builtins. (line 6)
29640 * scalblnf: Other Builtins. (line 6)
29641 * scalbn: Other Builtins. (line 6)
29642 * scalbnf: Other Builtins. (line 6)
29643 * scanf, and constant strings: Incompatibilities. (line 17)
29644 * scanfnl: Other Builtins. (line 6)
29645 * scope of a variable length array: Variable Length. (line 23)
29646 * scope of declaration: Disappointments. (line 21)
29647 * scope of external declarations: Incompatibilities. (line 80)
29648 * search path: Directory Options. (line 6)
29649 * section function attribute: Function Attributes. (line 573)
29650 * section variable attribute: Variable Attributes. (line 144)
29651 * sentinel function attribute: Function Attributes. (line 589)
29652 * setjmp: Global Reg Vars. (line 66)
29653 * setjmp incompatibilities: Incompatibilities. (line 39)
29654 * shared strings: Incompatibilities. (line 9)
29655 * shared variable attribute: Variable Attributes. (line 189)
29656 * side effect in ?:: Conditionals. (line 20)
29657 * side effects, macro argument: Statement Exprs. (line 35)
29658 * side effects, order of evaluation: Non-bugs. (line 199)
29659 * signal handler functions on the AVR processors: Function Attributes.
29661 * signbit: Other Builtins. (line 6)
29662 * signbitf: Other Builtins. (line 6)
29663 * signbitl: Other Builtins. (line 6)
29664 * signed and unsigned values, comparison warning: Warning Options.
29666 * significand: Other Builtins. (line 6)
29667 * significandf: Other Builtins. (line 6)
29668 * significandl: Other Builtins. (line 6)
29669 * simple constraints: Simple Constraints. (line 6)
29670 * sin: Other Builtins. (line 6)
29671 * sincos: Other Builtins. (line 6)
29672 * sincosf: Other Builtins. (line 6)
29673 * sincosl: Other Builtins. (line 6)
29674 * sinf: Other Builtins. (line 6)
29675 * sinh: Other Builtins. (line 6)
29676 * sinhf: Other Builtins. (line 6)
29677 * sinhl: Other Builtins. (line 6)
29678 * sinl: Other Builtins. (line 6)
29679 * sizeof: Typeof. (line 6)
29680 * smaller data references: M32R/D Options. (line 57)
29681 * smaller data references (MIPS): MIPS Options. (line 195)
29682 * smaller data references (PowerPC): RS/6000 and PowerPC Options.
29684 * snprintf: Other Builtins. (line 6)
29685 * SPARC options: SPARC Options. (line 6)
29686 * Spec Files: Spec Files. (line 6)
29687 * specified registers: Explicit Reg Vars. (line 6)
29688 * specifying compiler version and target machine: Target Options.
29690 * specifying hardware config: Submodel Options. (line 6)
29691 * specifying machine version: Target Options. (line 6)
29692 * specifying registers for local variables: Local Reg Vars. (line 6)
29693 * speed of compilation: Precompiled Headers. (line 6)
29694 * sprintf: Other Builtins. (line 6)
29695 * sqrt: Other Builtins. (line 6)
29696 * sqrtf: Other Builtins. (line 6)
29697 * sqrtl: Other Builtins. (line 6)
29698 * sscanf: Other Builtins. (line 6)
29699 * sscanf, and constant strings: Incompatibilities. (line 17)
29700 * statements inside expressions: Statement Exprs. (line 6)
29701 * static data in C++, declaring and defining: Static Definitions.
29703 * stpcpy: Other Builtins. (line 6)
29704 * strcat: Other Builtins. (line 6)
29705 * strchr: Other Builtins. (line 6)
29706 * strcmp: Other Builtins. (line 6)
29707 * strcpy: Other Builtins. (line 6)
29708 * strcspn: Other Builtins. (line 6)
29709 * strdup: Other Builtins. (line 6)
29710 * strfmon: Other Builtins. (line 6)
29711 * strftime: Other Builtins. (line 6)
29712 * string constants: Incompatibilities. (line 9)
29713 * strlen: Other Builtins. (line 6)
29714 * strncat: Other Builtins. (line 6)
29715 * strncmp: Other Builtins. (line 6)
29716 * strncpy: Other Builtins. (line 6)
29717 * strpbrk: Other Builtins. (line 6)
29718 * strrchr: Other Builtins. (line 6)
29719 * strspn: Other Builtins. (line 6)
29720 * strstr: Other Builtins. (line 6)
29721 * struct: Unnamed Fields. (line 6)
29722 * structures: Incompatibilities. (line 146)
29723 * structures, constructor expression: Compound Literals. (line 6)
29724 * submodel options: Submodel Options. (line 6)
29725 * subscripting: Subscripting. (line 6)
29726 * subscripting and function values: Subscripting. (line 6)
29727 * suffixes for C++ source: Invoking G++. (line 6)
29728 * SUNPRO_DEPENDENCIES: Environment Variables.
29730 * suppressing warnings: Warning Options. (line 6)
29731 * surprises in C++: C++ Misunderstandings.
29733 * syntax checking: Warning Options. (line 22)
29734 * system headers, warnings from: Warning Options. (line 590)
29735 * tan: Other Builtins. (line 6)
29736 * tanf: Other Builtins. (line 6)
29737 * tanh: Other Builtins. (line 6)
29738 * tanhf: Other Builtins. (line 6)
29739 * tanhl: Other Builtins. (line 6)
29740 * tanl: Other Builtins. (line 6)
29741 * target machine, specifying: Target Options. (line 6)
29742 * target options: Target Options. (line 6)
29743 * TC1: Standards. (line 6)
29744 * TC2: Standards. (line 6)
29745 * Technical Corrigenda: Standards. (line 6)
29746 * Technical Corrigendum 1: Standards. (line 6)
29747 * Technical Corrigendum 2: Standards. (line 6)
29748 * template instantiation: Template Instantiation.
29750 * temporaries, lifetime of: Temporaries. (line 6)
29751 * tgamma: Other Builtins. (line 6)
29752 * tgammaf: Other Builtins. (line 6)
29753 * tgammal: Other Builtins. (line 6)
29754 * Thread-Local Storage: Thread-Local. (line 6)
29755 * thunks: Nested Functions. (line 6)
29756 * tiny data section on the H8/300H and H8S: Function Attributes.
29758 * TLS: Thread-Local. (line 6)
29759 * tls_model attribute: Variable Attributes. (line 213)
29760 * TMPDIR: Environment Variables.
29762 * TMS320C3x/C4x Options: TMS320C3x/C4x Options.
29764 * toascii: Other Builtins. (line 6)
29765 * tolower: Other Builtins. (line 6)
29766 * toupper: Other Builtins. (line 6)
29767 * towlower: Other Builtins. (line 6)
29768 * towupper: Other Builtins. (line 6)
29769 * traditional C language: C Dialect Options. (line 200)
29770 * treelang <1>: Standards. (line 123)
29771 * treelang: G++ and GCC. (line 6)
29772 * trunc: Other Builtins. (line 6)
29773 * truncf: Other Builtins. (line 6)
29774 * truncl: Other Builtins. (line 6)
29775 * two-stage name lookup: Name lookup. (line 6)
29776 * type alignment: Alignment. (line 6)
29777 * type attributes: Type Attributes. (line 6)
29778 * type_info: Vague Linkage. (line 43)
29779 * typedef names as function parameters: Incompatibilities. (line 97)
29780 * typeof: Typeof. (line 6)
29781 * ULL integer suffix: Long Long. (line 6)
29782 * Ultrix calling convention: Interoperation. (line 150)
29783 * undefined behavior: Bug Criteria. (line 17)
29784 * undefined function value: Bug Criteria. (line 17)
29785 * underscores in variables in macros: Typeof. (line 42)
29786 * union: Unnamed Fields. (line 6)
29787 * union, casting to a: Cast to Union. (line 6)
29788 * unions: Incompatibilities. (line 146)
29789 * unknown pragmas, warning: Warning Options. (line 476)
29790 * unresolved references and -nodefaultlibs: Link Options. (line 79)
29791 * unresolved references and -nostdlib: Link Options. (line 79)
29792 * unused attribute.: Function Attributes. (line 654)
29793 * used attribute.: Function Attributes. (line 659)
29794 * User stack pointer in interrupts on the Blackfin: Function Attributes.
29796 * V in constraint: Simple Constraints. (line 41)
29797 * V850 Options: V850 Options. (line 6)
29798 * vague linkage: Vague Linkage. (line 6)
29799 * value after longjmp: Global Reg Vars. (line 66)
29800 * variable addressability on the IA-64: Function Attributes. (line 396)
29801 * variable addressability on the M32R/D: Variable Attributes. (line 277)
29802 * variable alignment: Alignment. (line 6)
29803 * variable attributes: Variable Attributes. (line 6)
29804 * variable number of arguments: Variadic Macros. (line 6)
29805 * variable-length array scope: Variable Length. (line 23)
29806 * variable-length arrays: Variable Length. (line 6)
29807 * variables in specified registers: Explicit Reg Vars. (line 6)
29808 * variables, local, in macros: Typeof. (line 42)
29809 * variadic macros: Variadic Macros. (line 6)
29810 * VAX calling convention: Interoperation. (line 150)
29811 * VAX options: VAX Options. (line 6)
29812 * vfprintf: Other Builtins. (line 6)
29813 * vfscanf: Other Builtins. (line 6)
29814 * visibility attribute: Function Attributes. (line 665)
29815 * VLAs: Variable Length. (line 6)
29816 * void pointers, arithmetic: Pointer Arith. (line 6)
29817 * void, size of pointer to: Pointer Arith. (line 6)
29818 * volatile access: Volatiles. (line 6)
29819 * volatile applied to function: Function Attributes. (line 6)
29820 * volatile read: Volatiles. (line 6)
29821 * volatile write: Volatiles. (line 6)
29822 * vprintf: Other Builtins. (line 6)
29823 * vscanf: Other Builtins. (line 6)
29824 * vsnprintf: Other Builtins. (line 6)
29825 * vsprintf: Other Builtins. (line 6)
29826 * vsscanf: Other Builtins. (line 6)
29827 * vtable: Vague Linkage. (line 28)
29828 * warn_unused_result attribute: Function Attributes. (line 707)
29829 * warning for comparison of signed and unsigned values: Warning Options.
29831 * warning for overloaded virtual fn: C++ Dialect Options. (line 379)
29832 * warning for reordering of member initializers: C++ Dialect Options.
29834 * warning for unknown pragmas: Warning Options. (line 476)
29835 * warning messages: Warning Options. (line 6)
29836 * warnings from system headers: Warning Options. (line 590)
29837 * warnings vs errors: Warnings and Errors. (line 6)
29838 * weak attribute: Function Attributes. (line 724)
29839 * whitespace: Incompatibilities. (line 112)
29840 * X in constraint: Simple Constraints. (line 112)
29841 * X3.159-1989: Standards. (line 6)
29842 * x86-64 options: x86-64 Options. (line 6)
29843 * x86-64 Options: i386 and x86-64 Options.
29845 * Xstormy16 Options: Xstormy16 Options. (line 6)
29846 * Xtensa Options: Xtensa Options. (line 6)
29847 * y0: Other Builtins. (line 6)
29848 * y0f: Other Builtins. (line 6)
29849 * y0l: Other Builtins. (line 6)
29850 * y1: Other Builtins. (line 6)
29851 * y1f: Other Builtins. (line 6)
29852 * y1l: Other Builtins. (line 6)
29853 * yn: Other Builtins. (line 6)
29854 * ynf: Other Builtins. (line 6)
29855 * ynl: Other Builtins. (line 6)
29856 * zero-length arrays: Zero Length. (line 6)
29857 * zero-size structures: Empty Structures. (line 6)
29858 * zSeries options: zSeries Options. (line 6)
29864 Node: G++ and GCC
\7f3740
29865 Node: Standards
\7f5805
29866 Node: Invoking GCC
\7f12932
29867 Node: Option Summary
\7f16702
29868 Node: Overall Options
\7f42898
29869 Node: Invoking G++
\7f51200
29870 Node: C Dialect Options
\7f52822
29871 Node: C++ Dialect Options
\7f63832
29872 Node: Objective-C and Objective-C++ Dialect Options
\7f81422
29873 Node: Language Independent Options
\7f90872
29874 Node: Warning Options
\7f92683
29875 Node: Debugging Options
\7f134197
29876 Node: Optimize Options
\7f163462
29877 Node: Preprocessor Options
\7f232241
29878 Ref: Wtrigraphs
\7f236205
29879 Ref: dashMF
\7f240962
29880 Ref: fdollars-in-identifiers
\7f249858
29881 Node: Assembler Options
\7f257719
29882 Node: Link Options
\7f258424
29883 Ref: Link Options-Footnote-1
\7f266676
29884 Node: Directory Options
\7f267010
29885 Node: Spec Files
\7f272377
29886 Node: Target Options
\7f291665
29887 Node: Submodel Options
\7f292971
29888 Node: ARC Options
\7f294720
29889 Node: ARM Options
\7f295910
29890 Node: AVR Options
\7f307118
29891 Node: Blackfin Options
\7f309251
29892 Node: CRIS Options
\7f311017
29893 Node: Darwin Options
\7f315239
29894 Node: DEC Alpha Options
\7f321552
29895 Node: DEC Alpha/VMS Options
\7f333030
29896 Node: FRV Options
\7f333415
29897 Node: H8/300 Options
\7f339799
29898 Node: HPPA Options
\7f340860
29899 Node: i386 and x86-64 Options
\7f350239
29900 Node: IA-64 Options
\7f367992
29901 Node: M32R/D Options
\7f371990
29902 Node: M680x0 Options
\7f375578
29903 Node: M68hc1x Options
\7f382811
29904 Node: MCore Options
\7f384379
29905 Node: MIPS Options
\7f385400
29906 Node: MMIX Options
\7f399462
29907 Node: MN10300 Options
\7f401944
29908 Node: NS32K Options
\7f403004
29909 Node: PDP-11 Options
\7f407549
29910 Node: PowerPC Options
\7f409386
29911 Node: RS/6000 and PowerPC Options
\7f409620
29912 Node: S/390 and zSeries Options
\7f435450
29913 Node: SH Options
\7f442498
29914 Node: SPARC Options
\7f445760
29915 Node: System V Options
\7f456391
29916 Node: TMS320C3x/C4x Options
\7f457225
29917 Node: V850 Options
\7f462750
29918 Node: VAX Options
\7f465895
29919 Node: x86-64 Options
\7f466442
29920 Node: Xstormy16 Options
\7f466656
29921 Node: Xtensa Options
\7f466945
29922 Node: zSeries Options
\7f470785
29923 Node: Code Gen Options
\7f470981
29924 Node: Environment Variables
\7f490240
29925 Node: Precompiled Headers
\7f497912
29926 Node: Running Protoize
\7f504463
29927 Node: C Implementation
\7f510800
29928 Node: Translation implementation
\7f512463
29929 Node: Environment implementation
\7f513037
29930 Node: Identifiers implementation
\7f513587
29931 Node: Characters implementation
\7f514641
29932 Node: Integers implementation
\7f517447
29933 Node: Floating point implementation
\7f519272
29934 Node: Arrays and pointers implementation
\7f522201
29935 Ref: Arrays and pointers implementation-Footnote-1
\7f523636
29936 Node: Hints implementation
\7f523760
29937 Node: Structures unions enumerations and bit-fields implementation
\7f525226
29938 Node: Qualifiers implementation
\7f527189
29939 Node: Declarators implementation
\7f527572
29940 Node: Statements implementation
\7f527914
29941 Node: Preprocessing directives implementation
\7f528241
29942 Node: Library functions implementation
\7f530346
29943 Node: Architecture implementation
\7f530986
29944 Node: Locale-specific behavior implementation
\7f531689
29945 Node: C Extensions
\7f531994
29946 Node: Statement Exprs
\7f536173
29947 Node: Local Labels
\7f540686
29948 Node: Labels as Values
\7f543665
29949 Ref: Labels as Values-Footnote-1
\7f545719
29950 Node: Nested Functions
\7f545902
29951 Node: Constructing Calls
\7f549796
29952 Node: Typeof
\7f552132
29953 Node: Conditionals
\7f555298
29954 Node: Long Long
\7f556189
29955 Node: Complex
\7f557690
29956 Node: Hex Floats
\7f560256
29957 Node: Zero Length
\7f561291
29958 Node: Empty Structures
\7f564568
29959 Node: Variable Length
\7f564984
29960 Node: Variadic Macros
\7f567751
29961 Node: Escaped Newlines
\7f570133
29962 Node: Subscripting
\7f570972
29963 Node: Pointer Arith
\7f571695
29964 Node: Initializers
\7f572263
29965 Node: Compound Literals
\7f572759
29966 Node: Designated Inits
\7f574921
29967 Node: Case Ranges
\7f578576
29968 Node: Cast to Union
\7f579259
29969 Node: Mixed Declarations
\7f580355
29970 Node: Function Attributes
\7f580861
29971 Node: Attribute Syntax
\7f615411
29972 Node: Function Prototypes
\7f626495
29973 Node: C++ Comments
\7f628276
29974 Node: Dollar Signs
\7f628795
29975 Node: Character Escapes
\7f629260
29976 Node: Alignment
\7f629554
29977 Node: Variable Attributes
\7f630871
29978 Node: Type Attributes
\7f644318
29979 Node: Inline
\7f657736
29980 Node: Extended Asm
\7f662440
29981 Ref: Example of asm with clobbered asm reg
\7f668526
29982 Node: Constraints
\7f682622
29983 Node: Simple Constraints
\7f683472
29984 Node: Multi-Alternative
\7f690000
29985 Node: Modifiers
\7f691717
29986 Node: Machine Constraints
\7f694405
29987 Node: Asm Labels
\7f718333
29988 Node: Explicit Reg Vars
\7f720009
29989 Node: Global Reg Vars
\7f721617
29990 Node: Local Reg Vars
\7f726167
29991 Node: Alternate Keywords
\7f728608
29992 Node: Incomplete Enums
\7f730036
29993 Node: Function Names
\7f730793
29994 Node: Return Address
\7f732983
29995 Node: Vector Extensions
\7f735780
29996 Node: Offsetof
\7f739282
29997 Node: Other Builtins
\7f740067
29998 Node: Target Builtins
\7f761172
29999 Node: Alpha Built-in Functions
\7f761841
30000 Node: ARM Built-in Functions
\7f764833
30001 Node: FR-V Built-in Functions
\7f771536
30002 Node: Argument Types
\7f772361
30003 Node: Directly-mapped Integer Functions
\7f774117
30004 Node: Directly-mapped Media Functions
\7f775199
30005 Node: Other Built-in Functions
\7f782231
30006 Node: X86 Built-in Functions
\7f783427
30007 Node: MIPS Paired-Single Support
\7f793819
30008 Node: Paired-Single Arithmetic
\7f795424
30009 Node: Paired-Single Built-in Functions
\7f796364
30010 Node: MIPS-3D Built-in Functions
\7f799028
30011 Node: PowerPC AltiVec Built-in Functions
\7f804397
30012 Node: SPARC VIS Built-in Functions
\7f905701
30013 Node: Target Format Checks
\7f907360
30014 Node: Solaris Format Checks
\7f907767
30015 Node: Pragmas
\7f908164
30016 Node: ARM Pragmas
\7f908731
30017 Node: RS/6000 and PowerPC Pragmas
\7f909349
30018 Node: Darwin Pragmas
\7f910090
30019 Node: Solaris Pragmas
\7f911157
30020 Node: Symbol-Renaming Pragmas
\7f912318
30021 Node: Structure-Packing Pragmas
\7f914940
30022 Node: Weak Pragmas
\7f916169
30023 Node: Unnamed Fields
\7f916944
30024 Node: Thread-Local
\7f918454
30025 Node: C99 Thread-Local Edits
\7f920538
30026 Node: C++98 Thread-Local Edits
\7f922550
30027 Node: C++ Extensions
\7f925995
30028 Node: Volatiles
\7f927567
30029 Node: Restricted Pointers
\7f930913
30030 Node: Vague Linkage
\7f932507
30031 Node: C++ Interface
\7f936163
30032 Ref: C++ Interface-Footnote-1
\7f940460
30033 Node: Template Instantiation
\7f940597
30034 Node: Bound member functions
\7f947609
30035 Node: C++ Attributes
\7f949152
30036 Node: Strong Using
\7f950792
30037 Node: Java Exceptions
\7f952041
30038 Node: Deprecated Features
\7f953437
30039 Node: Backwards Compatibility
\7f956416
30040 Node: Objective-C
\7f957771
30041 Node: Executing code before main
\7f958352
30042 Node: What you can and what you cannot do in +load
\7f960958
30043 Node: Type encoding
\7f963125
30044 Node: Garbage Collection
\7f966368
30045 Node: Constant string objects
\7f968992
30046 Node: compatibility_alias
\7f971500
30047 Node: Compatibility
\7f972378
30048 Node: Gcov
\7f978945
30049 Node: Gcov Intro
\7f979415
30050 Node: Invoking Gcov
\7f982131
30051 Node: Gcov and Optimization
\7f993982
30052 Node: Gcov Data Files
\7f996635
30053 Node: Trouble
\7f997749
30054 Node: Actual Bugs
\7f999289
30055 Node: Cross-Compiler Problems
\7f1000029
30056 Node: Interoperation
\7f1000443
30057 Node: Incompatibilities
\7f1008041
30058 Node: Fixed Headers
\7f1016191
30059 Node: Standard Libraries
\7f1017854
30060 Node: Disappointments
\7f1019226
30061 Node: C++ Misunderstandings
\7f1023584
30062 Node: Static Definitions
\7f1024403
30063 Node: Name lookup
\7f1025456
30064 Ref: Name lookup-Footnote-1
\7f1030234
30065 Node: Temporaries
\7f1030421
30066 Node: Copy Assignment
\7f1032397
30067 Node: Protoize Caveats
\7f1034204
30068 Node: Non-bugs
\7f1038166
30069 Node: Warnings and Errors
\7f1048789
30070 Node: Bugs
\7f1050553
30071 Node: Bug Criteria
\7f1051117
30072 Node: Bug Reporting
\7f1053327
30073 Node: Service
\7f1053719
30074 Node: Contributing
\7f1054538
30075 Node: Funding
\7f1055278
30076 Node: GNU Project
\7f1057767
30077 Node: Copying
\7f1058413
30078 Node: GNU Free Documentation License
\7f1077563
30079 Node: Contributors
\7f1099959
30080 Node: Option Index
\7f1130138
30081 Node: Keyword Index
\7f1255487